Untitled

Glow Discharge
Spectroscopies
MODERN ANALYTICAL CHEMISTRY
Series Editor: David M. Hereules
University 0/ Pittsburgh
ADVANCES IN COAL SPECTROSCOPY

Edited by Henk L. C. Meuzelaar
ANALYTICAL ATOMIC SPECTROSCOPY

William G. Schrenk
CHEMICAL DERIVATIZATION IN ANALYTICAL CHEMISTRY

Edited by R. W. Frei and J. F. Lawrence
Volume 1: Chromatography
Volume 2: Separation and Continuous Flow Techniques
COMPUTER-ENHANCED ANALYTICAL SPECTROSCOPY

Volume 1: Edited by Henk L. C. Meuzelaar and Thomas L. Isenhour
Volume 2: Edited by Henk L. C. Meuzelaar
Volume 3: Edited by Peter C. Jues
GLOW DISCHARGE SPECTROSCOPIES

Edited by R. Kenneth Marcus
ION CHROMATOGRAPHY
Hamish Small
ION-SELECTIVE ELECTRODES IN ANALYTICAL CHEMISTRY

Volumes 1 and 2
Edited by Henry Freiser
LIQUID CHROMATOGRAPHY/MASS SPECTROMETRY

Techniques and Applications
Alfred L. Yergey, Charles G. Edmonds, Ivor A. S. Lewis, and Marvin L. Vestal
MASS SPECTROMETRY: CLiNICAL AND BIOMEDICAL APPLICATIONS

Volume 1
Edited by Dominic M. Desiderio
PRINCIPLES OF CHEMICAL SENSORS

Jiri Janata
TRANSFORM TECHNIQUES IN CHEMISTRY

Edited by Peter R. Griffiths
A Continuation Order Plan is available for this series. A continuation order will bring delivery of
eacb new volume immediately upon publication. Volumes are billed only upon actual sbipment. For
furtber information please contact tbe publisber.
Glow Discharge
Spectroscopies
Edited by
R. Kenneth Marcus
Clemson University
Clemson, South Carolina
Springer Science+Business Media, LLC

L1brarll of Congr8ss Cata·loglng-ln-Publlcatlon Oata
Glow discharge spectroscoples I edlted by R. Kenneth Marcus.

p. cm. -- (Modern anallltical chemlstry)
Includes blbllographlcal references and Index.
1. Gas dlscharges. 2. Spectrum analysis. 3. Solid state physlcs.

4. Chemlstry. Analytlcal. I. Marcus. R. Kenneth. 11. Serles.
OC711.8.G5G58 1993
543' .0858--dc20 93-10300
CIP
ISBN 978-1-4899-2396-7 ISBN 978-1-4899-2394-3 (eBook)

DOI 10.1007/978-1-4899-2394-3
© 1993 Springer Science+Business Media New York

Original1y published by Plenum Press, New York in 1993.
Softcover reprint ofthe hardcover 1st edition 1993
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted

in any form or by any means, e1ectronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
To the true joys in my life:
Melanie, Drew, Courtney, and Kendalee
Contributors
J. A. C. Broekaert • Department ofChemistry, University ofDortmund,

D-4600 Dortmund 50, Germany
Sergio CaroU • Department of Applied Toxicology, Istituto Superiore di
Sanitä, 00161 Rome, Italy
Douglas C. Duckworth • Analytical Chemistry Division, Oak Ridge
National Laboratory, Oak Ridge, Tennessee 37831
Duencheng Fang • Materials Research Corporation, Orangeburg, New
York 10962
James M. Harnly • United States Department of Agriculture, Agricul-
tural Research Service, Beltsville Human Nutrition Research Center,
Nutrient Composition Laboratory, Beltsville, Maryland 20705
W. W. Harrison • Department of Chemistry, University of Florida,
Gainesville, Florida 32611
Kenneth R. Hess • Department of Chemistry, Franklin and Marshall
College, Lancaster, Pennsylvania 17604
Hubert Hocquaux • IRSID UNIEUX, BP 50, 42702 Firminy, Cedex,
France
F. L. King • Department of Chemistry, West Virginia University,
Morgantown, West Virginia 26506
R. Kenneth Marcus • Department of Chemistry, Howard L. Hunter
Chemical Laboratories, Clemson University, Clemson, South Carolina
29634
Terry A. Miller • Laser Spectroscopy Facility, Department ofChemistry,
The Ohio State University, Columbus, Ohio 43210
Edward H. Piepmeier • Department of Chemistry, Oregon State Univer-
sity, Corvallis, Oregon 97331
Bryan L. Preppernau • Laser Spectroscopy Facility, Department of
Chemistry, The Ohio State University, Columbus, Ohio 43210
vii
viii CONTRIBUTORS
Philip G. Rigby • School ofBiological and Chemical Sciences, University

of Greenwich, Woolwich, United Kingdom
Oreste Senofonte • Department of Applied Toxicology, Istituto Super-
iore di Sanitä, 00161 Rome, Italy
David L. Styris • Battelle, Pacific Northwest Laboratory, Richland,
Washington 99352
Michael R. Winchester • Inorganic Chemistry Research Division,
National Institute for Standards and Technology, Gaithersburg, Mary-
land 20899
Preface
One of the greatest challenges remaining in the area of analytical atomic

spectrometry is the development of more universal methods for the direct
analysis of solid materials. While the success of flame and furnace atomic
absorption spectrophotometries and inductively-coupled plasma optical and
mass spectrometries is undeniable, there still remain great demand and effort
directed toward improving the techniques of solid sampIe dissolution. It is
the dissolution step that often limits the ultimate analytical utility of such
methods. Dissolution procedures, despite their payback in terms of
presenting a homogeneous, easily manipulated (flow injection, chromatog-
raphy, etc.) sampIe to the spectrochemical source, can fall short in terms
of time constraints or loss of analytical quality due to sampIe dilution or
contamination. Irregardless of such possible problems, aqueous solution
nebulization will continue to be the predominant method of sampIe
introduction.
Different from the requirements of just a decade or two ago, contempor-
ary solid materials analysis requires greater powers of detection, speed, and
precision than presently available with standard are and spark technologies.
In addition, spatially (depth) resolved elemental profiles are required to
assess systems such as galvanized coatings or multilayered automotive glass.
It is for these reasons that glow discharge (GD) devices are receiving
increased interest within the analytical community. These reduced pressure,
inert atmosphere plasmas rely on a cathodic sputtering step to atomize solid
sampIes directly. Subsequent atomic excitation and ionization processes in
the adjacent plasma (negative glow) render the devices useful for atomic
absorption (AA), fluorescence (AF), emission (AE), and mass spectro-
metries (MS).
Glow Discharge Spectroscopies is a multiauthored volume having a goal
to assess the state of the art in this diverse and continuously expanding
ix
x PREFACE
area of atomic spectroscopy. Given the wide variety of source geometries,

powering schemes, and analytical applications, no single-authored volume
could do just service to the field. Each of the authors is acknowledged as a
leader in the particular area described in the respective chapters. While each
author has written in bis own individual style, most of the chapters begin
with some historical perspective of the general area and then deal extensively
with fundamental aspects relevant to the application at hand. Analytical
characteristics and figures of merit are detailed, with some visions of future
developments and applications concluding each chapter. It is hoped that this
volume will be useful as both a survey of the use of glow discharge devices
in atomic spectroscopy and as a technical reference for practitioners in the
field.
As editor, I would like to express my appreciation to the authors (each
of whom was my first choice for the respective topics) for their thorough
and thoughtful contributions. As this was my first time undertaking such a
challenge, I greatly appreciate the staff at Plenum Publishing under the
direction of Amelia McNamara who were most helpful and patient.
R. Kenneth Marcus
Clemson, South Carolina
Contents
,. Introduction
R. Kenneth Marcus
1.1. Rationale ........................................ 1

1.1.1. Scope of Elemental Analysis ........ . . . . . . . . . . 2
1.1.2. Aqueous Sampie Elemental Analysis. . . . . . . . . . . . 3
1.1.3. Direct Solids Elemental Analysis .............. 5
1.1.3.1. The Ideal Technique for Direct Solids
Elemental Analysis .. . . . . . . . . . . . . . . . . 5
1.1.3.2. Comparison of Solids Techniques. . . . . . 7
1.2. Glow Discharge Devices: Basic Operating Principles . . . . 10
1.3. Volume Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II
References ............................................ 14
2. Fundamental Plasma Processes
Duencheng Fang and R. Kenneth Marcus
2.1. Gaseous Discharges ................................ 17

2.2. Gk>w Discharges .................................. 20
2.2.1. Glow Discharge Processes .................... 20
2.2.2. Alternate Operating Modes .................. 24
2.2.2.1. Hollow Cathode Effect .............. 24
2.2.2.2. Radio-Frequency-Powered
Discharges ........... . . . . . . . . . . . . . 25
2.3. Atomization via Cathodic Sputtering . . . . . . . . . . . . . . . . . . 26
2.3.1. Cathodic Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . 26
xi
XII CONTENTS
2.3.2.
Sputter Yield ............................... 27
2.3.2.1. Mass of Sputtering Ion .............. 28
2.3.2.2. Angle ofIncidence .................. 29
2.3.2.3. Incident Ion Energy ................ 29
2.3.2.4. Target Material .................... 30
2.3.2.5. Target Temperature ................. 31
2.3.3. Sputter Rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.4. Differential Sputtering . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4. Physical Characteristics of Glow Discharge Devices .... 37
2.4.1. Positive Ion Energies ........................ 39
2.4.2. Electrical Characteristics. . . . . . . . . . . . . . . . . . . . . . 41
2.4.2.1.Planar, Diode Devices .............. 42
2.4.2.2. Hollow Cathode Devices ............ 43
2.4.3. Energetic Partic1e Densities and Energies. . . . . . . . 44
2.4.3.1. Electrons .......................... 44
2.4.3.2. Metastable Atoms .................. 52
2.5. Analyte Excitation and Ionization . . . . . . . . . . . . . . . . . . . . 56
2.5.1. Excitation.................................. 57
2.5.2. Ionization.................................. 59
2.6. Conc1usions ...................................... 62
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
References ............................................ 63
3. Atomic Absorption and Fluorescence

Spectroscopies
Edward H. Piepmeier
3.1. Introduction ...................................... 67

3.2. Fundamental Considerations ........................ 71
3.2.1. Characteristics of Low-Pressure Atomic
Spectra .................................... 71
3.2.1.1. Shapes of Spectral Line Profiles ....... 72
3.2.1.2. Influence on Working Curves ........ 75
3.2.1.3. Influence of Hollow Cathode Lamp
Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.2.1.4. Fluorescence Spectral Resolution and
Resonance Monochromators. . . . . . . . . . 77
3.2.2. Sampling and Transport to the Observation
Region .................................... 78
3.2.2.1. Experimental Studies ................ 78
3.2.2.2. Transport Theory .................. 80
CONTENTS xiii
3.2.2.3. Transient Sampies .................. 85
3.3. Practical Considerations ............................ 86
3.3.1. Methods Development ...................... 86
3.3.1.1. Choice of Spectral Line .............. 87
3.3.1.2. Plasma Gas ........................ 88
3.3.1.3. Discharge Conditions . . . . . . . . . . . . . . . . 89
3.3.1.4. Sampie Temperature ................ 89
3.3.1.5. Preconditioning the Surface
(Prebum) .......................... 91
3.3.2. Relative Precision for Atomic Absorption
Determinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.3.3. Background Correction ...................... 94
3.3.4. Pulsed Sputtering Cells ...................... 96
3.3.4.1. Pulse Duration . . . . . . . . . . . . . . . . . . . . . . 97
3.3.4.2. Peak Height or Peak Area? .......... 98
3.3.5. Multielement Determinations . . . . . . . . . . . . . . . . . . 99
3.3.6. Compensation for Sputtering Rate Variations .. 100
3.3.7. Determining Element Ratios with Only
One Standard .............................. 103
3.4. Applications ...................................... 104
3.4.1. Alloys ............... . . . . . . . . . . . . . . . . . . . .. 104
3.4.2. Powders.................................... 105
3.4.3. Dried Solutions ............................ 105
3.4.4. Isotopes.................................... 107
3.5. The Future.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 108
References ............................................ 110
4. Atomic Emission Spectrometry
J. A. C. Broekaert
4.1. Optical Emission Spectrometry ...................... 113

4.1.1. Principles .................................. 113
4.1.1.1. Optical Spectra. . . . . . . . . . . . . . . . . . . . .. 113
4.1.1.2. Line Intensities. . . . . . . . . . . . . . . . . . . . .. 115
4.1.1.3. Spectral Linewidths. . . . . . . . . . . . . . . . .. 120
4.1.2. Radiation Sources .......................... 122
4.1.2.1. dc Arcs ............................ 123
4.1.2.2. Spark Sources ...................... 124
4.1.2.3. Laser Plumes ...................... 127
4.1.2.4. Hollow Cathode Devices ............ 127
4.1.2.5. Planar Glow Discharges . . . . . . . . . . . . .. 128
xiv CONTENTS
4.1.3.Spectrometers .............................. 129

4.1.3.1. Dispersive Spectrometers ............ 129
4.1.3.2. Multiplex Spectrometers. ... .... ... . .. 133
4.1.4. Radiation Detectors ........................ 135
4.1.4.1. Photomultipliers.................... 135
4.1.4.2. Multichannel Detectors .............. 136
4.2. Glow Discharges with Planar Cathodes .............. 139
4.2.1. Electrical Characteristics.. . . . . . . . .. . . . .. .... .. 140
4.2.2. Sampie Volatilization ........................ 141
4.2.3. Excitation Processes . . . . . . . . . . . . . . . . . . . . . . .. 144
4.3. Analytical Performance ............................ 146
4.3.1. Power of Detection . .. . . . ... . . . ... .. . ... . . . .. 146
4.3.2. Matrix Interferences ........................ 150
4.3.3. Analytical Precision. . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.4. Analytical Applications ............................ 153
4.4.1. Analysis of Metals .......................... 153
4.4.1.1. Bulk Analysis ...................... 154
4.4.1.2. In-Depth Profiling .................. 155
4.4.2. Analysis of Dry Solution Residues ............ 157
4.4.3. Analysis ofElectrically Nonconducting Sampies. . 158
4.5. Lines ()f Research for Further ImDrovement .......... 160
4.5.1. Application of Cross-Excitatlon .............. 160
4.5.2. Improvements in the Use of the Spectral
Information ................................ 164
4.5.3. Glow Discharges as Atom Reservoirs .......... 168
4.6. Comparison with Other Methods for the Direct Analysis
of Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 169
4.7. Conclusions ...................................... 170
Acknowledgments ...................................... 171
References ............................................ 171
5. Glow Discharge Mass Spectrometry

F. L. King and W. W. Harrison
5.1. Introduction ...................................... 175

5.2. Mass Spectrometry ................................ 176
5.2.1. Fundamentals .............................. 177
5.2.2. Mass Analysis .............................. 177
5.2.2.1. Dispersion in Space (Sector MS) ...... 178
5.2.2.2. Dispersion in Time (Time-of-Flight MS) 179
5.2.2.3. Delective m/z Transmission
(Quadrupole Mass Filters). . .. . .... . .. 181
CONTENTS xv
5.2.2.4. Selective m/z Ejection (Quadrupole Ion

Traps) ........................... . 184
5.2.2.5. Fourier Transform Ion Cyclotron
Resonance MS ..................... . 185
5.2.3. Mass Spectrometry of Glow Discharge Plasmas .. 187
5.3. Atomization/Ionization and Quantitation ............. . 188
5.3.1. Glow Discharge Processes ................... . 188
5.3.1.1. Atomization ....................... . 189
5.3.1.2. Ionization ......................... . 190
5.3.2. Quantitation ............................... . 191
5.4. Glow Discharge Ion Sources ....................... . 191
5.5. GDMS Applications ............................... . 193
5.5.1. Solids Elemental Analysis ................... . 195
5.5.1.1. Bulk Metal Analysis ............... . 195
5.5.1.2. Semiconductors ................... . 199
5.5.1.3. Surface Analysis ................... . 203
5.5.1.4. Compacted Sampies ............... . 205
5.5.2. Solutions ................................. . 207
5.5.3. Nonelemental GDMS Applications ........... . 209
5.5.4. Future Directions in GDMS Applications ..... . 212
References 212
6. Hollow Cathode Discharges
Sergio Caroli and Oreste Senofonte
6.1. Introduction ...................................... 215

6.1.1. Historical Background ...................... 215
6.1.2. Present Impact on the Progress of Spectroscopy.. 218
6.2. Fundamental Aspects ............. ;................ 219
6.2.1. Working Characteristics of a Hollow Cathode
Discharge .................................. 219
6.2.2. Plasma Formation .......................... 221
6.2.3. Energy Distribution. . . . . . . . . . . . . . . . . . . . . . . . .. 223
6.2.4. Sputtering Process .......................... 225
6.2.5. Excitation Mechanism ....................... , 227
6.3. Main Variants of Hollow Cathode Tubes. . . . . . . . . . . . .. 228
6.3.1. Preliminary Comments ...................... 228
6.3.2. Selected Models ............................ 229
6.3.3. Boosted Models ............................ 237
6.4. Recent Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 240
6.4.1. Sampie Introduction ........................ 240
6.4.2. Analysis of Solids .......................... 242
xvi CONTENTS
6.4.3. Analysis of Liquids . . . . . . . . . . . . . . . . . . . . . . . . .. 242

6.4.4. Analysis of Gases. . . . . . . . . . . . . . . . . . . . . . . . . . .. 244
6.4.5. Special Uses ...... . . . . . . . . . . . . . . . . . . . . . . . . .. 244
6.4.5.1. Optogalvanic Spectroscopy .......... 244
6.4.5.2. Laser Technology .................. 245
6.4.5.3. Deposition and Etching ............ " 247
6.4.5.4. Beam Generation. . . . . . . . . . . . . . . . . . .. 247
6.4.5.5. Plasma Diagnostics . . . . . . . . . . . . . . . . .. 248
6.5. Future Trends .................................... 251
Acknowledgment ...................................... " 254
References ............................................ 254
7. Analysis of Nonconducting Sampie Types
Michael R. Winchester, Douglas C. Duckworth,

and R. Kenneth Marcus
7.1. Introduction ...................................... 263

7.2. Application of the Direct Current Glow Discharge . . . . .. 264
7.2.1. Introduction................................ 264
7.2.2. Sampie Preparation . . . . . . . . . . . . . . . . . . . . . . . . .. 265
7.2.3. Analytical SampIe Composition ............... 270
7.2.3.1. Choice of Host Matrix Material ...... 270
7.2.3.2. Relative Percentages of Sampie and
Host in Mixture .................... 272
7.2.3.3. Effect of ParticIe Size . . . . . . . . . . . . . . .. 273
7.2.3.4. Effect of Trapped Gases and Water . . .. 273
7.2.4. Atomic Transport in the Compacted Solid. . . . .. 275
7.2.5. Methods for Improved Quantitation .......... 280
7.2.5.1. Internal Standardization. . . . . . . . . . . . .. 281
7.2.5.2. Pulsed Glow Discharges . . . . . . . . . . . . .. 283
7.2.6. Illustrative Analytical Applications ............ 285
7.2.6.1. Atomic Absorption Spectrophotometry:
Analysis of Automotive Catalyst
Materials. .......................... 285
7.2.6.2. Atomic Emission Spectrometry:
Analysis of Geological SampIes . . . . . . .. 288
7.2.6.3. Atomic Mass Spectrometry: Analysis
of NucIear Materials ................ 291
7.3. Application ofrf-Powered Glow Discharges............ 291
7.3.1. Introduction................................ 291
7.3.2. Historieal .................................. 294
CONTENTS xvii
7.3.3. Theory of Operation ....................... . 295
7.3.4. Fundamental Considerations ................. . 296
7.3.4.1. Source Geometry ................... . 297
7.3.4.2. rf Power Coupling ................. . 300
7.3.4.3. Operating Frequency ............... . 301
7.3.5. Analytical Applications ..................... . 303
7.3.5.1. rf Discharges as Secondary Excitation/
Ionization Sources ................. . 304
7.3.5.2. The rf Glow Discharge as a Primary
Spectrochemical Source ............. . 305
7.3.5.3. rf Glow Discharge Mass Spectrometry .. 312
7.3.5.4. rf Glow Discharge Optical
Spectroscopies ..................... . 322
7.4. Conclusions ..................................... . 325
References 326
8. Thin Film Analysis
Hubert Hocquaux
8.1. Historical Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 329

8.1.1. Bulk Analysis .............................. 329
8.1.2. Surface Analysis ............................ 329
8.2. Principle of Surface Analysis by Glow Discharge
Sputtering ........................................ 330
8.3. Equipment........................................ 331
8.3.1. Glow Discharge Source ...................... 331
8.3.2. Spectrometer................................ 332
8.3.3. Electronics and Computer. . . . . . . . . . . . . . . . . . .. 334
8.4. Analytical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 336
8.5. Typical Operating Parameters for Surface Analysis. . . . .. 338
8.6. Shape of the Crater ................................ 340
8.7. Sampie Geometry .. ............ .. .... ... ...... . . . .. 342
8.8. Quantitative Surface Analysis with GD-OES .......... 344
8.8.1. Factors Affecting Sputtering Rate. . . . . . . . . . . . .. 346
8.8.1.1. Practical Considerations . . . . . . . . . . . . .. 346
8.8.1.2. Study of the Sputtering Rate Variation
with Respect to Discharge Parameters.. 349
8.8.l.3. Experimental Results ................ 349
8.8.1.4. Theoretical Predictions-Modeling .... 351
8.8.2. Concentration Calibration . . . . . . . . . . . . . . . . . . .. 355
8.8.2.1. Practica1 Approaches ................ 356
xviii CONTENTS
8.8.2.2. Theoretical Approach. . . . . . . . . . . . . . .. 358

8.9. Main Applications in Surface Analysis. . . . . . . . . . . . . . .. 364
8.9.1. Analysis of Coatings and Surface Treatment by
GD-OES .................................. 364
8.9.1.1. Hot Nitriding Treatment on Tool Steels 364
8.9.2. lonic Implantation .......................... 367
8.10 Limits and Advantages of GD-OES .................. 368
8.10.1. Limits .................................... 368
8.10.2. Advantages ................................ 369
8.11. Future Trends .................................... 370
References .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 370
9. Discharges within Graphite Furnace Atomizers
James M. Harnly, David L. Styris, and Philip G. Rigby
9.1. Introduction ...................................... 373

9.2. System Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 376
9.3. Fundamental Considerations ........................ 382
9.3.1. Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 382
9.3.2. Sheath Dimensions .......................... 384
9.3.3. Discharge Characteristics .................... 386
9.3.4. Voltage-Current Characteristics .............. 389
9.4. Operational Characteristics . . . . . . . . . . . . . . . . . . . . . . . . .. 392
9.4.1. Analytical Signals .......................... 392
9.4.1.1. Support Gas. . . . . . . . . . . .. . . . . . . . . . .. 392
9.4.1.2. Discharge Pressure .................. 392
9.4.1.3. Atomization Temperature ............ 399
9.4.1.4. Discharge Current .................. 402
9.4.2. Noise Sources .............................. 406
9.4.3. Interferences................................ 406
9.4.4. Figures of Merit ............................ 410
9.4.4.1. Detection Limits. . . . . . . . . . . . . . . . . . .. 410
9.4.4.2. Reproducibility .................... 412
9.4.4.3. Dynamic Range .................... 412
9.5. Applications ...................................... 413
9.5.1. Metal Determinations . . . . . . . . . . . . . . . . . . . . . . .. 413
9.5.2. Nonmetal and Molecular Determinations ...... 415
9.6. Simultaneous Multielement Determinations . . . . . . . . . . .. 416
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 418
References ............................................ 418
CONTENTS xix
10. Laser-Based Methods
Kenneth R. Hess
10.1. Introduction ...................................... 423

10.2. The Laser ........................................ 424
10.3. Supplemental Atomization .......................... 425
10.3.1. Introduction to Laser Ablation. . . . . . . . . . . . . . .. 426
10.3.2. Mechanisms of Laser Ablation. . . .. . . . . . . .. . .. 426
10.3.3. Laser Ablation Coupled to Glow Discharge
Devices .................................... 429
10.3.3.1. Laser Ablation of an Auxiliary Sampie 429
10.3.3.2. Laser Ablation of the Cathode Material 434
10.3.4. Fundamental Discharge Investigations. . . . . . . . .. 436
10.3.4.1. Mechanism of Formation of ArM+
Species ............................ 436
10.3.4.2. Sputter Redeposition Studies. . . . . . . . .. 438
10.3.5. Analytical Investigations. . . . . . . . . . . . . . . . . . . . .. 441
10.4. Supplemental Excitation ............................ 441
10.4.1. Atomic Fluorescence ........................ 442
10.4.1.1. Analytical Applications .............. 445
10.4.1.2. Diagnostic Studies by Laser-Induced
Fluorescence . . . . . . . . . . . . . . . . . . . . . . .. 449
10.4.2. Optogalvanic Effect Spectroscopy . . . . . . . . . . . . .. 453
10.4.2.1. Optogalvanic Studies of Discharge
Ionization Processes ................ 456
10.4.2.2. Other Applications of QGE
Spectroscopy ...................... 458
10.4.3. Laser-Enhanced Ionization. . . . . . . . . . . . . . . . . . .. 461
10.4.4. Resonance Ionization Mass Spectrometry ...... 463
10.4.5. Additional Laser-Based Techniques . . . . . . . . . . .. 472
10.4.5.1. Doppler-Free Spectroscopy .......... 472
10.4.5.2. Laser Atomic Absorption ............ 474
10.4.5.3. Degenerate Four-Wave Mixing . . . . . . .. 474
10.5. Conclusions ...................................... 476
References ............................................ 477
11. Laser-Based Diagnostics of Reactive Plasmas
Bryan L. Preppemau and Terry A. Miller
11.1. Introduction ...................................... 483

xx CONTENTS
11.2. An Overview of Reactive Plasmas ... . . . . . . . . . . . . . . . .. 485

11.2.1. Species Identity ............................ 485
11.2.1.1. Atoms ............................ 485
11.2.1.2. Radicals .......................... 486
11.2.1.3. Ions .............................. 486
11.2.1.4. Excited Molecular States ............ 487
11.2.2. Sampling Considerations for Concentration
Measurements .............................. 488
11.3. Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . .. 491
11.3.1. One-Photon LIF ............................ 492
11.3.2. Two-Photon and Multiphoton LIF ............ 494
11.3.3. Multiphoton Ionization ...................... 495
11.3.4. Optogalvanic Detection ...................... 496
11.3.5. Absorption Techniques ...................... 496
11.3.6. Comparison with Other Diagnostic Methods . . .. 497
11.3.6.1. Emission Spectroscopy .............. 497
11.3.6.2. Actinometry . . . . . . . . . . . . . . . . . . . . . . .. 498
11.3.6.3. Ex Situ Titration. . . . . . . . . . . . . . . . . . .. 499
11.3.6.4. Absorption Spectroscopy ............ 499
11.3.6.5. Charged Partic1e Methods . . . . . . . . . . .. 499
11.4. Species Detected and Studied . . . . . . . . . . . . . . . . . . . . . . .. 500
11.4.1. One-Photon LIF Detection .................. 500
11.4.2. Two-Photon LIF Detection .................. 501
11.4.3. OGS and REMPI Detection .................. 502
11.4.4. Plasma Characteristic Measurements by
Laser-Based Techniques . . . . . . . . . . . . . . . . . . . . .. 503
11.4.4.1. Electric Field Strength .............. 503
11.4.4.2. Photodetachment and Negative Ions . .. 504
11.4.4.3. Energy Content by Doppler Profiles. . .. 505
11.4.4.4. Particulates ........................ 505
11.5. Conc1usions and Future Trends . . . . . . . . . . . . . . . . . . . . .. 505
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 506
References ............................................ 506
Index .................................................... 509

1
Introduction
R. K enneth Marcus
1.1. Rationale
Driven by advances in materials research, government regulations, and inter-

disciplinary collaborations, analytical chemistry has been one of the most
active areas of chemical research over the last three decades. Fundamental
advances in the fields of environmental modeling, medicine, and materials
science are evidence that the discipline has met many of the challenges that
have been posed. In developing methods, analytical chemists have become
more adept at interpreting and implementing the findings of biologists and
applied physicists. The role that computers have had in revolutionizing
chemical instrumentation and techniques cannot be underestimated.
Advances in analytical chemistry have shown up in the vernacular of the
trade; units of quantitation have switched from weight percent to parts per
billion (and trillion). Simple functional-group analysis has been replaced by
methods of extraordinary selectivity and specificity. Each step toward better
sensitivity or specificity in analytical measurements can, in many cases, cause
more difficulties than provide answers. For example, an analytical method
can be too sensitive for many uses and environments. This volume addresses
some of the advances that have taken place in the area of elemental spectro-
chemical analysis.
Glow discharge devices were some of the earliest spectrochemical
sources, first used in fundamental spectroscopic studies of atomic structure in
the 1910sY-3) From that time until the 1960s, their application in analytical
R. Kenneth Marcus • Department of Chemistry, Howard L. Hunter Chemical Labora-

tories, Clemson University, Clemson, South Carolina 29634.
Glow Discharge Spectroscopies, edited by R. Kenneth Marcus. Plenum Press, New York, 1993.
1
2 CHAPTER 1
chemistry was very limited. Over the last 30 years, the applications of these
devices have grown slowly but steadily with commercial instrumentation
now being readily available. Their primary uses have been as line sources in
atomic absorption spectrometry and in bulk solids elemental analysis by
atomic absorption,· emission, and mass spectrometries. The forte of the
devices is their ability to allow the direct elemental analysis of materials in
the solid state. It is the purpose of this book to outline the developments in
analytical applications of glow discharge devices over the last two decades
and to highlight future trends as the techniques continue to evolve. Experts
in the various applications of glow discharge devices have contributed to
this volume and put their own applications into perspective with competing
and complementary methods. It is hoped that the reader will begin to gain
an appreciation for the fundamental processes occurring in glow discharges
and also the scope of the current and expected analytical applications.
1.1.1. Scope of Elemental Analysis
The challenges facing all of analytical chemistry can at times seem most
complicated in the area of elemental analysis. Obviously, the gross elemental
composition of a given system (material) defines its chemical and physical
characteristics and in the end, its utility. Beyond elemental quantitation,
additional information such as ionic speciation, isotopic composition, and
spatial distribution may be of importance. These other considerations take
e1emental analysis far beyond the determination of "X-in-Y." In many
instances, it is the presence and concentration of an alien element that is of
interest, making minor and trace analysis of utmost importance in many
applications.
Often, generation of the appropriate e1emental information is com-
plicated by the variety of sampie types that the atomic spectrometrist is
presented. Table 1-1 is a short compilation ofthe analytical systems (sampie
types) that often require elemental analysis. It is when the atomic spectro-
metrist is faced with this wide array of sampie matrices that one can see why
Table 1-1. Sampie Types Requiring Elemental Analysis

Metals and alloys
Geological specimens
Potable, waste, and ground waters
Glasses and ceramics
Biological specimens
Semi- and superconducting materials
Catalyst materials
Composites and polymers
Thin films
INTRODUCTION 3
sampIe preparation and introduction continue to be at the forefront of
spectrochemical research. In fact, one of the most important considerations
in deciding between the various atomic spectrometry techniques is their
respective susceptibility to matrix effects. In many cases, the sampIe form
will dictate the ultimate choice in methodology.
1.1.2. Aqueous SampIe Elemental Analysis
Going through Table I-I, it is first evident that there are a wide variety
of sampIe matrices that must be covered, making the possibility of using a
single analytical method or instrument for all of these systems extremely
unlikely. Municipal, industrial, and environmental regulations are such that
the third entry (waters) is by far the most prevalent of the sampIe types.
Human body fluids (e.g., serum, urine) must also be among the most com-
mon of elemental analysis matrices. Beyond basic elemental analysis, these
essentially aqueous sampIes may require that elemental concentrations be
broken down into specific oxidation states (i.e., speciation). SampIe prepara-
tion generally involves nothing more complex than an organic extraction
step or possibly analyte preconcentration with an ion exchange polymer or
the like. Advantages in direct solution analyses include ease in calibration,
sampIe homogeneity, and ease of sampIe delivery (automation and
throughput). Even so, the introduction of aqueous solutions continues to be
the "Achilles' heel" in atomic spectroscopy.(4) The relatively new approaches
incorporating flow injection(5) or ion chromatography(6) have become power-
ful tools for sampIe introduction particularly in dealing with matrix effects
and allowing on-line speciation. It is these advantages that have led to the
near-total dominance of solution-based analytical methods in the atomic
spectrometry market.
A number of excellent review articles have been written that are devoted
almost exclusively to those techniques that are best suited for solution-phase
sampIes. (7-9) The most widely applied of these methods are flame and furnace
atomic absorption spectrophotometry, and inductively coupled plasma
atomic emission and mass spectrometries. As is the case of the sampIe forms,
techniques that rely on aqueous sampIe nebulization have domina ted the
field since the early 1960s (as opposed to solids techniques). Flame atomic
absorption spectrophotometry (AAS) was the first commercially available
of this group and continues to be the most widely applied. SampIes are
introduced by pneumatic nebulization under continuous solution flow or by
discrete injection methods. Furnace AAS was first introduced as a supple-
mentary atomization source to the conventional flame, sold for those specific
applications where sampIe volumes are limited « 0.1 ml). SampIes origi-
nally in the liquid state are dried in the furnace and then atomized as residues.
Currently, the technique is acknowledged as being the most sensitive in terms
4 CHAPTER 1
of absolute analyte mass. Inductively coupled plasmas (lCP) are generated

by coupling radio frequency (multiples of 13.56 MHz) energy (1-2.5 kW) to
a flow of argon discharge gas. The resulting plasma is characterized by gas
temperatures that reach upwards of 5000 K. As such, the ICP is very efficient
at breaking down the most refractory species to free atoms. The atornic
emission mode has the distinct advantage of providing truly simultaneous
multielement analysis when direct reading spectrometers are employed. Since
its first commercial availability in the early 1980s, ICP mass spectrometry
has become the dass of the atomic spectrometry field in terms of "bulk
solution" sensitivity. It is this group oftime-tested solution analysis methods
that make aqueous sampie introduction the most popular means of elemental
analysis, regardless of the initial sampie form. It must be emphasized, in
discussions of the competing direct solids methods, that figures of merit for
these techniques are quoted in terms of mass per unit of solution after
",100-1000 x dilution [i.e., I ppm (Jlgjml) in solution is 0.01-0.1% by
weight in the solid].
Beyond the liquid matrix sampies presented in Table I-I, there are a
very wide variety of solid sampie types that require elemental analyses. These
matrices can be electrically conducting or nonconducting (to one degree or
another), organic or inorganic, thermally stabile or heat sensitive, and either
homogeneous or have a well-defined structure. From an atomic spectroscopy
standpoint, the possible mechanisms or steps required to generate a free
atom for analysis are greatly complicated by the physical properties of the
sampie. Indeed, if one is solely equipped for the analysis of aqueous sampies,
by what method can the sampie type be placed into solution? Certainly, the
number of methods and range of difficulties in sampie dissolution are as
wide as the number of possible matrices. (10) Sampie preparation can range
from simple nitric acid dissolutions to sodium or lithium fusions followed
by treatment with HF. The diversity in procedures is particularly apparent
in the analysis of refractory materials, glasses, and ceramics where acidic or
basic media may be emPIOyed depending on the compositionjrnineralogy.
In fact, the dissolution procedures for specialty alloys such as those used in
the aerospace industry (e.g., Al, Ti) can be very complex in their own right.
Fortunately, there is extensive documentation provided by instrument manu-
facturers and in the analyticalliterature which covers almost any case of
sampIe dissolution. In addition, the advent of rnicrowave technology has
added a great deal of power and speed to the sampIe preparation process,
often reducing digestion times from days to hours and rninutesY 1)
Despite any complexities found in solid sampie dissolution procedures,
the analyst gains all of the benefits of solution sampie analysis induding ease
in quantitation (e.g., calibration curves, standard additions) and the use of
chromatographic or flow injection introduction schemes. Possibly of great
importance is the inherent matrix homogenization of aqueous sampIes.
INTRODUCTION 5
Many of the complicating factors with regard to obtaining representative
sampling from solids are circumvented by placing the solid sampies into
solution. Finally, dissolution of asolid sampie leaves the analyst with a
matrix with which he or she is well accustomed.
1.1.3. Direct Solids Elemental Analysis
Given the many attributes of solution-based sampie analysis methods,

there are a number of factors that the analyst must consider in determining
the ultimate procedure for the analysis of a particular solid sampie type. This
decision goes far beyond the relative precision, accuracy, and sensitivities of
the respective methods. In spite of the wealth of sampie preparation methods,
the bottom line in many production laboratories is sampie tumaround time.
Therefore, ease of sampie preparation can be a determining factor for those
laboratories contemplating solution phase analyses of solid materials. The
classic example of the necessity of speed is in metallurgical foundries, where
analytical time frames are on the order of minutes rather than hours. In
high-purity material applications, time may not be as important as absolute
sensitivity. Sampie dissolution procedures dilute the apparent analytical con-
centrations of analytes and, in some instances, result in the introduction
of contaminant species at levels exceeding the trace impurities. The final
consideration in the choice of methodology in solids analysis is the desire to
obtain regio-specific information, either laterally (xjy) or depth resolved (z).
If such information is desired, then aqueous sampie preparation is practically
obviated. It is these special considerations that have led to the development
and continued use of direct solids elemental analysis techniques.
1.1.3.1. The Ideal Teehnique for Direet Solids Elemental Analvsis
As is the case for spectrochemical sources employed for solution-based

analyses, each of the so lids techniques has its unique analytical character-
istics. Clearly, the range of physical and chemical properties of solid sampies
places much more stringent demands on the applicable spectrochemical
methods than for liquid matrix sampies. The number ofpossible solid sampie
matrix types, each of which requires special sampie preparation and ana-
lytical standards, makes solid-sampling techniques much more specialized
than liquid-sampling methods. Borrowing in part from other atomic spectro-
metry reviews,(12,13) Table 1-2 lists aseries of characteristics that the ideal
spectrochemical source for solids elemental analysis would possess. In choos-
ing the most appropriate solids source for a given applicationjlaboratory,
each method should be assessed relative to the ideal case.
In going through the characteristics of the ideal source, the first six
entries refer to the atomization properties of the technique. As is the case
6 CHAPTER 1
Table 1-2. The Ideal Spectrochemical Source for Solids Elemental Analysis
I. Applicability to all possible sampIe matrices (bulk and particulate, electrically conductive
and nonconductive)
2. Easily controlled atomizationjexcitationjionization rates
3. Stoichiometric atomization for all elements within a sampIe
4. Nondestructive
5. Rapid, high-precision depth profiling
6. High degree of lateral resolution
7. No sampIe preparation
8. Easy to operate
9. UsefuVfor all elements under the same operating conditions
10. Low capital and maintenance costs
11. Total elemental coverage
12. Infinite linear dynamic range (1 atom to 100%)
13. Easy to quantitate (no matrix effects)
14. Multimode operation (e.g., AA, AE, MS)
15. Ability to analyze liquid and gaseous sampIes
with any sampie introduction scheme for atomic spectrometry, the success
or failure of the method can be determined solely by the atomization effi-
ciency of the device. In direct solids analysis, this is of course a taU order
because of the wide variety of possible matrices. Depending on the actual
atomization mechanism, the sampie may be an integral part of the method
of signal generation; as such, the electrical and mechanical characteristics of
the sampie may determine the applicability of a particular method, or at
least the rate at which sampie constituents are evolved. In the best of aU
worlds, the technique would be nondestructive. Unfortunately, because of
the limited sampling depth of charged particle/photon probing techniques,
surface erosion is required to generate depth-resolved analyses.
Characteristics 7 through 10 refer to the user friendliness of the method.
Given the availability of solution-based spectrochemical methods, any solids
technique must be more convenient to use than the corresponding sampie
dissolution scheme. This includes sampie preparation, optimization of instru-
ment response characteristics, and the actual purchase and maintenance
costs of the instrument. Items 11-14 deal most specifically with the basic
analytical characteristics ofthe techniques. While the desire for total elemen-
tal coverage is a noble one, it is almost an impossibility given the fact that
most atomic spectrochemical techniques employ support gas atoms/ions to
generate the analytical signal. The qualities of infinite dynamic range and
lack of matrix effects are the most daunting of the characteristics, but are at
the same time possibly the most desired for any analytical system. Tech-
niques of direct solids analysis do not have the luxuries of sampie dilution
or preconcentration inherent to solution analysis methods, nor is outright
matrix removal possible. The complete elemental makeup is continuously
INTRODUCTION 7
present during the analysis, with the matrix element(s) dictating the energy
absorption and atomization characteristics of all the constituents, inflicting
spectral and excitationjionization interferences as weIl. In short, the minor
and trace species would ideally be separated from the matrix by some mecha-
nisms before the actual analysis.
The final analytical quality would be the capability to operate in more
than one sampling mode. For example, in some instances only bulk, single-
element analysis might be required, making atomic absorption (AAS) a good
option. Of course, not all elements are weIl suited for study by AAS. Atomic
emission spectroscopy (AES) is very useful for multielement analysis, provid-
ing the capabilities for depth-resolved analysis when polychromators are
employed. As a final example, mass spectrometry (MS) is the most practical
option for those instances where isotopic information of constituent elements
is needed. Thus, the flexibility of the device and its ability to generate specific
types of information are favourable characteristics.
In the best of all worlds, the source could be operated as an auxiliary
"solids" source directly compatible with commercial "solution" instrumen-
tation. The final characteristic of the ideal solids source is the ability to
analyze solutions (residues) and gases in addition to solid materials. Admitt-
edly, this is turning the table on the solids device, but in the ideal world the
capability to accept aIl sampie types is desirable.
1.1.3.2. Comparison of Solids Techniques
In the real world, the analyst is often forced to make a number of

compromises from the ideals of Table 1-2. The analyst must be willing to
admit that no single technique can meet all ofhis or her analytical objectives.
The realistic points of comparison are far more traditional and must take
into account the current available technologies. Table 1-3 is a compilation
Table 1-3. Ten Points of Compari-

son for Solids Elemental Analysis
Techniques
I. Accuracy
2. Precision
3. Limits of detection
4. Dynamic range
5. Matrix effects
6. Elemental unifonnity
7. Sampie pretreatment
8. Analysis time
9. Depth profiling
10. Cost and complexity
8 CHAPTER 1
of ten practical points of comparison for direct solids elemental analysis

techniques. In essence, this list is the same as comparable compilations for
any set of analytical techniques. The only glaring difference is the preferred
ability to perform depth-resolved analyses.
Accuracy, precision, and limits of detection are as easily defined for
solids as they are in solution analyses. Entries 4-6, on the other hand, take
on new meaning in direct solids analysis. In the presence of matrix elements
ranging from 10's to 100%, dynamic range and matrix effects are virtually
out ofthe analyst's control. As mentioned previously, separation of analytes
from the sampie matrix along with its concomitant spectral and spectroch-
emical influences is not generally an option. In all of the techniques to be
discussed presently, the sampie matrix has a direct influence on the absolute
analyte sensitivity (signal per unit concentration). In the majority of
instances, the sampie matrix affects sampie introduction (atomization) rates
both between different sampies (matrices) and individual elements within a
given sampie. Altematively, in x-ray fluorescence, electron spectroscopies,
and secondary ion mass spectrometry, the matrix controls the yield of secon-
dary particles that reach the detector. The overabundance ofmatrix atoms in
a spectrochemical plasma will most probably affect the excitafion/ionization
characteristics of the device. It is this propensity for matrix effects of one
form or another that has limited the greater application of most direct solids
techniques.
If there are any straightforward arguments for direct solids analysis it
is the possible freedom from sampie pretreatment and subsequent reduction
in total analysis time. The bottom line is often "time is money." This point
can go a long way for solids methods as long as there are no large compro-
mises to be made in the quality of the analytical data. In many laboratory
settings, the fact that a sampie need not be dissolved is insignificant in
comparing the different solids techniques. In some instances, the actual ana-
lysis time (from sampie in to numbers out) can be a determining factor.
The ninth item in Table 1-3 is one of increasing importance. In many
applications, simple bulk analyses are not as important as the spatial distri-
bution of sampie constituents. Such requirements are no longer limited to the
semiconductor industry, but are spreading into fields such as the automotive
industry. In that field, depth-resolved elemental analyses are required of
plated bumpers, painted sheet metal, and tinted safety glass (note the wide
range of matrices and coatings). In the vast majority of such applications,
the materials are constructed as multilayered systems that require depth
profiling of "matrix" species along with the detection of trace species at the
same time. For example, analysis of a galvanized coating on a steel base
material would require both the determination of the thickness of that layer
(Zn) and also the presence of contaminant species such as oxygen that may
be located at the Zn/steel interface. Therefore, the requirements are the
INTRODUCT/ON 9
ability to make proper depth assignments and concentration determinations
simultaneously over the full range of trace to matrix concentrations. A final
set of depth profiling considerations is the resolution of the profiling and the
speed at which the profiles can be generated. These two quantities are for
the most part mutually exc1usive with the criteria varying widely with
applications.
The final points of comparison for solids elemental analysis methods
are the cost and complexity of the instrumentation. In general, these two
attributes are direct1y re1ated to one another. The range of costs is from lOs
to 100s of thousands of dollars, with the corresponding level of operator
expertise ranging from technicians ofmoderate skill to Ph.D.-level scientists.
Table 1-4 summarizes the most common commercially available types of
direct solids elemental analysis techniques along with their relative analytical
figures of merit and characteristics. As described in the previous section, the
salient points of comparison for the analytical chemist are sampie turn-
around time, freedom from matrix interferences, matrix compatibility,
sampie size, instrumentation complexity, ease of quantitation, and depth-
resolving capabilities. The reader is encouraged to refer to the scientific
literature and spectroscopy monographs for more detailed descriptions of
the respective methods.
A cursory comparison of the characteristics of the techniques in Table
1-4 indicates that there are a number of methods that are ideally applied in
Table 1-4. Teehniques for Direet Solids Elemental Analysis"

Dynamie Matrix
Teehnique Accuraey Preeision LOD range effeets
High-eurrent are AES ++ + ++ + +
High-voltage spark AES ++ ++ + + +
X-ray fluoreseenee +++ +++ + + +
Seeondary ion MS ++ ++ +++ +++ +
Laser ablation MS + + ++ ++ +
Laser ablation ICP-MS ++ ++ ++ ++ +
Glow diseharge AES/MS +++/++ +++/++ ++/+++ ++/+++ ++/++
Elemental SampIe Analysis Depth Cost and
uniformity pretreatment time profiling eomplexity
High-eurrent are AES ++ +++ +++ NA ++
High-voltage spark AES ++ +++ +++ NA ++
X-ray fluoreseenee + ++ +++ + +++
Seeondary ion MS + + + +++ +
Laser ablation MS + +++ ++ + ++
Laser ablation ICP-MS ++ +++ ++ + +
G10w diseharge AES/MS ++/+++ ++/+ ++/+ ++/++ ++/+
"+, unfavorable; ++. moderate; +++, favorable; NA, not applicable.
10 CHAPTER 1
certain analytical situations. For example, spark emission analyses are well
suited for rapid survey of metallurgical sampies with limits of detection
on the order of 1 ppm. They are limited by fairly severe matrix matching
requirements and a lack of depth-profiling capabilities. These capabilities
are nearly ideal for foundry applications. Secondary ion mass spectrometry
(SIMS) is characterized as having excellent depth-resolving capabilities
along with high absolute sensitivities. Unfortunately, SIMS is a rather slow
technique that requires a high level of operator skill. Thus, the technique
finds its major use in the analysis of materials associated with the production
of integrated circuits.
Of the techniques listed in Table 1-4, glow discharge spectroscopies
seem to have the most advantageous characteristics across-the-board. As
such, there is growing interest in the devices and their applications across a
broad range of analytical problems. Of current interest are the analysis of
bulk metals and alloys, the depth-resolved analysis of layered metallic sys-
tems, and the development of new methods for the analysis of nonconductive
materials such as geological materials, glasses, and ceramics. Throughout
this volume, the operating principles and applications of glow discharge
devices will be described in detail and compared with the competitive
methods listed here.
1.2. Glow Discharge Devices: Basic Operating Principles
Glow discharge devices would seem to offer a number of advantages

over other methods in the area of direct solids elemental analysis. In order
to familiarize the reader with the scope of this text, it is instructive to discuss
briefly the basic operational aspects of the devices. A much more detailed
description of glow discharge processes is presented in Chapter 2.
Glow discharge devices are reduced-pressure, inert-atmosphere, gaseous
conductors. (14,15) A depiction of their most simple geometry and structure is
presented in Fig. 1-1. A glow discharge is initiated by the application of a
Anode
Faraday Dark Spaee

~
6 i& Negative Glow
Cathode Dark Spaee
~~
~
Cathode
Figure 1-1. Schematic representation ofa simple gIow discharge device.

INTRODUCTION 11
sufficiently high voltage between two electrodes in contact with the discharge
gas (most typicaIly Ar). The potential difference (250-2000 V) causes the
breakdown of the discharge gas to form positively charged ions and free
electrons. The relative potentials on the cathode ( - ) and anode ( + ) result
in the establishment of electric field gradients such that positively charged
ions are accelerated to the cathode surface. The impinging ions transfer their
momentum to the surface (and lattice), setting off a cathodic sputtering
event. The products of the sputtering process are ejected atoms and smaIl
clusters of cathode material, ionic species, and secondary electrons. The
process of cathodic sputtering is the means of sampie atomization and the
basis for depth-resolved analyses. The secondary electrons sustain the dis-
charge through gas-phase ionization of sputtered material and discharge gas
atoms. Electrons are also efficient at producing excited state atoms of the
sputtered and discharge gas atoms. Evidence of these electron impact colli-
sions is seen in the characteristic luminous "negative glow" region. Impor-
tant as weIl in bulk plasma ionization are Penning-type collisions between
highly excited, metastable discharge gas atoms and neutral atoms of the
sputtered material. The result of these collisions is the formation of ions of
the sputtered atoms.
In summary, glow discharge devices are inherently capable of generating
a representative atomic population of asolid sampie and producing both
excited state and ionic populations of those atoms as weIl. Thus, as shown
in Fig. 1-2, the devices are directly applicable for bulk and depth-resolved
analyses by the traditional techniques of atomic absorption, emission, and
mass spectrometries. In addition, interaction of external light sources with
the sputtered populations aIlows for the use of the devices in atomic fluo-
rescence, optogalvanic effect, and resonance ionization spectroscopies. As
can be seen, glow discharge devices are indeed some of the most versatile of
the elemental analysis methods.(16)
1.3. Vo/urne Outline
In an effort to cover the most relevant applications of glow discharge

spectroscopies in the most informative way, the chapters ofthis volume have
been written by acknowledged research and application leaders in the
respective areas. It is they who are the most up-to-date regarding literature
coverage and can furnish the most insight into the nuts and bolts of source
operation in their particular fields. The chapters have been arranged so as
to build first on the fundamentals of glow discharge operation, discuss the
general application of the devices in the various atomic spectrometric modes,
and then look at specific fields of application. The authors have been encoUf-
aged to provide historical background in these areas as appropriate and also
12 CHAPTER 1
MASS SPECTROMETRY
hv
ATOMIC
ATOMIC EMISSION
FLUORESCENCE
LASER
RIMS lOGE
p
ATOMIC
ABSORPTION
CATHODE
,
t
OGE
Power
Supply
Figure 1-2. Analytical sampling modes of glow discharge sources.
to speculate on future directions in which they see advances being made. In

an effort to give completeness to the volume, spectroscopic methods of
interrogating molecular plasmas such as those employed in plasma etching
and deposition systems are also inc1uded.
The fundamental processes occurring within glow discharge devices are
covered in detail in Chapter 2. The authors, Duencheng Fang (Materials
Research Corporation) and R. Kenneth Marcus (Clemson University), dis-
cuss in detail the fundamental aspects of cathodic sputtering in terms of the
role that the choice of discharge gas and source operating parameters have
on sampie atomization. Methods for characterizing electron populations
in the plasma negative glow are described. Gas-phase collisional processes
inc1uding electron impact and Penning-type interactions are discussed with
reference to creating excited state and ionic populations of the sampie atoms.
INTRODUCTION 13
It is hoped that this chapter will provide for a better understanding of the
plasma processes that are manifested in the analytical applications.
Professor Edward Piepmeier (Oregon State University) describes in
Chapter 3 the application of glow discharge devices in atomic absorption
and fluorescence spectroscopies. In these applications, the devices are
employed as atomization sources for bulk solid and solution residue analysis.
The reduced-pressure, inert plasma is highlighted as being advantageous in
providing a nearly ideal environment for such studies.
The most widespread commercial application of glow discharge devices
is in the area of atornic emission spectroscopy. In Chapter 4, Professor J. A.
C. Broekaert (University of Dortmund) describes the fundamental aspects
of emission spectroscopy and how these plasmas generate useful emission
spectra. Also described is the evolution of the common Grimm-type cell
geometry and its many applications in bulk solid analysis. Methods ofmodi-
fying the basic design in order to optimize the source characteristics are also
discussed.
Chapter 5 concems the design considerations and applications of glow
discharge devices in the area of mass spectrometry. Professors Fred L. King
(West Virginia University) and Willard W. Harrison (University ofFlorida)
review the pertinent plasma processes responsible for the ionization of sput-
tered atoms and the roles of plasma parameters, operating modes, and mass
analyzers in producing quantitatively useful mass spectra. Applications in
the analysis of bulk metals, geological materials, and thin-film systems are
highlighted. Finally, the use ofthe devices in the area ofmolecular (organic)
mass spectrometry of gaseous or liquid sampIes is described.
Hollow cathode discharge geometries are the oldest and most widely
studied of the glow discharge sourees. Drs. Sergio Caroli and Oreste Seno-
fonte (Istituto Superiore de Sanita) outline, in detail, the historical develop-
ment of hollow cathode atomic emission devices in Chapter 6. Applied
primarily as line sources in atomic absorption spectrophotometry, the
devices have in fact been applied extensively in elemental analysis of bulk
solids, solution residues, and gases. The hollow cathode geometry is also
weIl suited for the analysis of metal filings and compacted nonconducting
sampIes.
One ofthe limiting features of glow discharge devices is the requirement
that the cathode (sampIe) be electrically conductive in nature. Drs. Michael
R. Winchester (National Institute for Standards and Technology) and
Douglas C. Duckworth (Oak Ridge National Laboratory), and Professor
R. Kenneth Marcus (Clemson University) describe in Chapter 7 two basic
approaches to the analysis of nonconducting sampIe types. First, methods
of compacting oxide powders in metal powder matrices are outlined, with
emphasis placed on sampIe preparation considerations. Second, the operat-
ing principles and prelirninary applications of radio-frequency-powered glow
14 CHAPTER 1
discharge devices are presented. The radio frequency discharges allow for
direct analysis of both conducting and nonconducting samples.
The ability of glow discharge devices to remove successive layers of
cathode material makes them applicable for depth-resolved elemental analy-
ses ofmetals and semiconductors. Dr. Hubert Hocquaux (lRSID UNIEUX)
describes this specialized area in Chapter 8. In particular, those source design
and operation considerations for atomic emission monitoring of such sys-
tems are detailed. The special requirements of optical spectrometers in these
applications are also described.
In Chapter 9, Drs. James Harnly (USDA-NCL), David Styris (Pacific
Northwest Laboratory), and Philip Rigby (University ofGreenwich) present
the applications of glow discharge atomic emission devices for the analysis
of solution residues, a growing area of interest. The techniques of furnace
atomization nonthermal excitation spectroscopy (FANES) and furnace ato-
mization plasma emission spectroscopy (FAPES) rely on sample volatiliza-
tion in a graphite furnace (tube) with subsequent excitation within the
hollow cathode discharge. As such, the techniques offer high-sensitivity,
multielement analysis of volume-limited samples.
The use ofhigh-intensity, monochromatic laser light sources has opened
up a number of analytical and diagnostic opportunities in the field of glow
discharge spectrometries. Professor Kenneth R. Hess (Franklin and Marsh-
all College) describes in Chapter 10 the wide variety of laser/plasma inter-
action schemes and the types of information they generate. The techniques
available inc1ude resonance ionization, optogalvanic effect spectroscopy, and
laser ablation of cathode material.
In Chapter 11, Bryan L. Preppernau and Professor Terry A. Miller (The
Ohio State University) review the optical techniques applied to those types
of glow discharges used in the semiconductor industry. These molecular
plasmas require extensive characterization of particle densities and energies.
The laser-based methods applied in these systems include single- and multi-
photon fluorescence (LIF), optogalvanic spectroscopy, and absorption
spectroscopy.
In the development of this volume, it was the editor's goal to put
together a comprehensive overview of the quickly expanding area of glow
discharge spectroscopies. It is hoped that the volume will be a useful refer-
ence source for those entering the field and practitioners alike.
References
1. F. Paschen, Ann. Phys. 50 (1916) 901.

2. H. Schuler, Z. Phys. 59 (1929) 149.
3. H. Schuler, and J. E. Keystone, Z. Phys. 72 (1931) 423.
4. R. F. Browner and A. W. Boom, Anal. ehern. 56 (1984) 786A.
5. J. F. Tyson, Spectrochirn. Acta Rev. 14 (1991) 169.
INTRODUCTION 15
6. A. Siriraks, H. M. Kingston, and J. M. RivielIo, Anal. Chem. 62 (1990) 1185.
7. M. L. Parsons, S. Major, and A. L. Forster, Appl. Spectrosc. 37 (1983) 41l.
8. G. Tölg, Analyst 112 (1987) 365.
9. W. Slavin, Spectroscopy 6(8) (1991) 16.
10. R. Bock, A Handbook 0/ Decomposition Methods in Analytical Chemistry, Wiley, New
York, 1979.
11. H. M. Kingston and L. B. Jassie, Introduction to Microwave SampIe Preparation: Theory
and Practice, American Chemical Society, Washington, D.C., 1988.
12. G. M. Hieftje, Fresenius Z. Anal. Chem. 337 (1990) 528.
13. J. D. Ingle and S. R. Crouch, Spectrochemical Analysis, p. 227, Prentice-HalI, Englewood
Cliffs, N.J., 1988.
14. A. M. Howatson, An Introduction to Gas Discharges, Pergamon Press, Elmsford, N.Y.
1976.
15. B. Chapman, Glow Discharge Processes, Wiley-Interscience, New York, 1980.
16. Analytical Atomic Spectrometry: Use and Exploration of Low Pressure Discharges (P. W.
J. M. Boumans, J. A. C. Broekaert, and R. K. Marcus, eds.), Spectrochim. Acta 46B, Nos.
2 and 4 (1991).
2
Fundamental Plasma Processes
Duencheng Fang and R. K enneth Marcus
2.1. Gaseous Discharges
When a sufficiently high voltage is applied across two electrodes immersed

in a gaseous medium, atoms and moleeules of that medium will break down
electrically, forming electron-ion pairs and permitting current to flow. The
phenomenon of current flowing through a gaseous medium is termed a
"discharge," also known as a plasma. The breakdown of the medium is
characterized by the transition of the gas from a poor electrical conductor
with resistivity (resistance x area/separation) of some 10 140hms' m to a
good conductor with resistivity (dependent on the particular conditions) of
~ 103 ohms· m. The potential difference between electrodes at which this
transition occurs is called the breakdown potential, Vb, which depends on
the identity and density of the gas, the electrode material and interelectrode
separation, and the degree of preexisting ionization.
Breakdown of this gas results in the formation of positively and nega-
tively charged ions and electrons. Electric fields developed between the elec-
trodes dictate that the positively and negatively charged species be
accelerated toward the cathode (-) and the anode (+ ),respectively. After
the initial breakdown of the support gas, collisional processes within the
discharge serve to produce more charged species. At this point the discharge
is said to be "self-sustaining" provided that a suitable voltage, Vn , which is
usually lower than the breakdown voltage, is continually applied.
Duencheng Fang and R. Kenneth Marcus • Department of Chemistry, Howard L.

Hunter Chemical Laboratories, Clemson University, Clemson, South Carolina 29634. Present
address of D .F.: Materials Research Corporation, Orangeburg, New Y ork 10962.
17
18 CHAPTER 2
Discharges may be classified as a function of their voltage-current

characteristics.(1) Figure 2-1 illustrates the most common types of discharges
and their operating regimes. Ofthe three major classifications, the Townsend
discharge, the glow discharge, and the arc discharge, only the latter two
have been applied extensively in analytical chemistry. In Fig. 2-1, Vb is the
breakdown voltage, Vn is the normal operating voltage, and Vd is the operat-
ing voltage of arc discharge.
The electrical characteristics of agas discharge can be best understood
by beginning with the Townsend discharge regime. This discharge is gen-
erally operated in the submillitorr pressure regime and is characterized by
having only a small degree of ion and free electron production. Following
the Townsend discharge is a transition region, resulting from the increased
energy exchange through collisions (owing to higher gas pressures), wherein
the electrical current increases while actually decreasing the required dis-
charge maintenance voltage.
After the transition region, a luminous glow forms between the elec-
trodes and is accordingly named a "glow discharge." At the onset of the
glow discharge regime, increases in the current do not change the current
density because the cathode surface is only partially covered by the dis-
charge; as such, no increase in voltage is required. This is classified as the
"normal" glow discharge regime. As shown in Fig. 2-2, as the current is
I I I I ~I
I I
(lJ
0> I (lJ
0> l' [I (lJ
H H 0::
I c: 1
0>
I I I
c: .c'u" .c'u" H
'"
(lJ
I .....0 I I -<Ul I .SI .c
UO>
c: H
I ....Ul
+J
I -<Cl
.....Ul
I ECl
"'
..... I .~I
U
....Ul
'" H Ul
(lJ
Ul.c'" (lJ Cl
c: u I C:O> E ~ I oc: 0~ I ~I u
~ Ul
'" c: I H 0
> 0· ... 0-< H
E-<Cl I H '"
E-<O:: I ZCJ I ~G I ~I -<
I I I I I
r ,, I I I I
I
,, I I I I
I I I I
- '.J I I
I
I
,- ~ -
I
I I I I I I I I I
-9 -7 -5 -3 -1
10 10 10 10 10 10
------.~ Current (A)
Figure 2-1. Operating regimes of various gas discharges. Adapted from Howatson. (1)
FUNDAMENTAL PLASMA PROCESSES 19
Dark
Discharge Abnormal Glow
Region Region
GI
:: Vb
.jl
.-t
o Normal Glow
:- Region
GI
tII
s..
~U
.....111
A Vn
-6 -5 -4
log(current, Al
Figure 2-2. Transition between normal and abnormal glow discharge operation.
further inereased, the discharge glow will eventually cover the entire eathode
surfaee. At this point, any inereases in discharge eurrent will result in an
inerease in eurrent density, requiring an inerease in the discharge voltage.
Discharges that display this type of inereasing i- V relationship are termed
"abnormal" glow discharges. It is the abnormal glow discharge mode that
is used most often in atomie speetroseopy. Analytieal glow discharge deviees
generally operate in reduced pressure (0.1-10 Torr), inert gas atmospheres
and at powers of less than 100 W.
As the eurrent is inereased further in the glow discharge, the eurrent
density beeomes so high that intensive heating of the eathode through bom-
bardment by filler gas ion speeies eauses thermal vaporization of the eathode.
Under these eonditions, the availability of high number densities of analyte
perturbs the potential fields and the i- V eharaeteristics of discharge become
"normal," i.e., the eurrent then inereases while deereasing the required dis-
charge voltage, as is the situation for a de are.(2)
Usually operating at atmospherie pressure, the de are is eharaeterized by
its large eurrents and bright discharge plasma. At typical operating eurrents,
10-1000 A,(3) the eathode surfaee is heated to the point that thermionie
eleetron emission beeomes a prominent eurrent-earrying meehanism. If the
eathode is made of some type of metal alloy, eathode heating will result in
vaporization of large amounts of material. Gas temperature in are discharge
20 CHAPTER 2
can be Up to 2 x 104 K with charged particle densities Up to 1016 cm - 3. The

combination of high vaporization rates and collisionally energetic plasma
has made dc arcs a mainstay in analytical spectrochemical analysis of metal-
lurgical sampies. (4-6)
2.2. Glow Discharges
Glow discharge devices have a rich history ofuse dating back to 1912.(7)
Since their introduction, experimenters have taken advantage of the glow
discharge's low operating powers and rich collisional environment for such
studies as atomic structure. (8) Another important early application of glow
discharges was in the first-generation mass spectrometers. (9) As will be dis-
cussed in detail in subsequent chapters, glow discharge devices have found
wide application in a number of fields of analytical spectroscopy including
atomic absorption, emission, fluorescence, and mass spectrometries. More
recent applications include laser-enhanced ionization and resonance ioniza-
tion mass spectrometry. In addition, the devices are now widely used in the
production of electronic devices and components in metal vapor lasers. The
following sections will give the reader insight into the fundamental processes
occurring in glow discharge devices and how they are utilized in analytical
spectroscopy.
2.2.1. Glow Discharge Processes
A glow discharge is initiated in a reduced-pressure environment when

the voltage applied between two electrodes exceeds the necessary energy to
cause breakdown of the rare gas, leading to the creation of electron-ion
pairs. The supplied voltage is usually 500 to 2000 V in a rare gas atmosphere
at apressure of approximately 0.1 to 10 Torr. The resultant operating cur-
rent, which is dependent on the gas pressure and the impedance in the power
supply, is usually in the range of 5 to 100 mA. The maintenance voltage is
dependent on the discharge current, fill gas identity and its pressure in the
source, cathode identity, and the particular electrode configuration. Figure
2-3 is a schematic diagram of a simple diode glow discharge source. Three
prominent regions are 0 bserved in most analytical glow discharge devices:
the cathode dark space, the negative glow, and the Faraday dark space.
While the two electrodes are specifically designated as the cathode and the
anode, in order to simplify instrumentation design the grounded source
housing is employed as the anode in most glow discharge devices.
In the vicinity of the cathode, electrons are repelled, resulting in the
creation of a positive space charge near the cathode. Consequently, the
majority of the potential difference between the two electrodes is dropped
(+) L ~
ANODE
FARADAY DARK
SPACE
.... '-
•
NEGATIVE
GLOW
CATHODE DAR K ......

SPACE
CATHODE
( -) {
Figure 2-3. Schematic diagram of a simple diode glow discharge configuration.
across a narrow region surrounding the cathode, known as the cathode dark
space (CDS) because of its noticeably low luminosity. Adjacent to the CDS
is the bright, collision-rich negative glow (NG) region. The visible emission
is the result of gas-phase excitation and ionization collisions and, therefore, is
where most analytical information is acquired. In addition, a glow discharge
plasma may exhibit Faraday dark space, positive column, anode dark space,
and anode glow regions. However, most analytical glow discharge sources
are designed in such a way that only the CDS and NG regions are observed
between the two electrodes because the other portions of the glow discharge
plasma offer little useful analytical information.
The eight distinct regions of anormal glow discharge plasma are shown
in Fig. 2-4 along with the potential and charged particle distributions. (10) In
a dc discharge, the negative potential of the cathode creates fields that accel-
erate positively charged gas ions to its surface. Ions colliding with the
cathode surface cause secondary electron emission, resulting in a net negative
space charge defining the Aston dark space. In the voltage range in which
most glow discharges operate, the secondary electron (r) emission
coefficient, ri, is on the order of 0.1 electronjion for argon ion
bombardment.(ll) The electrons, which have not yet been accelerated by
the cathode potential, can undergo inelastic collisions with gaseous species.
Emission from these species characterizes the cathode layer. Electrons that
pass through the cathode layer without experiencing any collisions may
22 CHAPTER 2
Aston Dark Cathode Dark

Space Space Anode Glow
1-) i\1 , Im,,,;:::, ,,,. •

t
}I"
t
Cathode Layer
Negative
Glow
Positive Column
Anode Dark
Space
Emission Intensity ~
r--J
Potential Distribution -,
'I'"~ ",,'''' r\ .' ,
' " S.." '""" lL?\ \2? <::>

,
Negative Charges t=:-----~----------~

Positive Charges
J
Negative Current
Density
Figure 2-4. Properties of the different regions of a dc glow discharge device.
acquire energies Up to that of the cathode fall(12) and as such have a low
probability (low cross section) for excitation or ionization collisions. This
region of low luminosity is the CDS. The CDS is characterized by its rela-
tively high positive charge density, p, such that if V is the potential fall
across the dark space and D its length, p = VI D 2 for a planar diode glow
discharge device. (13) The vast majority of the total discharge potential drops,
fairly linearly, (13) in this cathode fall region. The actual thickness of the CDS
is an integral consideration in source design. At the anode edge of the CDS,
a Iarge number of electrons exist because of multiplication via ionization
processes ; therefore, the highest density of ions is present at this, the begin-
ning of the NG. (14)
The onset of the NG region corresponds to the position where the fast
cathode e1ectrons have lost a large fraction of their energy through elastic
collisions with other particles and, therefore, a large population of electrons
with energies from ,,-,25 eV to near thermal values is found.(15) The actual
distance over which this energy loss occurs, the CDS thickness, can be
approximated as being three times the Debye length (AD)
(2-1)
where k is the Boltzmann constant, Te the electron temperature, So the

permittivity of free space, ne the electron density, and e the electron charge.
The resultant range of electron energies is low enough to allow electron
impact excitation and ionization collisions with the various atomic species
present.
The NG is the region of the discharge with which most people are
familiar and is of most importance in analytical chemistry applications. Two
general groups of electrons enter the NG region: fast secondary electrons,
which have not undergone collisional energy losses, and slow electrons,
which may be collisionally cooled secondary electrons or those created as a
result of an ionization reaction. Fast electrons are only capable of ionizing
collisions while the slow electrons may either excite atomic (or molecular)
species or ionize excited state species. The NG is characterized by its high
luminosity, which is the result of collisions of slow electrons having densities
of 109-10 11 cm- 3. The electron density is matched by an equal (order of
magnitude) number of positive ions making the NG an essentially field-free
region. The electric field strength in the NG region is negligible relative to
that in the dark space as reported by Aston. (13,16) The intrinsic electric field-
free characteristic in the NG region has been further confirmed for analytical
devices by electrical probe experiments of Fang and Marcus.(14) They used
a single cylindrical Langmuir probe system to characterize a planar diode
analytical glow discharge device (similar to Fig. 2-3). Their findings showed
that the plasma potential in the NG region is almost invariable, less than
1 V difference, in comparison with changes in discharge voltage between 800
and 1000 V.
Electrons diffusing from the anode end of the NG have experienced
enough excitation/ionization collisions to deplete their energies. This region
of low luminosity is called the Faraday dark space. In the Faraday dark
space, electrons begin to feel the effects of the positive anode potential and
are accelerated toward it, albeit with a relatively low field strength. As the
electrons gain energy, they are again capable of excitation collisions in the
positive column. Depending on the source size and discharge pressure, the
positive column may be the largest region of the discharge. Eventually,
electrons may be accelerated to the point where they are only capable of
ionization. This region, analogous to the CDS, is called the anode dark
24 CHAPTER 2
space. Likewise, as in the cathode region, dose to the anode surface there
exists an anode glow, the last region in the dc glow discharge.
The number and size of the regions in a glow discharge plasma depend
on the size of the vessel and pressure of the fill gas. In general, glow dis-
charges employed in analytical chemistry applications are made up only of
the three regions shown in Fig. 2-3. In actuality, a glow discharge can exist
without many of the discrete regions but never without a CDS. (10)
2.2.2. Alternate Operating Modes
To this point, discussions of the fundamental glow discharge processes

and electrical characteristics have been limited to relatively simple, diode-
type source geometries, i.e., a single metal cathode and its adjacent NG in
contact with a single counter electrode (generally the grounded vacuum
chamber). The use of dc voltage sources has also been assumed. There are
two variations of this model that should be mentioned here. The first, the
hollow cathode effect, deals with operation ofthe source with the NG region
trapped within a cathode cavity. The second, radio frequency-powered glow
discharges, entails maintaining the discharge with high-frequency potentials
rather than simple direct current. Both topics will be discussed in greater
detail in Chapters 6 and 7, respectively. It must be stressed that these modes
of operation still rely on the fundamental plasma processes that characterize
glow discharge devices in terms of analyte atomization, excitation, and
ionization.
2.2.2.1. Ho//ow Cathode Effect
By placing two planar cathodes (with a common anode) some distance

d apart, they may act as two discrete glow discharges. As d is decreased so
that the respective NGs coalesce, the operating voltage will decrease if in
constant-current operation or the current will increase if in constant-voltage
regulation. The ability to operate the resultant discharge at a voltage lower
than the sum of the distinct discharge voltages, for a constant current, is the
simplest case of what is generally termed the "hollow cathode effect." The
effect occurs when operating pressures allow a single NG to form between
two planar cathodes or inside of a cup-shaped cathode. (17)
The drop in maintenance voltage in hollow cathode geometries can be
attributed to enhanced electron production. In conventional (simple diode)
electrode configurations, the required electrons are generated by gas-phase
ionization near the boundary of CDS and NG or as a by-product of sputter-
ing. In a hollow cathode arrangement, secondary electrons ejected from the
cathode surface are accelerated toward the CDSjNG boundary where they
may ionize atoms or molecules, producing electrons. These electrons may
diffuse through the NG and eventually be drawn toward the opposite
CDS/NG boundary by its positive space charge, or possibly enter some
small distance into the opposite CDS. If the electron travels elose enough to
the opposing cathode wall, it will eventually be repelled backward with a
net gain in kinetic energy. In this way, electrons are trapped within the
cathode body and attain higher energies than in a diode electrode geometry,
resulting in enhanced electron multiplication through ionizing collisions. In
this way, the required discharge maintenance voltage is lowered relative to
two independent discharges. It is the enhanced production of electrons,
which can excite and ionize atomic species in the plasma, that leads to the
widespread use of hollow cathode discharges in atomic spectroscopy.
2.2.2.2. Radio-Frequency-Powered Discharges
As seen to this point, the flow of charged particles (current) is implicit in

glow discharge plasma operation. Positive ions are attracted to the cathode,
followed by the release of secondary electrons fed through the power supply
from the anode. As such, both electrodes must conduct electricity (electrons).
Unfortunately, solid analytical sampies (cathodes) are not only metals and
alloys, but also glasses, ceramics, and the like, which are electrically noncon-
ductive. As will be discussed in Chapter 7, analysis of these types of sampies
employing dc voltage sources requires modification of the sampie matrix to
become conductive in nature. This involves grinding and mixing the sampie
with a metal powder and pressing a composite conducting sampie.
To address the inability to atomize nonconducting matrices in glow
discharge sputter deposition systems, Wehner and co-workers(18) proposed
the use of high (radio)-frequency potentials to power the plasmas. Very
briefly, the placement of a high voltage on the surface of a nonconductor
induces a capacitor-like response where the surface acquires the applied
potential only to be neutralized by charge compensation by (depending on
the polarity) ions or electrons. The result is no net current flow and an
unsustained discharge. Rapid polarity reversals of voltage pulses allow for
rapid charge compensation and reapplication of the desired high voltage,
overcoming the inherent decay time constant. To achieve a "continuous"
discharge, pulse frequencies on the order of 1 MHz are required. A necessary
by-product of the capacitor-like response is the self (dc) biasing of the elec-
trodes such that the smaller of the two electrodes acquires an average nega-
tive bias potential sufficient to maintain the discharge processes, establishing
it as the cathode. Duckworth and Marcus(19) have demonstrated the utility
of this technology for mass spectrometric analysis of such materials as alloys,
oxide powders, and glass sampies.
26 CHAPTER 2
2.3. Atomization via Cathodic Sputtering
As an atomic spectrochemical source, atomization of the sampie is

the first step in analysis. The atomization in a glow discharge device is
accomplished by "cathodic sputtering." Cathodic sputtering makes glow
discharge devices useful in analytical spectrometry as atomization sourees,
for it provides a means of obtaining directly from asolid sampie an atomic
population for subsequent excitation and ionization.
2.3.1. Cathodic Sputtering
The first observation of metal deposits sputtered from the cathode of a

glow discharge device was reported in 1852. (20) At that time, cathodic sputter-
ing was considered a nuisance because it caused erosion of the electrodes
and led to undesired deposits that blackened glass walls and observation
windows of discharge tubes. This attitude changed greatly as interest in glow
discharge sputtering has been nourished over the years by experimentalists in
the fields of electrical engineering, material analysis, physics, and analytical
chemistry. (21) All of these disciplines are attempting to clarify, enhance, and
apply various aspects of glow discharge sputtering phenomena.
When a metal is bombarded by high-velocity (>30 eV) ions, atoms can
be ejected from the surface. The phenomenon is referred to as "sputtering. "
In a glow discharge, the positive gas ions in the NG region, having energies
corresponding to the neutral gas temperature in random motion (~kT),
may be accelerated across the cathode fall region if their motion brings them
close to the CDSjNG boundary. An energetic ion impinging on the solid
surface is either backscattered from a surface atom (a low-probability event,
estimated at 10-3) or penetrates the solid and transfers its energy to surface
atoms. The potential energy of the bombarding ion goes into effecting secon-
dary electron emission, while its kinetic energy and momentum are trans-
ferred to lattice atoms through a number of elastic and inelastic collisions.
This dissipation results in a collision cascade in the lattice as shown in Fig.
2-5, propagating in random directions in the vicinity of the collision site and
lasting about 10- 12 S(22) for a given ion impact. (A simple mental picture of
the resultant processes is a three-dimensional billiards break.) Recoil atoms
acquiring velocity vectors near the surface normal can escape the solid and
enter the gas phase if their energies are greater than the surface binding
energy. (23)
This bombardment results in the emission of atoms, secondary elec-
trons, ions (positive and negative), photons, and atom clusters(23) from the
cathode surface. However, neutral atoms make up the vast majority of the
ejected particle flux. In early glow discharge sputtering studies, Von
Hippel(24,25) found in spectrograms oflight takenjust at the cathode surfaces
Incident Ion Reflected Ions
CB and Neutrals Secondary

Electrons
Sputtered
Atorrs
Surface
••
•• • ••
•••••••
••••••••••
Collision • May terrninate or Result in the ejection
Sequences within the target of a target atom
Figure 2-5. The process of cathodic sputtering.
that the only spectral lines of the cathode material that appeared were the
resonance lines of the atom. From this he conc1uded that the material is
given off in an atomic state and arrives in the NG by a diffusion process.
These studies laid the groundwork for the use of glow discharges for bulk
so lids, elemental analysis.
It is the process of cathodic sputtering that is exploited in most glow
discharge applications as the means of sampie atomization. It is relevant,
therefore, to discuss the fundamental aspects of sputtering that have been
determined in high-vacuum ion beam experiments. The discussions that fol-
low are intended to illustrate basic phenomenology, and not to assess the
state of the art in high-vacuum sputtering theory.
2.3.2. Spulter Yield
The most basic measure of sputter efficiency is the sputter yield, the
ratio of the number of atoms sputtered from the surface to the number of
incident sputtering partic1es, usually expressed as the number of atoms per
ion. Sigmund(26) has presented a general theory of sputter yields deriving the
28 CHAPTER 2
following equation, semiquantitatively evaluating the physical parameters

affecting them. The sputter yield (S) is given by
(2-2)
where a is a function of the relative masses and the angle of incidence of

the incoming ion,(27) MI and M 2 are the respective masses of the ion and
sputtered atom, Eis the incident ion energy, and Uo is the surface binding
energy that must be overcome for sputtering to occur. The third term is
called the mass transfer term. Those parameters used in Eq. (2-2) will be
discussed in detail in the following subsections.
The practical (experimental) expression for sputter yield is defined as
the number of sputtered atoms per primary ion colliding with the solid
surface :(28)
S = 9.6 X 104 (W/ M . i+ . t) (2-3)
where W is the weight loss (in grams), M is the relative atomic mass of the
sputtered species (in grams), i+ is the ion current (in amperes), and t is the
sputtering time (in seconds). The bombarding ion current is related to the
total discharge current, i, by
(2-4)
where ri is the number of secondary electrons released, on the average, by

one ion. For argon, the gas most often used in glow discharge devices, this
value is approximately 0.1.
2.3.2.1. Mass of Sputtering Ion
In glow discharge devices, as weIl as most other techniques utilizing

sputtering, ions of noble gas atoms are used as the primary sputtering species.
These ions are quite advantageously applied in these applications because
of their chemical inertness, ease of ion formation, and availability of the gas
in high purity. However, for example, cesium ions have found widespread
use in organic mass spectrometry. (29)
The mass transfer term of Eq. (2-2), 4M1M 2 /(M I + M 2 )2, maximizes
as Mt!M 2 approaches unity. Therefore, Ar+ would be the best choice for
sputtering the first row transition elements, Kr + for the second, etc. The
other mass factor in Sigmund's theory, a, increases from 0.17 to 1.5 over a
target atom-to-sputtering ion mass ratio of 0.1 to 10, so that lower ion
masses are favorable. For this reason, 4°Ar+ shows better overall sputtering
characteristics than 84Kr+ or 13IXe+ even though the latter two would deliver
more kinetic energy, and momentum, at the cathode surface.(30,31)
2.3.2.2. Angle of In eiden ce
The angle of incidence between the incoming ion and the cathode sur-
face has a large effect on the subsequent sputter yield. (32) Figure 2-6 illustrates
the effect of bombarding angle of rare gas ion beams on a polycrystalline
copper target. The enhancement at angles away from normal incidence
comes from the increased probability of the collisional cascade propagating
back to the cathode surface. At severe angles (>80°) the incoming ion is
more likely to reflect off the surface without any penetration or momentum
transfer. The high angular dependency explains, in part, the general enhance-
ment in glow discharge sputtering rates as the cathode surface is sputter-
roughened. The angle of incidence also affects the relative elemental sputter-
ing yields in high-vacuum sputtering systems.(33)
2.3.2.3. Incident Ion Energy
Equation (2-2) states that the sputter yield of a given target material is
proportional to the energy of the bombarding ion. In practice this is not the
60 90
--+ a (degrees)
Figure 2-6. Effect of angle of incidence in ion beam sputtering.(32)

30 CHAPTER 2
case. As illustrated in Fig. 2-7, at relatively low ion energies, sputter yields
increase rapidly (nonlinearly) with energy up to about 100 eV.(34) From this
point, the sputter yield is seen to increase fairly linearly with energy, until a
plateau is reached at energies on the order of 1 keV. The leveling off of
sputter yields for ion energies larger than 1000 eV is the result ofion implan-
tation phenomena, where the ions begin to become embedded within the
lattice. The penetration depth for a l-keV Ar+ ion is roughly 10 A in Cu.
The exact ion energy at which these transitions occur are dependent on the
specific sputter ion-target atom pair.
2.3.2.4. Target Material

The largest variable in sputter yields, and most important in multiele-
ment analyses, is the identity of the target material. Equation (2-2) indicates
that sputter yields are dependent on the ion-atom mass ratio. Wehner(31)
has shown, though, that there is a more complex relationship than explained
3·0
•
c:0
"E
E
2·0
S
"0
Gi
'",
0>
'E
.!!
"'c.5
Cf)
1·0
1000
Ion energy (eV)
Threshold
energy
(Er)
Figure 2-7. Effect of incident ion energy on sputter yield; Cu target, Ar ions. (34)
by simple mass considerations. Figure 2-8 is a compilation of the sputter

yields obtained for various elements under sputtering conditions close to
those found in glow discharge devices, 400-eV Ar+. The author found that
the sputter yields seem to vary according to the d-shell filling. For example,
copper, which has nine d-shell electrons, has a higher yield than iron (six d
electrons) even though it has a larger mass and thus should sputter less
readily. The d-shell filling dependence is due to the fact that as the d shells
are filled the respective atomic radii decrease, increasing the atomic density
in the matrix. High atomic densities decrease the ion penetration depth,
resulting in more efficient energy transfer to the surface and more sputtering.
In the sputtering of multicomponent targets, the binding energy for
each element in the alloy is determined not only by the bulk composition but
also by the surface concentration, which differs from the bulk composition
because of differential sputtering (see Section 2.3.4). The ejection probability
of the atoms sputtered by momentum transfer in the collision cascade is
inversely proportional to its surface binding energy, (26) thus
(2-5)
where c and Co are the respective surface and bulk composition of each
component A and B, and Eb is the surface binding energy at that combina-
tion of bulk and surface compositions.
2.3.2.5. Target Temperature
At low bombardment energies «1 keV), the sputter yields tend to

decrease with increased temperature because of the annealing of more loosely
bound atoms on the surface (such as created in a previous ion impact) to
Ag
2·6
_ 2·4 Argon
>- 2·2
:: 2·0 Au
"
l
Cf) 1·8 Pd
> 1·6
~. 1·4
0
<t 1·2
er Ni Pt
Ci 1·0 Co G
Ru
.Al Fe· e Re Ir
u.
.
"0 0·8
Mo Os
]! 0·6 Hf W
S. V Zr
.
Be Nb Th
>- OA .1 Ti Ta
0·2 C
0 10 20 30 40 50 60 70 8090
Atomic number
Figure 2-8. Sputter yields for various elements under bombardment of 400-eV Ar ions.(34)
32 CHAPTER 2
positions of stronger binding. In addition, at surface temperatures approach-

ing the target melting point the "relaxation" of the metal lattice lowers the
binary momentum transfer efficiencies. At high bombarding ion energies
(> 10 keV) when metals are heated to within 250°C of their melting point,
"thermal sputtering" becomes superimposed on binary collisional sputtering
so that the apparent sputter yield increases with temperature. (35) Cathodic
sputtering in a glow discharge, with relatively high power densities, is prone
to suffer from thermal effects. For metals and some oxides, the sputter yield
is reported to be independent of temperature up to about 250°C below the
melting point.(35) For some oxides, the sputter yields have been reported to
be temperature dependent with "thermal" sputtering (i.e., a higher sputter
yield than that predicted by Sigmund's collision theory of sputtering)
reported to occur. (36) While the effects can be severe, most glow discharges
employ sampie cooling schemes that nearly eliminate thermal complications.
2.3.3. Spulter Rate
The analytically pertinent quantity describing the sputtering phenom-

enon is the sputter rate. The sputter rate of a system describes the amount
of cathode (sampie) material removed per unit time. This of course is a
direct result of the sputter yield under a given set of conditions. The sputter
rate of a sampie cannot be thought of as a sum of the rates of the sample's
constituents; the chemical (metallurgical) form of the sampie mayaiso play
an important role. For sputter-based techniques, the sputter rate may loosely
define the capabilities of the source in that it defines the rate of sampie
introduction.
Stocker(3?) has studied the atomization properties of a planar dc glow
discharge employing aseparate collector electrode to measure weight losses
in a neon-argon discharge gas mixture. He found that the sputter rate varied
as {ijp)2.5 (where i is the current and p is the discharge pressure) under a
variety of discharge conditions. Musha(38) obtained similar sputtering
characteristics using molybdenum electrodes in a parallel planar configura-
tion (approximating a hollow cathode discharge). By passing resonant radia-
tion from aseparate molybdenum discharge between the plates, absorbance
measurements were made. Measurements were also made using a single
cathode acting as a conventional glow discharge. Musha found that atomiza-
tion with the hollow cathode configuration varied with i 2.5, while that of the
planar discharge had an i 2.2-2.4 dependence. These results seem reasonable
in light of the fact that the bombarding ion energies should be similar at the
cathode surface even though the single-cathode discharge operates at a much
higher potential. A much greater difference was observed when only the
molybdenum atomic emission was monitored for the hollow and planar
cathode geometries. In the hollow cathode configuration, the emission
intensity varied with i 3.8 , while the corresponding dependence in the planar
cathode case was i 1.2-1.7. The enhancement in the hollow cathode emission
can be attributed to the increased electron activity in the NG region.
Fang and Marcus(39) studied in detail the sputter rate dependencies on
the discharge current and gas pressure in planar, diode glow discharge. As
shown in Fig. 2-9, the studies revealed adefinite i 2 dependence over a range
of discharge pressures. The relative slopes of the response functions show
much stronger dependence on the current-squared for the lower argon
pressures. This is most likely the result of the significantly higher voltages
of the low-pressure plasmas. Lower amounts of redeposition of sputtered
atoms are also a likely contributor. The authors also found that the sputter
rate is proportional to the discharge power (i. V) and the reduced power
[i( V - Vo)]. The reduced power is based on the discharge operation voltage
beyond that required to cause breakdown (Vo) of the gas, which is a function
of the source pressure. Though the sputter rates in Fig. 2-9 exhibit distinct
pressure dependencies, the data in Fig. 2-10 show that the use of reduced
power units removes the influence of pressure on the sputter rates. Plots of
the simple discharge power dependence, though linear, show definite dis-
charge pressure effects.
As illustrated in Fig. 2-8, sputter rates (yields) will certainly vary accord-
ing to the identity of the cathode (target) material. Furthermore, as will be
alluded to in the next section, sputter rates of multicomponent alloys of
different lattice structure (phase) can be different. In general, however, the
sputter rates across a wide variety of matrices are still within the range of
elemental values shown in Fig. 2_8.(14.28,40)
80
Cl 60 0 1.5 Torr
E
vi •
C
2.0 Torr
3.0 Torr
.2'" 40
• 4.0To"
j; 0 5.0 Torr
Cl
'ij
20
•
6
6.0 Tor,
8.0 Torr
~
1000 2000 3000 4000 5000

2
Current squared, mA
Figure 2-9. Effect of discharge current on sputter weight 10ss for OFHC (Cu) at various
discharge pressures ; Ar discharge gas. (39)
34 CHAPTER 2
70
60
GI
e 50 o 1.5 Torr
.,
.; 40 •
o
2.0 Torr
3.0 Torr
~
30 • 4.0 Torr
~ • 5.0 Torr
GI
"ii 20 • 6.0 Torr
~ 6 8.0 Torr
10
Reduced power, W
Figure 2-10. Effect of reduced power on sputter weight 10ss for OFHC at various discharge
pressures ; Ar discharge gas. (39)
2.3.4. Differential Sputtering
The sputter atomization step in glow discharges has certain advantages

over thermal volatilization methods such as arcs, sparks, and lasers. In
particular, elemental sputter yields over the range of bombarding ion
energies likely to be found in the devices vary by only a factor of 3-5/45 )
whereas elemental volatilities differ by many orders ofmagnitude. (41) Ideally,
the sputter yields (sampie introduction rates) would be uniform across the
periodic table. As is found in the sputtering of solids in high-vacuum environ-
ments [such as in secondary ion mass spectrometry (SIMS»), the sputtering
characteristics of the elements may be affected significantly when they are
incorporated in multicomponent alloys. Betz(42) and Betz and Wehner(43)
have reviewed extensively the work performed in this field. Complications
arise from the fact that atoms of different elements may sputter preferentially
from the cathode surface. This "preferential sputtering" is not as significant
as the disparities in atomization rates found in thermal vaporization pro-
cesses. The effects that alloy constituents have on the overall sputter charac-
teristics of a sampie are even more complex.
The general phenomenon illustrated in Fig. 2-8 has important conse-
quences for techniques utilizing cathodic sputtering for multielement ana-
lysis, such as glow discharge methods and SIMS. (44) In multielement analysis,
the rate at which each given element comes offthe surface during the analysis
time frame depends on its relative sputter yield. Since these rates differ from
element to element, the phenomenon of "differential sputtering" may occur.
Differential sputtering occurs on a cathode surface as elements with high
sputter yields are sputtered away first, leaving those of lower yields enriched
on the surface. A simplistic view of how differential sputtering might occur
in a simple binary (say Cu: Al) alloy matrix is shown in Fig. 2-11. Starting
with al: 1 (atom: atom) mixture on the surface, copper atoms will be sput-
tered away preferentially, leaving the surface enriched with aluminum atoms.
As sputtering continues, aluminum atoms will be sputtered in larger numbers
than the copper until a new "layer" of copper atoms is exposed, at which
point copper makes up the majority of the flux. In actuality, this process
occurs at a very rapid rate, with equilibrium (steady state) sputtering
conditions reached after sputtering through a few monolayers. In fact, the
Surface; t steady state
Surface; t x
Bulk SOlid; t = 0
Figure 2-77. Depiction of differential sputtering in a Cu-Al binary alloy (1: I). Solid circJes
are Cu.
36 CHAPTER 2
eonservation of mass dictates that a steady-state flux eomposition must be

established. In the steady state, the relative surfaee eomposition of a two-
eomponent system (eAI eB ) ean be ea1culated from the simple relationship
(2-6)
where P is the ejeetion probability of the respeetive atoms.

As mentioned previously the absolute, and differential, sputtering rates
for eaeh element in an alloy matrix may not be simply proportional to
the pure elemental sputtering rates of the elements. The target material
eomposition and strueture also play important roles when quantitative infor-
mation is of interest. Hammer and Shemenski (46) used argon and xenon ions
at energies between 1 and 4 keV to bombard aseries of one- and two-phase
brass alloys with eoneentrations of eopper ranging from 18 to 48% to study
differential sputtering phenomena. The authors found that all the speeimen
surfaces studied were depleted of zine (the higher-sputter-yield element) for
all sputtering eonditions. However, signifieant differenees in the relative
Cu: Zn ratios were found between the two phases. The steady-state surfaee
eomposition was dependent on ion energy for the a (fce) phase, and energy
and mass for the a + ß (bee) phase, with the relative zine eoneentrations
deereasing with inereasing ion energy and mass. Greater deviations from the
original bulk eomposition were found in the mixed phase alloys, along with
a more rapid change in composition as a function of depth immediately
beneath the surfaee. Therefore, for analytieal applieations, differential sput-
tering effeets must be eonsidered and eharaeterized for a given sampIe type.
An important result eoneerning the differential sputtering effeets in the
analytieal use of glow diseharges has been obtained by Fang and Mareus. (47)
The authors used Auger eleetron speetroseopy to study the surfaee eomposi-
tions of a brass alloy before and after sputtering along with adeposited thin
film that was collected from the NG region during sputtering. In this way,
the Auger speetrum of the thin film would be representative of the steady-
state flux into the NG region. The results of these analyses are shown in Fig.
2-12. As can be seen, the surface composition after sputtering indicates an
enrichment of the low-sputter-yield element, copper, relative to the bulk
sampIe. The composition of the deposited thin film shows a nearly identical
Cu: Zn peak height ratio to that of the bulk sampIe. This indicates and
ensures that onee the diseharge has reached steady-state sputtering condi-
tions, the plasma composition does reflect the bulk sampIe composition.
While the absolute rates at which different sampIes sputter will vary, the
eonservation of mass dictates that within a given sampIe sputtering is stoichi-
ometrie. Of course, the fact that the sampled volume does reflect the actual
sampIe composition is a desired quality for any analytical technique. This
work shows, as weIl as a previous work by Boumans, (28) that the sputter
Cl
UJ UJ
a:
UJ
"C
5a. w
Z a
:za: P
UJ
a.
C!l
z
ä: UJ
"C
UJ
~
~ W
a.
(Jl Z
...:. "C b
(Jl Lu
~
Cu
(Jl
::!E
0 UJ
"C
~
Cl
UJ W
t::
(Jl
Z
"C
~ Lu
UJ
Cl
800 900 1000 1100
KINETIC ENERGY (eV)
Figure 2-12. Auger electron spectra of a brass alloy (72% Cu: 28% Zn) taken (a) before and
(b) after sputtering, and (c) of a deposited thin film.(47)
yields for simple binary alloys can be approximated from the relative elemen-
tal sputter yields.
2.4. Physical Characteristics of Glow Discharge Devices
Glow discharge devices are characterized by a more complex set of

processes than the easily controlled high vacuum sputtering systems. Figure
2-13 illustrates the cumulative collisional processes occurring in a glow dis-
charge device. We concem ourselves here only with the factors related to
cathodic sputtering. In the glow discharge, electric field gradients and
ion-neutral collisions control ion trajectories and energies. The majority of
the potential between the electrodes (~) is dropped across the CDS, which
accelerates positively charged discharge gas (argon) ions to the cathode
surface with secondary electrons accelerated away from the surface. [The
38 CHAPTER 2
Negative glow
.0 .0
.+
• .* ~
• • •
~
.0
•
~
~
• (+1
•
0 l
0 <;)
0
....
..."
Cathode dark
•
oa
0
...
,<I
u "
J
(- 1
Figure 2-13. Fundamental collision processes in glow discharge devices.
current flow in the electric field-free region (NG) is reduced by coulombic

collisions of electrons and ions.(23,48)] In actuality, sputtering ions generally
have only a fraction of' kinetic energy (-40%) because of charge exchange
reactions with neutral gas atoms(49) and the proximity of ion formation to
the cathode surface. The ion bombardment will sputter-release atoms, atom
clusters, (23) and ions of the cathode material as weIl as secondary electrons
(y)(l1) and UV photons.(50) The negative bias of the cathode dictates that
positively charged ions (M+), which make up about 1% ofthe total particle
flUX,(51) will be returned immediately to the cathode surface. Conversely,
negative ions (M-) will be accelerated away. The vast majority ofthe particle
flux is made up of neutral atoms (Mo), generally leaving the surface in their
electronic ground state with kinetic energies Up to 5-15 eV.(52-54) In the
pressure regime of0.1-l0 Torr, atomic mean free paths are less than 0.1 mm.
As a result of these short mean free paths, the sputtered atoms quickly
lose their momentum through elastic collisions with other discharge species
(atoms and ions). In this way, sputtered neutral atoms may be knocked back
to, or redeposited, onto the cathode surface. Sputter weight loss studies by
Harrison and Bruhn(55) indicate that up to 95% of sputtered atoms in a glow
discharge environment may be retumed to the cathode surface. The use of
directed discharge gas flow to sweep atoms away from the cathode surface
will be discussed in Chapter 3.
While the analytical spectra (e.g., absorbance, emission) obtained from
glow discharge sources can be characterized as being "atomic" in nature, it
must be stated that a number of studies indicate that the majority of atoms
in the NG region seem to originate as "molecular" species. (56-58) Spatially
resolved atomic absorption profiles of sputtered species indicate that atom
densities are at a maximum directly above the cathode surface (as close as
possible for sampling), with a sharp decrease through the dark space region,
and finally a second maximum in the region ofthe CDSjNG interface. From
the second maximum, atom densities drop almost exponentially as would be
predicted from diffusional losses. A typical spatial profile of this phenom-
enon is illustrated in Fig. 2-14. The initial maximum (not seen here) seems
to be re1ated to re1eased free atoms with the subsequent loss indicating a high
degree of redeposition. The appearance of the second maximum indicates an
"injection" of atoms within the plasma. The most probable mechanism for
such an increase is the dissociation of sputtered clusters, most likely by
energetic electrons andjor metastable atoms, whieh are concentrated in this
region of the plasma.
2.4.1. Positive Ion Energies

The energy of the positive ions impinging on the cathode surface is
an important quantity because this will determine the sputtered atom and
secondary electron yields in the plasma. The energy of a bombarding ion
depends on the number of collisions it experiences between its point of
formation (e.g., NGjCDS interface) and the cathode. Positive ions carry a
substantial fraction of the total discharge current in the cathode fall. The
motion of the positive ions is limited by symmetrie charge exchange
(2-7)
Each collision produces a fast neutral and an ion with only thermal
energy.(34,59) At typieal discharge voltages, collisions between ions and atoms
will tend to be e1astic, so that no energy will be lost through excitation. The
40 CHAPTER 2
Cathode Limit of cathode Anode

surface dark space surface
11\
1·0 i 1000 VI t 1400 V
I
E
~
o
~
.x.
.~ 0·8
'0
r::
o
0<
~ 0·4
...
o
...<.>
r::
o
•
.a
~ 0·2
.a
«
A (cm)
Figure 2-14. Spatially resolved absorption profiles of "atomic" species from a Ni target.(s.)
major source of energy loss will be through symmetric charge exchange. (60)
The produced thermal ion then is accelerated by the fraction of the fall
potential existing between that point and the cathode. The frequency ofthese
collisions will be determined by the source pressure, the fall potential, and
the dark space thickness.
It is now generally accepted that positively charged particles passing
through agas do not necessarily retain their ionic identity, but by capturing
and losing electrons may lose and regain their charge a number of times. It
appears probable that a certain percentage of positive ions formed in the
NG actually do reach the cathode with the full velocity corresponding to a
free fall. However, a much larger proportion reach the cathode with veloci-
ties far less than the fall potential because of successive collisions.
The energy of bombarding ions is generally determined by mass spectro-
metric sampling through a small hole in the wall of the electrode of interest.
Bodarenko(61) has studied the energy distribution of argon discharge gas
ions in both abnormal, planar, and hollow cathode glow discharge config-
urations. As shown in Fig. 2-15, the singly charged argon ions from the
a
b
o 2000 3000 o
E, (eV)
Figure 2-15. Kinetic energy distributions for argon ion species in (a) plan ar and (b) hollow
cathode configurations. (61)
planar discharge (panel a) are centered at an energy of about one-sixth of

the cathode fall potential while those sampled from the base of a hollow
cathode (panel b) have energies eloser to the applied potential. This is likely
due to the reduced dark space distance in the hollow cathode discharge.
Howorka et al. (62) used a nonconducting orifice plate mounted at the
end of a cylindrical hollow cathode to map radially the argon ion density of
the NG. As was the case with electron density distribution measurements (see
Section 2.4.3.1),(63) they found that the ion densities concentrated toward the
center of the cathode at increasing pressures, although the maximum for
each pressure was not on the cathode axis as was the case with the electrons,
but more toward the respective NGjCDS interfaces.
2.4.2. Electrical Characteristics
As discussed in the previous section, the kinetic energy of sputtering

ions is intimately related to the potential at which a plasma is maintained.
This will, of course, affect the sputter yield for a particular target. The
second component affecting sputtering rates is the discharge current. At each
discharge pressure there is a unique current-voltage (i-V) relationship that
controls the sputtering rate. Therefore, for a given discharge geometry one
must be concerned with the i- V characteristics and the effects that they have
on absolute sputtering rates. While sputtering rate is the analytical variable
in atomic absorption and fluorescence applications, it must be kept in mind
that for atomic emission and mass spectrometry applications, discharge con-
ditions wi11 also affect the extent of analyte excitation and ionization.
42 CHAPTER 2
2.4.2.1. Planar, Diode Devices
The relationship between discharge current and voltage at a given source

pressure in a glow discharge device, along with the effects of redeposition,
deterrnine the rate of cathodic sputtering. High-vacuum sputtering studies
of copper with argon ions over the range of energies most likely occurring
in glow discharges indicate that the sputter yields (No. of sputtered atoms/
incident ion) increase nearly linearly with ion energy. (34) Likewise, one would
expect the overall sputter rate of a material to increase proportionally to the
sputtering ion current.
For glow discharges operating in an abnormal mode (as do most sput-
tering plasmas) at a constant pressure, increases in discharge current must
be accompanied by increases in operating voltage. Fang and Marcus(39)
have reported the relationships between discharge current, voltage, and gas
pressure in a planar, diode glow discharge device. Figure 2-16a is a plot of
the operating voltage of the glow discharge source as a function of current
for a range of argon pressures from 1.5 to 8 Torr. These curves c10sely
resemble those determined by Tong and Harrison(64) who employed a similar
source design to sputter chromium and niobium disks in a lower-pressure
(O.3-1.0/Torr) regime. Dogan et al.(65) and Boumans(66) observed these sorts
of relationships in Grimm-type discharges. It is c1ear that at lower source
pressures the dependence of operating voltage on current is greater. The
response of voltage to increasing current reflects increases in current density
at the cathode surface. The pressure effects are attributable to the more
efficient (higher frequency) collisional processes occurring at high pressures,
requiring lower initial secondary electron energies for plasma ionization to
maintain the discharge. (11)
It must be emphasized that the emission of secondary electrons from
the cathode surface, with their subsequent acceleration by the cathode fall
potential, is the major means of maintaining plasma ionization. The electron
energy requirement also leads to the dependence of CDS thickness on source
pressure. Under high-pressure (collision frequencies) conditions, electrons
lose their energy efficiently, terminating the dark space at shorter distances.
The inverse relationship between source pressure and discharge voltage and
dark space thickness is an important factor in glow discharge source design.
Figure 2-16b illustrates a more specific relationship between discharge
voltage and current. A plot of voltage versus the discharge current squared
(mA2) shows linear relationships for all examined discharge pressures. Simi-
lar plots have not been explicitly depicted in the literature, although the
comparable i- V curves of others suggest that they mayaIso follow an i 2
term. Fang and Marcus's explanation for this relationship involves an Ohm's
law analogy where the resistance of the plasma is controlled by the discharge
current (V=iRai 2 ).(39) The physical quantity, which does in fact change
a
1100
o 1.5 torT
900
> • 2.0 torT
o 3.0 torr
700 • 4.0 torr
• 5.0 torT
• 6.0 torT
500 .. 8.0 torT
3oo+---~--r-~~-,--~---r--~--~
o 20 40 60 80
Current, mA
1100
o 1.5torT
900
> • 2.0 torT
.; o 3.0 torr
01
700 • 4.0 torT
~
'0 • 5.0 torT
> • 6.0 torT
500 .. 8.0 torT
300
0 1000 2000 3000 4000 5000
Curnnt aquared, mA 2
Figure 2-16. Current- voltage characteristics for simple, diode glow discharge device at vari-
ous pressures (Cu target). (a) Voltage versus current ; (b) voltage versus current squared.(3')
proportionally with the discharge current, is current density. As the current

density at the cathode surface increases, more charge carriers are being
forced to the same conductive area, effectively causing a bottleneck. The
response to the increase in current density is an increase in resistance and
therefore a higher operating voltage.
2.4.2.2. Ho//ow Cathode Devices
Most studies dealing with the comparison of the i- V characteristics

between planar and hollow cathode configurations are approximated using
a parallel plane configuration. This approach allows experimenters to vary
the distance between the plates, analogous to varying the cathode diameter,
and source pressure to determine the degree of the hollow cathode effect.
Guntherschulze(67) took this approach using two planar cathodes (50 cm2)
44 CHAPTER 2
spaced 1 cm apart and wired such that he could operate either a single
cathode or both at the same time. These studies showed that the CDS length
of the single cathode in a 450-V hydrogen discharge was 14 electron free
paths (efp's) while the comparable two-electrode discharge had a pair of
CDSs of only 5 efp's each. The lengths of the two dark spaces varied less
with pressure than those of the single-cathode case, which varies inversely
with pressure. The shortened dark space in the hollow cathode reflects the
maintenance of higher electron energies and densities in that geometry rela-
tive to the planar cathode case.
Little and Von Enget<68) determined that the discharge voltage dropped
linearly across the CDS of a hollow cathode going from near zero at the
NG/CDS interface to the applied voltage at the cathode surface. (In this
particular case, the cathode was at a negative potential and the anode
grounded.) Kirichenko et al.(69) found that while the CDS length is indepen-
dent of pressure from 0.1 to 1 Torr, there is a decrease as the pressure is
increased to 2 Torr. They also determined that a decrease occurred as the
current was raised from 30 to 200 mA; the change being from 4 to less than
1 efp. Those studies indicate that an electron emitted from the cathode sur-
face will reach the NG after few or no collisions, which would lower the
voltage requirements for a given discharge current as compared with the
planar cathode case. In fact, most hollow cathode discharges operate in a
near-normal discharge mode with very little or no increase in operating
voltage as current is increased. We are not aware of any specific mathemati-
cal functions that predict current-voltage relationships in hollow cathode
discharges.
2.4.3. Energetic Particle Densities and Energies
The preceding discussions have dealt with those factors that determine
the sputtering (introduction) rate and identity of analyte species sputtered
from the cathode sampie. Once in the gas phase, the atoms (and small
clusters) are subject to elastic and inelastic collisions causing excitation,
ionization, and possibly association/dissociation. These reactions are
depicted schematically in Fig. 2-l3. In terms of performing emission or mass
spectrometries, the principal collisions of interest are those between analyte
atoms and electrons or with metastable (excited state) discharge gas atoms.
The following discussions deal with the densities and energies ofthese impor-
tant gas-phase collision partners.
2.4.3. ,. Electrons
Three electron groups are believed to exist in the NG region of a glow

discharge :(70) group I are the secondary electrons emitted from the cathode
surface (so-called fast or primary electrons), which gain kinetic energy
through the CDS and attain electron temperatures (Te) of -20-25 eV and
number of densities on the order of 106 cm - 3; group II are the secondary
electrons of gas-phase ionization collisions with electron temperatures of
-2-10 eV and number densities in the range of 107_108 cm - 3; and group III
(ultimate or slow electrons) are the electrons from either group I or group
II that have experienced several elastic and inelastic collisions with particles
in the plasma and thus have electron temperatures of only 0.05-0.6 eV,
having densities in the range of 109_10 11 cm-3.
Electron energy (temperature) and density measurements in athermal
equilibrium environment can be carried out, in principle, by optical methods
that include the slope or two-line methods for excitation temperature
measurements and H ß line broadening for electron number density
determinations.(71) However, none of these methods are applicable in
diagnosing a glow discharge plasma. The reasons are that (l) the glow
discharge plasma is not at thermal equilibrium or even local thermal equilib-
rium (LTE). This excludes the validity of excitation temperature measure-
ments. (2) H ß line broadening is only useful/practical for plasmas that have
electron number densities larger than 10 13 cm- 3 , which is much larger than
those determined in glow discharges by other methods.
Given the properties and chemical environment of the glow discharge,
electrostatic (Langmuir) probe techniques may be the best way to measure
localized electron temperature (Te), electron number density (ne), average
electron energy «E >), and electron energy distribution functions
(EEDF).(14,72,73) Langmuir probes have been used widely in diagnosing
nuclear plasmas. The electron temperatures and number densities found for
different glow discharge plasma types are listed in Table 2-1 along with the
particular measurement technique employed. As can be seen, the ne values
measured by H ß line broadening are much higher than those determined by
Langmuir probes because of the fact that the H ß line broadening method
can only generate electron number densities larger than _10 13 ern- 3.
A typical experimental configuration for a Langmuir probe system is
shown in Fig. 2_17.(14) Basically, a Langmuir probe is a small metal wire or
disk submerged in a plasma (or flame) to which a dc bias is imposed. Accord-
ing to the polarity and magnitude of the applied voltage, either positively or
negatively charged species will be attracted to its surface, causing a resulting
current to flow through the probe circuit. The magnitude of the current is a
function of the number of charged species incident on the probe, the applied
probe potential, and its surface area. Scanning the applied voltage across a
range of values, say from -12 to +12 V, generates a current-voltage
response curve ofthe form shown in Fig. 2-18. As the scan progresses, posi-
tive ions and then electrons are attracted to the probe, with the magnitude
of the current as a function of potential being indicative of the particles'
46 CHAPTER 2
Table 2-1. Electron Temperatures and Electron Number Density in

Low-Pressure Discharges(75)
Tc n.
Plasma Fill gas pressure Power (eV) (10 12 cm -3) Ref.
MI}>" 3.0 Torr Ar 25W 1.87 15.0 43
MI}>" 3.0 Torr He 25W 4.48 10.5 43
RF" 2.0 Torr Ar 50W 3.71 1.00 10
RF" 2.0 Torr Ne 50W 4.74 0.85 10
HCD (Ni)" 2.3 Torr Ar 10 mA x 325 V 1.21 28
HCD (Ni)" 1.9 Torr He 10mA x 325 V 1.38 28
MI}>" 3.0 Torr Ar + Hg 25W 2.73 19.5 43
RF" 2.0 Torr Ar + TI 50W 3.45 2.00 44
RFa 2.0 Torr Ar + Ti 50W 3.19 0.15 44
RF" 2.0 Torr Ar 50W 2.41 0.80 44
HCD (Cu)" 1.0 Torr Ar 25 mA x 350V 2.24 63
HCD (Ni)" 1.0 Torr Ar 25 mA x 350V 1.98 63
HCD (steel)" 1.0 Torr Ar 25mA x 350V 0.83 63
GD (AI)b,C 4.0 Torr He + Hg 8mAh 0.05d 0.82d 69
GD (AI)b,C 12.0 Torr He + Hg 8mAh 0.07d 0.22d 69
GD (Cut 2.0 Torr Ar 10mA x 620 V 0.27 0.16 75
GD (CU)b 3.0 Torr Ar 10mA x 530V 0.27 0.16 75
GD" 15.3 crn 3 Imin Ar 40mA x 800 V 0.08' 203 g 44
GD" 29.1 crn 3Imin Ar 60mA x 800 V 0.1(1 287g 44
GD" 35.5 crn 3 Imin Ar 80mA x 800 V O.l1f 319K 44
GD" 17.5 crn3 Imin Ar 40 mA x 1000 V 0.07' 201 g 44
aMeasured by double cylindrical probe.
bMeasured by single cylindrical probe,
cDischarge tube.
dUltimate electron.
<Grimm-type glow discharge lamp with unspecified cathodes.
fDoppler temperatures.
gHIl method.
hNo operating voltage has been given.
kinetic energies and densities. Detailed theory and pertinent equations deal-
ing with Langmuir probe operation and data evaluation are described in
Refs. 14, 72, and 73.
Fang and Marcus have developed a low-cost computer-controlled single
Langmuir probe system. (74,75) This system has been employed to determine
the axial dependences (away from the cathode surface) of electron tempera-
ture, electron number density, average electron energy, and electron energy
distribution function under various discharge conditions for a planar, diode
glow discharge device. In an argon glow discharge, they found that the
electron number density is proportional to the discharge current, while the
axial density distribution in the NO region is fairly constant as shown in
Figs. 2-19 and 2-20 respectively. The electron number densities are found to
be of the magnitude of 10 10 cm- 3 .(14) The electron energy distribution
functions in the NO region are shown in Figs. 2-21 and 2-22 as a function
Ar gas
Langmuir
--
probe
~
-W ater
Vacuum
Figure 2-17. Typical experimental configuration for Langmuir probe measurementsY4)
10
.....~
8
~
6
.
.§.
I-
.
.
z 4
w
a:
a:
:::l
u 2
..
0
...
-2
-s o s 10
APPLIED VOLTAGE
Figure 2-18. Typical current versus voltage plot obtained from a Langmuir probe experiment.
of the axial position and discharge current, respectively. The electron popu-
lations are centered at energies of less than 2 eV, showing a depletion of
higher-energy electron populations relative to those predicted from a
Maxwell-Boltzmann distribution (which would be characteristic of a system
at thermodynamic equilibrium). Deviations from Maxwellian behavior are
also evident from differences observed in the measured electron temperature
values and those of the average electron energy. In the dc glow discharge
48 CHAPTER 2
8.0
..
?:
C
GI
6.0
o
+
2 Torr e.n.d.
3 Torr e.n.d.
.
"C
4.0 o 2 Torr i.n.d .
E.,. • 3 Torr i.n.d.
ii: 2.0
0.0
4 8 12 16 20 24
DIscharge current, mA
Figure 2-19. Effect of discharge current and pressure on electron (e.n.d.) and ion (i.n.d.)
number densities for a simple, diode discharge configuration (Cu target).(14) (Units are
10 10 cm- 3 .)
3.0
..
?:
c
GI • •• • •
"C
~
2.0 •
0 0
• 0 e
•
.a 0 2 Torr 12 mA
E • 2 Torr 15 mA
"
c
1.0
0
• 3 Torr 15 mA
c
e
ü
GI
iii
0.0 +--..---r--..----,,-----r---.---r--I
4 8 12 16 20
Axial dlstance, mm
Figure 2-20. Axial profile of electron number density at various discharge conditions (Cu
target)Y4) (Units are 10 10 cm- 3.)
studied here, (14) the average electron energies were consistently higher than
those of the electron temperatures (",0.9 versus 0.3 eV). Based on the way
these quantities are derived, this relationship again points to a relative deple-
tion of higher-energy electrons. Because the average electron energy is
derived from the complete distribution function, the authors tend to believe
that it is a more realistic characteristic of the plasma excitation conditions.
The observed depletion of the high-energy e1ectrons is likely related to
the first excitation threshold of the gaseous atomic species. Figure 2-23
depicts the electron energy distribution functions for the case of sputtering
a copper cathode at discharge conditions of2 Torr argon pressure and 8 mA
50~----------------------------~
D 5.44mm
• 7.26mm
• 9.07 mm
LL • 13.6 mm
o
w o 18.1 mm
w
0.0 1.0 2.0 3.0 4.0

Electron energy, eV
50.-------------------------~
b
40
::i D 5.44mm
30 • 7.26 mm
.(
• 9.07mm
LL • 13.6 mm
o
w o 18.1 mm
w
10
0.0 1.0 2.0 3.0 4.0

Electron energy, eV
50~----------------------------~
c
::i D 5.44mm
.( • 7.26mm
• 9.07 mm
u:
o • 13.6 mm
w o 18.1 mm
w
0.0 1.0 2.0 3.0 4.0

Electron energy, eV
Figure 2-21. Axial dependence of e1ectron energy distribution funetion (EEDF) (Cu target).
(a) 2 Torr, 12 mA; (b) 2 Torr, 15 mA; (e) 3 Torr, 15 mA.(14)
50 CHAPTER 2
60
&&&M
&& &
a
50 &&ocPcP &&
o. 0
Aa..... 0 c 5mA
40 .0 &
:::)
ci
&oe
&0.000 000 ·.0 &
•0 8mA
10mA
..:-
30
• 12mA
eOO...
0·0 0 &
C
..... 0 .0 0 15mA
w 20
.0 &
w •• • 0 • & 18mA
o ccCl:t:ICb. 00 0 &
10 •C .
De.o.&· 0
C Q]~oQ&
t.!o..liIqo!&.
0
0.0 1.0 2.0 3.0 4.0
Electron energy, eV
60
oovuao
o 00 b
50 00
o.•••••• 0
:::) 40 •
0.000000 •
0
c 10mA
ci 0.00 •••••• 0 • 0
•0 12mA
a'
30 .... .0. 0
..:- ea
[l]l:t:I .0 •
15mA
C
w 20 c.. 0 •0 18mA
w ,C Ce 0 .0:. 21 mA
ce. • 0
10 De·O.o
.~=-.~ 0".
"b~
0
0.0 1.0 2.0 3.0 4.0
Electron energy, eV
Figure 2-22. Effect of discharge current and pressure on EEDF (Cu target). (a) 2 Torr; (b)
3 Torr.(14)
discharge current.(75) The respective plots represent the MaxweII-Boltzmann

distribution based on (A) 0.286 eV electron temperature (caIculated), (B)
average electron energy of 0.832 eV (caIculated), and (C) the actual EEDF.
According to the two- and three-electron group models for low-pressure gas
discharges developed by Vriens,(73) the discrepancy between the three plots
indicates non-MaxweIIian behavior of group III electrons. The intersection
point of the curves indicates a depletion of electrons above the first inelastic
threshold (required for atomic excitation), which in this case is ",1.2 eV.
This value is fairly reasonable considering the electronic excitation of the
atomic species (Cu and Ar) in this plasma. The plots indicate that the
electron temperature best describes the electrons of energies > 1.2 eV, while
the average electron energy represents the bulk of the electrons in the plasma.
In detailed studies by Fang and Marcus, (75) the role of cathode identity
on the charged particle characteristics has been studied in order to ascertain
10
•
i\
30
8 A 20
•
;:)
6 10
ci •
IL 0
C 4 0.0 0.4 0.8 1.2 1.6 2.0
w
w
2
.... .....
0
.........
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Electron energy, eV
Figure 2-23. Maxwell-BoltzInann distribution functions determined by (A) electron tempera-

ture (0.286 eV), (B) average electron energy (0.832 eV), and (C) experimentally determined
EEDF. Insert shows the full distribution of (A). Cu target; 2 Torr, 8 mA.(75l
the origins, or lack, of spectrochemical matrix effects. The studies employed

cathodes of copper, various brass alloys, stainless steel, titanium, molyb-
denum, and a compacted oxide sampIe. In considering the average electron
energies and densities as the most important analytical characteristics, it was
found that there are no appreciable matrix effects. The electron energies and
densities were found to vary by less than 15% when the effects of discharge
voltage and cathode surface area were taken into account (i.e., constant
power densities). It must be mentioned that the densities of gas-phase ions
(discharge gas and sputtered species) are of the same order of magnitude as
the electron populations, but generally two- to fivefold larger.
In a hollow cathode discharge, Borodin and Kagan(76) used a Langmuir
probe to study electron energies and densities in the NG and positive column
of a parallel plane discharge using helium as the fill gas. They determined
that the majority of electrons in the NG exist at 10w energies (~2 eV)
although there is a substantial population of fast electrons of 19-25 eV
kinetic energy. The higher-energy electrons were not as prominent in the
positive column. The observed energies, however, do not account for the
high degree of emission observed from high-Iying helium atom and ion
states.
Subsequent studies of hollow cathode discharges were performed by
Kagan and co-workers(77) to determine the extent to which electrons existed
at energies above 25 eV by employing an electrostatic energy analyzer to
sort the electrons. The authors found that electrons of higher energy do
indeed exist with a large number centered around the cathode fall potential.
These would be electrons formed at the cathode surface that had not experi-
enced inelastic collisions as they traversed the NG from side to side.
52 CHAPTER 2
Howorka and Pahl(63) found similar electron energy distributions in dis-

charges operating at lower pressures (0.1-0.7 Torr). As an extension ofthese
studies, a Langmuir-type probe was used to map the electric fields and
electron densities. They found that over the length of the hollow cathode
(7.4 cm long; 2 cm diameter), the field changes were on the order of
0.1 V ern-I indicating that the electron and ion densities were equivalent, as
would have been expected. The authors also studied the radial electron
distribution in an effort to explain observed spatial variations in hollow
cathode emission studies. Figure 2-24 illustrates the radial electron densities
as a function of discharge pressure. The density maximizes toward the
cathode walls at high pressures and. toward the middle at lower ones, imply-
ing that there may be CDS thickness dependency.
2.4.3.2. Metastable Atoms

In addition to their chemical inertness, ease of handling, and ready
availability, rare gas atoms have the fortuitous characteristic of high-lying,
long-lived metastable states. Metastable (triplet) electronic states have long
radiative lifetimes because radiative transitions to lower-energy levels are
spin-forbidden. Table 2-2 lists the energy levels and spectroscopic notation
for the metastable levels of the rare gas elements. As can be seen, the energy
levels for these states are quite high relative to the excitation and ionization
potentials of most other elements. In all cases, the lifetimes are longer than
the collisional lifetimes of gaseous atoms. Therefore, given the extended
lifetimes of the states, metastable atoms are effectively sources of potential
1.0
-0.7torr
___ 0.4 torr
" /"
/~ 1
I
- · - 0 . 2 torr
•• _••••• 0.1 torr
,,,-......
~),.,V "
- \
, -;- .\
I
I
I .'.
" •
• \
\
\
I •
~
~ 0.5
I
I '. \
•
·· .\,
" I \
I
I • I
I I
.\
I I
I
I I I
I I
I I
I
10 5 o
mm
Figure 2-24. Radial eIectron densities as a function of discharge pressure for a hollow cathode
configuration. (63)
Table 2-2. Metastable Energy Levels for Rare Gas Atoms
Energy
Spectroscopic notation Ionization potential
Atom (L-S) ern-I eV (eV)
He 2 1S 166272 20.6 24.5
23S 159850 19.8
Ne Ipo 134820 16.7 21.6
3P2
134043 16.6
Ar 3po 94554 11.7 15.8
3P2 93144 11.5
Kr 3po 85192 10.5 14.0
3P2 79973 9.9
Xe 3po 76197 9.4 12.1
3P2 67086 8.3
energy within the plasma, with number densities on the order of 10 11 _

10 12 cm- 3 .
The actual role of metastable atoms in glow discharge devices is still of
considerable debate.(ll,78-88) Their roles in analyte excitation and ionization
will be discussed in Section 2.5. Hess and co-workers(78-80) have studied
extensively the discharge parameters that affect the production and quench-
ing of argon and neon metastable atoms in simple, diode geometry devices.
The role of metastables in discharge maintenance has been established
through optogalvanic effect (OGE) spectroscopy in a hollow cathode where-
in the discharge currents are seen to decrease (in a constant-voltage mode)
as the metastable atom states are depopulated by a laser (see Chapter 10).(78)
Thus, the relative ease of ionization of metastable atoms, relative to direct
ionization from the ground state, is important for maintaining the total
discharge ionization. (81-83) The density of metastable atoms (as determined
by atomic absorption spectrometry) was shown to be reduced by use of
quenching species introduced into the plasmas. (79,80,84,88) Gases such as
methane, nitrogen, and mixtures of rare gases have clearly shown that colli-
sional energy transfer from a metastable atom to a target of lower ionization
potential is a very efficient process. These sorts of findings point out the
importance ofvacuum integrity in glow discharge devices, as residual atmos-
pheric gas can greatly perturb the source operation characteristics.
Larkins(89) has recently demonstrated that use of argon discharge gas with
as Httle as 140 ppm of water vapor can reduce sputtering rates as much as
77%, with the actual percentage being a function of discharge conditions
and the cathode material.
As would be expected, the production of metastable atoms is a function
of source pressure (collision frequency) and discharge voltage / current (elec-
tron density and energy). Hardy and Sheldon(90) have presented a theoretical
model for the production of helium, neon, and argon metastable atoms in
54 CHAPTER 2
low-pressure (0.01-1 Torr) glow discharge environments. Most important

for the discussions here is the breakdown of the metastable densities as a
cumulative set of loss mechanisms. Figure 2-25 illustrates the ca1culated
cross sections for metastable neon atom depopulation via diffusion to the
chamber walls (D), electron impact ionization (E), and quenching by colli-
sions with ground state atoms (TB) as a function of cell pressure. Compari-
sons between ca1culated density response functions and actual densities
determined by atomic absorption show much more extensive depopulation
at higher source pressures than predicted, with the density values peaking
at ",0.5 Torr. The authors suggest that the discrepancies seen for the higher
source pressures are likely the result of the nonideal spectroscopic sampling
of the plasma, rather than a flaw in their model. The observed response to
discharge pressure at constant discharge current (a maximum density occur-
ring over a limited pressure range) has in fact been observed by other
researchers as illustrated in Fig. 2-26a, which is typical for glow discharge
devices.(70) Likewise, the response to increases in discharge current at con-
stant pressure (Fig. 2-26b) reflects the increases in electron density and
energy as the voltage is increased. These effects are general to all glow
discharge sources and, therefore, must be elucidated for particular source
N
g 10-'
n
~
e
"0
Q.
Ql
~
Ul
~ 10-8
-.J
TB
10-12
10-2 10-1 10
PR (torr-ern)
Figure 2-25. Caleulated loss cross sections for metastable neon atoms lost by diffusion to
the chamber walls (D), electron impact ionization (E), and quenching collisions with ground
state atoms (TB). (90)
120,-------------------,
a
100
SO
iii
C
Cl
C;;
GI 60
>
~
Gi
Ir 40
20
0.2 0.4 0.6 0.8 1.0 1.2 1.4
Gauge Pressure (torr)
120
b
100
80
iii
C
Cl
Cii
GI
60
~GI
äi
Ir 40
20
0
0
Current (mA)
Figure 2-26. Effect of (a) discharge pressure (current = 3 mA) and (b) current (pressure=
0.8 Torr) on argon metastable atom absorption (811.5 nm) for a coaxial pin configuration;
brass target. Adapted from Smith et al. (79)
geometries to gain a more thorough understanding of source operating

characteristics.
If metastable discharge gas atoms play an important role in spectro-
chemical determinations, then the spatial distribution of the species within
the glow discharge source is also an important consideration. The distribu-
tion of metastable argon atoms in the modified Grimm-type discharge has
56 CHAPTER 2
been studied by Ferreira and co-workers. (87) The populations were moni-
tored by the measurement of the absorbance of argon metastable atoms of
the Ar(I) 696.5-nm transition. A typical spatial profile is shown in Fig. 2-27
for a range of discharge conditions when sputtering a Cr: Cu (l :99) cathode.
The plots reveal that the maximum metastable population exists at the
cathode surface, with a steep dec1ine throughout the CDS. In all cases, a
second maximum is seen at a distance just beyond the CDS/NG interface.
Similar types of profiles were presented by Doughty and Lawler(83) who
measured the metastable atom densities and electric field strengths in a He
glow discharge by optogalvanic spectroscopy. They found that the concen-
tration of metastable atoms had a maximum value at the end of the electric
field, i.e., the boundary of dark space and NG. Nearly identical plots to
those in the Grimm source have been presented for a diode geometry device
by Winchester and Marcus, (58) who offered the following explanation. The
.6
~ .a
/ \
\ / \
/ \
.5 \
/ \,
i
a:
G3
.4
~~\
\\,/
/:' "", "
~.3 \.~ ____'~_
~ '\~\ ,,---,
, \~~._--_///~-------~~--::
,~ __ • ..,..lY' _______ ... ------_
')L, ----.a--
... __ ----
--
). ------- -----_.... -
"-
"-
.I
-_J __ - r . - _ _ x
---.&
Conditions: 6OOV. 60mA 0--0: 6OOV. SOmA 0------0: 800V. 20mA x __ x:
800V. 40mA 11----11: 800V. 60mA _ - - _ : IOOOV. 40 mA ~ _____ ~_
2 3 'I ~ 6
DISTANCE FROM CATHODE (mm)
Figure 2-27. SpatiaI profile of argon metastable atom absorbance (696.5 nm) in a Grimm-
type configuration (Cr-Cu target).(87)
high density of metastable atoms at the surface is likely the result of neutral-
izing collisions of sputtering argon ions at the cathode surface. The metast-
able atoms produced may then diffuse into the bulk of the plasma, or more
likely be redeposited onto the cathode surface, thus the minimum within
the cns. The correlation between the second maximum and the CnSjNG
interface corresponds to the region of the discharge where secondary elec-
trons have lost sufficient energy to efficiently populate metastable levels. The
decrease in metastable atom densities at farther distances reftects the lower-
energy electron distributions in the NG as shown previously. The differences
in the profiles as a function of discharge conditions can most likely be
explained on the basis of electron energies and densities (production), as
weIl as the density of sputtered and discharge gas atoms (loss).
2.5. Analyte Excitation and lonization

At the press ures at which glow discharges operate, collisional processes
determine the observed excited and ionized states. In order for sputtered
target atoms to become electronically excited or ionized, there must be some
sort of inelastic collision between the atom and a particle with either kinetic
or potential energy. Inelastic collisions with plasma electrons of various
kinetic energies are termed collisions of the "first kind." Collisions with
massive particles, followed by subsequent potential energy transfer, are
grouped together as collisions of the "second kind." Listed in Table 2-3 are
the most likely excitation and ionization processes occurring in the NG
region. Of course the relative roles of these processes will be determined by
the actual discharge conditions and source geometry, which also dictate
number densities and energies of the respective collision partners, and the
physical cross sections of the fundamental processes.
2.5.1. Excitation
Atoms that complete the fortuitous journey to the NG are subject to
inelastic collisions with electrons, metastable atoms, and ions. Based simply
on number density considerations, these mechanisms for electronic excita-
tionshould be roughly equivalent. Geometrically speaking, the latter two
processes should be dominant. Electronic excitations due to electron impact
are more generally termed "thermal excitation." The key factor in assessing
the relative roles of these processes is the resonant nature of electronic
excitation. On this basis, electron impact excitation is clearly most favorable.
Electrons in the NG, a virtually field-free region, are usually of low energy
as shown in Figs. 2-21 and 2-22. The resonant character reftects the fact
that those energy levels existing at ::::;;3 eV above the ground state will be
58 CHAPTER 2
Table 2-3 Excitation and Ionization Processes in the Glow Discharge

I. Primary excitationjionization processes
A. Electron impact
MO + e- (fast) ..... M* + e- (slow)jM+ + 2e-
B. Penning collisions
11. Secondary processes

A. Charge transfer
1. Nonsymmetrie
Ar+ + MO ..... M+ (M H ) + Aro
2. Symmetrie (resonance)
X+ (fast) + XO(slow) ..... XO(fast) + X+ (slow)
3. Dissociative
Ar+ + MX ..... M+ + X+Aro
B. Associative ionization
Ar! + MO ..... ArM+ + e-
C. Photon-induced excitationjionization
MO + hv ..... M* jM+ + e-
D. Cumulative ionization
DMO, sputtered neutral; Ar:', metastable Ar atom (for example); X, any gas-phase atom.
preferentially populated. For this reason, atomic transitions in high UV and

visible regions of the spectrum dominate the emission from the NG.
In addition to the production of excited state analyte atoms, electron
impact is also responsible for the populating of the metastable levels of the
discharge gas atoms. A second pathway would be the recombination of
discharge gas ions leaving the atom in the metastable level, (87) though at
typical glow discharge pressures the required three-body collisions are
improbable. Collisions between the metastable gas atoms and sputtered neu-
tral atoms may lead to an internal energy transfer from the metastable to
the atom resulting in the excitation or ionization of the latter. Hess and co-
workers, (79) in studying the quenching of metastable discharge gas atoms,
have suggested that Penning-type collisions do contribute to the formation
of excited state ionic species. This process will likely have a resonance
dependence in that the excited state ion energy should be elose to that of
the metastable energy value or that of an excited state atom plus the meta-
stable energy.
In discussing the production of excited state ions, asymmetrie charge
exchange has been postulated to be a prominent mechanism. Steers and co-
workers(91.92) have studied the population of eu(H) states with hollow
cathode lamps, conventional Grimm lamps, and microwave-boosted Grimm
lamps using neon and argon as discharge gases. The authors argue that, on
the basis of state selectivity, direct production of excited state ions via charge
exchange from argon ions is a prominent mechanism. Analogous production
of higher excited states by neon was not investigated. Populating the states
via a two-step mechanism of Penning ionization followed by electron impact
excitation of the ion was said to be improbable. The excited state-Penning
ionization mechanism was not addressed. In a similar study, Wagatsuma
and Hirokawa(93) have used He-doped Ar and Ne discharge gases in Grimm-
type discharges to study charge exchange reactions. They also concluded
this to be a mechanism for the production of excited state ions. Again, the
extent ofthis process will be highly dependent on the identity ofthe discharge
gas and the sputtered atom in question. It is interesting to note that while
the high-current-density hollow cathode and Grimm geometries exhibit
asymmetrie charge exchange characteristics, the low-current diode geometry
employed by Hess(79) and most mass spectrometry applications does not
seem to have asymmetrie charge exchange as a primary excitation mecha-
nism. While these differences are quite interesting, it must be kept in mind
that ionic emission is a very minor component of glow discharge optical
spectra.
2.5.2. lonization
In principle, each of the collisional processes listed in Table 2-3 is a
viable candidate for analyte ionization in glow discharge mass spectrometry
so long as the ionization potential can be met. In many respects, knowledge
of the dominant ionization processes is more important than for analyte
excitation. In mass spectrometry, there is no analogy to degree of excitation
where simply different states are populated; in glow discharge environment
ionization is simply a question of "ion or atom." Once in an ionic state,
the analyte can be detected mass spectrometrically. Therefore, the primary
concern is the production of a maximum number of ions with as much
elemental uniformity as possible.
When looking at the source of kinetic and potential energies for glow
discharge ionization in the NG (ions cannot be extracted from the dark
space), the mechanisms of electron impact, Penning ionization, and charge
exchange are again the most likely candidates. Referring to the electron
energy distributions shown in Fig. 2-21 and 2-22, it is fairly clear that there is
not a sufficient electron population in the 5-10 eV range to produce extensive
ionization. Given that electron and ion densities are both on the order of
60 CHAPTER 2
1011 cm- 3 , the populations of these higher-energy electrons are not likely
greater than '" 107_108 cm -3. Vieth and Huneke(94) have used glow discharge
mass spectrometry to measure the relative populations of singly and multiply
charged argon, matrix, and argide ions in an Ar glow discharge for different
sampie matrices. They found that the measured relative densities of multiply
charged argon ions (from + 3 to +6) are in good agreement with the Lovett
rate model(95) for electron impact ionizationjthree-body recombination with
an electron temperature of about 7 eV. The 7-eV electrons were said to be
the result of a double-Penning collision of the form
Ar,! + Ar,! ~ Ar+ + Ar + e-(7.1-7.6 eV) (2-8)
However, there is no experimental evidence for the existence of electron

populations of these energies. Regardless of the exact mechanism, such
highly charged ions (>M 2+) are extremely rare for analyte species.
Penning-type collisions have generally been assigned as the dominant
ionization mechanism of sputtered atoms in most types of glow discharge
devices.(78,79,82,85,86) Coburn and Kay(85) demonstrated the importance ofPen-
ning ionization in the rf sputtering of a europium monoxide (EuO) cathode
(containing 6% Fe) in both argon and neon discharges. Eu+ and Fe+ signals
were observed in both discharges, but an 0+ peak was observed only in the
neon discharge. The authors explained this by the difference in the metastable
energies of argon and neon. The energy of metastable argon atoms is
sufficient to Penning-ionize Eu and Fe (IP = 5.68 and 7.83 eV, respectively)
but unable to Penning-ionize oxygen atoms (IP = 13.55 eV), which can be
Penning-ionized by the metastable neon atoms. Their experiments also
implied that the contribution of electron impact ionization and dissociative
charge-transfer ionization is less significant relative to the Penning
ionization.
Coburn and co-workers(86) extended the previously mentioned studies
employing simultaneous mass spectrometric and atomic absorption measure-
ments to relate sputtered copper atom and Ne metastable densities to Cu +
signals. Figure 2-28 illustrates the relationship between the product of the
atom x metastable density and ion signalover a range of operating powers
and pressures in an rf glow discharge. The direct relationship is strong
evidence for a Penning ionization mechanism since that process would be
first-order in the respective collisional partner concentrations. Indirect evi-
dence of Penning ionization has been obtained by Hess and co-workers(79)
who related the depopulation of neon and argon metastable atoms through
methane quenching and reductions in analyte ion signals measured optically.
Laser sources have been useful tools for elucidating the possible role of
metastable atoms in glow discharge ionization. Harrison and co-workers(82)
used a combination of optogalvanic spectroscopy and mass spectrometry to
10.2
Cu' Nem (arbitrary unitsl
Figure 2-28. Cu + versus Cu.Nem as a function of rf power for a diode configuration at a

pressure of 20 mTorr.(86)
illustrate that depopulation of neon metastable levels in a hollow cathode

discharge affected discharge voltage characteristics and reduced the moni-
tored copper ion signal. These findings support a Penning mechanism. Hess
and Harrison(78) also used a tunable laser to depopulate metastable atom
levels in a coaxial pin discharge geometry and monitored the effects by
optogalvanic spectroscopy and mass spectrometry. In the neon discharge,
they found that depopulation of the metastables caused a general reduction
in all measured ion signals, while similar studies with argon showed that
many species with ionization potentials greater than metastable energies of
argon were not affected by laser depopulation. Figure 2-29 depicts the use
of gated detection to monitor the decrease of analyte ion signal in the time
frame of the metastable state depopulation.
In terms of quantitating the extent of Penning ionization, no definitive
experiments have been performed. Although most researchers believe that
Penning ionization is the dominant mechanism in the NG region, electron
impact ionization is still likely to play a role in analyte ionization. It is
difficult to conc1ude whether electron impact ionization will compete with
the Penning ionization, since secondary electron populations exist at two or
three orders of magnitude less than those of thermal electrons and the life-
times of metastables are long enough to dominate the ionization events.
Vieth and Huneke(96) have deduced through elemental quantitation studies
that Penning ionization accounts for between 50 and 95% of the overall
62 CHAPTER 2
...J
<
z
C!l
Cii
z
Q
w
>
~w
a:
Cu
60 62 64 66 68 70 72
M/e
Figure 2-29. Mass spectrum produced by the subtraction of the steady-state glow discharge
ion signal from the ion signal generated during laser metastable depopulation [Ne(I)
594.48 nm]; Ne discharge, 0.4 Torr, 3 mA.(78)
analyte ionization. The given range reflects the relative cross sections of
elements for electron impact ionization.
The final mechanism that might be a meaningful contributor to NG
ionization is asymmetric charge exchange. Given the resonant nature of the
charge exchange process, as discussed in the previous section, it seems quite
unlikely that charge exchange from (presumably) discharge gas ions is a
major contributor to the singly ionized analytes present in glow discharge
mass spectra. Hecq and Hecq(97) have, however, implicated charge exchange
from doubly charged argon discharge gas ions to the formation of doubly
charged ions of sputtered atoms.
2.6. Conclusions
In this chapter, the fundamental plasma properties of glow discharge

devices have been presented in an effort to give current and future prac-
titioners insight into those parameters that affect analytical source per-
formance. While glow discharges are easily controlled and operate under
rather mild conditions (in terms ofpower and gas temperature), the spectro-
chemical behavior is the result of a very complex set of collisional processes.
The cathodic sputtering (atomization) step presents real advantages over
thermal vaporization methods such as those found in arcs and laser/solid
interactions. Elemental sputtering yields were shown to vary by only a factor
of 3 to 5. For this reason, nominal differences are found between the sputter
rates ofvastly different sampie matrices. Even so, a truly outstanding feature
of the sputtering process is the fact that once steady-state sputtering condi-
tions are met (l to 5 monolayers), sputter rates for all sampie components
are the same and the sampie introduction is stoichiometric.
The gas-phase collisions most prominent in glow discharge devices are
those of electron impact, Penning, and charge transfer reactions. Based on
number density and energy considerations, electron impact is clearly the
dominant atomic excitation mechanism. Under discharge conditions that
involve large current densities (Grimm and hollow cathode sources), asym-
metric charge exchange from discharge gas ions is a likely mechanism for
production of excited state ions. In lower-density sources, such as those
employed in mass spectrometry, excited state ions seem to be produced via
Penning ionization of excited state atoms.
Ionization in glow discharge devices can be classified according to the
region of the plasma in which the ions are formed. Discharge gas ions, which
are the incident particles in the sputtering process, are formed by direct
electron impact ionization in the vicinity of the CDSjNG interface. In this
region, secondary electrons emitted from the cathode surface have been
cooled by collisions with heavy particles. In the NG the electron populations
at atomic ionization energies are negligible; therefore, collisions of the
second kind are dominant. (Ionization of discharge gas ions is most likely
the result of electron impact ionization out of atomic metastable levels.)
Most experimental evidence points to the Penning ionization mechanism as
the dominant reaction in the production of analyte ions.
Throughout the remainder of this volume, the respective chapters will
emphasize the fundamental processes that underlie the specific spectrochem-
ical application.
Acknowledgment
Many of the results presented here from Clemson University are based
on work supported by the National Science Foundation under Grant CHE-
8901788, whose support is gratefully acknowledged.
References
1. A. M. Howatson, An Introduction to Gas Discharges, Pergamon Press, Elmsford, N.Y.,

1976.
2. J. A. C. Broekaert, J. Anal. At. Spectrom. 2 (1987) 537.
3. A. von Engel, Electric Plasmas: Their Nature and Applications, Taylor & Francis, New
York, 1983.
64 CHAPTER 2
4. P. W. J. M. Boumans, Theory of Spectrochemical Exeitation, Hilger & Watts, London,

1966.
5. L. Paksy and I. J. Lakatos, Spectrochim. Acta 38B (1983) 1099.
6. L. Radermacher and H. E. Beske, Spectrochim. Acta 37B (1982) 769.
7. F. Paschen, Ann. Phys. 50 (1916) 90l.
8. H. Schuler, Z. Phys. 59 (1929) 149.
9. F. W. Aston, Mass Spectra and Isotopes, 2nd ed., Longmans, Green, New York, 1942.
10. E. Nasser, Fundamentals ofGaseous Ionization and Plasma Electronics, Wiley-Interscience,
New York, 1971.
11. B. Chapman, Glow Discharge Processes, Wiley-Interscience, New York, 1980.
12. K. Wiesemann, Phys. LeU. 29A (1969) 691.
13. F. W. Aston, Proc. R. Soc. London Sero A 84 (1911) 526.
14. D. Fang and R. K. Marcus, Spectrochim. Acta 45B (1990) 1053.
15. P. F. Knewstubb and A. W. Tickner, J. Chem. Phys. 36 (1962) 684.
16. F. W. Aston, Phi/os. Mag. 46 (1923) 211.
17. M. E. Pillow, Spectrochim. Acta 36B (1981) 821.
18. G. S. Anderson, W. N. Meyer, and G. K. Wehner, J. Appl. Phys. 33 (1962) 2991.
19. D. C. Duckworth and R. K. Marcus, Anal. Chem. 61 (1989) 1879.
20. W. R. Grove, Phi/os. Trans. R. Soc. London 142 (1852) 87.
21. W. D. Westwood, Prog. Surf Sci. 7 (1976) 71.
22. A. Benninghoven, Surf Sei. 53 (1975) 596.
23. R. F. K. Herzog, W. P. Poschenrieder, and F. K. Satidiewics, Radiat. Eff. 18 (1973) 199.
24. V.A. Von Hippei, Ann. Phys. (Leipzig) 80 (1926) 672.
25. V.A. Von Hippei, Ann. Phys. (Leipzig) 81 (1926) 1043.
26. P. Sigmund, Phys. Rev. 184 (1969) 383.
27. H. F. Winters, in: Topics in Current Chemistry No. 94, Plasma Chemistry III (S. Veprek
and M. Vervgopalen, eds.), Springer-Verlag, Berlin, 1980.
28. P. W. J. M. Boumans, Anal. ehem. 44 (1972) 1219.
29. M. A. Rudat and C. N. McEwen, Int. J. Mass Spectrom. Ion Phys. 46 (1983) 351.
30. G. K. Wehner, Phys. Rev. 108 (1957) 35.
31. G. K. Wehner, Phys. Rev. 102 (1956) 690.
32. H. A. Oeschner, Physics (N. Y.) 261 (1973) 37.
33. J. Roth, J. Bohdansky, and W. Eckstein, Nucl. Instrum. Methods Phys. Res. 218 (1983)
751.
34. G. Carter and J. S. Colligan, Ion Bombardment of Solids, Chapter 7, Heinemann, London,
1968.
35. R. S. Nelson, Phi/os. Mag. 11 (110) (1965) 291.
36. R. Kelly and N. Q. Lam, Radiat. Eff. 19 (1973) 39.
37. B. J. Stocker, Br. J. Appl. Phys. 12 (1961) 465.
38. J. Musha, Phys. Soc. Jpn. 17 (1962) 1440.
40. M. Dogan, K. Laqua, and H. Massmann, Spectrochim. Acta 26B (1971) 631.
41. G. F. Weston, Cold Cathode Discharge Tubes, Iliffe, London, 1968.
42. G. Betz, Surf Sei. 92 (1980) 283.
43. G. Betz and G. K. Wehner, in: Spuuering by Particle Bombardment II, (R. Behrisch, ed.),
Springer-Verlag, Berlin, 1983.
44. K. Wittmack, Nucl. Instrum. Methods 168 (1980) 343.
45. G. K. Wehner, in: Methods and Phenomena: Their Applications in Seience and Technology
(S. P. Wolsky and A. W. Czanderna, eds.), Vol. 1, Elsevier Scientific, New York, 1975.
46. G. E. Hammer and R. M. Shemenski, Surf Interface Anal. 10 (1987) 355.
47. D. Fang and R. K. Marcus, J. Anal. At. Spectrom. 3 (1988) 873.
48. V. L. Ginsburg, J. Phys. (USSR) 8 (1944) 253.
49. G. F. Weston, Cold Cathode Discharge Tubes, Iliffe, London, 1968.
50. N. H. Tolk, 1. S. T. Tsong, and C. W. White, Anal. Chem. 49 (1977) 16A.
51. J. M. Schroeer, R. N. Rhodin, and R. C. Bradley, Surf Sei. 37 (1973) 571.
52. R. V. Stuart and G. K. Wehner, J. Appl. Phys. 35 (1964) 1819.
53. V. 1. Veksler, Sov. Phys. JETP 11 (1960) 235.
54. S. Komiya, K. Yoshikawa, and S. Ono, J. Vac. Sei. Technol. 14 (1977) 1161.
55. W. W. Harrison and C. G. Bruhn, Anal. Chem. 50 (1978) 16.
56. A. J. Sterling and W. D. Westwood, J. Phys. D 4 (1971) 246.
57. N. P. Ferreira and H. G. C. Human, Spectrochim. Acta 36B (1981) 215.
58. M. R. Winchester and R. K. Marcus, 16th Annu. Meet. Fed. Anal. Chem. Spectrosc. Soc.
1989, Paper No. 327.
59. D. K. Doughty, E. A. Den Hartog, and J. E. Lawler, Appl. Phys. Lett. 46 (1985) 352.
60. F. W. Aston, Proc. R. Soc. London Sero A 104 (1923) 565.
61. A. V. Bodarenko, Sov. Phys. Tech. Phys. (Engl. Transi.) 18 (1973) 515.
62. F. Howorka, W. Lindinger, and M. Pahl, Int. J. Mass Spectrom. Ion Phys. 12 (1973) 67.
63. F. Howorka and M. Pahl, Z. Naturforsch. Teil A 27 (1972) 1425.
64. S.-L. Tong and W. W. Harrison, Anal. Chem. 56 (1984) 2028.
65. M. Dogan, K. Laqua, and H. Massmann, Spectrochim. Acta 26B (1971) 631.
66. P. W. J. M. Boumans, Anal. Chem. 44 (1972) 1219.
67. A. Guntherschulze, Z. Tech. Phys. 11 (1930) 49.
68. P. F. LittJe and A. Von Engel, Proc. R. Soc. London Sero A 224 (1954) 209.
69. V. 1. Kirichenko, V. M. Tkachenko, and V. B. Tyutyunnik, Sov. Phys. Tech. Phys. (Engl.
Transl.) 21 (1976) 1080.
70. W. Stern, Beitr. Plasmaphys. 9 (1969) 59.
71. C. T. J. Alkemade, T. Hollander, W. Snelleman, and P. J. T. Zeegers, Metal Vapours in
Flames, Pergamon Press, Oxford, 1982.
72. F. F. Chen, in: Plasma Diagnostic Techniques (R. H. Huddlestone and S. L. Leonard,
eds.), Academic Press, New York, 1965.
73. L. Vriens, J. Appl. Phys. 45 (1974) 1191.
74. D. Fang and R. K. Marcus, J. Anal. At. Spectrom. 5 (1990) 569.
75. D. Fang and R. K. Marcus, Spectrochim. Acta 468 (1991) 983.
76. V. S. Borodin, and Y. M. Kagan, Sov. Phys. Tech. Phys. (Engl. Transi.) 11 (1966) 131.
77. V. S. Borodin, Y. M. Kagan, and R. 1. Lyaguschenko, Sov. Phys. Tech. Phys. (Engl.
Transl.) 11 (1967) 887.
78. K. R. Hess and W. W. Harrison, Anal. Chem. 60 (1988) 691.
79. R. L. Smith, D. Serxner, and K. R. Hess, Anal. Chem. 61 (1989) 1103.
80. M. K. Levy, D. Serxner, A. D. Angstadt, R. L. Smith, and K. R. Hess, Spectrochim. Acta
46B (1991) 253.
81. R. B. Green, R. A. Keller, G. G. Luther, P. K. Schenck, and J. C. Travis, Appl. Phys. Lett.
29 (1976) 727.
82. K. C. Smyth, B. L. Bentz, C. G. Bruhn, and W. W. Harrison, J. Am. Chem. Soc. 101
(1979) 797.
83. D. A. Doughty and J. E. Lawler, Plasma Processing, Mater. Res. Soc. Symp. Proc. 68
(1986) 141.
84. N. P. Ferreira, J. A. Strauss, and H. G. C. Human, Spectrochim. Acta 37B (1982) 273.
85. J. W. Coburn and E. Kay, Appl. Phys. Lett. 18(10) (1971) 435.
86. E. W. Eckstein, J. W. Cobum, and E. Kay, Int. J. Mass Spectrom. Ion Phys. 17 (1975)
129.
87. J. A. Strauss, N. P. Ferreira, and H. G. C. Human, Spectrochim. Acta 37B (1982) 947.
88. L. G. Piper, J. E. Velazco, and D. W. Stetzer, J. Chem. Phys. 59 (1973) 3323.
66 CHAPTER 2
89. P. L. Larkins, Spectrochim. Acta 46B (1991) 291.

90. K. A. Hardy and J. W. Sheldon, J. Appl. Phys. 53 (1982) 8532.
91. E. B. M. Steers and R. J. Fielding, J. Anal. At. Spectrom. 2 (1987) 239.
92. E. B. M. Steers and F. Leis, J. Anal. At. Spectrom. 4 (1989) 199.
93. K. Wagatsuma and K. Hirokawa, Spectrochim. Acta 46B (1991) 269.
94. W. Vieth and J. C. Huneke, Spectrochim. Acta 45B (1990) 941.
95. R. J. Lovett, Spectrochim. Acta 37B (1982) 985.
97. M. Hecq and A. Hecq, J. Appl. Phys. 56 (1984) 872.
3
A tomie Absorption and
Fluoreseenee Speetroseopies
Edward H. Piepmeier
3. 1. Introduction
The analysis of alloys, powders, and dried solutions by atomic absorption

and fluorescence spectrometry (AAS and AFS) using sputtering cells for
atomization is receiving wider notice in the literature as the advantages of
these methods come to be appreciated. For example, compared with emission
spectrometry, these methods have higher spectral resolution and fewer spec-
tral interferences. Rapid multielement AAS analyses are now practical
because of recent technological advances. With recent developments; sensi-
tivities approach and in some cases exceed flame and graphite furnace
methods, and improvements are continuing.
AAS for determining chemical elements in solution was introduced as
a viable analytical method in 1955 by independent and simultaneous papers
by Alkemade and Milatz(l·2) who demonstrated the determination of Na in
solutions introduced into a flame and discussed practical considerations,
and by Walsh(3) who discussed in detail the theoretical and experimental
considerations of the method. The direct analysis of metals using a sputtering
chamber as a convenient atomic absorption cell was first proposed by Russell
and Walsh in 1959(4) in a paper describing perceptive observations of the
differences in the spectra emitted in different directions from a hollow
Edward H. Piepmeier • Department of Chemistry, Oregon State University, Corvallis,

Oregon 97331.
67
68 CHAPTER 3
cathode lamp. Their observation that atomic fluorescence was occurring in

the lamp eventually led to the use of AFS to observe the atomic vapor in a
sputtering cello (5) Their prediction that the sputtering chamber method would
be useful "particularly for analyses requiring very high sensitivity" is con-
tinuing to materialize as a better understanding is being obtained of ways
to improve sensitivity.
Figure 3-1 shows a block diagram of the main components of atomic
absorption and fluorescence instruments. Modulated or pulsed hollow
cathode lamps are commonly used as the primary radiation source for atomic
absorption measurements although tunable lasers are used in special cases.
A modulated Grimm-type glow discharge has been used as the primary
radiation source for multielement simultaneous reading AAS because of the
ease of changing the cathode to select the elements of interest. (6) Although
hollow cathode lamps are used to excite atomic fluorescence, tunable-laser-
excited fluorescence provides the best detection limits. (7)
The wavelength isolation system may be a tunable monochromator,
an interference filter, or a resonance atomic fluorescence monochromator
(RAFM; consisting of a cIoud of atoms of the element of interest, which
absorbs and then fluoresces radiation of that element(8»). Fluorescence
measurements have been made without a wavelength isolation system, (5) but
background interferences were apparent. After the appropriate wavelength
has been isolated, a photomultiplier converts the radiation into an electrical
signal for processing and eventual readout.
RAFM
Photomultiplier
:
AAS
Sputtering Wavelength
Primary ............
Radiation
Cell ......... Isolation
Device
··········1 Photomultiplier I
System
Source
~
i ..., Gated
Signal
Processor
Pulsed or AFS
1
Modulated Primary
~
Power Radiation
Supply Source
Data
I Handling
and
Readout
System
Figure 3-7. Block diagram of the main components of basic sputtering cell atomic absorption
(AAS) and atomic fluorescence (AFS) spectrometers. The RAFM photomultiplier is used when
the wavelength isolation device is a resonance atomic fluorescence monochromator.
A TOMIC ABSORPTION AND FLUORESCENCE SPECTROSCOPIES 69
Several types of sputtering cells are used to convert asolid sampie into
an atomic vapor for the atomic absorption or fluorescence measurements.
In 1960, Gatehouse and Walsh(9) used a stagnant-gas sputtering chamber
containing a sampie in the form of a hollow cathode. Although the time
required for an analysis was much longer than for other routine analytical
atomic spectrometry methods, Goleb(lO) found the narrow absorption lines
of a low-pressure hollow cathode sputtering cell useful for the isotopic
analysis of uranium sampies. Goleb and Brody(ll) applied tbis method to
solutions by drying them on the inside of the hollow cathode. In 1971,
Gandrud and Skogerboe(12) used flat spectroscopic-grade graphite and alu-
minum platrodes as replaceable cathodes on which sampIe solutions were
dried and inserted into the sputtering cell, which had a flowing gas stream.
An analysis rate of 15-20 determinations/h is possible with their system,
and detection limits range from 1 to 10 ng.
In 1973, Gough et al.(5) used a sputtering cell for atomic fluorescence
determinations. Flat sampIes were now easily mounted on the side of the
cell by pressing the sampIe against an O-ring surrounding a replaceable silica
annulus with a l-cm hole. The O-ring provides a vacuum seal and the l-cm
hole defines the location of the discharge on the sampIe surface. Surrounding
the hole is a ledge on the annulus that provides a 0.02-cm gap between the
sampIe and the annulus to prevent deposits that might otherwise electrically
short the sampIe to the anode. A flow-through gas control system aids
the rapid interchange of sampIes. In 1976, Gough(I3) reported an atomic
absorption cell (Fig. 3-2) similar to the fluorescence cell, but the gas flow
was specifically designed to improve the absorption signal by increasing the
transport of the sputtered atoms into the observation zone, which was
1-2 cm from the flat sampie.
In 1987, Bernhard(14) reported the discovery that gas jets that strike the
sputtering surface significantly increase the sampling rate as well as the
absorption signal in a sputtering chamber. This discovery led to a renewed
interest in sputtering cells for the analysis of alloys, powders, and dried
solutions by AAS and AFS. A commercial atomic absorption cell (Fig.
3-3), based on this principle and named the Atomsource(14) (Analyte Corp.,
Medford, OR), has further stimulated wider applications of gas-jet-enhanced
sputtering because it can be mounted on many of the atomic absorption
spectrometers that have appeared in analyticallaboratories since the pioneer-
ing work of Walsh and his co-workers over three decades ago. For the 15
elements studied, the characteristic concentrations that produce an absorb-
ance of 0.0044 for the Atomsource are comparable to, or better than, those
for a flame, assuming for the flame that 1 g of alloy is dissolved in 100 ml
of solution.(15)
The increased sampling rate of a gas-jet-enhanced sputtering cell has
recently been applied by Chakrabarti et a1Y 6 ) to the analysis of dried solu-
tions deposited on flat cathodes pressed against the sampIe port of the
70 CHAPTER 3
Water
Cooling
Sampie
(Cathode)
35mm
a
~ Anode
u'TO Pump
O.lmm recess
~01~DJ
b -I 6mm I--
Figure 3-2. Schematic diagram of (a) Gough's sputtering chamber showing the replaceable
hollow silica annulus (b) that admits high-velocity gas into the chamber. Reprinted with permis-
sion from D. S. Gough, Anal. ehern. 48 (1976) 1926, copyright 1976, American Chemical
Society.
Atomsource. They used a high-electrical-power pulsed mode of operation

for high atomization rates and obtained results that are comparable to those
of the highly sensitive graphite furnace atomic absorption spectrometry
(GFAAS) method for some elements. Absolute detection limits range from
0.2 to 600 pg. They predict that "the pulsed mode of atomization can be
made at least as sensitive as GFAAS for most elements."
Although lasers are still relatively very expensive radiation sourees, the
high irradiances that are possible and their narrow spectral bandwidths make
them attractive for special applications. Winefordner and co-workers have
calculated limiting noise levels that should eventually allow the detection of
only several thousand atoms (around 10-6 pg) in the observation region in
a sputtering cell.(7,l7,l8) Even so, detection limits of only 0.5 pg for Pb and
20 pg for Ir in dried aqueous solutions have been demonstrated with their
system, which used a hollow cathode sputtering cell at 30 mA, 500 V, and
0.7 Torr.(l7) The difference between the detection limits for atoms in the
ATOMIC ABSORPTION AND FLUORESCENCE SPECTROSCOPIES 71
c WCJ
a. ceramic arrestor
b. main body
c. cathode plate
i. insulating material
j. 6 jet nozzles, anode
s. sampie
t. traps to shield windows
from sputtered atoms
v. vacuum system ports
w. water cooling channel
Figure 3-3. Schematic diagram of the Atomsource showing the six nozzles that direct high-
velocity gas jets onto the sputtering surface of the sampIe. The isometrie views of the nozzles
and arrestor are adapted by permission from D. S. Gough, P. Hannaford, and R. Martin Lowe,
Anal. ehern. 61 (1989) 1652, copyright 1989, American Chemical Society.
observation region and the amounts in the dried solution sampIes can be
reduced by more efficient atomization and more efficient temporal and spa-
tial probing of the sputtered atoms by the pulsed laser; ideally all of the
sampIe atoms should be in the observation region when the atoms are
observed. The increased sampling rate of a gas-jet-enhanced sputtering cell
and high-current pulses of the type used by Chakrabarti and co-workers
should help to improve detection limits.
The rest of this chapter considers some of the fundamental aspects of
glow discharge AAS and AFS, followed by practical considerations, applica-
tions, and a look to the future.
3.2. Fundamental Considerations
3.2.1. Characteristics of Low-Pressure Atomic Spectra
The spectral profiles of the atomic absorption lines in low-pressure

sputtering cells have a significant influence on spectral interferences and on
the slopes and shapes of atomic absorption and fluorescence analytical
working curves. Detailed information about spectral line profiles can help
72 CHAPTER 3
determine the shapes of analytical working curves so that appropriate mathe-

matical models can be used to fit these curves to determine concentrations.
Detailed information also helps predict how changes in operating conditions
of the primary radiation source and the atom cell influence these curves, so
that we can obtain better control of the operating conditions, and more
efliciently update working curves as operating conditions drift with time.
Factors that contribute to the spectral profile of an atomic absorption line
will now be reviewed followed by the influences that spectral profiles have
on analytical working curves and spectral resolution.
3.2.1.1. Shapes of Spectral Line Profiles
The shapes of atomic absorption line profiles vary significantly from

one spectralline to another, even for different lines of the same element.
Unfortunately, specific experimental and theoretical information about spec-
tral profiles is available for relatively few analytically useful atomic lines,
but more information continues to become available as the need for it is
recognized. The theory of atomic spectral line profiles has been covered in
more detail in several books(19-25) and papers.(26,27)
a. Hyperjine Components and Isotope Shifts. In general, the atomic

absorption line profile may have several hyperfine components caused by
different isotopes having slightly different energy levels, and by splitting of
otherwise degenerate electron energy levels by nuclei that have a net nuclear
spin (i.e., nuclei with an odd number of protons andjor neutrons). The
overall absorption profile for a line is the sum of the spectral profiles of each
of the hyperfine components.
The isotope shift for a handful of light and very heavy elements is large
enough to form the basis for the determination of different isotopes using
sputtering cells. For the lighter elements the relatively large isotope effect is
due to the large relative differences in nuclear mass, and for the very heavy
elements the isotope effect is mainly due to the nonuniform way in which
nuclear charge is distributed over the volume of a nucleus. (28) That part of
the isotope shift caused by mass difference can be estimated by
L\1t = -q').. Ma - Mb
(3-1)
1836 MaMb
where L\1t is the wavelength shift (isotope shift) for a line at wavelength It
for two isotopes with masses M a and Mb in daltons, and q is a constant that
depends on the type of mass effect and the type of electrons involved in the
energy level transition. For the normal mass effect, q = 1, and for the specijic
mass effect involving d electrons, q varies from -20 to + 16.(26) For a mass
difference of 1 at 33 daltons, these isotope wavelength shifts are on the
order of -0.004 to +0.003 nm. The isotope shift caused by the nonuniform
distribution of nuc1ear charge is comparable in size to the mass effect for
intermediate elements and dominates for heavy elements (>140 daltons).(29)
About two-thirds of the stable atomic nuc1ei have a net nuc1ear spin that
causes hyperfine splitting, which typically amounts to 10- 3 nm.
The relative intensities of these hyperfine components vary, depending
on the relative abundances ofthe different isotopes and on Russell-Saunders
coupling. Although the relative intensities can often be predicted, the posi-
tions of the hyperfine components along the wavelength axis usually must
be determined experimentally.
b. Broadening. Each ofthe hyperfine components is individually broad-

ened by naturallifetime, Doppler, and collisional (or pressure) broadening
processes. For example, the vertical bars in Fig. 3-4 show the hyperfine
components for the 403.3-nm Mo(I) line while the profile with several peaks
corresponds to the low-pressure or Doppler-broadened profile. The broad
profile corresponds to the atmospheric pressure, high-temperature Doppler-
and pressure-broadened profile. Alllines are infiuenced by natural lifetime
broadening, which has a spectral profile in the shape of a Lorentzian function
with a width at half-height of
(3-2)
where Ti and Tj respectively represent the lifetimes of the upper and lower
energy states for the transition, cis the velocity of light (in nm S-I), and A
is in nanometers. For a resonance line, where the ground state lifetime is very
long, the term in brackets becomes the Einstein probability for spontaneous
Figure 3-4. Shapes of the Mn(I) 403.3-nm

spectral line in a low-pressure glow discharge
(four-peaked curve) and an atmospheric-
pressure flame (single-peaked curve). The
unbroadened hyperfine components are indica-
Wavelength
ted by vertical Iines.
74 CHAPTER 3
emission, A ij in photons per seeond. A strong line, with a I-ns lifetime (Aij =
109 S-I), at 300 nm has a naturallifetime half-width of only 5 x 10- 5 nm.
Although this is narrower than half-widths caused by other broadening
proeesses, the Lorentzian funetion has long tails, whieh beeome signifieant
in extreme eases as we will see.
The main broadening proeess in a low-pressure sputtering eell is the
Doppler effeet. The Doppler effeet eauses the wavelength at which an atom
absorbs to be shifted to the blue (shorter wavelength) for atoms traveling
with a velocity eomponent v (ern S-I) toward the deteetor, while a redshift
oeeurs for atoms traveling in the opposite direetion. To a first-order approxi-
mation the blueshift is given by
(Ao - 1\,) = Aov/c (3-3)
where Ao is the wavelength at whieh a stationary atom absorbs, I\, is the

wavelength absorbed by an atom moving with a velocity v (in em) relative
to the deteetor, and c is the velocity of light (in em S-I). The absorption
profile and the emission profile are shifted by the same amounts and in the
same direction.
Net Doppler shifts in emission and absorption have been observed in
pulsed hollow eathode lamps(30,31) and they may be present in other pulsed
atomizers where a rapid gas expansion occurs. They mayaiso be present in
atomizers where the gas is rapidly flowing toward or away from the detector.
When the translational energy of a population of atoms is in thermal
equilibrium so that the velocities of the atoms are given by the Maxwell-
Boltzmann distribution, the result is a Doppler-broadened spectral profile
that is deseribed by a Gaussian funetion. The width of the Gaussian peak
at half-height is given by
~I\,D = 7.16 X 10-7 Ao(T/M)I/2 (3-4)
where T is the temperature (in K), M is the atomic mass (in amu), and ~I\,D
has the same units as Ao (e.g., nm). A line at 300 nm for an atom with a
mass of 60 at a temperature of 1000 K has a Doppler width of 9 x 10-4 nm,
more than an order of magnitude greater than its naturallifetime width even
if the line is relatively strong.
Although eollisional broadening is usually negligible in sputtering eells,
it is eomparable in width to Doppler broadening in atmospherie-pressure
flames, fumaces, and electrical plasmas that may be used as light sourees
for atomie fluoreseence. Equation (3-2) may be used to deseribe eollisional
broadening. In this ease the term in braekets is inereased by adding the
collision rate (in S-I) for eollisions that are elose enough to interrupt the
eoherence of the interaetion between the emitted or absorbed radiation field
and the atom. These collisions may be adiabatic (no energy transfer) or
nonadiabatic (quenching or exciting).
When collisional broadening is comparable in width to Doppler broad-
ening, the overall profile for each hyperfine component is often described by
the Voigt function.(24) This function assurnes a Lorentzian profile for each
set of atoms with a common Doppler shift, and adds these profiles together
as though the population of atoms with common Doppler shifts were distri-
buted in a Gaussian manner. The Voigt function is only a first-order
approximation(25) but is adequate to fit experimentally determined spectral
profiles. (32) The Lorentzian and Voigt functions have long tailing wings rela-
tive to the Gaussian function, which drops off rapidlyon either side of
its center. Therefore, there will tend to be more spectral overlap in high-
temperature atmospheric-pressure ceHs than in low-pressure sputtering ceHs
where pressure broadening is negligible and the profile is dominantly
Gaussian in shape (Fig. 3-4). In extreme cases, long Lorentzian-type tails
caused by the minor amount of collisional and naturalline broadening that
occurs in a low-pressure sputtering ceIl may become significant, such as in
the emission spectral profile of self-reversed hollow cathode lines used for
background correction, (30,31,33) and in the absorption spectrum when a laser
beam is used to saturated a spectral transition to maximize atomic fluoresc-
ence signals. (19,34)
3.2.1.2. /nf/uence on Working Curves
Compared with the relatively broad spectral profiles in atmospheric

Harnes and fumaces, the spectral profiles in low-pressure sputtering cells are
narrower and closer in shape to the spectral profiles of hollow cathode lamps
that are commonly used as primary light sources to determine absorbance.
Because of the greater spectral overlap between the spectral profiles of the
lamp and the absorption cell, more photons from the source lamp are
absorbed per atom in a low-pressure sputtering ceH than in an atmospheric
flame or fumace. This tends to improve the initial slope of an absorbance
or fluorescence versus concentration working curve.
On the other hand, because the spectral profiles are narrower, the
atomic absorption analytical working curve at higher absorbances tends to
bend more in the direction of the concentration axis. The bending occurs
because the absorptivity (or absorption coefficient) is not the same for aIl
ofthe wavelengths ofthe photons that can pass through the monochromator
to the detector. (This is sometimes known as the polychromatic effect,
although different absorptivities as weIl as photons of different wavelengths
must also be involved.)
This author has studied these effects on atomic absorption working
curves for 17 lines be10nging to eight elements. (35) The spectral profiles of
76 CHAPTER 3
the primary light source and the absorption cell were assumed to have the
same shape. In general, the initial slope is higher but the bending is greater
for lines that appear to have a single Doppler-broadened hyperfine compo-
nent (e.g., Ca 422.7 nm) compared with lines that are wider because they
consist of several noticeable but heavily overlapping Doppler-broadened
peaks (e.g., Mn 403.3 nm, Fig. 3-4). Analyticallines ofmany elements tend
to have the latter type of shape.
Svoboda et al. (36) have studied the theoretical shapes of fluorescence
analytical working curves. This study was intended primarily for atmos-
pheric pressure flames and so the authors do not consider the influence of
hyperfine structure or cases where Doppler broadening is dominant. How-
ever, they discuss the considerations for several atomizer cell geometries and
cell illumination geometries.
While studying isotopes of lithium, Goleb and Yokoyama(37) found
large variation in the isotopic composition of some reagent-grade lithium
salts. Since absorption spectral profiles may change if the isotopic composi-
tion of the element varies from sampIe to sampIe, such variations may pro-
duce errors and appear as matrix effects.
3.2.1.3. /nf/uence of Ho//ow Cathode Lamp Current
When the absorption cell and primary radiation source both have the
same spectral profile, the atomic absorption analytical working CUrves are
more sensitive to changes in the spectral profiles than if the absorption cell
had a broader spectral profile than the primary radiation source.
Bernhard(14) and Chakrabarti et al. (16) have found that the lowest possible
hollow cathode lamp currents give the best slope to the atomic absorption
working curve, but not necessarily the lowest detection limits, because of the
additional noise present at the lower lamp intensities. Attempts to increase
intensity in order to decrease shot noise(38) by increasing the current of a
hollow cathode lamp usually result in spectral broadening of the emission
line and a decrease in slope of the absorbance working curve.(I3,14,16) The
increase in broadening occurs because more atoms are sputtered from the
surface of the hollow cathode and diffuse out of the hollow cathode into the
optical path where they are no longer excited, but absorb light that was
emitted from inside the hollow cathode. Since the center of the spectral
profile absorbs photons more readily than the wings, the intensity in the
center of the profile does not increase as rapidly as the intensity in the
wings, and the profile broadens. This self-absorption may become so bad at
sufficiently high currents that self-reversal occurs, where the intensity at the
center of the spectral profile drops below the wing intensities.
In an extreme-but useful-<:ase of self-reversal, the intensity in the
center of the spectral profile drops to near zero in a region comparable in
width to the absorption cell spectral profile. The intensity in the far wings
of the spectral profile of the radiation source passes on through the absorp-
tion cell to be used to determine background absorption. (39)
Self-absorption in a high-current hollow cathode lamp can be reduced
by exciting the atoms that have diffused out of the excitation region inside
the hollow cathode. This additional excitation outside the hollow cathode
can be done with an auxiliary discharge such as a dc, (40) rf, (41) or
microwave(42) discharge. The Analyte 16 (Analyte Corp., Medford, Oreg.)
is a commercially available jet-enhanced sputtering cell atomic absorption
spectrometer that uses an rf pulse to improve the intensity and spectral
profile of commercial hollow cathode lampS operating in a pulsed mode.
Chakrabarti et al.(16) found that atomic absorption sensitivity depends
on the design and manufacture of the lamp and its age, as weH as on current.
Wagenaar et al.(43) measured the line profiles ofseveral different lamps under
different operating conditions, and their results can help to explain the
dependency of sensitivity observed by Chakrabarti and co-workers. As far
as age is concerned, it may be that an older lamp has a lower atomization
rate and therefore less self-absorption for the same operating current.
For an atomic fluorescence working curve, the initial slope will increase
as the primaxy light source intensity increases, until self-reversal in the pri-
mary light source becomes so obvious that the intensity in the center of the
spectral profile begins to decrease. As self-reversal increases, eventuaHy there
will be less light available in the center ofthe absorption (fluorescence excita-
tion) spectrum and the slope will decrease as lamp current is increased. If
scattering is a problem, the increase in self-absorption broadening of the
spectral profile of the primary source will increase the scattering contributed
by photons in the wings of the profile more than the increase in tluorescence
signal, and the detection limits will eventually deteriorate.
3.2.1.4. Fluorescence Spectral Resolution and Resonance

Monochromators
In some cases, atomic fluorescence will have better spectral resolution

or selectivity than atomic absorption when both use a hollow cathode lamp
as the primary radiation source. Consider a sufficiently low-pressure, low-
temperature cell, in which an excited atom undergoes no perturbing colli-
sions that change its Doppler velocity before it fluoresces (emits). In such
cells, better spectral resolution will occur for spectrallines where Doppler
broadening dominates over broadening caused by hyperfine structure. Under
these conditions, the fluorescence will have a spectral profile given by the
profile of the absorbed radiation, which is narrower than the profile of the
absorption line because preferential absorption occurs for those photons
that have wavelengths near the peak of the absorption profile. (5) This
78 CHAPTER 3
improvement in spectral resolution will only become apparent when the

fluorescence detection system has high enough spectral resolution to observe
it. Such a system would be based on an echelle monochromator, an interfero-
meter, or a resonance atomic fluorescence monochromator.
In cases where broadening of the line caused by hyperfine structure is
comparable to the Doppler broadening, the fluorescence peaks for each
hyperfine component may be narrower, but their overlap will still cause the
overall width of the line profile to be about the same. Then there will be
relatively little improvement in spectral resolution.
When a laser is used to excite fluorescence, an irradiance that is high
enough to begin to saturate the energy level transition is usually used. (7)
Although this increases the fluorescence, the absorption spectral profile is
broadened and the spectral resolution is reduced.(19) Although a potentially
interfering line that did not interfere at low irradiances may interfere at high
irradiances, such interferences are expected to be few if direct line fluoresc-
ence and an appropriate wavelength isolation system are used. Direct line
fluorescence occurs from the same upper level as the absorption/excitation
line, but to a different lower energy level, so that the fluorescence has a
different wavelength than the excitation. It is rare that an interfering element
would have both nearly the same excitation wavelength and nearly the same
fluorescence wavelength.
3.2.2. Sampling and Transport to the Observation Region
For atomic absorption and fluorescence measurements, it is desirable

to have the maximum number of ground-state atoms in the optical path,
and to minimize emission because of its contribution to noise. Therefore,
modem cells are designed to transport the sputtered atoms away from the
cathode so they can be observed at some distance from the glowing discharge
region that appears near the cathode. (As we will see, further reduction in
emission interference is obtained by pulsing the sputtering current and mak-
ing the measurement shortly after the current has been tumed off and the
emission subsides, but while a significant fraction of the atoms are still
present.)
3.2.2.1. Experimental Studies

McDonald(44) studied diffusion of beryllium and copper in a sputtering
cell to show the differences that occur when atoms differ by a factor of two
in their diffusion coefficients (because of differences in atomic mass and
radius). Absorbances were spatially resolved and made at different times up
to 20 ms after the sputtering pulse was shut off. Their results show that Be
atoms diffuse more rapidly from the light path than Cu atoms. Measure-
ments during steady-state sputtering show that along the center of the dis-
charge the Be/Cu absorbance ratio changes from 1.7 at a distance ofO.5 cm
from the sampie surface to 0.83 at a distance of 3.0 cm.
Winefordner and co-workers(45) used laser-excited fluorescence to study
diffusion ofNa atoms sputtered off a NaCl pellet. An axial scan at 2.9 Torr
shows a shallow maximum 0.2 cm from the surface of the sampie, indicating
the presence of a minimum nearer the surface. Others have observed similar
behavior during absorption measurements, and explain this behavior by the
presence of clusters of the analyte atoms.{46-48) These authors propose an
additional explanation that the minimum in Na atom population is caused
by a higher degree of ionization in the dark space near the cathode, caused
by energetic electrons in the cathode fall, compared with other locations.
After the sputtering pulse was switched off, the temporal decay ofthe sodium
density produced a decay in the fluorescence signal that dropped to e -I of
its initial value in 1.84 ms. When this value was used in a diffusion model,
a diffusion coefficient of 160 ± 50 cm2 S-I was found for Na atoms in Ar at
1.5 Torr.
Gough(13) was the first to use a flowing gas cell instead of a static cell
to move sputtered atoms into an observation region some distance from the
cathode. The gas enters the cell with high velocity through a hollow annulus
located 0.05 cm above the surface of the sampie (Fig. 3-2). The sputtered
atoms that diffuse into the rapidly moving gas are swept into the observation
zone about 1 to 2 cm from the sampIe. Absorbances were improved by over
an order of magnitude.
A decade later, Bemhard(14) reported the discovery that gas jets pointed
directly at the surface of the sampie cause craters to form where the gas jets
impinge on the surface. Sputtered atoms at the surface are immediately
entrained in the rapidly moving gas and diffusion of the atoms back onto
the sampie surface is significantly reduced. In the Atomsource, six gas jets
are pointed toward the center of a circIe and also at an angle to the sampIe
so that they impinge on its surface. This design forms a column of gas
containing entrained sputtered atoms that is directed away from the surface
and into the observation zone, producing a significant increase in sampIe
erosion and in the absorption signal.
In the cell used by Gough, the gas flow is perpendicular to the optical
axis of the absorption path. This provides an absorption path length of
1-2 cm. He recognized that the path length and therefore the absorption
could be increased by measuring absorption "along the axis of the plume of
atoms swept down the center ofthe sputtering cell by the fast gas flow .... "
In the Atomsource, the column of rapidly moving gas is split into a "T" so
that the entrained atoms are bent in the direction of the optical path to
increase the absorption path length.
80 CHAPTER 3
Gough et al.(49) studied the Atomsource and atomizer cells that have
observation regions in the shape of a straight tube, a tube with a 90° bend,
and a T-shaped tube. The T-shaped tube and the tube with the 90° bend
both show a higher absorbance measured along the length of the tube than
when measured across the tube. A comparison of absorption measurements
made through the sides of the straight tube and the 90° glass tube at various
points along the tubes shows that the sputtered atoms are efficiently swept
around the 90° bend. Their results show that for a given gas mass flowrate,
there is an optimum diameter for the tube when a 90° bend in used. Atoms
are more efficiently swept along the tube and around a 90° bend by the
higher gas velocity of a smaller-diameter tube, but there is also a faster loss
of atoms by diffusion to the tube wall compared with a larger-diameter tube.
(Comparison ofthe 90° bend with the T-shape configuration is difficult from
this report because different discharge voltages were used, and tubes of
different diameters were not studied to find the optimum tube diameter for
the T shape.)
The T-shaped glass tube that intro duces gas through an annulus parallel
to the sampIe surface, and the T -shaped Atomsource that uses gas jets that
impinge on the sampIe surface show comparable absorbances for a relatively
low cell current of 25 mA, cell voltage of 430 V, and agas flowrate of 0.3
liter min-). The authors conclude that the enhanced absorption sensitivity
of the Atomsource over the cell design of Gough(13) mainly results from the
increased absorption path length under these operating conditions. Further
studies are needed to determine whether this is also true for the higher
currents normally used with the Atomsource, since the gas jets of the Atom-
source and the gas annulus of Gough produce distinctly different sampIe
erosion characteristics.
3.2.2.2. Transport Theorv
The theory of gas transport helps us understand the influences of

diffusion and entrainment so that optimum operating conditions may be
chosen, and, eventually, improvements in cell design may be made, especially
for special applications.
Ferreira and Human(46) used a differential equation model to study
diffusion and transport in the Grimm-type glow discharge that uses a flowing
gas in a cylindrical cell where the sputtered atoms enter the cell uniformly
across one end of the 0.8-cm-diameter cylinder. Experimental determinations
of spatially resolved number densities of Cu and Cr atoms were used to fit
this mathematical model. Diffusion coefficients ranged from 240 to
320 cm2 s-) and carrier gas velocities from 400 to 730 cm s-). Their results
show that most of the sputtered atoms are removed from the plasma within
0.5 cm of the cathode. In our laboratory this same differential equation has
been used with boundary conditions that were modified so that the sputtered
sampie atoms enter through a sampling disk (i.e., at the surface of the
sampie) that may have a smaller diameter than the cylinder. This model
more dose1y approximates the geometry of flow cells like the Atomsource,
although it does not indude any bends in the direction of the gas flow. The
solution to the differential equation with these boundary conditions is given
in Boumans. (50) It was assumed as a boundary condition that the concentra-
tion of sputtered atoms is zero at the wall of the cylinder where rapid
deposition occurs. (The boundary condition for zero concentration at the
wall was effected by letting kiD = 100 in Eq. 9.27 in Boumans.(50»)
a. Model Limitations. It is worthwhile to briefly consider the influence

of the assumptions made in this model before considering the results. The
assumption that the concentration of sputtered atoms is zero at the wall is
a conservative, worst-case condition, because it leads to the fastest loss rate
and lowest concentration throughout the cell.
The model considers the flow in the cylinder to be uniform across the
diameter of the cell rather than laminar, which would produce a parabolic
velocity profile that is highest in the center and lower near the wall. Turbu-
lence tends to make the average flowrate more uniform across the cross
section of the cylinder.
Most of the calculations were done with a diffusion coefficient D =
120 cm2 s-I, corresponding to 5 Torr(50); others with D = 30, corresponding
to 20 Torr. These values produce slower diffusion than in Ref. 46, but are
doser to the values determined in Ref. 45. The pressures are typical of those
in the Atomsource. The cell pressure is assumed to be independent of flow
velocity, which may not be true if the vacuum system is operating at full
capacity. A velocity of 1000 cm S-I in a cylinder of radius 0.381 cm (radius
of a typical annulus) corresponds to a flowrate of 0.8 liter min- I measured
at atmospheric pressure.
The model assumes the analyte atoms enter the cell in a uniform manner
through the sampling disk and immediately attain the velocity of the carrier
gas. The mass flow of the analyte into the cell is assumed to be constant so
that the transport characteristics of the gas flow can be studied independently
of any influence that the gas flow has on sampie erosion. Since the mass
flow is constant and the carrier gas immediately dilutes the analyte, the
concentration of the analyte just off the surface of the sampling disk is
inversely proportional to the velocity of the carrier gas.
Figure 3-5 shows radial concentration profiles 0.3 cm from the surface,
corresponding to the mouth of the arrestor region, for two values of the
diffusion coefficient and two gas ve1ocities. The diameter of the cylinder
equals the diameter of the sampling disk, which corresponds to the geometry
in this part of a typical cell containing an annulus. The lower two curves are
82 CHAPTER 3
'1-; 3 FiguTe 3-5. Calculated radial profiles

(J for the concentration of sputtered
~ atoms 0.3 cm downstream from a flat
I 1---__
Q sampIe surface for diffusion coefficients
§ 2 D = 30 and 120 cm2 s-I and carrier gas
velocities v of 400 and 1I 00 cm S-I. The
o sputtered atoms are assumed to enter
eS uniformly across the open end of a
x cylindrical cell that has a radius of
§ 0.381 cm, where the concentration
eS 0 +--r---r--r--r--r--"--T---I drops to zero. Concentration is in num-
- 0 0.1 0.2 0.3 0.4 ber density per cm3 and Q is the number
Radius (ern) of atoms per second that enter the cell.
for the higher diffusion coefficient D = 120 cm2 S-1 (5 Torr), and show that
the two velocities 400 and 1100 cm S-1 produce about the same profiles at
this distance from the surface, with the lower-velocity curve showing the
lower concentration. Since the concentration just off the surface of the disk
is inversely proportional to velocity, the lower-velocity curve would be
expected to have the higher concentration. The reason it does not is that the
higher initial concentration has been offset by diffusion losses to the walls
that occur during the longer time that the gas spends in the annulus region
at the low flowrate.
The concentration in the center of the 1100 cm s-1 profile has dropped
to 43.7% of its initial value at the sampIe surface. Integration over the
circular cross section at 0.3 cm from the surface shows that 80% of the
analyte is deposited on the walls before the gas leaves the mouth of the
annulus. This loss increases to 93% for the lower velocity.
The two upper curves in Fig. 3-5 for a smaller diffusion coefficient (at
20 Torr instead of 5 Torr) show a greater difference. There is much less loss
of analyte to the walls than for the higher diffusion coefficient. The lower
velocity (top curve) shows a higher concentration because ofless dilution as
expected; the loss by diffusion has not yet overwhelmed the increase in
concentration produced by less dilution at the lower velocity.
These results indicate that there may be an optimum velocity that bal-
ances both (l) the increase in concentration caused by less diffusion loss to
the wall at higher velocities and (2) the decrease in concentration that occurs
because of the dilution effect with higher velocities. This optimum will now
be considered.
b. Optimum Velocity. The five lower curves in Fig. 3-6 show how the
concentration varies with gas velocity at five different points along the axis
of the cello The cylinder radius was increased to 1.0 cm to more accurately
represent the geometry of flow cells such as the Atomsource at larger dis-
tances from the sampie surface. Each curve has a maximum caused by the
Figure 3-6. Calculated sputtered
atom number density versus carrier "7 5 4
Integral
gas velocity at five distances z from .!!l.
the sampie surface. The cylinder ~ 3 z=l em
radius was increased to I cm while 0
the entry of atoms is stilllimited to a fä
~
2em
disk with radius 0.381 cm. The inte- 2l 2
gral of the number density from I to § 3em
4em
5 cm along the cylinder is shown as ~
a dashed line whose values have been 8 Sem
scaled arbitrarily so that it can be ö
plotted on the same graph. Q is the
1000 2000 3000 4000
number of atoms per second that
enter the cell. Gas Velocity, v (em/s)
balance between more dilution and less diffusion loss as flow velocity
increases. The flow required for the maximum concentration increases as
distance from the surface increases.
The analytical absorbance signal is proportional to the integral of con-
centration along the optical path. In a flow cell where the carrier gas bends
90°, the absorbance is measured along a path that begins at some distance
from the sampIe surface, say I cm, and ends at a window. To simulate this,
the top curve in Fig. 3-6 is proportional to the absorbance that would be
measured along the axis of the cell from 1 cm to 5 cm. A maximum in
absorbance appears for agas velocity of about 2700 cm S-I. Although the
geometry is somewhat different in areal cell, a similar maximum would be
expected in areal cell where losses by diffusion at low velocity and dilution
at high velocity also exist.
c. Radial Profiles for a Large Cylinder. Figure 3-7 shows how diffusion
rapidly spreads the analyte so that a large fraction of it is outside the radius
5 ,----------------------------,
Figure 3-7. Calculated radial pro-

file for the concentration of sputtered
atoms 0.01 cm downstream from a
flat sampie surface disk of radius
0.381 cm located at one end of a cylin-
drical cell with a radius of 1.0 cm. The
Disk Radius ~
diffusion coefficient is 120 cm2 S-I and
the carrier gas velocity is 160 cm S-I.
Q is the number of atoms per second o 0.2 0.4 0.6 0.8 1.0
that enter the cell. Radius (em)
84 CHAPTER 3
of the sampling disk even at a downstream distance of only 0.01 ern when
the diameter of the cylinder is larger than the diameter of the sampling disko
The radial profiles beyond 2 cm downstream have the general shape of the
profiles in Fig. 3-5. The radial concentration profiles for small distances from
the disk surface do not have the parabolic shape of Fig. 3-5, but have an
infiection point near the radius of the disko Since the concentration at the
disk radius is not zero, as it would be if the cylinder were the same size as
the disk, the concentration gradient from the center to the wall is lower, and
there is correspondingly less diffusion of the analyte to the wall. Conse-
quently, there is less loss of analyte whenever a larger-diameter cylinder is
used.
d. Axial Concentration Gradients. The log-linear plots in Fig. 3-8 show

that the concentration along the axis of the cylinder decreases exponentially
with downstream distance for three different cylinder radü from 0.381 cm to
1.0 cm. The mass fiow of the argon gas is the same (0.8 liter min- I at
760 Torr) for these cases and the gas velocity decreases with the square of
the radius. For the smallest-diameter cylinder (curve A), the concentration
drops about three orders ofmagnitude from a downstream distance ofO.3 cm
to 2.4 cm because of diffusion losses to the walls. For the l-cm cylinder
(curve C), the axial concentration is generally about an order ofmagnitude
higher and drops off less than two orders of magnitude in the same distance.
The larger-diameter cylinder decreases diffusion losses, and in this case also
increases concentration because of reduced dilution caused by the lower
carrier gas velocity.
The infiuence of velocity can be eliminated for the sake of discussion
by adding makeup gas to increase the velocity for the I-ern cylinder from
160 cm s-I to 1100 cm s-I. The added makeup gas causes the curve to change
from C to D, which shows a considerable improvement in downstream
~r----------------------------.
1:2 ·3
c
o D
~ -4
C Figure 3-8. Influence of cell radius
Q)
o
c
on axial concentration. Line A, cell
§ -5
Q. radius =0.381 cm, carrier gas velo-
Cl B city = 1100 cm S-I; line B, radius =
g -6 A 1.0 cm, veIocity = 160 cm S-I; line
C, radius = 1.0 cm, velocity = 1100
-7 L -_ _- ' -_ _ _ _' - -_ _- ' -_ _ _ _' - -_ _- ' -_ _- - ' cm S-I; line D, radius = 0.45 cm,
o 0.5 1.0 1.5 2.0 2.5 3.0 velocity = 788 cm s -1. Diffusion
Downstrearn Distanee. z (ern) coefficient = 120cm2 S-I.
concentrations. Clearly the change in cylinder radius from 0.381 cm (curve
A) to 1.0 cm while maintaining the same gas velocity (curve D) causes a
dramatic change in concentration, and emphasizes the significance of losses
caused by diffusion to the walls when rapid deposition occurs there.
These studies of the influence of gas transport can be summarized as
folIows:
1. Most of the analyte is deposited on the walls around the sampling

disk before it has a chance to enter the observation region.
2. Largei~diameter tubes help to reduce analyte losses caused by deposi-
tion on the wall.
3. The concentration decreases exponentially with distance from the
sampie.
4. An increase in pressure causes an increase in concentration, because
the diffusion losses are reduced due to the lower diffusion coefficient.
5. Higher flowrates increase the absorption signal, but eventually a
maximum is reached.
It should be kept in mind that gas flowrates and pressure may influence
the sputtering process as well as the transport processes. For example, Kim
and Piepmeier(51) found a maximum in absorbance when the carrier gas
flowrate was changed, but the cell voltage and sampie erosion rates also
changed with flowrate. In any case, these maxima are important because
they provide a relatively stable operating region where absorbance does not
change with minor changes in flowrate.
3.2.2.3. Transient Samp/es
When the mass of the sampie (such as a dried solution) is limited so

that a transient absorbance peak occurs, additional considerations must be
taken into account. Ideally the largest analytical signal would be obtained
by quickly atomizing the entire sampIe and transporting all of the atoms to
the observation region. Regardless of whether or not the ideal case can be
approached, peak duration is important because a high, short (narrow) peak
has a larger signal-to-noise ratio than a low long (wide) peak of the same
area when background noise is the main source of noise, because more
background noise is present during the long peak. This suggests that atomiz-
ation should occur as quickly as possible, and that the atoms should be
transported by the carrier gas to the observation region as fast as possible
to minimize diffusion losses, but that the carrier gas should not transport
the atoms out of the observation region. Although the fundamental theory
of this transient transport problem remains to be studied, it seems clear, in
86 CHAPTER 3
the face of diffusion losses, that there will be an optimum carrier gas flowrate,
and that it may be helpful to shut off the carrier gas flow at some optimum
point in time near the maximum of the analytical signal.
Chakrabarti and co_workers(l6,52-54) have been able to fit an equation
derived from kinetic theory to transient absorbance peaks obtained using
dried solutions deposited on the cathode of the Atomsource. The model
assumes that the rate ofloss of atoms from the observation region is propor-
tional to the concentration of atoms [B]/ in the observation region at any
time t, and similarly that the rate of entry of atoms into the observation
region is proportional to the concentration of atoms [A]t that can enter the
observation region at time t. The proportionality constants are first-order
rate constants k 1 and k 2 and the resulting equation is
(3-5)
where [A]o is the initial concentration of atoms that can enter the observation
region at t = O. This equation describes a peak that quickly rises in an
exponential manner with a rate corresponding to the larger of the two rate
constants, and that exponentially falls more slowly at a rate corresponding
to the smaller of the rate constants. Values can be obtained for the two rate
constants by fitting this equation to the shape of an experimentally deter-
mined absorbance peak. However, since [A]o is not known and since the
same peak shape is produced simply by interchanging the values for the two
rate constants, the shape of the peak alone does not reveal whether the rate
constant for entry of atoms into the observation zone is larger or smaller
than the rate constant for the loss of atoms. Which rate constant is smaller
is determined from other measurements, such as the rate of loss when the
power was suddenly turned off during a pulse.
It was concluded that for low-power sputtering conditions, the rate of
entry of atoms into the observation region was the limiting step, whereas
for higher-power sputtering conditions, the rate of loss was slowest. Even
so, since the ratio of the entry rate constant to the loss rate constant is ideally
large, but was experimentally found to be only in the range of 1.3 to 7.0 for
14 elements, they concluded that there is considerable room for improving
the sensitivity of the transient method, for example, by increasing the entry
rate by increasing the sputtering rate.
3.3. Practical Considerations
3.3.1. Methods Development
Although instrument manufacturers usually have recommended sets

of operating conditions for a variety of sampies, an analytical laboratory
sometimes encounters other types of sampIes or requirements. Additional
considerations that are helpful in choosing a set of operating conditions are
therefore worthwhile discussing.
3.3.1.1. Choice of Spectral Une
Experience with the Atomsource has shown that a spectralline that is

commonly used for flame or graphite furnace AAS may not be the best line
for use with a low-pressure sputtering cell.(55) One reason for this may be
differences in background emission interferences. Another reason may be
the increased influence that the spectral profile of the emission line from the
hollow cathode lamp has on the slope and shape of the analytical working
curve in low-pressure cells as we have seen.
When pulsed operation is used, lines may vary in their pulsed mode
behavior. Demers et al.(56) show the redistribution in ground-state and nearly
ground-state lines of U that occurs in the few milliseconds after a sputtering
pulse is turned off in a hollow cathode absorption cello
The best atomic absorption lines for a sputtering cell are usually among
the lines that have a lower energy level at or within a few hundred kaysers
(wave numbers, cm -I) ofthe ground state ofthe element and have the highest
gfvalues, typically above 1.0, wherefis Ladenburg's oscillator strength and g
is the statistical weight of the upper energy level (also known as the degeneracy,
the number of electrons that can occupy the upper energy level). Tables such
as those in Ref. 57 often list the product of g and f, rather than individual
values. The absorption coefficient is proportional to the gf value times the
square of the wavelength. (19) Therefore, for two spectral lines with widely
different wavelengths, the gfvalues found in a table should be multiplied by the
square of the wavelengths before a comparison of the strengths of absorption
(absorption coefficients) is made. On the other hand, the error in comparing
the gf values alone directly from a table is only a factor of 4 when comparing
a line at 200 nm with one at 400 nm, and it would not be uncommon to find
the gfvalues to be in error by that amount anyway.
Additional care is necessary when tables of gA values are used to estimate
the absorption coefficient, where A is the Einstein transition probability per
second. Although gA values in tables are appropriate when estimating emis-
sion intensities (particularly for atom cells in thermal equilibrium), the absorp-
tion coefficient is proportional to the fourth power ofthe wavelength times the
gA value. Therefore, estimating relative absorption coefficients by comparing
gA values alone, without first multiplying by the fourth power of the wave-
lengths, will produce an error of a factor of 16 when comparing a line at 200 nm
with one at 400 nm. The values of gf and gA are related by the equation
(3-6)
where the wavelength A is in angstroms.

88 CHAPTER 3
A more accurate comparison also requires dividing each gf or gA value

by the statistical weight gl of the lower energy state. Unfortunately, these
values are not always readily available. Information about the hyperfine
structure ofthe lines being compared is also needed for an accurate compari-
son. Again this information is usually not available. Therefore, it may be
worthwhile to choose more than one line from a table and check the experi-
mental performance of each one.
Fluorescence is a two-step process involving absorption followed by
emission. Therefore, some function of the product of the absorption
coefficient times the Einstein probability for emission is compared when
choosing spectrallines. When the same wavelength is used for the absorption
and fluorescence steps, it is appropriate to compare the square of the gf
values (divided by gl squared when known). When the lines are different, a
comparison is made of the product of ;'}gffor the absorption (excitation)
wavelength times gA for the fluorescence wavelength. These comparisons
for fluorescence do not take into account quenching of the excited state.
However, true quenching is uncommon in sputtering cells that use noble
gases.
The choice of lines for fluorescence also depends on the relative intensi-
ties of the lines emitted by the primary light source. Fortunately, the strong-
est absorption line for the analyte often corresponds to one of the stronger
emission lines from a hollow cathode lamp. However, when a laser is used
to excite fluorescence, the choice of lines may be limited by the relative
intensities of the lasing wavelengths available for that particular laser. When
background correction is required, the choice of line may be influenced by
the performance of the background correction system, which may vary from
line to line. Although background has not been a problem for metal alloy
analysis with the Atomsource, this is not true for cells that do not direct gas
jets direcdy at the sampie surface. Even with the Atomsource, background
may be found when sampies, such as organic materials (made into pellets
with a conducting matrix), are used that contain major concentrations of
elements that form gaseous molecules, for example C2 and CN, that absorb
in the visible and ultraviolet.
3.3.1.2. Plasma Gas
The quality of the plasma gas is important. The effect of 17 ppm water
vapor in argon plasma gas on the determination of Cr in low-alloy steel is
to decrease the absorbance by 7% and double the noise levet<49) compared
with runs using argon containing 3 ppm water. It is best to keep atmospheric
gases out of the sputtering cell and its gas and vacuum tubing so that
moisture does not adsorb on the walls and then outgas during an analysis.
When in storage the cell should be evacuated. During sampie changing, a
positive pressure of argon in the cell along with an accompanying flow of gas
out of the sampie port helps to keep the atmospheric gases out. Larkins(58)
describes agas control unit for sputtering cells and presents important
considerations in its design.
3.3.1.3. Discharge Conditions
High sputtering currents increase the atomic population and the

analytical signal. For a given current, decreasing the pressure increases
sputtering because the sputtering ions attain a higher kinetic energy due to
the increase in their mean free path and the accompanying increase in cell
potential that is necessary to keep the same current. Fang and Marcus(59)
have found that the sampie erosion rate for a flat cathode is proportional
to the electrical power (current times voltage) applied to a cell above a
particular threshold power that varies with pressure (Fig. 3-9.) Chakrabarti
et al. (16) have found that controlling the power to the Atomsource instead
of controlling current produces more reproducible results.
However, care must be used when lowering the pressure to increase
voltage and power in order to increase sputtering, because reduced pressure
increase losses by diffusion. Also, if the vacuum system is running at maxi-
mum capacity, then the pressure is reduced by reducing the flow of carrier
gas into the cell chamber. The reduction in mass flow of the carrier gas
reduces the transport of sputtered atoms from the surface of the sampie into
the observation zone. Therefore, there may be an optimum pressure that
will provide the best analytical signal.
If detection limits are a consideration, then high power is desirable.
Although a high sputtering power increases the atomic population and the
analytical signal, it also increases sampie heating (because most of the power
is dissipated near the cathode) and tends to increase background emission.
Although most instruments are able to compensate electronically for back-
ground emission, the noise in the background emission signal cannot be
eliminated and still contributes to the total noise in the analytical signal.
Therefore, when the background emission signal approaches the transmitted
emission signal of the primary source in atomic absorption, it may be worth-
while to determine if an increase in sputtering current still increases the
signal-to-noise ratio. A similar study of signal-to-noise ratio is worthwhile
when the background emission signal approaches the scattering signal of the
primary light source in atomic fluorescence.
3.3.1.4. SampIe Temperature
High temperatures produced by sampie heating may cause a change in

the sputtering rate and preferential diffusion of atoms from the bulk to the
90 CHAPTER 3
80
A
Ci 60
t a 1.5 Torr
•• • 2.0 Torr
40 • 3.0 Torr
-
.!
• 4.0 Torr
.t::.
Cl
•
D
5.0 Torr
6.0 Torr
'i
~
20
• 8.0 Torr
0
0 10 20 30 40
Power(W)
60
eil
t a
•
1.5 Torr
2.0Torr
•• 40
• 3.0Torr
-
.!
.t::.
CII
'i 20
o
•
D
4.0Torr
5.0Torr
6.0Torr
~ • 8.0Torr
5 10 15
Redueed power (W)
Figure 3-9. Effect of sputtering discharge power on sampIe erosion of a Hat cathode at
various pressures. (A) Weight loss versus applied power; (B) weight loss versus reduced power,
i( V - Vo), where Vo is the threshold voltage at which a plot of weight loss versus discharge
voltage intercepts the voltage axis. Reprinted with permission from Spectrochim. Acta 43B, D.
Fang and R. K. Marcus, Parametric evaluation of sputtering in a planar g10w discharge-1.
Sputtering of oxygen-free hard copper (OFHC), copyright 1988, Pergamon Press.
surface of the sampIe. (60) Although it might be possible to compensate for

sputtering rate drift, this adds an additional variable to the technique, and
the thermal characteristics of sampIes may contribute to matrix effects if
compensation is not complete. Preferential diffusion for a sampIe that is hot
near its surface will cause the bulk material to become depleted in some
elements, and the sputtered atomic vapor will then no longer be representa-
tive of the sampIe. Preferential diffusion could be a cause of matrix effects
for sampIes that have different diffusion characteristics.
SampIe temperature can be reduced by cooling the sam pIe. Although
cooling has been done using dry ice and liquid nitrogen for special
applications,(61) water cooling is most commonly used. The Atomsource
cools the sampie by pressing the front surface of the sampie against a
water-cooled copper plate surrounding the sputtering region. With this con-
figuration only one sampie surface needs to be flat. Other methods include
a water-cooled plate pressed against the flat back of a sampie. (5) The latter
method may be helpful particularly for sampies with poor thermal conductiv-
ity if the sampies are thin relative to the diameter of the sputtering region
(e.g., some sampies of superconductor materials). The cooling effect of gas
jets in a gas-jet-enhanced sputtering cell remains to be determined.
The rise in temperature during sputtering of a water-cooled sampie
was studied experimentally by Gough et al.(5) Although the temperature
continued to rise for several minutes after the discharge was turned on,
stable fluorescence signals were obtained in 1-2 min.
Pulsing the sputtering discharge while maintaining the same average
power, increases the instantaneous sputtering rate during the pulses without
changing the average sampie temperature. However, further studies are
needed to determine the influence of the transient temperature rise during
the pulses.
3.3.1.5. Preconditioning the Surface (Preburn)
The best accuracy and reproducibility are usually obtained for an ana-
lysis of the bulk of a sampie when the sampie surface is preconditioned by
sputtering before the analytical measurements are made. Preconditioning
removes surface contaminants and roughness, and allows the surface compo-
sition to adjust to an equilibrium composition that eompensates for the
differences in sputtering rates of different elements. When the equilibrium
composition is reaehed, the surfaee is proportionally rieher in elements with
lower sputtering rates, and the composition of the sputtered material is
therefore the same as that of the original sample.(5,62-64)
All surfaces pick up contamination from the atmosphere and other
things with whieh they are in eontaet. Maehining or polishing methods
used to prepare aflat surface may cause contamination. Preconditioning the
surface by sputtering helps to remove contamination so that the analysis is
representative of the bulk material.
Sometimes the surface may be rough enough to influence the gas flow
patterns when gas jets strike the surface as they do in the Atomsource.
Therefore, preconditioning will usually be done at a low gas mass-flowrate
so that the gas jets do not cause preferential erosion of the surface where
the jets strike the surface. In a gas-jet-enhanced sputtering system where the
vacuum system is removing gas at its maximum capacity, a low gas mass-
flowrate is produced by restricting the entry of gas into the cello The decrease
in gas flow is accompanied by a reduction in cell pressure, which produces
92 CHAPTER 3
an increase in cell potential and in sputtering rate for a given current. Because
the influence of the gas jets is minimal at the low flowrate, this high sputtering
rate is uniform across the sampie surface and is desirable for preconditioning
the surface. Although the sampie erosion rate may decrease because of
increased redeposition at low gas flows, the high· sputtering rates will cause
the surface to quickly smooth out and come to equilibrium and lose
unwanted adsorbed gases. On the other hand, surface contamination by
solids may require a longer removal time because of the lower erosion rates.
After preconditioning, the gas mass-flowrate is increased to increase the rate
of sampie erosion under each gas jet and to increase the transport of the
sputtered atoms away from the surface to the observation zone for the
analytical measurements. These are general guidelines, and systematic
studies may be needed to obtain the very best results.
3.3.2. Relative Precision for Atomic Absorption Determinations
Because of the exponential relationship between concentration and the

fraction of light transmitted in an absorption measurement, there is an opti-
mum transmittance or absorbance at which the best relative precision in
concentration is obtained. Although this optimum is commonly presented
in textbooks when discussing spectrophotometric measurements with solu-
tions, it is rarely considered in flame or furnace AAS, probably because most
of these determinations are near the detection limit or do not require the
highest obtainable precision. However, under proper conditions the preci-
sion of emission and absorption measurements using sputtering cells can
be better than 0.1%.0 4 ) Since sputtering ceHs are commonly used for the
determination of major and minor elements in a sampie as weH as trace
elements, it is worthwhile to briefly review what is required to obtain the
best relative precision for an atomic absorption measurement.
Figure 3-10 shows a plot ofthe relative uncertainty, or relative standard
deviation (RSD), in concentration versus absorbance for the Mn(I)
403.3-nm line in an atmospheric flame (lower curve) and in a low-pressure
sputtering cell (upper curve). These curves were ca1culated from a model
that assumes that shot noise is the limiting source of noise. The noise in the
blank is arbitrarily assumed to be 0.01%; a different value would simply
shift the curves to a different vertical location and their shapes would not
change. At low absorbances, decreasing the concentration causes the RSD
to become worse because the analytical signal decreases while the back-
ground noise in the analytical signal remains the same. Background noise
for an absorption measurement corresponds to noise in the beam from the
primary radiation source, almost all of which is being transmitted to the
detector at low absorbances (high transmittances). At high absorbances, the
fraction of the beam from the primary radiation source that reaches the
1.00
o
CD
CI:
. 010'--------'.---~-----'-------'
o .500 1. 00 1. 50 2. 00
Absorbance
Figure 3-10. Relative standard deviation (RSD) in concentration versus absorbance for the
Mn(I) 403.3-nrn line in an atmospheric-pressure flame (lower curve) and a low-pressure sputter-
ing cell (upper curve), for measurements where the RSD in the blank is 0.01 and the uncertainty
in the spectrophotometer readings is limited by shot noise. Reprinted with perrnission from
Spectrochim. Acta 448, E. H. Piepmeier, The influence of the spectral profile of the source lamp
for low pressure atomic absorption cells, copyright 1989, Pergamon Press.
detector is small, and low intensities are difficult to measure because of shot
noise. Increasing the concentration reduces in an exponentially fast manner
the fraction of primary source radiation reaching the detector, and the RSD
continues to become worse as concentration increases.
Between these two extremes, at the bottom of the "U" -shaped curve
lies an optimum absorbance where the relative uncertainty in absorbance or
concentration is lowest. The minimum varies with the source of noise and
the shape of the analytical working curve. (38) These curves have been studied
in detail in our laboratory(35) for 17 lines that have widely different hyperfine
structures, since hyperfine structure controls the shape of the analytical
working curve for a given primary radiation source and absorption cello
The shapes of the spectral profiles of the primary radiation source and the
absorption cell were assumed to be the same. The differences in the curves
in Fig. 3-10 are caused by the greater curvature of the analytical working
curve in a low-pressure sputtering cell compared with the curvature of the
working curve in an atmospheric-pressure cello For a given RSD in the
blank, the location of the minimum along the absorbance axis is generally
in the range of 0.40-0.45 for a low-pressure cell if the noise is background
or dark current limited and is in the range of 0.62-0.80 if the noise is only
limited by shot noise as in Fig. 3-10 (the best situation). The best RSD
94 CHAPTER 3
in concentration is three times worse than the RSD of the blank for the
background- or dark current-limited case and for the shot noise-limited case
is two times worse.
It is best to choose an analytical line and absorption cell operating
conditions to obtain an absorbance in these ranges whenever the highest
precision is desired. However, the differences in blank noise for different
analytical lines must also be considered since the RSD of the sampie is
directly proportional to the RSD of the blank. In some cases it may be better
to choose a line that gives an absorbance on one side of the optimum range
if that line has a better blank RSD than a noisier line that causes the
absorbance to fall with the optimum range.
3.3.3. Background Correction
Background emission from the anode glow and positive column of a

glow discharge are usually eliminated by positioning the anode at the edge
of the negative glow. Unfortunately, this does not eliminate all emission,
which interferes with atomic absorption and fluorescence measurements. The
background emission signal from the sampie cell can be compensated by
modulating the primary radiation source and synchronizing the detector with
the modulation frequency. The detection system measures the background
emission signal when the primary radiation source is off and subtracts this
signal from the total signal when the primary light source is on.
Although the background emission signal is eliminated, its noise is
not. (38) If background emission noise is limiting the ability to detect small
concentrations of analyte, then background emission that is near but not at
the wavelength of the analysis line may be reduced by narrowing the spectral
bandpass of the detection system (e.g., by narrowing the slit width of the
monochromator). This will improve the detection limit until the photon rate
of the analysis line is reduced to the point where shot noise in the analytical
signal is the dominant source of noise.
Background emission at the wavelength of the analysis line (e.g., analyte
emission) may be reduced by changing the operating conditions of the dis-
charge. For example, emission may be reduced by increasing pressure, if the
accompanying loss in sputtering rate can be tolerated. An alternative is to
pulse the discharge and make the analytical measurement after the pulse
when the emission dies down, but while the atomic vapor is still present as
was first done by Yokoyama and Ikeda(65) for atomic absorption measure-
ments. This is also commonly done with fluorescence measurementsY 8,45,66)
The Analyte 16 atomic absorption spectrophotometer uses a similar pulsed
mode of operation to eliminate emission with the Atomsource.(I4,67) Ifback-
ground emission is still present in the pulsed mode, care must be exercised
in measuring the background emission at a different time than the absorption
signal because both signals are varying with time and an over- or undercom-
pensated correction may occur.(67)
Molecules and small particles cause background absorption, fluor-
escence, and scattering.(I3,44,68) McDonald(69) discusses Mie scattering and
measured the beam attenuation from 300 to 450 nm for a silver sampIe in a
glow discharge. He estimates that clusters of atoms may contain 30% of the
sputtered atoms. In other papers he reports background absorbances up to
0.002 near 234.8 nm and 222.6 nm for a Be-Cu alloy, (44) and from 0.004 to
0.006 for silver sampIes, 0.001-0.002 for copper sampIes, and less for other
alloys.(70) Gough(13) reports broadband beam attenuations of 0.25-1% for
iron, copper, and aluminum sampIes and 2% and more for silver. He attri-
butes the attenuation with silver sam pIes to be caused by the aggregation of
metal atoms in the discharge to form particles with diameters around 5 nm.
Patel and Winefordner(71) observed laser-excited fluorescence from CuO
and diatomic molecules of copper and lead in a simple planar geometry glow
discharge at 4.5 Torr and 25 mA. Broadband scattering attributed to large
metal clusters was also observed. No evidence for background absorption
has been reported for metal alloys run on a gas-jet-enhanced sputtering cello
Since aggregate formation is expected to be highest near the surface where
the concentration of atoms is highest, perhaps the flow characteristics of the
gas jets striking the surface reduce aggregate formation or direct the heavier
aggregates out of the column of carrier gas that transports the atoms to the
observation zone. However, until more is known about this, the presence of
molecu1es and small particles may be anticipated when nonmetallic sampIes
are analyzed, particularly those with an oxide or organic matrix.
The same methods that are used for other atomic absorption cells may
also be used to correct for background absorption in a low-pressure sputter-
ing cello These include the use of a continuum source, lines near the analysis
line, self-reversed hollow cathode lines (e.g., the Smith-Hieftje method), and
1ines shifted by the Zeeman effect for atoms in a magnetic field. These
methods are discussed in more detail in textbooks.(e. g., 72,73) The
Smith-Hieftje method(39) has been shown to cause rollover (bending back
down toward the concentration axis) of the analytical working curve at
high absorbances for flame atomic absorption when stray light or other
nonabsorbable spectrallines reach the detector.(33,74,75) Similar problems are
expected in low-pressure cells. Nonabsorbed lines and their influence can
sometimes be reduced by reducing the spectral bandpass of the detection
system, which then also changes the shape of the working curve and may
increase the noise level.
Zeeman splitting methods will work if the magnetic field is placed
around the hollow cathode lamp. However, a Zeeman method that places
the sputtering cell in the magnetic field will distort the plasma. This may be
beneficial if the magnetic field is stationary, but achanging magnetic field at
96 CHAPTER 3
the cell, a technique often used for background correction, may produce an
unacceptably changing background absorption. Tie-Zheng and Stephens(76)
found that placing the magnetic field around a hollow cathode lamp pro-
duced a worse signal-to-noise ratio than placing the magnetic field around
the atom cell (a flame) when high lamp currents were used for atomic fluo-
rescence. This was attributed to the emitting and the self-absorbing regions
in the lamp not being subjected to the same magnetic field strengths. Clearly,
Zeeman background correction methods need to be studied experimentally
for low-pressure sputtering cells.
Because spectrallines with minimal hyperfine structure are narrower in
a low-pressure sputtering cell than in atmospheric flames and fumaces, there
will be less overlap of the background correction radiation with the absorb-
ing analyticalline. Therefore, for these lines all of the correction methods
may be able to compensate over a larger range of background absorption
than when they are applied to higher-pressure cells.
On the other hand, all of these background correction methods assume
that the background absorption is the same for the analyticalline as for the
background correction wavelengths. Although this essentially may be true
for broadband absorption caused by small particles and large molecules, the
absorption spectra for small molecules may be more highly resolved at low
pressures than at atmospheric pressure, and the background will be more
structured than smooth. This increases the possibility that the background
produced by small moleeules will not be completely or properly compen-
sated, and care in studying background correction must be exercised just as
it must be for atmospheric cells. (73)
In atomic fluorescence, small particles and molecules produce back-
ground scattering, which ultimate1y limits detection limits. Scattering may
also occur off of walls and other parts of the instrument that are within the
optical aperture of the detector. Such scattering is usually eliminated, when
possible, by using direct-line fluorescence, where the excitation and fluoresc-
ence wavelengths are different, but where they share a common upper energy
state.
Background fluorescence mayaIso occur from windows, lenses, and
other parts of the instrument. This can be minimized by designing the system
so that these parts are not excited by the primary light source or its reflec-
tions, and by the use of materials that are manufactured specifically to
minimize fluorescence.
3.3.4. Pulsed Sputtering Cells
The analytical signal can be increased by increasing the sputtering rate.

However, the highest electrical power that can be used in a sputtering cell
to increase the sputtering rate is ultimately limited by overheating of the
sampIe. Overheating tends to prevent surface equilibration, and will eventu-
ally convert the discharge into an are. Overheating mayaIso crack fragile
sampIes.
To avoid overheating, short, high-power sputtering pulses may be used
to produce higher instantaneous concentrations of analyte atoms than is
possible with a dc current. (14,16) Although the temperature rise during a pulse
may be high, if the pulses are spaced far enough apart, cooling between
pulses occurs, especially in the presence of gas jets, and the average tempera-
ture rise can be limited to tolerable levels.
Capturing the analytical signal at its highest value, rather than observ-
ing its average value, requires a pulsed detection system that is synchronized
with the sputtering pulses. Since a shorter signal integration time or response
time tends to raise the noise level, pulsed primary radiation sources (e.g.,
pulsed hollow cathode lamps) that produce higher spectral radiances than
dc sources are commonly used to minimize shot noise.
A pulsed mode of operation not only helps to improve detection limits
by increasing the analytical signal, but also allows emission noise to be
reduced. Emission noise can be reduced by measuring the analytical signal
after the discharge pulse is turned off and the emission dies down, but before
the atoms dissipate. Although the pulsed mode of operation is somewhat
more complex than the dc mode, the lowest detection limits reported so far
have been achieved by pulsing the sputtering cello (16)
3.3.4.1. Pu/se Duration
There are generally three different cases to consider with a pulsed sput-
tering cell: the case where the sampIe lasts for many pulses,(14,18) the case
where the sampIe lasts for a limited number of pulses, and the case where
the sampIe is consumed during one pulse.(16) A metal alloy or pressed pellet
will repeatedly generate atom clouds for many pulses. This situation allows
preconditioning sputtering of the sampIe to improve the accuracy and preci-
sion of the analysis. It also provides the greatest flexibility in making
measurements that compensate for drift in the intensity of the primary light
source, background emission, and background absorption. The time between
pulses can be lengthened as necessary to ensure that these measurements can
be made independently for each pulse.
When the sampIe is a thin film or a dried solution deposited on the
cathode, the analytical signal may last for only a limited number of pulses,
and the envelope of the sequence of pulses may appear in the shape of a
peak rather than a steady plateau. In this case, any time difference between
the time that the analytical signal is observed and the time that a background
signal is observed will require proper interpolation of the results to produce
accurate background correction.
98 CHAPTER 3
When the sampie is consumed during one current pulse, the response
time of the detection system must be fast enough to keep up with the
changing signal if the most sensitive results are to be obtained. Emission
background correction may be possible by switching rapidly back and forth
between the total signal and the background signals (emission and absorp-
tion). Piepmeier and de Galan(77) have studied the effect of a slow detection
system that uses a modulated light source and detection system to observe
a transient absorption peak. They modified the general expression for the
atomic absorption signal to allow for possible time-dependence of the source
emission and sampie absorption spectralline profiles. A table in their paper
summarizes the results for several cases of practical interest. They found
that deviations from Beer's law may occur even with a perfect1y monochro-
matic source if the sampie absorption is time-dependent.
3.3.4.2. Peak Height or Peak Area?
When a peak is observed, there is the question of whether to use the

peak height or the peak area as the analytical signal. This question has been
actively discussed for years in graphite furnace AAS. Absorbance peak area
is recommended if the width of the peak changes from sampie to sampie or
run to run because of matrix effects, since the area of the peak often remains
proportional to concentration. However, this assumes that the response time
of the detection system is fast enough to give an accurate representation of
absorbance belore the integrations occur, that all ofthe analyte in the sampie
is eventually observed, and that each atom of analyte, when it is observed,
contributes in the same proportion to the analytical signal. Confusion about
whether area or peak height is better will undoubtedly occur when all of
these requirements are not met.
Chakrabarti et al.(16) have shown that dried solutions of different salts
sputter at different rates. Ohls(78) has shown that peak width depends on the
amount ofthe same element present in a dried solution. These results indicate
that peak width is expected to depend on the sampie matrix for dried com-
plex multielement solutions.
The use of peak area should have a signal-averaging advantage similar
to increasing the time interval over which the analytical signal is observed.
However, if the integration is carried too far into the wings of a peak, then
the noise in the baseline begins to cause deterioration in the signal-to-noise
ratio. If the area is to be used and the shape of the peak does not have a
mathematical model to which it could be fit, then the locations of the be gin-
ning and end of the peak may not be clear in the presence of noise. The
influence of baseline noise on the choice of how much of a transient absorb-
ance peak to include in the integral to obtain the best signal-to-noise ratio
has been considered in detail by Piepmeier .(79)
3.3.5. Multielement Determinations
The determination of more than one element in a sampie is often
desired. Various approaches to multielement atomic absorption determina-
tions have been reviewed by Salin and Ingle.(80) There are three basic
approaches to multielement determinations, depending on the available
instrumentation: simultaneous measurements of all elements of interest,
rapid sequential measurements of all elements while the sampie is being
sputtered, and the determination of one element at a time for all sampies,
then the next element for all sampies, and so on. The last approach is used
with most atomic absorption spectrophotometers because they cannot easily
measure the absorbance at more than one or two wavelengths during sampie
atomization. A rapid sequential method to determine 4 elements is described
by Salin and Ingle. (81) Light from four hollow cathode lamps is combined
into one beam with semitransparent mirrors, passed through the absorption
cell and into a polychromator. Light from each of the exit slits is reflected
to the same photomultiplier. The lamps are pulsed in rapid sequence so that
the signals for each lamp can be distinguished. Rapid sequential determina-
tions of up to 24 elements are possible during the sputtering of one sampie
with the Analyte 16 atomic absorption spectrophotometer because of its
ability to change reproducibly the hollow cathode lamps and the wavelength
settings of the monochromator within a few seconds.(15)
Simultaneous measurements are needed when the sampie is limited and
atomizes in a transient manner. Simultaneous atomic absorption measure-
ments are discussed by Mavrodineanu and Hughes.(82) Basically, a grating
is used to combine the beams from several hollow cathode lamps, pass the
combined beam through the absorption cell, and then another grating is
used to disperse the different wavelengths to separate photomultiplier tubes.
A modification of this approach is used in a Perkin-Elmer portable graphite
fumace atomic absorption spectrophotometer for wear metal analysis. (83)
This instrument uses a semitransparent flat mirror instead of a grating to
combine the beams from two multielement hollow cathode lamps. Instead
of several hollow cathode lamps, Ohls et al. (6) used a Grimm-type glow
discharge as the primary radiation source for multielement atomic absorp-
tion because of the ease of selecting and changing the elements by choosing
an alloy of those elements as the flat cathode. As an alternative to spatially
dispersing different wavelengths with a grating, separate wavelengths in a
combined beam can be detected with resonance detectors in series with each
other, each consisting of a resonance atomic fluorescence monochromator
to isolate a wavelength along with a photomultiplier tube to detect the
fluorescence. (84) Although a continuum light source, such as a xenon lamp,
might be used with a resonance atomic fluorescence monochromator, the
low spectral radiance of such sources and the high potential for scattered
100 CHAPTER 3
light from all of the unused wavelengths would make this an unlikely choice
for an instrument that would have good detection limits.
Multielement atomic fluorescence is somewhat less complicated because
the beams from each hollow cathode lamp do not have to be combined, but
can be focused from separate directions onto an appropriately designed atom
cell. A polychromator with a photomultiplier for each wavelength can be
used todetect the fluorescence if background emission is a problem and
simultaneous determinations are needed. Alternatively, separate photomulti-
pliers with interference filters in front of them could be positioned around
the atom cell to observe the fluorescence simultaneously, as was done on an
inductively coupled plasma AFS instrument manufactured by Baird Corpor-
ation (Bedford, Mass.). Background emission can be eliminated by pulsing
the atom cell and observing the fluorescence after cell emission has subsided.
Rapid ~quentia1 observations can. be made with a single photomultiplier
that observes the rapid sequence of pulses of fluorescence that occur when
each of the hollow cathode lamps is pulsed in rapid succession.
3.3.6. Compensation tor $puttering Rate Variations
Because sputtering rates do not vary from one type of sampie to the
ne~ nearly as much as vaporization rates vary for thermal methods of
atomization, it is possible to use the same analytical working curves for
sampIes that vary more widely in composition. For example, the same
working CUrves can be used with the Atomsource for some high-alloy and
low-alloy steel sampies. (85) On the other hand, Fang and Marcus(64) have
shown that the weight loss for aseries of brass sampies varies with the
concentration of copper. In such cases, improved analytical results can be
obtained by compensating for differences in sampie erosion rates. (70,86)
One method of compensation is to weigh the sampie before and after
the analysis as suggested by Gough et al.(5) However, the loss of material
during a typical analysis time is only in the microgram range. Even with a
microgram balance, such a small weight loss may be difficult to determine
because of the possibility of contamination during handling of the sampie.
To avoid these problems, Chakrabarti et al.(16) continued sputtering the
sampie long after the analysis time in order to obtain a weight loss that
could be accurately measured.
Winchester et al.(67) used an internal reference (or internal standard) of
BaC03 in pressed copper pellets to determine Pt and Rh in a-alumina cata-
lysts. McDonald(70) used an internal reference of Ni in a pressed püwder
disk containing 1% Ni, 30% copper slag, and 69% silver für the analysis of
copper slags. Ni was used because the silver lines were almost totally
absorbed. Erosion rates varied by a factor of five for these sampies. Aseries
of Ni compounds was analyzed for Ni by mixing with copper powder and
pressing the mixture into disks. Copper was used as the internal reference.
The absorbances were kept below 0.15 to ensure linearity of the working
curves. The exceIlent analytical results obtained by these studies show the
usefulness of an internal standard in compensating for differences in sputter-
ing rates.
When the concentration of none of the elements is known, an individual
element cannot be used as an internal reference. In this case it may be
possible to use the entire sampIe as an internal reference; that is, measure
the absorbances of all of the elements in the sampIe and use for an internal
reference the sum of the concentrations of all of the elements, which of
course has a known value of 100%. This method was used by Jäger(87) for
the Grimm discharge emission analysis of gold sampIes, but is applicable to
absorption and fluorescence methods if the analytical working curves are
linear and pass through the origin ofthe graph. For one standard, h, usually
selected because it has relatively high concentrations of all of the elements,
values of
kelt,h = Celt,hl Aelt,h (3-7)
are calculated for each element, elt, where Celt,h is the concentration of
element elt in standard h, and Aelt,h is the absorbance of element elt in
standard h. Then for each standard, the sum Sstd over all elements of the
absorbances A for each element times these kelt,h values is calculated as a
measure of total concentration
Sstd = L kelt,h Aelt,std (3-8)

eil
Then the corrected absorbance A' for each element in each standard is
calculated as
A' - kelt,hAelt,std
elt,std - S (3-9)
sld
The analytical working curve for each element is a plot of the corrected
absorbance for that element in each of the standards versus concentration.
To determine the concentration of an element in a sampIe sam, values of
Ssam and Adt,sam are calculated for that sampIe and used with the working
curve for that element to determine the concentration. This method has been
used by Bernhard and co-workers(86) to analyze tool steels, which vary widely
in composition and have different sputtering erosion rates.
When the absorbances are high enough that the working curves are not
linear, then using this method that assurnes linear working curves will still
102 CHAPTER 3
improve the analytical results, but even better results will be achieved if
account is taken of the nonlinear shape of the working curve for each ele-
ment. This can be done by using an appropriate mathematical model (e.g.,
a polynomial equation) for each of the nonlinear working curves, and using
the absorbances obtained with the standards to determine the unknown
parameters in these mathematical models. This produces a set of simulta-
neous nonlinear equations that must be solved. For example, we have used
a second-order polynomial equation for the shape of each of the working
curves to obtain a set of equations (one equation for each element in each
standard) of the following form:
(3-10)
where bl,elt and b 2,elt are the first and second-order coefficients of the polyno-
mial for the working curve for element elt, and RStd represents the sampie
erosion rate for standard std as well as the rate at which the sputtered
material for that standard is transported to the observation zone. The values
of bl,ell> b 2,ell> and R Std are the unknown parameters for which this set of
equations must be solved in order to construct the analytical working curves.
To analyze a sampie sam, the absorbances Aelt, sam for all of the elements
in the sampie are determined. These values and the known values of bl,elt
and b2 , elt [determined by solving Eq. (3-10) for the standards] are used to
solve the set of simultaneous equations (one equation for each element)
represented by Eq. (3-10) where the subscript std is replaced by sam to find
the values of Rsam and the concentrations of all of the elements in the sampIe.
We have solved this set of equations by using a Simplex search routine
to minimize the sum of the squares of the deviations of the experimental
absorbances from the working curves. By using a quadratic model for each
ofthe working curves, better results were achieved for the same set of experi-
mental data for tool steels that was analyzed by assuming linear working
curves. (86) The sum of the squares of the deviations from the working curves
of the measured absorbances for all ten elements determined in all six stan-
dards for the linear method was 0.00257. For the global fit to the quadratic
working curves, this was reduced to 0.00019. An example of the
improvement in working curves using the nonlinear global fit to all of the
data is shown in Fig. 3-11. Although this method gives better results, the
Simplex search finds different minima when the initial starting conditions
for the search are different. The minima all give results that are better than
the linear fit, but we are searching for a better way to solve the equations
along with a better mathematical model for the working curves.
A modification of this procedure involves determining the shape of the
analytical working curves by some independent means such as another set
of standards all of which have the same sampie erosion rate. Values of bl,elt
0.3
0.2
0.1
Ql
o 0
c 0 2 4
o
.0
.,
o'-
.0 0.3
<:
Figure 3-7 7. Curves for Mn in tool steel resulting from 0.2
global fits to all of the data in a total analysis of the
standards used to compensate for variations in sputter- 0.1
ing rates. The upper curve shows the hetter fit that results
when it is assurned that the working curves for all of 0
0 4
the elements are quadratic polynomials. The lower curve
results from assuming linear working curves. Mo Concentration, %
and b2,elt would then be known and only values of R Std would have to be
determined from the set of equations represented by Eq. (3-lO). This method
assumes that the shape ofthe working curve does not change. For example,
it assumes that the shape is controlled by the spectral profiles of the absorp-
tion line in the sputtering eell and the emission line of the primary radiation
source, and that these spectral profiles do not change. However, since the
emission profile of the primary source changes with lamp current and with
age, the parameters describing the shape of the working curve may have to
be updated from time to time.
3.3.7. Determining Element Ratios with Only One Standard

McDonald(70) has shown that only one standard containing the elements
of interest can be used to determine the ratio of two elements in a variety
of sampIes that have widely different sputtering erosion rates, provided the
working curves are linear. Since ratios are being determined within the same
sampIe, the sputtering erosion rates of the standard or sampIe need not be
known. The standard is used to determine the proportionality factor Fij
between the ratio of the absorbances A and the ratio of the concentrations
C of two elements i and j :
(3-11)
Onee determined, the proportionality factors Fij are valid for a wide variety
of sampIes. This method is based on the idea that the absorbance of an
104 CHAPTER 3
element is related to its concentration by proportionality factors that are

constants or that can be readily held constant in a sputtering chamber. (70)
These factors include the oscillator strength of the line, the spectral profiles
of the absorption line in the sputtering cell and the emission line of the
primary radiation source, the fraction of atoms in the ground state, wave-
length, the atomic weight, the diffusion coefficient, and in a flowing cell, a
carrier gas transport factor that helps determine the fraction of atoms that
reach the observation zone. The factor Fij is the ratio ofthese proportionality
factors for the two elements.
Also assumed is the absence of preferential effects, such as preferential
sputtering, preferential ionization, or preferential agglomeration of the ele-
ments. As we saw earlier, although preferential sputtering occurs initially,
its effects are eliminated (if the surface temperature is not too high) after
an equilibrium surface layer has been established during preconditioning
sputtering. However, ionization differences may be significant and agglomer-
ation may be different for different elements.
If the concentration of only one element in a sampie is known, then the
concentrations of the other elements for which Fij is known can be deter-
mined with this method. This comes close to the proposal by Walsh in
his seminal paper in 1955(3) that atomic absorption could be used for the
determination of concentrations without standards.
3.4. Applications
This section includes a brief summary of the types of analyses that have
been done with sputter cells using AAS and AFS. The list continues to grow
as the advantages of these methods continue to be recognized.
3.4. 1. AI/oys
In 1960 Gatehouse and Walsh(9) used a hollow cathode sputtering cell

to determine Ag in copper. Gough(l3) lists atomic absorption detection limits
ranging from ll!g/g to 800 I!g/g for 16 elements in alloys of Fe, Al, Cu,
and Zn. McDonald(70) used an internal standard to determine Cr in aseries
of low-alloy steels, high-speed steels, and product steel. He found that one
sampie was not homogeneous and that the sputtering cell correctly deter-
mined the concentration in the center of the sampie, which was confirmed
by another analytical method to be different from the certified bulk value.
Ni was determined in steel sampies, brass, bronze, and an aluminum alloy
by using Cu as the internal standard, since its concentration was known in
all of the sampies. Only one reference sampie of steel was used to determine
the calibration factor for the Ni-to-Cu ratio. The calibration factor then
remains the same for all of these different sampies because surface equilibra-
tion is readily achieved during the preburn.
Ohls(78) was the first to publish the results of an atomic absorption
analysis of alloys using the Atomsource. Elements inc1uded Cr, Cu, Mn, Nb,
and Zr in steels and La in zinc. Bernhard and Kahn(15) determined Al in seven
varied nickel-based alloys using only one standard with the Atomsource.
Detection limits for 15 elements in common alloys range from 1 to 5 pgjg
for the Atomsource. (88)
Gough et al.(5) used nondispersive atomic fluorescence to determine
Si, Mn, Ni, Cr, and Cu in low-alloy steels. Gough and Meldrum(66) used
nondispersive atomic fluorescence to determine Cr, Mn, Ni, and Mo in low-
alloy steels. Detection limits for 12 elements ranged from 1 to 100 pgjg for
various metals in low-alloy steel, aluminum, brass, and copper sampIes.
Winefordner and co-workers(18) used laser-excited nonresonance atomic
fluorescence to determine Pb in copper with a 0.1 pgjg detection limit.
3.4.2. Powders
A method of preparing pressed copper pellets for the analysis of powder

sampIes has been studied in detail by Winchester and Marcus. (89) Winchester
et al. (67) used the rapid sequential multielement capability of the Analyte 16
with an Atomsource to determine Pt and Rh in a-alumina catalysts by AAS
with an internal reference. The catalyst was mixed with copper powder and
BaC03 as an internal reference, and pressed into an electrically conducting
pellet. The internal reference reduced the deviations from the analytical
working curve by an order of magnitude.
McDonald(70) used Cu as an internal reference in pellets of pressed
copper powder to determine Ni in powdered Ni compounds. Cu was deter-
mined in powdered copper slags pressed into silver powder pellets with 1%
Ni added as an internal reference. Slags are weIl known for the lack of good
standards because of their wide variation in composition, but the internal
reference method alleviates this problem.
3.4.3. Dried Solutions
Before drying a solution on the surface of a cathode, the cathode is

often preconditioned by sputtering. However, because of the small amount
of dried material that is usually used, there is not enough sputtering time
prior to the analysis to precondition the dried material. As we sawearlier,
preconditioning by sputtering brings the surface to an equilibrium condition
that compensates for preferential sputtering. Therefore, the lack of precondi-
tioning may cause some elements to atomize more quickly than others,
706 CHAPTER 3
producing peaks of different durations. Even so, detection limits for some
elements are better than those for graphite fumace AAS.
Goleb and Yokoyama(37) found that uniform deposits of lithium hy-
droxide solutions could be prepared in a hollow cylindrical copper electrode
by placing 0.1 ml of solution in the copper electrode and adding about 0.1 ml
of acetone to it while the electrode was rotating at 60 rpm under a heat
lamp. Goleb and Brody(ll) determined Na by atomic absorption after drying
solutions of NaF on the surface of a hollow cathode. Li and Mg produced
severe matrix effects. McCamey and Niemczyk(90) studied eight elements in
a hollow cathode absorption cello Matrix effects were generally not observed
for I: 1 ratios of potential interferants to analyte, but became noticeable for
ratios of 100: 1 in most cases. Additional information about hollow cathode
cell applications appears in Chapters 6 and 7.
Gandrud and Skogerboe(12) used flat spectroscopic-grade graphite and
aluminum platrodes as cathodes on which sampie solutions were evaporated
and inserted into the sputtering cello Cell current was 50 mA and cell pressure
was 3 Torr. Peak widths at half-height were 12 s but the peaks had tails
lasting several minutes. Detection limits ranged from I to 10 ng for eight
elements.
Chakrabarti et al. (16) analyzed dried solutions deposited on flat cathodes
pressed against the sampie port of the Atomsource. Power regulation was
found to give more reproducible results than current or voltage regulation.
They used a high-electrical-power pulsed mode of operation for high atomiz-
ation rates and found that an instrumental time constant of 2 ms on one of
the faster commercial atomic absorption spectrophotometers is too slow to
respond to the 5-ms peak of one of the elements. They obtained results that
are comparable to those of the highly sensitive GFAAS method for some
elements. Absolute detection limits for 14 elements range from 0.2 to 600 pg.
OhlS(78) showed that elements like Zr could be determined although they
are not suitable for determinations by graphite furnace atomization because
they form refractory carbides. The detection limit for Zr using 5 J.Ll of solu-
tion was 100 ng using the Atomsource. Although peak height was linearly
related to concentration up to 500 ng, a 5000-ng sampie produced a much
broader peak than the lower quantities of Zr.
Chakrabarti et al.(16) found that matrix effects for Fe caused by the
presence of a Ni(N0 3h matrix with a matrix-to-analyte ratio up to 100 to
1 could be eliminated by using peak area rather than peak height. In other
cases this did not work. The severe matrix effects caused by alkali and alkali-
earth chlorides on the determination of metals using dried solutions could
not be compensated by using peak area. These differences were attributed to
the differences in electrical conductivity of the dried deposits; nonconducting
deposits produce large matrix effects because they allow a positive charge to
build up that repels sputtering ions.
The problem of charge buildup on insulating sampIes has been studied

by Drobyshec and Turkin.(61) They showed that pulses could be used to
sputter insulators by studying the emission of Ga, Gd, Fe, and Y when an
insulating monocrystalline film of Y 3FesÜl2 on a Gd3Gasü12 substrate was
subjected to successive sputtering pulses. They also developed a model to
predict how long sputtering will continue before significant positive charge
builds up. In an example for a 0.5-mm partic1e, their model predicts that
sputtering will continue for 0.3 s at 10 mA, or for 1.7 ms at 2 A before
sputtering is stopped by charge buildup. The latter time is comparable to
the shortest pulses observed for dried solutions by Chakrabarti et a1Y6) and
indicates that atomic absorption pulses could be used for the analysis of
insulating materials, inc1uding dried solution residues that are insulating.
Winefordner and co-workers(18) used nonresonance atomic fluorescence
to determine Pb in aqueous solutions. They used a frequency-doubled
nitrogen pumped dye laser with a spectral bandwidth ofO.015 nm and a 5-ns
pulse having a typical energy of 5 pJ. The linear dynamic range is six orders
of magnitude. The detection limit is 20 pg for 5 pI of aqueous solution
deposited on a graphite cathode.
3.4.4. Isotopes
In 1955 Walsh(3) suggested that AAS might be suitable for the isotopic
analysis of certain elements. As mentioned earlier, the isotope shift for a
handful oflight and very heavy elements is large enough for isotopic analysis
using sputtering cells. Larkins(28) determined the major isotopes of Pb using
flames as weH as the flowing sputtering ceH of Gough. (13) Hannaford and
Walsh(91) determined the isotopes of boron. Goleb(IO·92) used water-cooled
hollow cathode lamps and a water-cooled hollow cathode sputtering cell for
the isotopic analysis of uranium. He confirmed that heavier sputtering ions
produced a larger absorption signal for a given sputtering current. Krypton
was used in the hollow cathode lamps and xenon in the absorption cello
Goleb and Yokoyama(37) used a similar water-cooled system for the isotopic
analysis of Li. They found large variation in the isotopic composition of
some reagent-grade lithium salts.
Doppler-free saturated absorption spectrometry(93) extends isotopic
analysis to other elements and has been used for the isotopic analysis of
Ta,<94) V,<9S) Y,(94,96,97) and Zr.(91,94). With this technique, the beam from a
highly monochromatic tunable laser is split into two beams and the resulting
strong and weak beams are passed through the sputtering cell in opposite
directions. The strong laser beam is modulated and saturates the line while
the weak beam detects the modulated decrease in absorption produced by
the saturation. Since both laser beams have the same wavelength, only those
108 CHAPTER 3
atoms that have a zero velocity component along the direction of propaga-
tion of the beams produce a signal. Therefore, a Doppler-free spectrum
results when the wavelength of the laser is scanned. Further experimental
details are discussed by Hannaford and Walsh.(9l)
Doppler-free saturated fluorescence methods are briefly reviewed by
Hannaford(98) who used Doppler-free intermodulated fluorescence to study
the isotopes of molybdenum. Two counterpropagating laser beams are used
to excite the atoms, and their fluorescence is detected. Gough and
Hannaford(94) studied the hyperfine structure of the 578.1-nm line of 181Ta,
and Gough et al.(95) studied the hyperfine structure of 51V.
3.5. The Future
One of the most rapidly advancing areas of study is the atomic absorp-
tion analysis of dried solution residues using gas-jet-enhanced sputtering
with high-power sputtering pulses. Chakrabarti et al.(16) have achieved detec-
tion limits that riyal those of graphite fumace AAS (GFAAS) for some
elements, particularly those that are difficult to study with GF AAS, and
further improvements are continuing to be made. Although matrix effects
appear to be less severe than those of GFAAS, they may be worse than for
the sputtering cell analysis of alloys where a prebum allows the surface to
come to equilibrium before analytical measurements are made. In any case
a better understanding of matrix effects and sputtering processes in the
analysis of dried solution residues is needed.
Although gas jets help to reduce redeposition of sputtered atoms and
help to transport atoms into the observation region, the continued flow of
gas also moves atoms out of the observation region. A higher population of
atoms might be obtained by synchronizing gas-jet pulses with high-power
sputtering pulses soas to transport a large number of atoms into the observa-
tion zone and then to stop the gas flow when the atom population has
reached its maximum. Synchronized gas-jet pulses might have another
advantage when dried solutions are used. A low-current prebum with the
jets off could bring the surface to equilibrium; sampie loss would be minimal
because of redeposition of the sputtered atoms back onto the surface. Surface
equilibration would help to reduce those matrix effects that are caused by
the differences in sputtering rates of different elements and forms of elements.
The direct sputtering of electrically nonconducting (insulating) mat-
erials, discussed in more detail in Chapter 9, would have many applications.
Although insulating powders are now analyzed by incorporating them into
pressed pellets of conducting powders, it would be much easier to simply
dust a powder onto a cathode surface and atomize the material with short
pulses. As more is leamed about surfaces, it should become easier to choose

the type of surface on which to retain a particular type of powder.
Duckworth and Marcus(99) have shown that an rf discharge sputtering
cell can produce mass spectrometer ion signals for a glass sampIe that are
comparable to those produced by metals in dc sputtering cells. Although
their results do not directly extrapolate to the generation of a population of
ground-state atoms, the idea looks very promising for the direct sampling
of insulating materials for AAS and AFS.
Sputtering of insulating materials using high-current pulses has been
discussed and demonstrated by Drobyshev and Turkin.(61) Direct-current
pulses with a controlled duty cycle would produce efficient sputtering during
each current pulse, while each pulse would be followed by sufficient time to
dissipate the electrical charge that has built up on the insulating sampIe.
Perhaps a reversed current pulse or negatively charged gas-jet pulse could
be used to help dissipate the accumulated charge. Some important advances
in the rapid sputtering of insulating materials for AAS and AFS may be
near.
Laser-excited fluorescence has the potential of reaching detection limits
that are many decades lower than those that have been demonstrated so far
by either AAS or AFS. To date only stagnant-gas cells have been used as
atomizers. A combination of improvements of the type suggested above and
those being used for atomic absorption measurements (e.g., gas jets and
higher-power sputtering pulses) should help to improve detection limits and
may lead to wider applications. However, improvements in laser reliability
are needed to move the tunable laser methods out of the special applications
laboratory and into the routine workplace.
Because sputtering cells use a carrier gas that is relatively transparent
at vacuum ultraviolet wavelengths, they are good atomizers for the determin-
ation of elements such as C, P, and the halides whose strongest spectrallines
are below 200 nm. Although emission has been used for such determinations,
atomic absorption or fluorescence would be encouraged by the availability
of hollow cathode lamps that are designed with windows and excitation
conditions that emphasize the vacuum ultraviolet lines.
A properly designed gas-jet flow clearly helps to improve atomic absorp-
tion signals. Although transport processes are one reason for the
improvement, little is known about how jets that strike the surface of the
sampie influence the spatial distribution of the flow of sputtering current.
More information about the influences of the jets may lead to further
improvements in sampIe erosion.
As we have seen, the spatial resolution that is inherent in the size of the
sampIe ports of sputtering cells of Gough and the Atomsource has already
been used to demonstrate differences in the bulk concentration and the
concentration in the center of a large sampIe. Analyses with this spatial
110 CHAPTER 3
resolution may eventually lead to improved or more homogeneous products.

Even higher spatial resolution is possible by reducing the diameter of the
jets andj or bringing the jet nozzle closer to the surface. In our laboratory a
jet has been used to drill a O.05-cm hole through a copper strip. As more is
understood about gas-jet-enhanced sputtering, a microscopic atomizer may
be possible with spatial resolution and control comparable to those of a
laser microprobe.
The low detection limits that can be achieved with sputtering cell atomic
absorption and fluorescence methods will make them complementary to
graphite furnace methods for the analysis of solutions and small sampies.
Because collisional broadening is so low, spectral resolution is so high, and
sampie containment is relatively easy, sputtering cell atomizers will continue
to find use for isotopic analyses, especially using Doppler-free laser methods.
Because of their high precision, multielement capability, and ease of use,
atomic absorption and fluorescence with sputtering cell atomizers will
become important analytical methods for a wide variety of sampies, includ-
ing superalloys and other new "high-technology" non-, semi-, and super-
conducting materials.
References
1. C. T. J. Alkemade and J. M. W. Milatz, J. Opt. Soc. Am. 45 (1955) 583.

2. C. T. J. Alkemade and J. M. W. Milatz, Appl. Sei. Res. B4 (1955) 289.
3. A. Walsh, Spectrochim. Acta 7 (1955) 108.
4. B. J. Russell and A. Walsh, Spectrochirn. Acta 15 (1959) 883.
5. D. S. Gough, P. Hannaford, and A. Walsh, Spectrochirn. Acta 28B (1973) 197.
6. K. Ohls, J. Fiock, and H. Loepp, Fresenius Z. Anal. Chem. 332 (1988) 456.
7. B. W. Smith, J. B. Womack, N. Omenetto, and J. D. Winefordner, Appl. Spectrosc. 43
(1989) 873.
8. J. V. Sullivan and A. Walsh, Spectrochim. Acta 21 (1965) 727.
9. B. M. Gatehouse and A. WaIsh, Spectrochirn. Acta 16 (1960) 602.
10. J. A. Goleb, Anal. Chern. 35 (1963) 1978.
11. A. Goleb and J. K. Brody, Anal. Chirn. Acta 28 (1963) 457.
12. B. W. Gandrud and R. K. Skogerboe, Appl. Spectrosc. 25 (1971) 243.
13. D. S. Gough, Anal. Chern. 48 (1976) 1926.
14. A. E. Bernhard, Spectroscopy 2 (1987) 24.
15. A. E. Bemhard and H. L. Kahn, Arn. Lab. 20 (1988) 126.
16. C. L. Chakrabarti, K. L. Headrick, J. C. Hutton, Z. Bicheng, P. C. Bertels, and M. H.
Back, Anal. Chern. 62 (1990) 574.
17. M. Glick, B. W. Smith, and J. D. Winefordner, Anal. Chern. 62 (1990). 157.
18. B. M. Smith, N. Omenetto, and J. D. Winefordner, Spectrochim. Acta 39B (1984) 1389.
19. E. H. Piepmeier, in: Analytical Laser Spectroscopy (N. Omenetto, ed.), Wiley, New York,
1979.
20. I. I. Sobel'man, Introduction to the Theory 0/ Atornic Spectra, Pergamon Press, Elmsford,
N.Y.,1972.
21. H. G. Kuhn, Atornic Spectra, 2nd ed., Academic Press, New York, 1969.
22. C. Candler, Atomic Spectra and the Vector Model, Hilger & Watts, London, 1964.
23. R. G. Breene, Jr., The Shift and Shape of Spectral Lines, Pergamon Press, Elmsford, N.Y.,
1961.
24. A. C. G. Mitchell and M. W. Zemansky, Resonance Radiation and Excited Atoms, Cam-
bridge University Press, London, 1961.
25. C. T. J. Alkemade and R. Herrrnann, Fundamentals of Analytical Flame Spectroscopy,
Wiley, New York, 1979.
26. L. de Galan and H. C. Wagenaar, Methodes Phys. Anal. 3 (1971) 10.
27. C. F. Bruce and P. Hannaford, Spectrochim. Acta 26B (1971) 207.
29. H. Kopferrnann, Nuclear Moments (English translation by E. E. Schneider), Acadernic
Press, New York, 1958.
30. G. J. De Jong and E. H. Piepmeier, Spectrochim. Acta 29B (1974) 179.
31. E. H. Piepmeier and L. de Galan, Spectrochim. Acta 30B (1975) 263.
32. H. C. Wagenaar and L. de Galan, Spectrochim. Acta 28B (1973) 157.
34. J.W. Hosch and E. H. Piepmeier, Appl. Spectrosc. 32 (1978) 444.
35. E. H. Piepmeier, Spectrochim. Acta 44B (1989) 609.
36. V. Svoboda, R. F. Browner, and J. D. Winefordner, Appl. Spectrosc. 26 (1972) 505.
37. J. A. Goleb and Y. Yokoyama, Anal. Chim. Acta 30 (1964) 213.
38. J. D. Ingle, Jr., Anal. Chem. 46 (1974) 2161.
39. S. B. Smith, Jr., and G. M. Hieftje, Appl. Spectrosc. 37 (1983) 419.
40. J. V. Sullivan, Anal. Chim. Acta 105 (1979) 213.
41. P. B. Famsworth and J. P. Walters, in: Improved Hollow Cathode Lamps for Atomic
Spectroscopy (S. Caroli, ed.), Wiley, New York, 1985.
42. S. Caroli, A. Alimonti, and F. Petrucci, in: Improved Hollow Cathode Lamps for Atomic
Spectroscopy (S. Caroli, ed.), Wiley, New York, 1985.
43. H. C. Wagenaar, I. Novotny, and L. de Galan, Spectrochim. Acta 29B (1974) 301.
44. D. C. McDonald, Anal. Chem. 54 (1982) 1052.
45. C. Van Dijk, B. W. Smith, and J. D. Winefordner, Spectrochim. Acta 37B (1982) 759.
47. A. J. Stirling and W. D. Westwood, J. Phys. D. 4 (1971) 246.
48. A. J. Stirling and W. D. Westwood, J. Appl. Phys. 41 (1970) 742.
49. D. S. Gough, P. Hannaford, and R. Martin Lowe, Anal. Chem. 61 (1989) 1652.
50. P. W. J. M. Boumans, Theory of Spectrochemical Excitation, p.297, Plenum Press, New
York, 1966.
51. H. J. Kim and E. H. Piepmeier, Anal. Chem. 60 (1988) 2040.
52. C. L. Chakrabarti, K. L. Headrick, J. C. Hutton, B. Marchand, and M. H. Back, Spectro-
chim. Acta 44B (1989) 385.
53. C. L. Chakrabarti, K. L. Headrick, J. C. Hutton, and M. H. Back, XXVI Colloquium
Spectroscopicum Internationale, Invited Paper No. 4 (1989) 10.
54. C. L. Chakrabarti, K. L. Headrick, P. C. Bertels, and M. H. Back, J. Anal. At. Spectrom.
3 (1988) 713.
55. A. E. Bemhard, personal cornmunication.
56. Y. Demers, J.-M. Gagne, and P. Pianarosa, J. Anal. At. Spectrom. 2 (1987) 59.
57. C. H. Corliss and W. R. Bozrnan, Experimental Transition Probabilitiesfor Spectral Lines
of Seventy Elements, National Bureau of Standards Monograph 53, U.S. Govemment
Printing Office, Washington, D.C., 1962.
58. P. L. Larkins, Anal. Chim. Acta 132 (1981) 119.
60. G. S. Anderson, J. Appl. Phys. 40 (1969) 2884.
112 CHAPTER 3
61. A. I. Drobyshev and Y. I. Turkin, Spectrochim. Acta 36B (1981) 1153.

62. E. Gillam, J. Phys. Chern. Solids 11 (1959) 55.
63. W. L. Patterson and G. A. Shirn, J. Vac. Sei. Technol.4 (1967) 343.
64. D. Fang and R. K. Marcus, J. Anal. At. Spectrorn. 3 (1988) 873.
65. Y. Yokoyama and S. Ikeda, Spectrochirn. Acta 24B (1969) 117.
66. D. S. Gough and R. J. Meldrum, Anal. Chern. 52 (1980) 642.
67. M. R. Winchester, S. M. Hayes, and R. K. Marcus, Spectrochirn. Acta 46B (1991) 615.
68. B. M. Patel and J. D. Winefordner, Spectrochim. Acta 41B (1986) 469.
69. D. C. McDonald, Anal. Chern. 54 (1982) 1057.
71. B. M. Patel and J. D. Winefordner, Appl. Spectrosc. 40 (1986) 667.
72. J. D. Ingle, Jr., and S. R. Crouch, Spectrochernical Analysis, Prentice-Hall, Englewood
Cliffs, N.J., 1988.
73. B. Welz, Atornic Absorption Spectrornetry, 2nd ed., VCH Verlagsgesellschaft, Weinheim,
1985.
74. L. de Galan and M. T. C. de Loos-Vollebregt, Spectrochirn. Acta 39B (1984) 1011.
75. M. T. C. de Loos-Vollebregt and L. de Galan, Spectrochirn. Acta 41B (1986) 597.
76. G. Tie-Zheng and R. Stephens, J. Anal. At. Spectrorn. 1 (1986) 355.
77. E. H. Piepmeier and L. de Galan, Spectrochim. Acta 31B (1976) 163.
78. K. Ohls, Fresenius Z. Anal. Chern. 327 (1987) 111.
79. E. H. Piepmeier, Anal. Chern. 48 (1976) 1296.
80. E. D. Salin and J. D. Ingle, Jr., Appl. Spectrosc. 32 (1978) 579.
81. E. D. Salin and J. D. Ingle, Jr., Anal. Chern. 50 (1978) 1737.
82. R. Mavrodineanu and R. C. Hughes, Appl. Opt. 7 (1968) 1281.
83. W. Niu, R. Haring, and R. Newrnan, Arn. Lab. 19 (1987) 40.
84. J. V. Sullivan and A. Walsh, Appl. Opt. 7 (1968) 1271.
85. A. E. Bemhard, E. H. Piepmeier, H. J. Kim, and H. L. Kahn, 39th Pittsburgh Conference
and Exposition on Analytical Chemistry and Applied Spectroscopy, 1988, Paper No. 809.
86. E. H. Piepmeier, H. J. Kim, C. E. Crandall, and A. E. Bemhard, 16th Annual Meeting of
the Federation of Analytical Chemistry and Spectroscopy Societies, 1989, Paper No. 920.
87. H. Jäger, Anal. Chim. Acta 60 (1972) 303.
88. Analyte Corp., Medford, Oregon, brochure, 1990.
89. M. R. Winchester and R. K. Marcus, Appl. Spectrosc. 42 (1988) 941.
90. D. A. McCamey and T. M. Niemczyk, Appl. Spectrosc. 34 (1980) 692.
91. P. Hannaford and A. Walsh, Spectrochirn. Acta 43B (1988) 1053.
92. J. A. Goleb, Anal. Chirn. Acta 34 (1966) 135.
93. T. W. Hänsch, I. S. Shahin, and A. L. Schawlow, Phys. Rev. Lett. 27 (1971) 707.
94. D. S. Gough and P. Hannaford, Anal. Chern. 55 (1985) 91.
95. D. S. Gough, P. Hannaford, R. M. Lowe, and A. P. Willis, J. Phys. B 18 (1985) 3895.
96. R. J. McLean, P. Hannaford, H.-A. Bachor, P. T. H. Fisk, and R. J. Sandeman, Z. Phys.
D I (1986) 253.
97. D. S. Gough, P. Hannaford, and R. J. McLean, J. Phys. B 21 (1988) 547.
98. P. Hannaford, Conternp. Phys. 24 (1983) 251.
99. D. C. Duckworth and R. K. Marcus, Anal. Chern. 61 (1989) 1879.
4
Atomic Emission Spectrometry
J. A. C. Broekaert
4.1. Optical Emission Spectrometry
4.1.1. Principles
4. 1. 1. 1. Optical Spectra
Optical emission spectrometry is one of the oldest physical methods of ana-

lysis, enabling multielement determinations at the level of major elements,
minor elements, as weIl as trace elements. Its historical development is closely
related to the progress in basic understanding of the nature of atomic spectra
and the structure of matter. The optical emission spectra of atoms and ions
(for a discussion see Refs. 1, 2) stern from transitions between the outer
electron shells of the chemical elements which give rise to line spectra where
the wavelength of the lines relates to the energy difference of the levels
concemed according to:
IlE= heiA (4-1)
The atomic levels as known from Bohr's atomic model relate to the
atomic quantum numbers. The main quantum number is described by
E = _ 21CJ1.Z2e4
(4-2)
n2h2
J. A. C. Broekaert. Department of Chemistry, University of Dortmund, D-4600

Dortmund 50, Germany.
113
114 CHAPTER4
with J.l being the reduced mass of the electron and the nucleus, e the charge
of the electron, h Planck's constant, and n = 1,2,3, .... The 1 quantum
number is defined by the quantization of the impulse moment of the electron
as:
ILI = ~ .Jl(l + 1) (4-3)

21C
and has values of / = 0, 1,2, ... ,n - 1.
The magnetic quantum number is related to the quantization of the
impulse momentum in the orientation of an external field and is defined by:
h
L =-m)
z 21C (4-4)
with m) = ±1, ±(l- 1), ... ,0.

The spin quantum number is introduced for describing the splitting of
the atomic spectral lines in a magnetic field and is defined by:
h
ISI = - .JS(S + 1) (4-5)
21C
with S the spin momentum and for the component inthe field direction Sz:
withms = ±~ (4-6)
The impulse momentum Land the spin momentum S combine to a total

angular momentum J of which
IJI = ~ .Jj(j + I) (4-7)

21C
withj = 1± s the internal quantum number.
The atomic energy levels can be described by the quantum numbers and
indicated by a term symbol:
(4-8)
When two valence electrons are present, coupling of the angular momentum
of both electrons gives rise to a total angular momentum and the coupling
of the spin momentum of both electrons to a total spin momenum (L - S
or Russell Saunders coupling) and total quantum numbers are defined as:
(4-9)
A TOMIC EMISSION SPECTROMETRY 115
and
(being 0 or I) (4-10)
and
IJI=~.JJ(J+ 1) withJ=L-S, ... ,L+S (4-11)

21r
Also here the terms can be indicated as
(4-12)
and radiative transitions are only allowed when !in is an integer number,
!iL = 0, ± 1, and !iJ = 0, ± 1, but not for J = O. The complexity of the
atomic spectra is reflected by the number of terms relevant for the element
concerned.
The spectral lines of relevance in atomic spectrometry occur between
160 and 800 nm. As the line spectra are characteristic ofthe emitting element,
the presence of lines of known wavelength in a spectrum is a criterion for
the presence of the element above a detectable concentration level. This
forms the base for emission spectrometry as a tool for qualitative analysis.
4.1.1.2. Line Intensities
The line intensities emitted by the so-called radiation source, the glow
discharge in this work, directly relate to the number density of the atoms or
ions present, as given by the expression for the radiance in terms of the
energy emitted per unit of space angle:
(4-13)
where I nm is the radiation density (W jcm2 • sr), nm is the number density of

the excited-state atoms or ions, h is Planck's constant (6.623 x 10-27 erg· s),
Anm is the Einstein transition probability for spontaneous emission (S-I),
and Vnm is the frequency of the emitted line (S-I).
In the radiation source, free atoms which are generated by thermal
dissociation of the sampie material or entrained from aseparated atomizer,
are excited or ionized and excited additionally. Several collisional or other
processes may be responsible for delivering the required energy to the analyte
species. When the excitation energy of the excited state is Ern, the population
116 CHAPTER4
of this level is given by the Bo1tzmann equation provided the source is in

so-called thermal equilibrium and is given by:
nm _ gm -Ern/kT
- --e (4-14)
no go
where nm and no are the populations of the excited level and the ground
state, respective1y, gm and go are their statistica1 weights, k is Bo1tzmann's
constant (1.38 x 10- 16 erg/K), and T is the temperature in degrees Kelvin.
When, instead of no, a summation for all levels is made, the Bo1tmann
equation reduces to:
n = n ~ e- Em / kT (4-15)
m Z(T)
where Z(T) is the partition function and n is the total number density for
all levels of the ana1yte atoms. According1y, the expression for the line
intensities becomes:
n - - e-Em/kTA nmh Vnm

I nm -_gm (4-16)
Z(T)
However, radiation sources often are very hot or they are e1ectrica1
discharges in which the analyte substance is partially ionized. The degree of
ionization aj for a certain element j according to the reaction:
(4-17)
where naj, nij , and n e are the atom, the ion, and the e1ectron number densities,
respective1y, can be expressed as:
a. = n·1J
(4-18)
J
naj + nij
and = ajnj and naj = (1 - aj )nj.
nij
According1y, the intensities for atom and ion 1ines are given by:
(4-19)
and
(4-20)
ATOMIC EMISSION SPECTROMETRY 117
According to the Saha equation, the electron partial pressure Pe in a

plasma which is in so-called thermodynamical equilibrium is given by:
S .(T) =PijPe = (2nm)3/2(kT)5/2 2Zij e-Eij/kT

ru h3 (4-21)
P~ Z~
where Pij and Paj are the partial pressures of the ions and the atoms of the
element in the plasma, respectively. With the aid of Eq. (4-21) and the
temperature, the degree of ionization of an element can be calculated from
the intensity ratios of ion and atom lines as expressed by Eqs. (4-19) and
(4.20). This, however, only applies provided the plasma is in local thermal
equilibrium.
In electrical discharges under reduced pressure, as schematically shown
in Fig. 4-1, free electrons are accelerated in the vicinity of the cathode.
They ionize by electron impact and may cause secondary emission. Energy
exchange between electrons and ions formed may result from elastic colli-
sions, charge transfer, and recombination according to the reactions:
A+e -t A*+e electron impact (4-22)
A + e -t A +* + 2e (4-23)
A * + B -t A + B* collision of the second kind (4-24)
A * + B -t A +B H Penning ionization (4-25)
A + + B -t A + BH charge transfer (4-26)
A++e-tA* recombination (4-27)
r
Electron impact
+ e = •.•+:e=.·
Charge transfer
Radiative
Recombination
• + O·~ ••
L+
Collision of 2 nd kind
Om~ ...
Cathodic •••••••• •••• Penning effect
sputtering
FigurB 4-1. Excitation processes in discharges under reduced pressure.
118 CHAPTER4
All of these collision-dependent processes are hampered by the low number

densities of all partners as weIl as by their high energies, resulting from the
high electrical field in the relevant parts of the discharge as illustrated by Fig.
4-2. The collision processes occur especially in the vicinity of the cathode, as
weIl as in the negative glow. Absence of local thermal equilibrium will be
found, as the distribution of the kinetic energies due to the high electrical
field can no longer be described by a Maxwell distribution function and as
the mean kinetic energies of the different plasma components will be differ-
ent. Thus, one will have to distinguish between electron temperatures, which
have to be detefmined by probe measurements, kinetic temperatures of
atoms and ions, and excitation temperatures depending on the lines (transi-
tions) considered. The process (4-25) is particularly interesting in the case
when noble gases such as argon and helium are employed as the discharge
gas. They have metastable states (e.g., 11.5 and 11.7 eV for argon) which
can only decay by collisions and may ionize and excite atoms released from
the cathode in one step.
The energy distribution of the charged particles in a glow discharge has
been described by a Druyvenstein function:
dn = 1.039y'l7 e-0.548u'2 du' (4-28)

n
where u' = E/kT. In the case of glow discharges, even the existence of two
groups of electrons has been proposed, namely, a group of slow electrons
involved in recombination processes and a group of highly energetic elec-
trons mainly involved in excitation. Also, the Saha equation has to be
replaced by the much more complex Corona equation. Accordingly, intensity
ratios of atomic and ionic lines, as weIl as the relations with the e1ectron
number densities become complex.
b
Figure 4-2. (a) Geometry and (b)
potential distribution of a dc electrical
discharge under reduced pressure. (1)
Aston dark space; (2) Hittorf dark
space; (3) negative glow; (4) Faraday
dark space; (5) positive column; (6)
5 6 A anode region.
In the excitation of molecular spectra, however, the Boltzmann function

still reasonably well describes the population of the rotational-vibrational
states in the different electronic levels. Both for oxides such as AIO+, TiO+,
YO, ... and radicals stemming from molecular impurities (OH, CNi,
Ni, ... ) the electronic states CI:, 2I:, 2n, ... ) have energy differences of
1-10 eV. They have a vibrational fine structure (v = 0, 1, 2, 3, ... ) and a
rotational hyperfine structure (j = 0, 1,2,3, ... ). The energy difference
between two vibrational states of the same electronic level is ",0.25 eV and
between two rotational levels of the same vibrational states ",0.005 eV.
Accordingly, the radiation emitted from transitions between rotational and
vibrational levels of the same electronic states lies in the infrared region.
However, when two different electronic levels are involved, the emission of
radiation is at visible or UV wavelengths. This is the case for the
2I:_2I:(0,0) CN bands, different OH bands, Ni bands, etc. They conse-
quently will complicate the optical spectra of glow discharges, when these
gases are used as discharge gas or when they are present as impurities in the
noble gases used as discharge gas.
Transitions between rotationallevels are possible when Ilj = 0, ± 1 and
the P, Q, and R branches correspond to + 1, 0, and -1, respectively. For a
transition from an upper level m to a lower level n, the intensity of the
corresponding rotational line is given by:
(4-29)
where Nm is the population of the level m and V nm is the frequency of the

rotationalline. A nm , being the transition probability for dipole radiation, is
given by:
(4-30)
where i and k indicate the degeneracy of the levels n and m, Rnjmk is a matrix
element of the electrical dipole momentum, and gm is the statistical weight
of the vibrationallevel. Nm is given by the Boltzmann equation
N = N ~ e-E,/kT (4-31 )
m Z(T)
where Er is the rotational energy of the excited electronic level and relates
to the quantum number of the upper level as:
Er = hcBv,J'(J' + 1) (4-32)
120 CHAPTER4
where B v ' is the rotational constant of the excited electronic and vibrational
level. For a 2~g_2~U transition, IRnjmkl2 is given by (J + J' + 1) where J" is
the rotational quantum number of the lower level. Accordingly,
In I nm = C _ hcBv.J'(J' + 1)
(4-33)
J' + J" + 1 kT
By plotting In[Inm/(J' + J" + 1)] versus J(J + 1) for aseries of rotational

lines, a rotational temperature can be determined which in the case of low-
pressure discharges is a good approximation for the gas temperature. As
sho)VIl for end-on measurements in a hollow cathode glow discharge, values
of 600-1000 Kare obtained (Fig. 4-3). (3)
4.1.1.3. Spectral Linewidths
The linewidths of atomic emission lines result from several broadening

mechanisms. (4) The natural width of spectral lines results from the finite
cathode
wall
1800
I I
.~I
• ~ I
1600
· " i
o JX II
1400
-~
x-J(~
11
I
I 1 Figure 4-3. Rotational temperature
distribution in a hollow cathode glow
I I discharge. (3) i: 150 mA; graphite hol-
1200 ~r-r-r-r-r-r-r-r-r-r-r---~
I I low cathode with 4.71-mm interna)
diameter; x: 530 Pa, 0: 800 Pa, . :
2 0 1060 Pa argon; 2I:_2I:(0,0) CN
Distance I mm 388.3 nm.
lifetime of excited levels from which radiative transitions to lower levels are
allowed (about 10-8 s) and is given by:
1
AVN=-- (4-34)
27rT
Naturallinewidths are on the order of 10-2 pm.

As the emitting atoms and ions have a velocity component in the
direction of observation, spectrallinewidths also include a Doppler broaden-
ing component. Its full width at half-maximum (FWHM) in terms of the
frequency is given by:
(4-35)
where c is the velocity of light, Vo is the frequency of the line maximum, R

is the gas constant, and M is the atomic mass. The Doppler broadening
strongly depends on the gas temperature. For a hollow cathode glow dis-
charge the Doppler linewidth of the Ca(I) 422.5-nm line for instance
ranges from 0.8 to 2 pm in the case oftemperatures between 300 and 2000 K.
Pressllre, or Lorentz broadening results from the interaction of emitting
and foreign atoms. Its contribution to the frequency FWHM is given by:
27rRT(_I- + _I) (4-36)

MI M2
where MI and M 2 are the atomic masses, N the concentration of the foreign
atoms, and (:TL the collision cross section. The pressure broadening is low in
the case oflow-pressure discharges. For instance, in a discharge operated at
9 mbar, in which the kinetic temperature is 300 K, the pressure broadening
of the Ca(I) 422.6-nm line is only 0.02 pm, and negligible compared with
the Doppler broadening. From the latter, the gas temperature can be deter-
mined in the case of glow discharges. The values are on the same order of
magnitude as the rotational temperatures found from the molecular spectra.
As in every plasma discharge, self-absorption occurs in glow discharges.
However, because ofthe optically thin plasma, it is much less than in atmos-
pheric-pressure plasmas. With some anode-cathode geometries, a cloud of
ground-state atoms may be present in front of the emitting zones. Accord-
ingly, self~reversal may occur especially for matrix resonance lines, as shown
by the profile of a copper line measured for a Grimm-type glow discharge
122 CHAPTER4
Wavalength, Dm
327.40 327.38
:s
•, 3.53
~ 2.5
.pt
• 2
,
CI
~ 1.5
CI
H
1
I.J
0.5
~
30535 30537
WavenUlllber, cm- i
Figure 4-4. Spectralline profile of the Cu(I) 324.7-nm line emitted by a Grimm-type gIow
discharge and recorded by Fourier transform atomic emission spectroscopy.(5) Grimm-type
gIow discharge lamp with floating restrictor (diameter: 8 mm); end-on observation; copper
sampIe; 3.5 Torr argon pressure; 50 mA, 1020 V.
lamp with the aid of high-resolution Fourier transform spectrometry (Fig.

4.4). (5)
The optical spectra in glow discharges, according to the above-described
principle, will contain mainly atomic and less ionic lines. Because of the
absence of local thermal equilibrium, the kinetic gas temperatures will be low
(1000-2000 K) relative to the electron temperatures measured with probes
(above 20,000 K). Molecular bands may occur as a result ofwater vapor or
molecular gases present in the filler gases, which in most cases are noble
gases. They may lead to considerable spectral interferences and their removal
calls for the use ofhigh-purity gases and effective evacuation ofthe discharge
lamp subsequent to sampIe change. The analyte lines have low physical
widths as pressure broadening is minimal, by which spectral interferences of
atom and ion lines are minimal. For matrix atomic resonance lines, self-
reversal may occur, but only at rather high analyte number densities. Accord-
ingly, the linear dynamic range obtained still will be considerable.
4.1.2. Radiation Sources

For atomic emission spectrometry a wide variety of radiation sources
have been developed and used for practical analysis (Fig. 4-5). For the
analysis of liquids, fiames were used as emission sources beginning with the
early work ofBunsen and Kirchhoff, and still playa role in the determination
of the alkalines. This work led to the development of electrical plasmas with
a geometry and stability similar to those of flames but with much higher
_ARC
_SPARK
_ FLAME
_PLASMA SOURCES Q, ~
ICP O~O D~~~. CMP~ ~ MIP 0

0111111 0 ~
;1:
_LOW PRESSURE DISCHARGES
GDL(~. HC _ I
_GRAPHITE FURNACE
cJ b
_LASER PLUME
~
Figure 4-5. Radiation sources for atomic spectrometry.(6)
temperatures. These plasma sources [inductively coupled plasma (ICP) and

microwave-induced plasma (MIP) sources] now are routine tools for atomic
emission spectrometric(6) and mass spectrometric analysis of liquids and
solids subsequent to sam pie disssolution. Dc Arc and spark sources have
developed into powerful sources for the direct analysis of solids.
4.1.2.1. de Ares
These sources are operated in the so-called abnormal part of the cur-
rent-voltage characteristics (see Fig. 4-6). Here, an intensive heating of the
sampie contained in the anode takes place and volatilization results from
thermal evaporation. In the cathode region there may even be an enrichment
of analyte material by which an intensification of the analytical signal may
occur (cathode-Iayer effect). The volatilization may be effective for distilling
volatile elements selectively from a refractory matrix. However, one also has
124 CHAPTER4
..
'0 ..
C Cl
on 0
~
c
o
--
..cO
c..c 'ü; GI
~ ~ CCl
OC
..
o '-
:=: f'O'O ~O
I-~
2'Vb 1-------....
g "
" ,,
1~
,,
,
-----------------~--_r
10- 5 10-3 10-1 10

-Curren t (A I
Figure 4-6. Current-voltage characteristics of dc electrical discharge. (32) (Courtesy of the

American Chemical Society.)
to take into account the matrix effects, which oceur as a result of anions
(such as halogens) producing volatile compounds and causing volatilization
interferences or carrier effects because of the presence of volatile matrix
elements. Accordingly, the aeeuraey of DC are emission speetroseopy is
rather limited. Nevertheless, for survey analysis it still is one of the most
sensitive multielement methods, as shown by the example of traee analysis
in U 30 8 (Table 4-1).(7)
4.7.2.2. Spark Sources
Spark emission spectrometry is a powerful method for multielement

analysis of metals and alloys as required for production control and product
control in the steel industry. (8) Detection limits for modern medium-voltage
unidirectional sparks are at the microgram per gram level. Especially for light
elements, spark atomic emission spectrometry is a necessary complement
to x-ray fluorescence spectrometry for metallurgical analysis. In sparks, a
capacitor is charged by a tank circuit and discharged across an analytical
gap. The sampIes are mounted as flat disks opposite a tungsten electrode
(point-to-plane geometry) with an interelectrode distance of 2-4 mm. Often
the diameter of the spark impact on the sampie is limited to avoid wandering
across the sampIe, whieh would lead to irreproducible results. Mostly uni-
directional sparks are used. Whereas earlier on high-voltage sparks (voltage
2-5 kV) were used, modern sparks almost always are operated at medium
Table 4-1. Deteetion Limits Obtained
by de Are Analysis Atomie Emission Spec-
trometry of U 30 8 (Cathode Region
Method) (7)
Element CL (pg/g)
B 0.1
Ba 1.0
Bi 0.3
Cd 0.1
Co 1.0
Ga 1.0
Ge 0.3
In 0.3
K 1.0
Li 1.0
Mg 0.1
Mo 0.1
Na 1.0
Nb 4.0
Pb 0.5
Si 1.0
Sr 1.0
Zn 10.0
voltage (500 V). The repetition rate (100-1000 Hz) can be varied electron-
icaHy but also the energy dissipation can be changed by changing the resist-
ance, the capacitances, and the inductances in the discharge circuit. Material
ablation in spark discharges partially results from cathodic sputtering due
to the highly energetic particles produced in the spark channel. However,
because of the intense bombardment, the sampie is locally molten (Fig.
4-7), by which thermal evaporation contributes considerably to material
volatilization. The contribution ofboth can be varied by changing the circui-
try parameters. Regardless, the contribution of thermal evaporation leads
to selective volatilization and related matrix effects. Accordingly, matrix
correction ca1culations with extensive software as weH as a careful selection
of the standard sampies for calibration are required. With sparks, analysis
times down to 20 s can be obtained. This is possible by the application of
high-energy presparking, by which the metaHograpical structure of the
sampie is destroyed. When using internal standardization, aprecision of 1%
in the concentration range of the minor elements can be realized. Consider-
able innovation in this method, however, is still taking place. For instance,
by a proper selection of the analytical zone, as is possible with the aid of
quartz optical tibers, the matrix effects can be greatly reduced. Electronic
control of the waveform mayaiso lead to improvements of the precision and
126 CHAPTER4
Figure 4-7. Burning spot obtained in spark emission spectroscopy. 500 Hz monoalternance
spark in argon; sarnple: aluminum.
of the power of detection. The latter also is still attainable by using cross-
excitation with high-frequency or microwave energy. (9)
Fumaces and electrothermal devices are of special interest as thermal
atomizers in combination with other sources for signal generation, such as
glow discharges. This combination is known as furnace atomic nonresonance
emission spectrometry (FANES)(IO) (Fig. 4-8) and is discussed in aseparate
chapter in tbis book. Combinations with plasma sources at atmospheric
pressure also have been studied extensively and shown to be useful for
Figure 4-8. Principle of FANES (furnace

atomic nonresonance emission spectro-
metryYO) (Courtesy of Pergamon Press.)
(I) carrier-gasport; (2) pump port; (3)
electrical connector; (4) anode; (5) remov-
able lid for sampie injection; (6) graphite
electrode; (7) graphite furnace and hollow
cathode; (8) window; (9) water-cooled
vacuum vessel; (10) water-cooled part of
the vacuum vessel and rotation arm for
changing the graphite tube; (li) pivot; (12)
gasket.
analyzing microaliquots of solutions as weIl as powdered organic sampies.
Only in a limited sense can they also be used as direct emission source. (11)
In this case, only elements of which the most sensitive lines have low excita-
tion energies, such as the alkali elements, can be determined down to the
trace element level.
4.1.2.3. Laser Plumes
Laser-generated plumes have been used both as atomizers and as direct

emission sources in atomic emission spectrometry. A review has recently
appeared. (12) These techniques, of which the principles stem from the 1960s,
have gained new interest since stable and reliable Nd-YAG laser sources
with high modal purity have recently become availabe. Both atomic emission
and dedicated mass spectrometers for work with pulsed sources are available.
The fact that there is a large need for the direct and microdistributional
analysis of compact electrically nonconductive sampIes such as ceramics has
spurred this development. Volatilization by laser ablation in combination
with other signal generation techniques such as ICP-OES and ICP-MS, and
also direct emission spectrometry (e.g., see Ref. 13) enables it to obtain
detection limits down to the microgram per gram range. Laser-induced
atomic fluorescence at the laser plume leads to an increase of the power of
detection by at least a factor of 10.0 4 )
4.1.2.4. Ho/low Cathode Devices
Glow discharge sources have become of special interest as sources for

at01nic emission spectrometry. Trus applies especially to hollow cathode
sources as described by Schüler and Gollnow in 1935.0 5 ) They still are
important as emission sources and are treated in another chapter in this
book. The sampIe here is placed in a hollow cathode and volatilized partly
by cathode sputtering, as weIl as by thermal evaporation when the cathode
is not cooled. Its unique features lie in the fact that the analyte vapor has a
very long residence time in the negative glow of the plasma which fills the
cathode cavity and thus is very efficiently excited. Moreover, background
intensities are low because the electron number densities are low and accord-
ingly high-intensity ratios of line to background are obtained. (16) Hollow
cathodes for this reason are the most sensitive glow discharge emission
sources. The absolute detection limits are in the subnanogram range. How-
ever, no steady-state analyte volatilization is obtained. This may be a draw-
back with regard to a flexible multielement determination. However, in the
case of a hot hollow cathode, a selective volatilization occurs and this fact.
may be analytically used for performing a sensitive determination ofvolatile
elements in refractory matrices, as shown by Thomton. (17) This approach is
128 CHAPTER4
still very useful for the determination of P, As, Se, Bi, etc. in high-tempera-
ture alloys, (18) for which atomic emission spectrometry using a hot hollow
cathode is still a unique approach for routine use.
4.1.2.5. Planar Glow Discharges
Glow discharges with flat cathodes first became of interest for atomic
emission spectrometry with the lamp introduced by Grimm in 1968.0 9 ) The
Grimm-type lamp (Fig. 4-9) is a constricted type of glow discharge, of which
the discharge is confined to the sample which is taken as the cathode and is
water-cooled. Confinement is achieved in the classical Grimm-type lamp by
keeping the distance between the anode tube and the cathode block below
the mean free path length of the electrons, which is at the pressure of opera-
tion (some 100 Pa) below a few tenths ofa millimeter. Moreover, the vacuum
in the anode-<:athode interspace is lower than in the discharge region itself,
which is achieved by the use of a dual outlet pump with a larger throughput
for the interelectrode space. Accordingly, the sampie, which must be
vacuum-tight and electrically conducting, is volatilized by cathodic sputter-
ing only and volatilization interferences do not OCCUf. The sampie material
is excited in the negative glow of the low-pressure discharge. Because of
the absence of thermal evaporation as well as the separation of analyte
volatilization and the excitation in the negative glow (which is away from
the most energetic cathode fall), interelement effects are low and it may be
possible to analyze a large variety of sampies with the same calibration
teflon sheoth I
t to pump 1
Figure 4-9. Principle of the Grimm-type glow discharge lamp. (32) (Courtesy of the American
Chemical Society.)
function. Because ofthe consistency ofthe material volatilization and excita-

tion processes with time, rapid simultaneous but also very flexible multi-
element determinations of bulk constituents with glow discharge sources are
possible. Moreover, because of cathodic sputtering the sampie is ablated
layer by layer and the source is suitable for performing in-depth profiling,
as is discussed in aseparate chapter in this book. The Grimm-type glow
discharge is now a routine tool for the direct atomic emission spectrometric
analysis of solids and will be treated in detail.
4.1.3. Spectrometers
4.1.3.1. Dispersive Spectrometers

Dispersive spectrometers used in atomic emission spectrometry with
glow discharges normally make use of a diffraction grating as the dispersive
element. For sequential multielement determinations, monochromators are
used whereas simultaneous multielement analyses are performed with
polychromators. As the physical widths of the analytical lines in glow dis-
charge atomic emission spectrometry are on the order of 2 pm, optical spec-
trometers with a high power of detection are required for resolving the
physical linewidths and obtaining maximum signal-to-background ratios.
Therefore, a practical resolution of 200,000 at 300 nm would be required.
Accordingly, the use ofhigh-resolution gratings as weIl as spectrometers with
sufficient focal length fand narrow slit widths is required. The theoretical
resolution obtained with a diffraction grating is given by:
Ro=Nm (4-37)
where N is the total number of grooves (grating constant, linesjmm x width)

and m is the order. NormaIly, holographie gratings, with which a uniform
response over a broad spectral range can be obtained, are used; the groove
density is beyond 1800jmm and the width is at least 60 mm. The reciprocal
linear dispersion of the spectrometer is given by:
(dzjdJ,,)-1 = fmja cos () cos ({J2 (4-38)
where ({J2 is the angle between the grating normal and the diffracted beam
and () is the angle between the diffracted beam and the radiation detector.
In order to keep the spectral slit width below the contributions of diffraction
and of the physical widths, the entrance and exit slits of the spectrometer
should often be in the micrometer range. However, tbis is not possible
130 CHAPTER4
Tab/e 4-2. Background Radiances iu a

Grimm-Type Glow Discharge Lamp
Wavelength Photocurrent
(nm) (A)(xl0- 9)
200 2.8
220 8
240 5
260 4
280 4.5
300 5
320 5.5
340 8.4
360 7
380 5
400 5
420 7
440 5.5
460 4
"Grimm-type lamp with floating restrictor
(diameter: 8 mm), 0.35-m McPherson monochrom-
ator (Czerny-Turner mounting, grating: 1200 linesi
mm, width: 50 mm, slit widths: 15 Jlm, IP28 photo-
multiplier with dark current of 2 x 10-9 A), copper
sampIe, 4 Torr, 50 mA, I kV.
because of stability requirements. Moreover, as shown by the results in Table

4-2, the background radiances of a glow discharge lamp measured with a
rather quick (high throughput) spectrometer are already on the same order
of magnitude as the dark current of a good photomultiplier and the signal-
to-background ratios would be determined by dark current limitations.
Therefore, typical slit widths are at least 30/lm.
For sequential spectrometers both the Ebert (Fig. 4-10a) and the
Czerny-Turner (Fig. 4-1 Ob) mountings are used. In the first case, optical
aberrations especially when using narrow slits and a focallength >0.5 mare
minimized by using curved slits. With a Czerny-Turner mounting this prob-
lem does not occur as two mirrors with slightly different focal length and
slightly different incident angles can be used. For measurements at VUV
wavelengths purging the spectrometer with nitrogen or argon can be applied
to yield access to wavelengths down to 165 nm. Below these wavelengths,
quartz lenses and windows have to be replaced by LiF. In order to minimize
the number of reflecting surfaces, a Seya-Namioka setup with a concave
diffraction grating (Fig. 4-lOc) may be useful. Wavelength selection is per-
formed by turning the grating in all cases. By using high-precision grating
drives which make use of optical encoders, an angular precision of 10-4
degree becomes possible. For a grating with 2400 groovesjmm, this results
• I
e I
• I
a I
Figure 4-10. (a) Ebert, (b) Czemy-

Turner, and (c) Seya- Namioka mounting
for monochromators. Se: entrance slit;
~ : exit slit; m: mirror ; mJ : mirror of
entrance collimator; m2: mirror of
camera; gp: plane grating; g.,: concave
grating.
in a wavelength presetting precision of 1 pm according to:
sin cp = A/2a cos cp (4-39)
where cp = 10-4 degree, m = 1, and a = (2400)-1 mm. In order to accurately

peak wavelengths, searching procedures are used where a scan across the
line is made, the intensities measured are digitized, and the apparatus is
automatically set at the maximum. This of course presumes that line intensi-
ties sufliciently differ from the background features, which is not the case in
trace analysis. For the latter aim, spectral apparatus with fixed optical ele-
ments, as is the case with polychromators, should be used.
132 CHAPTER4
Polychromators used for simultaneous measurements have a Paschen-

Runge setup (Fig. 4-11) where the concave grating, the entrance slit, and
many exit slits are on the Rowland circIe. These instruments have a focal
length between 0.5 and 1 m and are used in most commercial equipment. By
providing a computer-controlled displacement of the entrance slit, spectral
windows around the analyticallines provided can be scanned. Accordingly,
background and interference correction is possible. As a displacement of the
entrance slit can be made without mechanica1loss, rapid sequential analyses
can be performed by displacement of the detector in the focal plane. Then,
random access to the provided slits or access to any wavelength is possible.
In the latter case, one can provide equidistant exit slits in a mask at every
2-nm wavelength difference for instance and adjust the wavelength in
between by displacement of the entrance slit. By performing computer-con-
trolled displacement of detector and entrance slit, rapid access to any wave-
lengths is possible.
holographie grating
BAIRD
Speetrovoe 1000 steel masks with
exit slits
stepping motor
Figure 4-77. Paschen-Runge mounting with concave grating and background measurement
facilities for polychromator.
Instead o( one dispersive element, multiple dispersion elements can be

applied so as to realize high spectral resolution with compact instruments.
When using an echelle grating (with a grating constant of up to 150 linesj
mm) a prism can be used and both crossed dispersion and parallel dispersion
can be applied. An echelle spectrometer with a 0.5-m Czerny-Turner mon-
ochromator, a quartz prism for cross-dispersion, and an echelle grating with
791inesjmm being used in the 75 order and higher, allows a reciprocallinear
dispersion of 0.15 nmjmm and is commercially available. The entrance slit
heights are only some 0.1 mm, but nevertheless a sufficient radiant
throughput for many purposes is still obtained. Also, instruments with paral-
lel dispersion are commercially available.
4.7.3.2. Multiplex Spectrometers
Apart from dispersive spectrometers, multiplex spectrometers such as

Hadamard transformation spectrometers(20) and especially interferometers
using Fourier transform processing(21) are useful for work with glow dis-
charges. The latter especially applies to glow discharges delivering time-
stable signals, as will be discussed further. This especially relates to the noise
properties as weIl as to the relatively line-poor emission spectra of these
sources by which multiplex disadvantages are much lower than in the case
of inductively coupled plasmas, for instance. (100) For the latter aim a Michel-
son interferometer (Fig. 4-12) can be used to produce the interferogram.
With the aid of a beam splitter the radiation is split into two parts which
are each directed to a mirror. When shifting the mirror in one of the side
( Predispersion )
Interferometer
Detector
FT of data in
computer
FiguTe 4-12. Principle of Fourier transform emission spectrometry.

134 CHAPTER4
arms, the interference for each wavelength is given by:
f(x) = B( 0')[1 + cos(21l'O'x)] (4-40)
where x is the optical path difference, 0' is the wave number of the radiation
(ern-I), fis the intensity measured with the detector, and B is the radiation
density of the source. The frequency f of the signal on the detector for a
radiation with frequency v is a function of the velocity D with which x is
changed:
f= O'D (4-41)
by which optical frequencies are shifted to the audio range and their varia-
tions can be measured with conventional detectors. A polychromatic radia-
tion source accordingly gives an interferogram where the intensity of each
point is the sum of all values resulting from Eq. (4-40). The central part
contains the information at low resolution and the ends contain information
at high resolution. The latter also depends on the registration time, the
spectral bandwidth and the number ofrepetitive scans. By applying a Fourier
transform on the signal for each point of the interferogram
+OO f+oo
f(x) = f B(O') dO' + B(O') cos(21l'O'x) dO' (4-42)
-00 -00
+OO
=C+f B( 0') cos(21l'O'x) dO' (4-43)
-00
with C a constant term being subtracted before the transformation one

obtains:
B(O') = f+OO
-00
f(x) cos(21l'O'x) dx (4-44)
The spectrum thus is obtained in wave numbers. By digitizing the interfero-

gram, a quick Fourier transform is possible and by multiplying with a block
function a limited but highly accurate displacement L of the arm allows
one to collect sufficient information for obtaining aresolution which is
proportional to (2L)-1 as weIl as with the number ofmeasurement steps. In
addition, repetitive scanning is applied where a reference laser is used to
obtain sufficient precision in mirror positioning. The technique has been
applied at IR wavelengths for some time and is now used for spectra in
the visible and UV region as systems for a sufficiently precise mechanical
displacement and high-capacity computers for the calculation of complex

spectra have just recently become available. Fourier transform spectrometry
is attractive for sources with low radiance and rather line-poor spectra but
high stability, and therefore is uniquely suitable for use in combination with
glow discharges.
4.1.4. Radiation Detectors
4.1.4.1. Photomultipliers
For atomic emission spectrometry, photomultipliers are most com-
monly used for radiation detection. In a photomultiplier the photons strike
a photocathode, from which photoelectrons are released and accelerated to
a number of dynodes mounted in a cascade geometry and with an anode as
the last stage. The current obtained at the anode (Ia) depends on the radia-
tion flux N,p as:
Ia = IeO(V) (4-45)
with
Ie = Nee (e = 1.6 x 10- 19 A . s) (4-46)
in which
Ne = N,pQE(A) (4-47)
O(V) is the amplification factor (to up to 106 ) and QE(A) (up to 20%) the
quantum efliciency. Together with the dark current, being I a when no radia-
tion comes on the detector (~l nA), they are characteristic of the type of
photomultiplier. Photomultiplier currents are amplified and integrated on a
capacitor, which is read out prior to analog-to-digital conversion. By using
a preamplifier with several capacitors, a high dynamic range can be realized,
as is required for glow discharge atomic emission sources having large linear
response functions (Fig. 4-13). Repeated readout with the aid of rapid Aj
D conversion under control of a personal computer now is standard in
atomic emission spectrometric instrumentation.
Photomultipliers which can be used at wavelengths between 160 and
500 nm often have an S-20 photocathode having a quantum efficiency of 5-
25%, the multiplication factors are 3-5 at dynode voltages between 50 and
100 V, and they have 9-15 dynodes. For selected types, the dark current
may be below 100 photoelectronsjs (Id < 10- 10 A). Red-sensitive photo-
multipliers have a much higher dark current which can be lowered by
136 CHAPTER4
12bit
Figure 4-13. Photomultiplier tube and measurement system. (129) (Courtesy of Verlag Chemie
GmbH.) a: photomultiplier tube; b: current amplifier; c: measurement amplifier for dc signal
and for signal integration; d: conversion of sign; e: analog-to-digital conversion.
cooling. For wavelengths below 160 nm, MgF2 must be used instead of
quartz as window material. For the low UV, "solar blind" photomultipliers
which are not sensitive to radiation of longer wavelengths (>350 nm) are
often used. They are particularly useful in polychromators where the ana-
lytical lines at longer wavelengths are measured in the first order and the
VUV and UV lines below 300 nm in the second order.
4.1.4.2. Multichannel Detectors
Multichannel detectors with parallel input have made considerable pro-

gress for atomic emission spectrometry. They include vidicon, SIT-vidicon,
photodiode array detectors, and image dissector tubes. (22) However, more
recently, charge-coupled devices (CCD) and charge injection devices
(CID)(23) have been introduced. Photodiode arrays (PDA) may comprise
matrices ofup to 512 and even 1024 individual diodes (e.g., Reticon) which
have individual dimensions of 10 pm in width and up to 20 mm in height.
The charge induced by photoelectric effects gives rise to photocurrents which
are sequentially fed into a preamplifier and a multichannel analyzer. There-
fore, they are rapid sequential devices. They have the advantage that memory
effects due to the incidence of high radiant densities on individual diodes
(lag) as well as cross talk between individual diodes (blooming) are low.
Their sensitivity in the UV and the VUV range is low compared with photo-
multipliers. They nevertheless presently are of interest for atomic emission
spectrometry as new scintillators with sufficient stability are available and
as they can be coupled with microchannel plate detectors giving an additional
amplification. Signal-to-noise ratios can be considerably improved by cool-
ing the array with the aid of a Peltier element or liquid nitrogen. Commercial
instrumentation is now available and has been successful (e.g., EG&G:
OMA 4; Spectroscopy Instruments Inc.). They have replaced silicon
intensified target (SIT) vidicon detectors, where the charge is integrated in
the individual pixels and the simultaneously collected information is read
out sequentially with the aid of an electron beam. When using an electrostatic
image intensifier (Fig. 4-14), these devices for many purposes are quite sensi-
tive and permit bidimensional resolution, and thus allow spatially resolved
measurements ofspectral segments (for an example in the case ofmicrowave-
induced plasmas, see Ref. 24). Photodiode arrays have been used successfully
in a segmented echelle spectrometer (Plasmarray, Leco CO.)(25) where spec-
tral segments around the analyticallines are sorted out by primary masks
subsequent to spectral dispersion with an echelle grating and are detected
after a second dispersion onto a diode array. Accordingly, more than ten
analyticallines and their background contributions can be measured simul-
taneously. The spectrometer is used succesfully in the case of inductively
coupled plasmas and glow discharges (e.g., see Fig. 4-15), (26) but with restric-
tions due to detector noise limitations at low wavelengths.
Image dissector tubes (Fig. 4-16) make use of an entrance aperture
behind the photocathode, by which the photoelectrons stemming from
different locations of the photocathode can be scanned and measured after
amplification in the dynode train, as is the case in a conventional photomulti-
plier. The system has been used in combination with an echelle spectrometer
with crossed dispersion for flexible rapid sequential analysis. (27) However,
due to the limited size of the photocathodes available and stability problems,
the technique has not made a breakthrough.
radiation
silicon
target
fibre optic photoelectron scanning
plate imaging + electron
multiplication beam
Figure 4-14. Early SIT vidicon system. (Courtesy of SSR Instruments, Ine.)
138 CHAPTER4
11000
>-
>-
.... ~
cn ~ ~
zw ~
.
~ ~
>-
..
0
z
.... ~
'"
~ "!
..
N X
~
.~
...,
~
N
:::
o ec ec eC
~
~
N '"0 .5
.,., "! "': ~
S e
c '"
-S.,
:;; 0
N ~
ec ": N M
..'-:
.~
'"
.~ ~
'"
r- U
.
M
N
M
u"
W
0
.
.,'"
0 M
1024
PIXEL
Figure 4-15. Spectral-segmented spectrum of glow discharge. (26) Grimm-type glow discharge
lamp with floating restrictor (8-mm diameter): 50 mA, 3.5 Torr, 2 kV; Plasmarray spectrometer
(Leco); steel sampie: 217A (Research Institute, CKD, Prague, Czech Republic).
j
I "-.. /
LJ I~~
I
I
"-../
Figure 4-16. Image dissector tube. (I) photocathode; (2) accelerating electrode (mesh); (3)
deflection coils; (4) photoelectron beam; (5) aperture; (6) electron multiplier.
Charge-coupled devices (Fig. 4-17) allow the readout of the induced

charge after it passes apreset value and can be used as bidimensional
detectürs(28) by which they are ideally suited für cüupling with an echelle
spectrometer. As their sensitivity is high, they may be expected to become
real alternatives to photomultipliers für atomic emission spectrometry. An
evaluation of the different radiation detectors with respect to their feasibility
for work with glow discharges is given in Table 4-3.
parallel
phases
detector
element
output
amplifier
serial
phases
Figure 4-17. Charge-coupled device (CCD) detector. (28) (Courtesy of the Society for Applied
Spectroscopy.) Layout of a typical three-phase CCD. Photogenerated charge is shifted in
parallel to the serial register. The charge in the serial register is then shifted to the amplifier.
Tab/e4-3. Evaluation of Some New Radiation Detectors for Atomic Emission Spectrometry
Sensitivity
versus
Detector Dimensions Spectral region photomultiplier Ref.
Vidicon 12.5 x 12.5 mm With scintillator down Poorer especially 22
bidimens. a to 200 nm at < 350 nm
dynamics< lOZ
lagjblooming
diode array Up to 25 mm, up to Especially for >350 nm Poor in UV 25
1024 dynamics 103
Diode 12.5 mm unidimens. 200-800 nm Similar to PM 22
array - dynamics > 103
MCP
Dissector Up to 60mm 200-800 nm Similar to PM 27
tube bidimens. a dynamics> 103
CCD Up to 13 x 2mm 200-800 nm Similar to PM 23
bidimens. a dynamics > 105
·Coupled with crossed-dispersion echelle spectrometer possible.
4.2. Glow Discharges with Planar Cathodes
The Grimm-type glow discharge source mentioned in Section 4.1.2.5

and its further developments are the most important types of glow discharges
employing planar cathodes now in use. Their analytical performance is
140 CHAPTER4
elose1y re1ated to their e1ectrica1 properties, the nature of material vo1atiliza-

tion and of the relevant excitation processes.
4.2.1. Electrical Characteristics
The elassica1 glow discharge is operated with argon as working gas. Its
pressure is maintained at 150 to 600 Pa and it is continuous1y bled into the
1amp through a need1e valve and pumped away. As the working gas flow
and pressure are ofparamount importance both for the sampie vo1atilization
and for the signal generation processes, the pressure measurement should be
made as elose as possible to the discharge plasma. A capacitance transducer
as available from several manufacturers (e.g., MKS Baratron, Phillips-
Granville) should be used to be as independent as possible from the nature
of the working gas used. The electrical characteristics of a elassica1 glow
discharge lamp are abnormal because of the obstructed nature of the dis-
charge. They depend on the anode-cathode interspace (on the order of some
0.1 mm) but also on the burning spot diameter. For a burning spot diameter
of 8-10 mm, and argon as working gas, burning voltages at discharge
currents between 40 and 200 mA range from 0.5 to 2 kV but depend con-
siderably on the nature of the gas, its pressure, and the cathode material
(Fig. 4-18). (29) With krypton and neon the burning voltage is higher than in
the case of argon, which in the former case is due to the lower mobiIity of
the heavier krypton ions and in the latter case to the higher ionization energy
of neon compared with argon. Further, the burning voltage depends on the
filler gas pressure, decreasing with increased pressure as the ion production
due to collisions becomes more effective and more charge carriers become
available. The dependence of the burning voltage on the sampie matrix
analyzed is considerable, as shown by the diagrams obtained with stee1,
copper, and aluminum sampIes (Fig. 4-19). (30) This is due to the fact that a
considerable part of the current is carried by analyte species and the ablation
rate depends on the sampIe matrix. Indeed, at an ablation rate of some
milligrams per minute, the seeding of the plasma with analyte atoms or
ions is on the order of 1%. This is considerably higher than in the case
of atmospheric-pressure discharges such as inductively coupled plasmas for
solution analysis where analyte concentration is on the order of 0.01% at
maximum.
As the sputtered material is deposited at the edge of the burning crater,
where the fie1d strength is somewhat higher and suction of the pump is
effective, the burning time of a elassical Grimm-type lamp is limited to about
15 min before electrical instabiIities occur. The inner wall of the anode tube
as weIl as the anode-cathode interspace have to be eleaned regularly because
of sputtered atom deposition. For a given discharged lamp geometry and
type of sampie matrix, two of the parameters (current, voltage, and pressure)
8
V 107
.
13·3
1600 + I
I
\/
1400
J
1200
.
~
2-
::l 1000
800
600
/+
400
20 40 60 80 100 120 140
J [mA]
Figure 4-18. Current-voltage characteristics of Grimm-type glow discharge lamp. (29) (Cour-
tesy of Pergamon Press.) Buming voltage UB as a function of the current J with argon as
discharge gas and asteei sampie. Argon pressure in mbar. Curves indicating equal power are
drawn.
may be selected and the third one then is fixed. Normally the pressure is
kept at a preselected value and constant-current or constant-voltage opera-
tion, where the voltage or the current respectively then vary, is selected. One
also can operate the lamp at constant power, by imposing small voltage
corrections when current changes occur. It should also be mentioned that at
the initiation of the discharge the voltage and current may oscillate for a
certain time, which is due to the release of molecular gases from the sampie
surface. Eventual breakdowns of the discharge can easily be prevented by
providing a large inductor (1-2 H) in the discharge circuit.
4.2.2. SampIe Volatilization

In the case of a restricted type of glow discharge with a cooled and flat
surface, the sampie volatilizes as a result of cathodic sputtering only. The
theory of cathodic sputtering (e.g., see Ref. 31) can be weIl applied so as to
qualitatively describe the influence of various discharge conditions on the
sputtering. A quantitative description, however, is hampered by the fact that
sputtering models have been developed for high-vacuum sources where the
impacting particles exist as monoenergetic beams with a well-defined angle
~
.....
ci
,; ~
~
i
so
630
::i
ci
960
~
;;;
j
//~i :''''60 .
", .....,.:._ -J-'+A- ~ / Curr.nt,mA ~
,/ 1.1'1 I I ~--7 d
_ t-.,L--....,k..L I 1/ /
I I... ... I 1-,1---1'- ~
l-J'_ _ _ 1 I ./1 / <:
a ""q " / -"-7'-'---.' ..
dfi{&:
,..:; __ 1/
600-'7--,. -r-- -!-~_L.
,. : /+
/ :s
q'- 670 7---~--T---:/ ;60 ~ '~ sy'. :Currem. mA so
---,!~+-,L-::.-:+! -l-:;:/
~10 /' .;r/ : /'1.-/1 J_ . . . .
_-7/__ ~.,I'--iJl- I " : I ...
/ 25-1 ·~'490
... /
/ ,1// ",,1--;L
/,r..... -
~~'C:'-___ L_-
b~ 7 -_-7'---.. I ,. I
,jj 600,./ ... / : / __

// 1/ 1
q<!'.670--~--"'---,L--"'--1lXi 110
o
Figure 4-19. Response surfaces of GD-AES for different metal matrices. (30) Glow discharge lamp with floating restrictor
~
diameter of 8 mm. (a) Steel matrix: spectralline Fe(I) 371.9 nm; (b) aluminum matrix: spectralline AI(I) 396.1 nm; (c) ."
copper matrix: spectralline Cu(I) 324.7 nm. The values listed near each data point indicate the GDS operating voltage at iil~
that point on the response surface. ...
with respect to a monocrystalline target. In an analytical glow discharge, the

pressure is on the order of some 100 Pa and accordingly, due to collisions,
ions and also neutrals impact under different angles. Backdiffrision of sput-
tered atoms is also considerable. Furthermore, the sampIe material is mostly
polycrystalline and structural defects as weIl as chemically inhomogeneous
grains may be present.
The general theory of cathodic sputtering has been covered extensively
in Chapter 2 and this treatment will be confined to its use for understanding
sampie volatilization in a glow discharge lamp with a planar and cooled
cathode. The cathodic sputtering in a glow discharge with a flat cathode can
be weIl explained by the model of impulse theory. After the removal of the
molecular gases and impurities which are adsorbed on the surface and the
formation of a stable burning crater surface, we obtain sputtering equilib-
rium. At this point, the discharge penetrates with a constant velocity into
the sampIe and the composition of the removed material equals the bulk
composition of the sampIe. From this time onwards, reliable bulk analyses
can be made. The preburn required depends on the nature and the pretreat-
ment of the sampIe. For instance, at 90 W with an argon discharge gas the
preburn times required are: for zinc, 6 s; brass, 3- 5 s; steel, 20 s; aluminum,
40 s.(29) Accordingly, spectrometry with glow discharges normally is slower
than spark emission spectrometry and thus less suitable for rapid production
control. However, when applying high-energy preburns the speed of analysis
can be considerably increased. The preburn time also depends on the filler
gas used: it has been found to be shorter in krypton than in argon which
again is shorter than in neon. The burning crater has a structure (Fig. 4-20)
Figure 4-20. Bum crater obtained by glow discharge optical emission spectrometry of a
compact metal sampIe (5 min buming at 50 W).
144 CHAPTER4
whieh results from the presenee of different erystals with different orientation
and thus different ablation by sputtering and also from ehemieal inhomo-
geneities. This strueture limits the in-depth resolution when profiling is per-
formed. This is diseussed in detail in Chapter 8.
The material volatilization depends on the different working para-
meters. The dependenee on the sampIe material ean be explained by impulse
theory. At 100 Wand 1330 Pa in argon, the sputtering rate inereases in the
order C < Al < Fe < steel < Cu < brass < Zn. As the discharge eharaeter-
istics depend on the filler gas, it is diffieult to compare results obtained with
different filIer gases. O2 and N 2 are hardly used as diseharge gases as they
give rise to oxides and nitrides, respeetively, and moleeular band emission
would eomplicate the spectra. For noble gases, the sputtering rate (for alumi-
num at 90 W) increases in the order neon< argon< krypton, which again
agrees with impulse theory. Helium gives a very poor ablation and ean only
be used as an additive. In the ease of argon the sputtering rate is on the
order of 0.3-0.5 mg/min and the influence of pressure on the sputtering rate
at constant power is given by:
q= c/JP (4-48)
where c depends on the cathode geometry, the power used, and the sampIe
material, as investigated by Boumans. (32) At low pressure, the burning volt-
age indeed is high and energy losses due to collisions are minimal.
4.2.3. Excitation Processes

In a Grimm-type glow diseharge the plasma is not in loeal thermal
equlibrium. Excitation temperatures of 5000 and > 10,000 Kare cited,
depending on the lines used for temperature measurement. This would refer
to the existence of two groups of eleetrons, namely, high energetie eleetrons
for excitation and slow electrons involved in radiative reeombination. How-
ever, these values result from end-on measurements which give no aceount
of spatial inhomogeneities in the plasma. The gas temperatures in the plasma,
as derived from Doppler profile measurements or estimated from rotational
temperatures measured, are below 1000 K. The Doppler broadening of the
lines thus is low and the physical widths of the analyticallines are 1-2 pm,
as obtained by interferometric measurements(33) and by recent measurements
using Fourier transform spectrometry (Table 4-4).(30) The measurement
error in these data is about 0.5 pm. A tendeney toward increasing linewidths
can be observed as the glow diseharge souree power is increased, espeeially
for the Fe(l) 371.9-nm and the Cr(l) 425.4-nm lines. The glow diseharge
source power can be increased either by lowering the pressure while main-
taining the current constant or by raising the current while maintaining the
Table 4-4. Physical Widths of Atomic Lines Emitted by Glow Discharges Operated under
Different Working Conditions and Obtained by (a) Fourier Transform Spectrometry(30) and
(b) Interferometry(33)
(a) Steel sampie (CKD Research Institute) with 23.9 mgjg Cr and 1.3 mgjg Mo
Glow discharge working
conditions Physicallinewidth (pm)
Pressure Current Power Fe Mo Cr Cr
(Pa) (mA) (Wjcm 2 ) 371.9 nm 379.8 nm 359.3 nm 425.4 nm
430 50 200 2.1 1.6 1.6 1.9
530 50 120 1.8 1.7 1.7 1.6
600 50 115 1.5 1.7 1.5 1.4
70 270 1.6 1.6 1.9 1.7
60 190 1.6 1.6 1.7 1.6
30 50 1.2 1.3 10
15 20 1.2
(b) Glow discharge lamp operated at 1 kV, argon pressure variable; line: Cr 425.4 nm
Cr in steel Current Physical linewidth
(mgjg) (mA) (pm)
5.6 25 1.3
50 1.6
100 1.65
200 2.1
21 25 1.4
50 1.65
250 25 1.4
50 1.6
100 1.7
200 2.15
pressure constant; both seem to produce equivalent results and both are
known to increase the sampie ablation rates. (32) The observed increase in
linewidth may be due to self-absorption, which is very strong in the case of
resonance lines. (5)
Excitation mostly takes place in the negative glow where the number
density of free electrons is high. Mainly electron impact, but possibly also
Penning ionization, contributes to the excitation. The number density of
metastables in the negative glow is on the order of 10 12 cm-3 (34) as deter-
mined by absorption measurements, whereas 1016 cm- 3 argon atoms and
10 11 -10 13 cm- 3 sputtered analyte atoms are present.(35) Steers and Leis (36)
recently showed that charge transfer processes mayaiso playa role in excita-
tion in a Grimm-type lamp.
As the electron number densities are considerably lower than in atmos-
pheric-pressure plasmas (ne = 10 14 versus 10 16 ), the radiation density of the
146 CHAPTER4
continuum, which is caused by the interaction of free electrons (bremsstrah-

lung) and mainly by the interaction offree and bound electrons (recombina-
tion continuum), is given by:
r2
Iv = Kn n - - e-hv/kTe dv
dv e r T. 1/ 2 (free-free) +
e
1 Z4
K' - n n - /U; - hv)/kTe (free-bound) (4-49)
/ e Z t~/2
where nr is the concentration of ions with charge r, Z is the charge of the

nuc1eus, and ~ is the ionization energy of the term with quantum number
j. As ne is considerably lower than in atmospheric-pressure plasmas, the
background continuum will be accordingly lower and high signal-to-back-
ground ratios can be expected.
4.3. Analytical Performance
Because of the material volatilization by cathodic sputtering and the

excitation in a low-pressure discharge, glow discharges with a Hat cathode are
very suitable sources both for sequential and for simultaneous multielement
determinations in electrically conducting solids. The optimization will
inc1ude a careful consideration of the electrical parameters, the working gas
and its pressure as weIl as the sampie preparation. The optimization, how-
ever, must be made in view ofthe analytical performance sought, and accord-
ingly, differences can be expected when optimization aims at the highest
power of detection, minimal interferences, or the highest analytical precision.
4.3. 1. Power of Detection
Optimization toward a high power of detection in the case of the

Grimm-type glow discharge requires measures regarding high radiant out-
put. Indeed, as reflected by background measurements performed with a
high-speed monochromator (Table 4-2), especially at low wavelengths the
background signals obtained are hardly higher than the dark current of
common photomultipliers. Accordingly, dark current noise as weIl as
counting statistics limitations may curtail the power of detection. Indeed,
the total statistical error Cf (lu ) is given by:
(4-50)
where CTp is the photon noise, CTD the dark current noise, CTr the flicker
noise, CTv the amplification noise, and CTc the contribution from the counting
statistics. A photon noise contribution from the counting statistics. A photon
noise contribution of 5 x 10- 4 at 200 nm is only slightly beyond the contribu-
tion of the dark current (3 x 10- 4) and a decrease of the latter by a factor
of five would improve the noise level by 30% and the detection limit accord-
ingly. This applies when taking smooth spectral background into account.
When blank contributions or spectralline wing contributions become impor-
tant, the detection limit is deteriorated according to:
(4-51)
where Ix is the net signal for a concentration c, Iv the total signal for
background and dark current, I B the net blank contribution (i.e., the contri-
bution stemming from interfering lines) and CT r(lB + Iv) and CTr(lV) the
respective standard deviations. (37) From Eq. (4-50), it can be seen that blank
contributions or signals from interfering lines do influence the detection
limits even when these contributions are highly stable.
The principal part of optimization toward a high power of detection,
however, is related to optimization for the highest line-to-background
intensity ratio. In order to get maximum line intensities: (1) the material
ablation should be high; this is achieved as discussed earlier at high buming
voltage, and can be realized at relatively low gas pressure, which in addition
has the advantage of minimizing collisions by which the energy of the inci-
dent species, and thus also the ablation rate, would decrease; and (2) the
excitation efficiency, however, also should be as high as possible, which is
achieved at higher working gas pressures, as then the number densities for
the exciting species and thus the number of collisions will be high.
The significance of the working pressure is c1early shown in many
studies (e.g., Fig. 4_21).(29) Both the material ablation and the excitation
efficiencies increase with current density (Fig. 4-22). (32) Apart from the line
intensities, both optimization parameters also considerably influence the
background intensities, as both the electron number densities and the analyte
number densities change. (38) Furthermore, each sampIe matrix has its own
current-voltage characteristics, which in addition depend on the construc-
tion of the lamp (anode tube diameter and anode-cathode interspace as
weH as gas throughput of the lamp), and accordingly further influences the
optimum values for the working parameters. Apart from these parameters,
the possibility of selective excitation processes (e.g., for copper see Ref. 38)
makes an optimization for each particular line necessary. This also applies
when using different working gases.
148 CHAPTER4
~
a10
>-
~
j 8
....c
6
2
Figure 4-21. Influence of working pressure on
line intensities. (29) (Courtesy of Pergamon Press.)
8 12 16 20 Conventional Grimm-type glow discharge lamp
Press ure , mbar operated in argon; 70 mA, aluminum sampie.
Cu 3274-
50
Figure 4-22. Influence of current density on

intensity of Cu 324.7 nm at different penetra-
.w=::---:l:----::!::-----:~---d1OO tion rates (J.lmjmin) for a copper target. (32)
(Courtesy of the American Chemical Society.)
Detection limits obtained with a Grimm-type glow discharge in the case

of metallic sampIes are in the microgram per gram range, as shown for steels
and copper sampIes (Table 4-5). (39,44,48) The values found in the literature
often are higher than expected, which may be due to the fact that polychrom-
ators are used on which the most sensitive lines are not always available or
Table 4-5. Detection Limits of Glow Discharge Optical Atomic Emission Spec-
trometry for the Analysis of Steel(44) and Copper Samples(57)
Detection limits (jJg/g)
Element/line Steel with Steel without
(nm) microwaves microwaves Copper
Ag(l) 338.3 1.8
AI(I) 396.2 0.1 0.4
As(l) 190.0 <10
B(I) 209.1 0.3 0.8
208.9 0.4 0.9
Cr(l) 425.4 0.05 0.2
Cu(l) 327.4 0.3 1.5
324.8 0.9 2
Mg(l) 285.2 1.5 2
Mg(II) 279.6 0.9 1.3
Mn(l) 403.1 0.2 1
Mo(l) 386.4 0.8 1.5
Nb(I) 405.9 0.6 4
Ni(l) 232.0 0.1 0.5
Si(l) 288.2 0.4 3
Ti(l) 364.3 0.6 3
V(I) 318.4 3
Zr(l) 360.1 1.5 8
of which the practical resolution is low because of stability reasons (e.g., for
steels, see the values in Ref. 40).
Several means are available to improve the power of detection. They
include
1. Innovation in the lamp concept itself. Here, measures that improve

the ablation rate as weIl as those that increase the excitation efficiency
can be taken. For the first part, the use ofmagnetic fields to increase
the density of the sputtering beam(41) increases the current density
by further constricting the buming spot, and improving the cooling
of the lamp can be advantageous. For the second part, the use of
cross-excitation by a dc discharge, (42) a high-frequency(43) or a micro-
wave discharge(43.44) (see Seetion 4.5.1) has been shown to be
effective.
2. Improvement of the speed of the spectrometer. This is particularly
useful to overcome detector noise limitations which may become
limiting at low wavelengths as already discussed. As the spectra of
glow discharges are almost less line-rich than those of c1assical
sources such as the dc are, this can be achieved by employing spectro-
meters of rather low practical resolution which have large spectral
150 CHAPTER4
bandwidths, insofar as spectral interferences do not take place. A

straightforward approach is the use of Fourier transform spectro-
metry, as discussed in Section 4.5.2.
4.3.2. Matrix Interferences
Because of the sampIe volatilization by cathodic sputtering, matrix

interferences due to different volatilities of the elements are absent. There-
fore, matrix effects may result from differences in elemental sputtering rates
from one alloy to another and even from the influence of differences in the
metallographie structure from one sampIe to another. For a minimization
of these matrix effects, measurements should be performed after a sputtering
equilibrium is obtained. When igniting the discharge, the impurities present
on the sampIe surface enter the discharge plasma and may influence the
excitation conditions. Further, during the initial discharge stage the preferen-
tially sputtered elements are enriched in the material leaving the surface
until, as a result of a drop in the bulk concentration for these elements and
of backdiffusion, a dynamic equilibrium is reached. At this point, the line
intensity signals become constant and matrix influences are minimal.
The prebum time in the case of a Grimm-type glow discharge is on the
order of 20-30 s for metal sampIes. (45) Accordingly, and because of the more
difficult sampIe change as compared with sparks, glow discharges are less
useful for rapid production control as required in the steel industry. They
are excellent for product control. The prebum time can be decreased by all
measures which increase the ablation rate. This can be achieved by high-
energy prebum which can be done by a controlled decrease of the discharge
gas pressure, by which the buming voltage at constant current increases, or
by an increase of the discharge current at constant discharge gas pressure.
It has been found that for the different matrices the preburn times correlate
with the sputtering rates. Electrically nonconducting powder sampIes can be
analyzed by glow discharge lamp spectrometry by mixing them with a metal
powder such as copper and briquetting vacuum-tight, electrically conducting
pellets, as described by EI Alfy et al. (46) As reported in the literature, the
preburn times in this case are considerably longer (> 1 min) but also show
the same dependence on the discharge parameters. (47)
The magnitude of the matrix effects in glow discharge emission spectro-
metry is exemplarily shown for the case ofmetals in Fig. 4-23 (48) (aluminum
alloys) and for the case of electrically nonconducting powders in Fig. 4_24(49)
(silicate rocks). For aluminum alloys, matrix effects were found to be con-
siderably lower than in spark emission spectrometry. In the latter case, a
strong curvature and even a splitting into several curves occurred. It should
be emphasized that, in any case, referring to a matrix line is required to
compensate for differences in sampie ablation rates. When analyzing pul-
/~6T3
" 62591
0.3
~595
062409
" 6165 AO-O

"ne
0.2
062406
/65
126" K,,?,65A
29-:Z-72
51373
0 64
6195AO-O 62
0.1 ../o~'10225
A'A
+6252 AO-O
0.1 0.2 0.3 0.4 0.5
Figure 4-23. Analysis of aluminum sampies by glow discharge emission spectrometry. (48)
(Courtesy of GAMS.) Conventional Grimm-type glow discharge lamp; restrictor diameter:
7mm; 125W, l.l kV; 11 mbar argon; analyticallines: Si(I) 251.6 nm and AI(I) 236.7nm.
5.10 4
~
"
2.104
10 4
5.10'
2 ·10'
10' I
0.5 3 5 10 20 30 50 100
Concentration , % I W:W J
Figure 4-24. Calibration curves for the determination of minor and major elements in slags
after pulverization and mixing with copper powder by glow discharge optical emission
spectrometry. (49) (Courtesy of Dr. S. EI Alfy.) S: synthetic powder mixtures.
152 CHAPTER4
verized mineralogical sampies, it has been found possible to calibrate using

synthetic mixtures of metal oxides.
4.3.3. Analytical Precision
The types and sources of noise occurring in the case of glow discharges
have to be investigated to discover the ultimate precision achievable, and to
find ways for its further improvement. Noise power spectra reflect the types
of noise present in an analytical system, and results of rapid Fourier
transform of the direct signal of a Grimm-type glow discharge have been
recorded.(30) The authors found that most noise is white noise and that flicker
noise displaying a 1/1 frequency dependence is hardly present. However,
discrete noise frequency bands also occur. It was found that some contribu-
tions may stern from the vacuum system and others from the power line
frequency (Fig. 4-25). Thus, both of the last points require care with respect
to the instrument design. From comparative analyses of Cu/Zn sampies, it
was found that the analytical precision of glow discharge emission spectros-
copy approximates that of x-ray fluorescence spectrometry and is similar to
solution analysis for major elements in the same sampies with ICP optical
emission spectrometry, provided that line intensity ratios are used as ana-
lytical signals (Table 4-6). (50)
Frequency, Hz
Figure 4-25. Amplitude noise spectra

of gIow discharge optical emission spec-
trometry with a Grimm-type glow dis-
~ 2 b
oj
charge lamp. (30) Steel standard sampIe
"0
218A (Research Institute CKD, Czech
.~
ä. Republic); i: 50 mA; argon pressure :
~ 1 3.5 Torr; burning voltage: 900 V;
~
Q.
0.35-m McPherson monochromator;
line: Fe(I) 371.9 nm. (a) Without needle
o t:::::::::::::::;~~:::=~::::::::::~~:::::::::;::; valve between gIow discharge lamp and
o 20 40 vacuum pump; (b) with needle valve
Frequency, Hz between gIow discharge lamp and pump.
Table 4-6. Analytical Precision Obtainable in Glow Discharge Atomic Emis-
sion Spectrometry, X-ray Fluorescence Spectrometry, and ICP Atomic Emission
Spectrometry(50)a
Concentration Standard deviation

SampIes range for Cu over the regression
Method (brass) (mg/g) (mg/g)
GD-AES Alloy 500-900 2.1
XRF Alloy 500-900 0.6
ICP-AES Solution 100-900 2.1
'Calibration according to the method of binary ratios.
The use of an internal standard is known from the early days of ana-
lytical emission spectrometry. With I qp and I sr the intensities of an analytical
line of the element X and a reference line of the element R, respectively, the
intensity ratio can be written as:(51)
(4-52)
In order to compensate efficiently for fluctuations in sampie introduction as

weH as for changes in excitation temperature, the excitation energies Es and
Eq , and the partition functions Q should be similar. In any case, either two
atom or two ion lines should be selected, for otherwise ionization would
falsify the number densities of the radiating analytical or reference species.
For the same reason the ionization energies of both should be as similar as
possible.
The use of reference signals is indispensible when the highest analytical
precision is required. It compensates both for differences in ablation rate
from one sampie to another as weH as for smaH differences in excitation
conditions and smaH fluctuations of the source. This also applies to the
determination of major elements as shown by the procedure used by Jäger
for the analysis of gold. (52)
4.4. Analytical Applications
4.4. 1. Analysis of Metals

For both bulk and in-depth analysis of metal sampies, glow discharge
atomic emission spectrometry with a Grimm-type glow discharge lamp is
very powerful and many systematic studies and applications to diverse matri-
ces have been described in the literature. (108)
154 CHAPTER4
4.4.1.1. Bulk Analvsis
Multielement analyses for major, minor, and trace elements have been
described for most industriaUy important matrices. In methodological
studies, the sputtering and ablation rates of various metals in a Grimm-type
glow discharge have been investigated. (32) Under analytically relevant condi-
tions, ablation rates range from 0.1 (Al) to 3 mg/min (Zn) for a target area
of 0.5 cm2, a current of 180 mA, and a voltage of 500-600 V. Accordingly,
preburn times of at least 30 s are required to reach sputtering equilibrium.
In the first paper of Grimm, (19) the strong dependence of the voltage on the
metal matrix was shown and therewith the necessity to relate relative intensi-
ties to concentration ratios in the case of the determination of major ele-
ments, as shown by the example of determination of Ni in binary Fe-Ni
alloys (250-1000 mg/ g).
The working pressure and the related sputtering and excitation proper-
ties of the source have also been shown to strongly inftuence the analyte line
intensities, whereas the low spectral background is much less influenced. The
properties of the source are also known to vary considerably with the dis-
charge gas used. As is known from sputtering studies (Chapter 2), argon
represents a good compromise as sputtering gas. However, neon or argon-
helium mixtures have also been used. (53,54) In the latter case, the sputtering
rate decreases with increasing amounts of helium, although the net line
intensities for many lines increase. (55) The properties of glow discharge lamp
spectrometry were criticallY compared with those of spark emission by Durr
alld Vandorpe. (56) Whereas in the first case rectilinear calibration curves
were obtained for nickel and sulfur in steel, the latter suffered from curved
calibration curves (Ni) or even from splitting. For Sn, however, the glow
discharge calibration was also curved. With a demountable type of glow
discharge lamp, where different anode-cathode geometries and a floating
restrictor can be used, (57) fine-tuning of the pressure and flow conditions in
the lamp measurably influences the freedom from interferences stated above.
This has been shown mainly for Cu/Zn and Cu/Zn/Pb alloys. For steel
analysis, the Grimm-type lamp can be used to advantage product control,
as shown by Rademacher and de Swardt(45) who reported preburn times on
the order of 20-30 s for Si, Mn, P, C, and S. For C, 156.1 nm was found
most advantageous. RSDs from 0.5 to 2.5% were found in the 0.01 to
10 mg/g range. With a magnetically focused glow discharge lamp, Kruger
et al. (40,41) succeeded in increasing the power of detection by a factor of 2-5.
This is partly due to the increase in ablation rate (up to twofold). However,
it mayaIso be that the residence time of electrons, and accordingly their
number density, in the negative glow, increases by which an improvement
in excitation takes place. The lamp, however, also showed promise for non-
ferrous metals as here no rinsing problems occur. (128) In the investigation of
iron-based alloys, Wagatsuma and Hitokawa(58) reported that chromium

oxide films on the surface influence the sputtering yield, as reflected by the
change of chromium and iron line intensities. Results for ternary Fe/Ni/Co
alloys are presented. By applying voltage modulation of the lamp and lock-
in amplification of the analytical signals, the signal-to-noise ratios can be
improved by a factor of 20-50 in the case of low-alloyed steels. (59) Apart
from steels, cast iron can also be analyzed. For C, here, the C 247.86-
nm line is proposed. However, for cast irons with different metallographie
structure, not only the signals but also the intensity ratios are different.
Further working at increased burning voltage was shown to shorten the
preburn time considerably. (60)
For the analysis of noble metals by glow discharge emission spectrome-
try, techniques for increasing the precision were developed; they are based
on the use of the sputtering rate as reference signal and allow it to obtain
in gold analyses RSDs down to 0.1% in the major element range (for Au).
Further cast and rolled sampies could then be analyzed with the same calib-
ration curves. (61) The same principle allowed RSDs below 0.09% to be
obtained for silver in silver analyses, whereas for Au, Pd, Pt, and Cu, the
RSDs were on the order of 5% at the 1-2% (w/w) range.
For Cu-Al alloys, Dogan(62) found an RSD at a concentration ofO.06%
(1 - w), with w the concentration for a concentration range of 0.5-
99.5% (w/w) ofCu or Al in binary alloys. This shows that, especially in the
case of simultaneous emission spectrometry, a high analytical precision for
major elements is attainable. Also, in other papers on the analysis of Cu
alloys, (63) RSDs of 0.2% have been reported for the major elements. Never-
theless, proper measures must be taken to correct for the influence of minor
elements on the emission intensities. Yamada et al. (64) reported that for
copper alloys, the metallurgical structure considerably influenced the sput-
tering rate; changes with the concentration of aluminum [1-12% (w/w)]
were discussed. It was found that after a preburn time of 100 s, however,
structural effects no longer affected the AI(I)/Cu(I) line intensity ratios.
4.4.1.2. /n-Depth Profi/ing
Due to the layer-by-Iayer ablation in glow discharges with flat cathodes,

in-depth profiling was found at an early stage to be one of the most impor-
tant analytical features of glow discharges, as shown in 1973 by Belle and
Johnson(65) and in 1978 by Berneron. (66) This important feature is discussed
in detail in Chapter 8, so only abrief outline will be given here.
For in-depth profiling, it is important to calibrate the penetration rate
as a function of the working conditions as well as to know its dependence
on the material composition. The latter point may be difficult especially as
the main composition in the case of coatings may vary considerably with
156 CHAPTER4
the sampling depth. For solving this problem, approaches that make use of
reduced sputtering coefficients, as discussed and developed by Bengtson, (67)
may be helpful. As shown by Bemeron (68) the problem of converting the
measured intensities of in-depth resolved concentrations is still more difficult.
Here, an attempt could be made to calibrate intensity ratios by comparing
them with values obtained for signals from bulk analysis of sampies of
known composition. Adequate software, which uses these data as a function
of the sputtering conditions (current density, working gas pressure, and
voltage) to calculate the in-depth variations of elemental concentrations, is
under development (see Chapter 8).
For obtaining sufficient resolution in in-depth profiling, the radiant
density per ablated layer should be high. (69) This necessitates operating the
lamp at relatively low buming voltage, to achieve low sputtering rates, and
at relatively high working gas pressure, to yield a high excitation efficiency.
Typical values are 400-500 V buming voltage at apressure of 4-5 mbar
argon. At a discharge current of some 10 mA, ablation rates for steels were
4 nm/s, as reported by Waitlevertch and Hurwitz(70) in early glow discharge
work. De Gregorio and Savastano(71) reported the use of discharge currents
of 20-100 mA, a buming voltage of 1 kV, and a time resolution of 0.1 s for
steels. However, it should be clear that the optimum conditions strongly
depend on the construction of the lamp used. The analytical figures of merit
of glow discharge emission spectrometry for in-depth profiling have to be
compared with those of other techniques for in-depth profiling such as secon-
dary ion mass spectrometry (SIMS), as shown by Takimoto et al. in the case
of galvanized steels.(72) For SIMS, the in-depth resolution is much better
than for GD-OES (in the nm range). However, in the production control of
technical surfaces, it is often necessary to investigate thicker layers, for which
SIMS is too expensive and time-consuming. Often it is sufficient to accurately
control the thickness of the layer to have semiquantitative information on
its elemental composition.
Most applications reported stern from the steel industry. Koch et al.(73)
reported that for bulk analysis, glow discharges have not made a break-
through in the steel industry. However, for surface characterization of steels
involving the investigation of oxide and nitride layers with a thickness greater
than 0.1 11m, GD-OES is of use. (74) In their work, they used a buming
voltage of 1000 V and a discharge current of 100 mA and reported sputtering
rates of 25 nm/s for iron oxide layers. The problem of quantization was
experimentally attacked by Wagatsuma and Hirokawa, (75) who studied the
sputtering behavior of binary alloys so as to obtain sputtering constants that
could be used in calibration programs. In the case of silver-copper alloy
surfaces, (54) they found that the metallurgical structure of the sampies played
an important role regarding the length of the transition times prior to steady-
state sputtering. They also observed that redeposition of the silver grains
with copper may influence the sputtering yield after some time, which further
may falsify in-depth determination from sputtering times. Ohashi et al. (76)
used a glow discharge with floating restrictor and remote electrode to record
in-depth profiles of steels. In the case of measurements at 4 mbar, 50 mA,
and 200-400 V for sampies of differing nature, layers between 1 and 10 nm
still could be characterized using a time resolution of 1 s for the
measurements.
Progress in the use of glow discharges for in-depth profiling will result
from an improvement of the flatness of burning craters. As shown by inves-
tigations with copper alloys in a demountable glow discharge lamp by
Ko, (57) the use of a floated restrictor for this aim is very helpful. Furthermore,
the radiant output per ablated layer in the glow discharge still can be
increased considerably. This is indicated by the self-reversal found in the
line profiles of copper resonance lines measured with Fourier transform
spectrometry(77) which relates to a high number density of ground-state
atoms in the glow discharge plasma. As shown by recent measurements, the
self-reversal of the resonant lines emitted by the glow discharge vanishes
when applying cross-excitation with microwaves as described by Leis
et al. (44) This was indeed found to considerably enhance the net line intensi-
ties (Table 4-7).
4.4.2. Analysis of Dry Solution Residues
The analysis of dry solution residues is known from work with hollow
cathodes enabling extremely low absolute detection limits. This especially
Table 4-7. Signal Enhancement in Glow Discharge Atomic Emission Spectro-

metry Resulting from Microwave Boosting(44)a
Gain factor
Element/line dc current
(nm) (mA) 300 Pa 420 Pa 600 Pa
Cr 425.4 10 27 21 33
20 23 10 23
30 23 14 4
40 9
Cu.327.4 10 58 124
20 43 83 81
30 56 48
40 32
50 27
60 25
Si 288.2 10 36 44 39
20 22 14 16
30 11 10
'Steel sampIes, 40 W microwave power.
158 CHAPTER4
applies when the technique is used as the determination step in a multistage

combined procedure where the analyte is isolated by chemical procedures
and presented as a small-volume multielement concentrate.
Büger and Fink(78) tried to directly introduce liquids in the hollow
cathode glow discharge plasma but found strong interferences of band
spectra. Harrison and Prakash (79) succeeded in analyzing liquids by hollow
cathode emission spectrometry subsequent to their evaporation in steel or
graphite cathodes and obtained detection limits down to 0.1 .ug/ml, requir-
ing less than 0.5 m1 of sampie. For rare earths, similar results were obtained
with copper(80) and graphite cathodes. (81) For the alkali elements, extreme1y
low absolute detection limits (pg range) have been reported recently. (82)
However, the dispensing of smallliquid aliquots to plane surfaces is much
easier as reported by Brackett and Vickers. (83) Provided the necessary optimi-
zation and choice of carriers is made, a handy method with high power of
detection and multielement capacity can be developed, as has been shown
for the case of a Grimm-type glow discharge used as ion source for mass
spectrometry recently. (84)
4.4.3. Analysis of Electrically Nonconducting SampIes
Electrically nonconducting powder sampies have for some time been

analyzed with glow discharges after mixing with a metal powder and briquet-
ting compact and vacuum-tight pellets of good thermal conductivity.
Toward this aim, EI Alfy et al. (46) proposed a mixture of 3-4 parts of ductile
copper powder with the powder sampie. In order to keep sampie consump-
tion low (25 mg), a small amount ofthis mixture is embedded in pure copper
powder and pressed at 104 N/cm2 with a hydrau1ic press (Fig. 4-26). It
should be mentioned that sampies such as Ah03 can be easily analyzed,
provided no crystal (adsorbed) water is present, but that materials such as
Si02 tend to swell subsequent to pressing and make it difficult to get stable
pellets. The powder particle size must be small enough to effect a stable
material ablation. From sieving analysis, EI Alfy(49) reported no problems
with pellet stability when the powder particle size was below 25 .um.
The behavior and optimization of the Grimm-type glow discharge in
the case of pellets for powder analysis have been investigated by a number
of groups. As this is covered in detail in Chapter 7, only some lines of
research and their trends will be mentioned here. EI Alfy et al. (46) described
the determination of the minor compounds in ores and minerals and showed
that a calibration could be performed with synthetic standards prepared by
mixing the analyte oxide powders in a calcium carbonate matrix (Fig. 4-24).
The prebum times, however, were on the order of 2 min. Si, Ca, Mg, Mn,
Fe, Al, Ti, Na, K, P, S, and C have been determinated in a wide variety of
e1ectrically nonconducting substances (e.g., rocks, minerals, ores, glasses,
1 - - - - - 25mm - - - - l
copper • sampie copper
Figure 4-26. Briquetting of pellets for the analysis of electrically nonconducting powders by
glow discharge atomic emission spectrometry.
slags, cement). It was found that pronounced structures occurred in the

burning craters. Different theories were developed to explain their form.
Jäger and Blum(85) proposed the distortion of the local electrical field in the
vicinity of electricaHy nonconducting grains to be responsible for changes
in the local sputtering rates. However, preferential sputtering due to the
orientation of the crystallites with respect to the incident ions and atoms, as
weH as redeposition at favored places, may playa role. This has been studied
in more detail by electron probe microanalysis(86) and it was suggested that
partial melting at the sampie surface may occur hut did not influence the
volatilization of the nonconducting grains. Mai and Scholze(47) mentioned
that the preburn times required in the case of the discharge power of 100 W
are below 1 min. They also suggested that diffusional transport to analyte
from the nonconducting grains to their surface might play a role in the
material volatilization. Sputtering of these surfaces indeed could occur as a
result of surface conduction of redeposited material. XPS signals for CuO
and Cu supported this theory.
For the analysis of electrically nonconducting powders, glow discharge
emission spectrometry is a viable alternative to x-ray fluorescence spectrome-
try, as the light elements such as B, Be, and Li are easily determined. Further-
more, the power of detection for these elements extends to the microgram
per gram range and the precision in terms of the RSD relative to a reference
160 CHAPTER4
line is at the 1% level or better within a large range of concentration. As

the analysis of ceramic powders becomes more and more important, many
elements (including the light ones) have to be determined in Ah03, AIN,
SiC, ShN4 , and Zr02 matrices(87) and these applications of glow discharges
will become more important.
Recently, the use of boosted glow discharges for the analysis of electri-
cally nonconducting powders has also been described. Lomdahl et al. (88)
used the dc-boosted glow discharge lamp developed by Gough (89) to
determine down to 0.002% (w/w) NiO and 0.0015% (w/w) CuO in copper
and nickel binders, respectively, and achieved an RSD of 1-2%. Minerals
and rocks(90) were also analyzed. SampIes were mixed 5%-95% (oxide to
host) with copper powder and pellets of 6-mm diameter were pressed at
109 N/cm2. The method was used to determine major to trace compounds
[e.g., down to 0.004% (w /w) K] in rocks. The intensity enhancement realized
by the auxiliary dc discharge is essential for penetrating the trace element
concentration level.
Further progress lies in the use of an rf glow discharge with which
compact electrically nonconducting sampIes such as ceramics can be ana-
lyzed directly, as recently proposed by Duckworth and Marcus(91) first for
mass spectrometry and later for optical emission spectrometry(92) as dis-
cussed in detail in Chapter 7.
4.5. Lines of Research for Further Improvement
Further improvement in the analytical figures of merit of glow discharge

optical emission spectrometry may be realized by improving the glow dis-
charge itself and also by improved use of the spectral information.
4.5.1. Application of Cross-Excitation
The idea of boosting a glow discharge to achieve a higher radiation

output, overcoming shot-noise limitations and also increasing the line-to-
background intensity ratios, has been pursued for some time. Indeed, the
need for high-intensity primary line sources for atomic absorption and
atomic fluorescence spectroscopy forced the research on increasing the radia-
tion output of hollow cathode lamps. This could be achieved by pulsing the
sources and also by boosting the dc hollow cathode discharges. Indeed,
plasma spectroscopic investigations revealed that in glow discharges a lot of
the volatilized material is present in the form of ground-state atoms, and by
increasing the excitation efficiency the number densities of excited atoms still
can be substantially increased.
For the glow discharge sources in emiSSIOn spectrometry, several

methods for cross-excitation have been investigated. First, a dc discharge
has been put across the glow discharge plasma, an idea that was mainly
pursued in Australia by Gough, Sullivan, and others. In the work of
Lowe, (42) a dc discharge of several hundred milliamperes was put across the
glow discharge plasma and found to produce a considerable increase in line
intensities. McDonald(93) found excitation temperatures for Fe(l) and Fe(II)
lines in the range of 4000 K and electron densities of 5 x 10 13 while the
degrees of ionization were 81.3% for Ca, 19.5% for Cr, and 5.5% for Mg.
Gough and Sullivan(94) combined a source using a secondary discharge (up
to 750 mA at 120 V) on anormal glow discharge (up to 100 mA at 900 V)
with a polychromator and analyzed steels. When using mainly re sonant lines
for analysis, aseries of elements could be determined down to the microgram
per gram level. The secondary discharge made use of a heated filament
(0.3 V, 10 A) as an electron source. This emitter, however, as well as crystals
of rare earth compounds used later are easily poisoned by the air contact
during sampie change as well as by deposition, which has made the lamp
difficult to operate. Other approaches make use of electrodeless discharges
for boosting. Ferreira et al. (43) used a high-frequency discharge and reported
an enhancement between 10- and 20-fold when superimposing a 13.6-MHz
discharge on a 25-mA glow discharge. However, when the current was
increased to 100 mA, the difference became negligible. (95)
The use of microwave boosting seems to be more effective. Coupling
with an antenna was used by Ferreira et al. (43) for a Grimm-type glow
discharge and by Caroli et al. (96) for a hollow cathode glow discharge. Later
it was recognized that the efficiency of cross-excitation could be increased
greatly by integrating a Beenakker TM olO cavity with the low-pressure dis-
charge, as reported for the hollow cathode discharge by Caroli et al. (97) and
for the Grimm-type glow discharge lamp by Leis et al. (44)
The lamp used by Leis et al. (44) is schematically represented in Fig. 4-27.
The GDL part is essentially the same as Grimm's original Iamp with the
exception that it is relatively thin so that the integrated microwave cavity
part is as elose to the cathode as possible. However, for construction reasons
(water cooling, gas inlet and outlet, O-rings) the distance from the cathode
to the inner wall of the microwave resonator is still 16 mm. The exchangeable
anode tube has an internal diameter of 8 mm and is at ground potential by
electrical contact with the walls of the cavity. A quartz tube, also of 8-mm
internal diameter, separates the inner low-pressure region from the outer
part of the cavity, which is filled with air at atmospheric pressure. The gas
inlet, the illumination of the I-rn sequential spectrometer, and readout sys-
tem are described in Ref. 44.
The lamp can be opera ted with or without microwaves and has always
been used in a current-regulated mode. When microwaves are superimposed,
162 CHAPTER4
coaxia gas inlel porl

line
waler cooled
calhode block
coupling
loop
anode body
sümple
quartz
wind:w "l
quartz tube
ancde tube
Isclaler
TM 010 -resonator
vacuwm
pump
Figure 4-27. Glow discharge lamp with integrated microwave cavity.(44)
a microwave discharge which forms one optically thin plasma with the posi-
tive column can be observed. A microwave discharge occurs, depending on
pressure and current of the dc discharge, at 30 W (at 300 Pa and 10 mA)
and from 60 W onwards, it can be sustained without the glow discharge.
The addition of the microwave discharge considerably decreases the burning
voltage and a most favorable tuning is obtained at a forward power of 40 W
(reflected power 15 W). The long-term stability (after more than one day)
is limited by deposits of sputtered material on the quartz tube, influencing
the microwave coupling, and on the observation window.
The burning craters obtained with microwaves tend to be somewhat
deeper in the center, their depth determined by changes in burning voltage,
which is a function of the microwave power applied. The erosion rates are
on the order of 10-30 nm/s, increasing with the applied anode currents.
These values are 2-5 times lower than in a conventional glow discharge.
Accordingly, the preburn times will be long (the signals fully reach their
equilibrium value after 3 min). Considerable enhancement in net line
intensity especially for lines with rather low excitation energies (resonant
lines) has been found. For phosphorus and sulfur lines, for instance, the
enhancement was lower. At low glow discharge currents, the enhancement
was highest, which is similar to the observations of Walters and Human for
the rf-boosted lamp. (95) Also, when comparing the background equivalent
concentrations obtained under optimized conditions (16 mA, 40 W, 300 Pa)
with and without supplementary microwave excitation, a gain is still
obtained. The RSDs both for the line intensities and for the spectral back-
ground at integration times of 8 s are 1% or lower. The detection limits
obtained for stee1 sampies in the case of the microwave-boosted discharge
are 3-5 times lower than in the case of a conventional glow discharge lamp
operated under optimized conditions (Table 4-5). The linearity of the cali-
bration curves extends over three decades of concentration with no curvature
observed at the high-concentration side (Fig. 4-28). It is interesting to remark
that the se1f-reversal, which can be observed for resonance matrix lines [see
Fourier transform spectra for Cu(I) 327.4 nm], is no longer observed in the
case of the microwave-boosted lamp. (77) The glow discharge lamp with inte-
grated microwave cavity thus is a suitable tool for the analysis of electrically
conducting solid sampies. For many elements in steel, this constitutes pro-
gress in the power of detection down to the submicrogram per gram level.
Further measurements showed that this also applies to alurninum and copper
107 Cr I 425.4nm
AI I 396.2 nm
10'
Figure 4-28. Calibration curves

for live elements in steel sampIes
measured with the microwave-sup-
ported glow discharge lamp. (44)
Each sampIe was measured only
once with an integration time of 10 3 10'
3 s. Concentrction. 1J9 19
164 CHAPTER4
matrices. (98) The high radiance, which is achieved even at a reduced erosion
rate, also is favorable for in-depth profiling, although under the conditions
used, the crater is not as flat as with a conventionallamp. In addition, the
low heating of the sampie is a suitable basis for the analysis of metals and
alloys of low melting point such as lead.
Another approach is furnace nonthermal emission spectrometry, devel-
oped by Falk et al. (10) and discussed in aseparate chapter of this book. Here,
progress lies in the complete separation of sampie volatilization in a furnace
and excitation in the negative glow of a dc discharge. Liang and Blades(99)
follow a similar approach using capacitive powering of the plasma. Further
lines of improvement in the power of detection of glow discharge emission
sources may be found in side-on observation of the glow discharge plasma.
This can be done easily with the aid of quartz fibers. Indeed, preliminary
measurements performed with a glow discharge operated at 25 W /cm2 ,
350 Pa for the Fe 371.9-nm line revealed that the signal-to-background ratio
for side-on measurements at 2 mm above the cathode was 12 compared
with 7.5 for end-on measurements. Of course, optimization of the working
conditions here may bring still further progress. Time-resolved measure-
ments in pulsed glow discharges may yield progress in the power of detection.
Indeed, especially due to the long lifetime of the exciting species (especially
of the argon metastables) the excitation capacity may remain high while a
decrease in background continuum may take place.
4.5.2. Improvements in the Use of the Spectrallnformation
The radiant output of the Grimm-type glow discharge lamp with a flat
cathode is very stable. Moreover, the optical emission spectra consist prim-
arily of neutral atomic lines, of which the resonance transitions dominate.
Accordingly, emission spectra of the glow discharge source ordinarily con-
tain fewer emission lines than do spectra of such sources as the inductively
coupled plasma (lCP). Furthermore, the relatively high analyte-vapor con-
centration in the source (up to 1% relative to the argon density) caused by
the high ablation rate from the cathode surface (on the order of a few
milligrams per minute for a Cu matrix) leads to favorable analyte-to-argon
line emission intensity ratios.
These strengths make the glow discharge source particularly attractive
as a potential source for Fourier transform spectrometry (FTS). In previous
studies of FTS-AES, where an ICP was used as a source of atomic emission,
a loss in the power of detection and a substantial dependence of sensitivity
on the type and concentration of the matrix were observed. (100) In contrast,
the stability of the GDS and the line-poor character of its spectra should
yield signal-to-noise ratios in FTS-AES that are comparable to those
realized in wavelength-dispersive spectrometers. Finally, the high optical
conductance of the FTS and the advantage of working with physically
resolved linewidths (readily available with FTS) guarantee the highest degree
of sensitivity(IOI) and the lowest level of systematic errors. Initial experiments
have been performed at the Los Alarnos Fourier Transform Spectrometer
Facility(l 02, 103) with a Grimm-type glow discharge lamp using a floating
restrictor. The glow discharge plasma was imaged 1 : 1 on the 8-mm entrance
aperture of the FTS instrument. Emission spectra over the spectral range
300-600 nm were calculated from interferograms recorded during a mirror
displacement of approximately 12 cm. The sampling rate of the interfero-
meter was 5000 Hz and approximately 7 min was required for each scan.
The signal was filtered by low- and high-pass filters before digitalization.
The FTS operating conditions provided a spectral resolution of 0.043 cm- I
(at 300 nm, the spectral bandwidth was 0.39 pm).
The high resolution of the FTS provides a means by which the widths
(FWHM) of matrix and minor elements lines can be measured directly and
simultaneously. Moreover, the dynamic range of the FTS is sufficient to
measure at the same time matrix lines, argon lines, and trace element lines.
The spectral widths of a number of analytical lines under different GDS
operating conditions are listed in Table 4-4. At discharge currents of 15-
70 mA and argon pressures of 430-600 Pa, the observed linewidths were on
the order of 1-2 pm, in good agreement with results reported by West and
Human, (33) and the measurement error is about 0.5 pm. Understandably,
the linewidths are lower than those obtained with sources operated at atmos-
pheric pressure. (101,104,105) A tendency toward increasing spectrallinewidths
can be observed as the glow discharge operating power is increased,
especially for the Fe(I) 371.9-nm and the Cr(I) 425.4-nm lines. The GDS
power can be increased either by lowering the pressure while maintaining
the eurrent eonstant or by raising the eurrent while maintaining the pressure
constant; both seem to produce equivalent results and both are known to
increase the sampIe ablation rates. (32) The observed increase in linewidth
with operating power may be due to self-absorption.
With spectrometers of sufficiently high resolution, matrix and spectral
interferences are ordinarily not significant in glow discharge spectrometry.
Consequently, it is possible to estimate detection limits by extrapolating an
analyte atomic line intensity (I) to a level equal to three times the standard
deviation of the spectral background adjacent to the line. Because of the flat
noise distribution in FTS, a valid estimate of the noise can be obtained as
the root mean square (Nrrns ) noise in a spectral region elose to, but distinct
from, the spectralline of interest. Following this procedure (see Fig. 4-29),
a detection limit of 30 pg/g is obtained for the FTS determination ofmolyb-
denum [Mo (I) 379.8 nm] in low-alloy steel (IMo = 7.74 X 106, I avg =
2.00 X 106, Nrrns = 1.00 X 106, and [Mo] = 50 pg/g).
It should, however, be emphasized that the value stated for Mo is only
an estimate as no serious attempts were made to optimize the Mo signal.
Nevertheless, it suggests that the sensitivity of GDS-FTS approaches that
166 CHAPTER4
['001.0
.-I
iC
o
.-I Mo I 379.8 nm
.......
,/
....~.,
Sideband
i
.jJ
0.5
I:l
H
--- - Average
0.04-~----~----~----~--~~-
379.83 379.81 379.79
Wavelength, nm
Figure 4-29. High-resolution record of spectral structure in the vicinity of Mo(I)
379.8 nm.(30) SampIe: steel (216A) (CMo: 50 jlg/g); i: 50 mA; argon pressure 470 Pa; buming
voltage: 1500 V; high-resolution FTS (0.043 ern-I). Detection limit for Mo: 30 jlg/g.
obtainable from a GDS in combination with a high-resolution wavelength-

dispersive monochromator. (44) The latter might be limited by dark current
rather than by line-to-background intensity ratios. This behavior is verified
by the spectral background signals obtained with a small (0.35 m, slit width:
15 J.lm, spectral bandpass: 0.03 nm) monochromator (at 200 nm:
2.8 x 10-9 A; at 300 nm: 5 x 10-9 A; at 400 nm: 4 x 10-9 A), which are
hardly above the dark-current level of the photomultiplier (2 x 10-9 A). In
addition, these results indicate that the multiplex disadvantage is not a seri-
ous limitation for FT-GDS-AES. Moreover, because the entire spectrum is
available at high resolution, multiline calibration could further enhance the
precision and the reliability of the analytical data. Figure 4-30 shows the
molybdenumjiron pair observed in the analysis of low-alloy steel and docu-
ments the analytical performance of the technique.
The use of new radiation detectors is of interest for improving the
analytical performance of GD-OES. The use of silicon photodiode arrays as
detectors in GD-OES was described by Bubert and Hagenah. (\06) They rec-
ognized the strengths of the available arrays for work in the visible and IR
region and determined the alkaline elements as weH as Ba, La, and Tl having
their most sensitive lines in this region. An array with 20 individual diodes
with the corresponding electronics enabled the simultaneous measurement
of the line and background intensities and thus aHowed rapid sequential
analyses. In the spectral region investigated, both detector noise and fiuctua-
tions of the radiation source were found to limit the power of detection.
With the modern arrays which have up to 1024 diodes, the spectral range
a Fe I 379.851 nm
"Mo I 379.825 nm
"
b
379.90 379.85 379.80

Wavelength, nm
Figure 4-30. Iron and molybdenum peaks [rom steel sampies (Research Institute CKD,
Czech Republic) obtained with FTS.(30) i: 50mA; argon pressure: 470 Pa; burning voltage:
1340 V; (a) sampie 218A (CMo: I.3mgjg); (b) sample224A (CMo: 4.1 mgjg); (c) samp1e221A
(CMo: 14.8 mgjg).
covered and the number of pixels available for simultaneous data acquisition
are greatly enhanced. These types of detectors have especially regained inter-
est since spectrally segmented spectrometers have become commercially
available (PRA Inc.). (\07) Here a primary dispersion element with the aid of
an echelle grating is applied and with an exchangeable mask the spectral
regions around the analytical lines of interest are isolated and recombined.
With a second grating the radiation is spectrally resolved again with the lines
and their spectral environment imaged on the photodiode array. In this way,
up to ten analyticallines can be simultaneously detected together with their
spectral background intensities. With the array used, detector noise limita-
tions may still occur, as shown by Hieftje and Brushwyler(\09) for the ICP
and found in initial experiments with the GDS (Fig. 4_15).(26) Here, progress
168 CHAPTER4
may he expected both from the detectors combining a photodiode array and
a microchannel plate as weil as from CCD and CID devices. (110)
4.5.3. Glow Discharges as Atom Reservoirs
Glow discharges with flat cathodes not only are suitable sources for
atomic emission spectrometry but also perform weil as atom reservoirs. Their
use in atomic fluorescence dates back to combinations with a glow discharge
as primary source, as in the resonance fluorescence detector studied by Butler
et al., (1lI) Human et al., (112) and Buhert. (1l3) The full power of detection of
this approach, however, is realized only when laser-induced fluorescence
spectrometry is employed. For lead, detection limits of 20 pg have been
found by Smith et al. (114) when dispensing 5-Jll aliquots on a flat copper
cathode. In hollow cathodes, the feasibility of performing single atom detec-
tion has been described. (115) When soft discharge conditions are used, it is
even possible to detect diatomic molecules. (116) The use of glow discharges
as atom reservoirs for laser-enhanced ionization (LEI)(117) and resonance
ionization mass spectrometry (RIMS)(118) has heen described as weil.
Special interest has been generated for the use of glow discharges as
sputtering cells enabling direct atomic absorption spectrometric analysis of
metals and alloys (see Chapter 3). Gough(89) descrihed the use ofjet-assisted
sputtering to produce high analyte number densities in a glow discharge
plume and determined minor and trace elements in steel [0.08% (w/w) Cr].
Detection limits were shown to be at the 0.0001 to 0.001% (w/w) level.
Larkins(119) improved the gas-control unit and reduced the moisture intro-
duction in the cell by which both the analytical precision and the power of
detection could be further improved. McDonald (120) further discussed the
use and selection of an internal standard to increase the analytical precision.
He also studied the diffusion of analyte atoms in a glow discharge source by
atomic absorption measurements. (121) The technique recently became of
interest when a commercial system was introduced by the Analyte Corpora-
tion. K.im and Piepmeier(122) studied the influence of the gas-jet parameters
on the material ablation and used electron microscopy to investigate the
analyte removal mechanism. Sputtering studies of metal (123) and also of
electrically nonconducting powders(124) for atomic absorption purposes will
promote this technique further, which Ohls et al. (125) have recently shown
to be of use for multielement analyses in steel sampies, including refractory
elements such as Zr. Banks and Blades(126) and Broekaert et al. (130) also
studied the possibilities of the source for optical emission spectrometry. It
was found that despite an enhancement ofthe sputtering rates, the intensities
of Cu(I) resonant lines are not enhanced in contrast to those of Cu(II) and
Zn(I) lines, which may relate to self-reversal.
The experience gained with glow discharges using flat cathodes

as sources for optical atomic spectrometry has also been very valuable
regarding progress in their use as ion sources for mass spectrometry (see
Chapter 5).
4.6. Comparison with Other Methods for the Direct

Analysis of Solids
For bulk analysis, glow discharges have their strength in the material
volatilization mechanism of cathodic sputtering in which the volatilization
interferences are low compared with classical spectrochemical sources such
as arcs and sparks. In comparison with x-ray fluorescence methods, GD-
AES has the advantage of low matrix effects and in addition allows trace
determinations even for the light elements down to the microgram per gram
range. Here, improvement through the use of cross-excitation and side-on
observation still might shift the limits by one order of magnitude. GD-AES
also enables high-precision work from the trace element to the major element
Table 4"8. Comparison of Glow Diseharge Atomie Emission Spectrometry with Other
Methods for the Direct Analysis of Solid SampIes
Bulk analysis
Power of Matrix Analysis
Method detection Precision effects throughput
de are AES + + +a + + ++
Spark AES + ++ + +++
Spark ablation ICP-AES ++ ++ +++ ++
Glow diseharge AES ~ ++ +++ +
Glow discharge mass speetrometry +++ ++ +++ +
Spark mass spectrometry +++ + + +
Laser AES ++ ++ + ++
X-ray fluorescenee + +++ + +++
In-depth profiling
Power of Analysis
Method deteetion Precision Resolution throughput
Secondary ion mass speetrometry +++ +++ +++ +
Auger electron spectrometry + ++ +++ +
Sputtered neutrals mass speetrometry +++ +++ +++ +
Glow diseharge AES ++ ++ + +++
Glow diseharge mass spectrometry +++ +++ +++ ++
a+, poor; + +, good; + + +, very good.
170 CHAPTER4
level. The use of internal standardization allows it to approach the perform-

ance of x-ray fluorescence spectrometry, with the improvement of the spec-
tral information acquisition by multiplex detectors with parallel input
allowing further improvement. Glow discharges, however, are more complex
to operate than the above-mentioned sources. This sterns from their opera-
tion at reduced pressure, which complicates sampie preparation and inter-
change, as weIl as the instrumentation, and decreases the speed of analysis.
The latter has until now hampered the use of GD-AES for product control
in metallurgy. High-energy preburning might lower the analysis times, but
only to a small extent. The strengths of glow discharges for bulk analyses
therefore lie in high-precision product control for a wide variety of sampies.
Particular fields of application here are copper analysis, which at the trace
element level in some cases may present difficulties with modern spark emis-
sion spectrometry, precious metals analysis, and the forthcoming interest in
powder analysis.
With respect to microdistributional analysis, GD-AES offers possibilities
for in-depth profiling, but not for laterally resolved measurements. The latter
is the field of surface scanning techniques such as Auger electron spectro-
scopy, laser ablation in combination with different types of spectrometry
[e.g., mass spectrometry (LAMMS)], proton-induced x-ray emission
(PIXE), electron probe microanalysis (EPMA), secondary ion mass spectro-
metry (SIMS), and scanning Raman spectrometry, which all differ in
principle, lateral resolution, and sampie requirements. (127) GD-AES, how-
ever, enables in-depth profiling with a depth resolution of 1-10 nm, a good
power of detection (down to the trace element level), and a fair precision
(RSD below 10%). Quantization with respect to both the in-depth scale
as well as the determination of in-depth concentration profiles is making
considerable progress. It can be used for a control of thicknesses of coatings
as required by the metallurgical industries producing technical surfaces. It
is quicker and easier to handle than SIMS and has found its way to the
routine industrial laboratories. Here, alternative techniques such as laser
sources enabling reproducible in-depth profile recording also exist. The latter
are also of use for analyses of compact ceramics, by direct emission
spectrometry(I3) as weIl as the more sensitive laser-induced fluorescence
technique. (14) Important progress in this area of work is possible with glow
discharges operating as rf plasmas, enabling a reproducible and sufficiently
intensive ablation of electrically isolating solids. (91,92)
4.7. Conclusions
Optical emission spectrometry with glow discharges is now an estab-

lished method for the direct analysis of solid sampies of widely different
origin and composition. It will further develop together with the use of glow
discharges as atom reservoirs for atomic absorption or atomic fluorescence
spectroscopy and as ion sources for mass spectroscopy.
In particular, the improvements in sampie ablation and excitation as
weIl as in the use of the spectral information are likely to give rise to a wide
diversity of instrumentation for practical spectrochemical analysis and to
contribute to the solution of analytical tasks in the most widely different
fields of work.
Acknowledgments
This study has been supported by the Bundesministerium für Forschung

und Technologie, Bonn and by the Ministerium für Wissenschaft und For-
schung des Landes Nordrhein-Westfalen, Düsseldorf.
References
1. P. W. J. M. Boumans, Theory 01 Spectrochemical Excitation, Hilger & Watts, London,

1966.
2. J. D. lngle, Jr. and S. R. Crouch, Spectrochemical Analysis, Prentice-Hall, Englewood
Cliffs, N.J., 1988.
3. J. A. C. Broekaert, Bull. Soc. Chim. Belg. 86 (1977) 895.
4. J. Junkes and E. W. Salpeter, Ric. Spettrosc. 2 (1961) 255.
5. J. A. C. Broekaert, K. R. Brushwyler, and G. Hieftje, unpublished work.
6. J. A. C. Broekaert, Anal. Chim. Acta 196 (1987) 1.
7. R. Avni, in: Applied Atomic Spectroscopy (E. L. Grove, ed.), Vol. 1, Chap. 4, Plenum
8. K. Slickers, Automatic Emission Spectroscopy, Brühl Druck & Pressehaus, Giessen, 1977.
9. D. M. Coleman, M. A. Sainz, and H. T. Butler, Anal. Chem. 52 (1980) 746.
10. H. Falk, E. Hoffmann, and C. Lüdke, Spectrochim. Acta 39B (1984) 283.
11. H. Falk, E. Hoffmann, C. Lüdke, J. M. Ottaway, and S. K. Giri, Analyst (London) 108
(1983) 1459.
12. L. Moenke-Blankenburg, Loser Microanalysis, Wiley, New York, 1989.
13. F. Leis, W. Sdorra, J. B. Ko, and K. Niemax, Mikrochim. Acta II (1989) 185.
14. W. Sdorra, A. Quentmeier, and K. Niemax, Mikrochim. Acta II (1989) 201.
15. H. Schüler and H. Gollnow, Z. Phys. 93 (1935) 611.
16. S. L. Mandelstam and V. V. Nedler, Spectrochim. Acta 17 (1961) 885.
17. K. Thornton, Analyst (London) 94 (1969) 958.
18. B. Thelin, Appl. Spectrosc. 35 (1981) 302.
19. W. Grimm, Spectrochim. Acta 23B (1968) 443.
20. P. J. Treado and M. D. Morris, Anal. Chem. 61 (1989) 723A.
21. L. M. Faires, Anal. Chem. 58 (1986) 1023A.
22. Y. Talmi, Multichannel Image Detectors, American Chemical Society Symposium Series
102 (1979).
23. J. V. Sweedler, R. F. Jalkian, and M. B. Denton, Appl. Spectrosc. 43 (1989) 953.
172 CHAPTER4
24. G. Heltai, J. A.C. Broekaert, F. Leis, and G. Tölg, Spectrochim. Acta 45B (1990) 301.
25. N. Furuta, K. R. Brushwyler, and G. M. Hieftje, Spectrochim. Acta 44B (1989) 349.
26. J. A. C. Broekaert, K. R. Brushwyler, and G. M. Hieftje, unpublished work.
27. A. Danielson, P. LindbIom, and E. Södermann, Chem. Ser. 6 (1974) 5.
28. R. B. Bilhom, P. M. Epperson, J. V. Sweedler, and M. B. Denton, Appl. Spectrosc. 41
(1987) 1114.
29. M. Dogan, K. Laqua, and H. Maßmann, Spectrochim. Acta 26 (1971) 631.
30. J. A. C. Broekaert, C. A. Monnig, K. R. Brushwyler, and G. M. Hieftje, Spectrochim.
Acta 45B (1990) 769.
31. M. Kaminsky, Atomic and Ionic Impact Phenomena on Metal Surfaces, Springer, Berlin,
1965.
33. C. D. West and H. G. C. Human, Spectrochim. Acta 31B (1976) 81.
36. E. B. M. Steers and F. Leis, J. Anal. At. Spectrom. 4 (1989) 199.
37. A. Aziz, J. A. C. Broekaert, and F. Leis, Spectrochim. Acta 36B (1981) 251.
38. E. B. M. Steers and R. J. Fielding, J. Anal. At. Spectrom. 2 (1987) 239.
39. M. Dogan, K. Laqua, and H. Maßmann, Spectrochim. Acta 27B (1972) 65.
40. R. A. Kruger, R. M. Bombelka, and K. Laqua, Spectrochim. Acta 35B (1980) 589.
41. R. A. Kruger, R. M. Bombelka, and K. Laqua, Spectrochim. Acta 35B (1980) 581.
42. R. M. Lowe, Spectrochim. Acta 31B (1976) 257.
44. F. Leis, J. A. C. Broekaert, and K. Laqua, Spectrochim. Acta 42B (1987) 1169.
45. H. W. Rademacher and M. C. de Swardt, Spectrochim. Acta 30B (1975) 353.
46. S. EI Alfy, K. Laqua, and H. Maßmann, Fresenius Z. Anal. Chem. 263 (1973) I.
47. F. Mai and H. Scholze, Spectrochim. Acta 41B (1986) 797.
48. J. B. Ko and K. Laqua, 18th Colloq. Spectrosc. Int., Grenoble, Conferences et Tables
Rondes, Vol. 11, pp. 543-548, 1975.
49. S. EI Alfy, Ph.D. dissertation, University of Dortmund, 1978.
50. J. A. C. Broekaert, R. Klockenkämper, and J. B. Ko, Fresenius Z. Anal. Chem. 316 (1983)
256.
51. W. B. Bamett, V. A. FasseI, and R. N. Kniseley, Spectrochim. Acta 23B (1968) 643.
53. K. Wagatsuma and K. Hirokawa, Anal. Chem. 56 (1984) 2024.
56. J. Durr and B. Vandorpe, Spectrochim. Acta 36B (1981) 139.
57. J. B. Ko, Spectrochim. Acta 39B (1984) 1405.
59. K. Wagatsuma and K. Hirokawa, Bunko Kenkyu 33 (1984) 320.
60. M. Fujita, J. Kashima, and K. Naganuma, Anal. Chim. Acta 124 (1981) 267.
62. M. Dogan, Spectrochim. Acta 36B (1981) 103.
63. R. A. Kruger, L. R. P. Butler, C. J. Liebenberg, and R. G. Böhmer, Analyst (London)
102 (1977) 949.
64. T. Yamada, J. Kashima, and K. Naganuma, Anal. Chim. Acta 124 (1981) 275.
65. C. J. Belle and J. D. Johnson, Appl. Spectrosc. 27 (1973) 118.
66. R. Bemeron, Spectrochim. Acta 33B (1978) 665.
67. A. Bengtson, Spectrochim. Acta 40B (1985) 631.
68. R. Bemeron and J. C. Charbonnier, Anal. Proc. 17 (1980) 488.
69. A. Quentmeier and K. Laqua, in: 13. Spektrometertagung, Düsseldorf(K.-H. Koch and
H. Massmann, eds.), pp. 37-49, de Gruyter, BerJin, 1981.
70. M. E. Waitlevertch and J. K. Hurwitz, Appl. Spectrosc. 30 (1976) 510.
71. P. De Gregorio and G. Savastano, Ann. Ist. Super. Sanita 19 (1983) 613.
72. K. Takimoto, K. Suzuki, K. Nishizaka, and T. Ohtsubo, Quantitative Analysis of Zinc-
Iron Alloy Galvanized Coatings by Glow Discharge Spectrometry and Secondary Ion Mass
Spectrometry, Nippon Steel Technical Report No. 33 (1987).
73. K.-H. Koch, M. Kretschmer, and D. Gruenenberg, Arch. Eisenhüuenwes. 54 (1983) 395.
74. K.-H. Koch, D. Sommer, and D. Grünenberg, Mikrochim. Acta 11 (1985) 1.
76. Y. Ohashi, Y. Yamamoto, K. Tsunoyama, and H. Kishidaka, Surf Interface Sei. 1 (1979)
53.
77. F. Leis, J. A. C. Broekaert, and E. B. M. Steers, 1990 Winter Conference on Plasma
Spectrochemistry, St. Petersburg, Fla., Paper WP18.
78. P. A. Büger and W. Fink, Fresenius Z. Anal. Chem. 244 (1969) 314.
79. W. W. Harrison and N. J. Prakash, Anal. Chim. Acta 49 (1970) 151.
80. J. Mierrzwa and W. Zyrnick, Anal. LeU. 21 (1988) 115.
81. J. A. C. Broekaert, Spectrochim. Acta 35B (1980) 225.
82. J. Y. Ryu, R. L. Davies, J. C. Williams, and J. C. Williams, Jr., Appl. Spectrosc. 42 (1988)
1379.
83. J. M. Brackett and T. J. Vickers, Spectrochim. Acta 37B (1982) 841.
84. N. Jakubowski, D. Stüwer, and G. Toelg, Spectrochim. Acta 46B (1992) 155.
85. H. Jäger and F. Blum, Spectrochim. Acta 29b (1974) 73.
86. I. Brenner, K. Laqua, and K. Dvorachek, J. Anal. Atm. Spectrom. 2 (1987) 623.
87. J. A. C. Broekaert, T. Graule, H. Jenett, G. Tölg, and P. Tschöpel, Fresenius Z. Anal.
Chem. 332 (1989) 825.
88. G. S. Lomdahl, R. McPherson, and J. V. Sullivan, Anal. Chim. Acta 148 (1983) 171.
89. D. S. Gough, Anal. Chem. 48 (1976) 1926.
90. G. S. Lomdahl and J. V. Sullivan, Spectrochim. Acta 39B (1984) 1395.
92. M. R. Winchester and R. K. Marcus, J. Anal. At. Spectrom. 5 (1990) 575.
93. D. C. McDonald, Spectrochim. Acta 37B (1982) 747.
94. D. S. Gough and J. V. Sullivan, Analyst (London) 103 (1978) 887.
95. P. E. Walters and H. G. C. Human, Spectrochim. Acta 36B (1981) 585.
96. S. Caroli, A. Alimonti, and F. Petrucci, Anal. Chim. Acta 136 (1982) 269.
97. S. Caroli, O. Senofonte, N. Violante, and L. Di Simone, Appl. Spectrosc. 41 (1987) 579.
98. F. Leis, J. A. C. Broekaert, and E. B. M. Steers, Spectrochim. Acta 46B (1991) 243.
99. D. C. Liang and M. W. Blades, Spectrochim. Acta 44B (1989) 1049.
100. L. M. Faires, Spectrochim. Acta 40B (1985) 1473.
101. K. Laqua, W.-D. Hagenah, and H. Waechter, Fresenius Z. Anal. Chem. 225 (1967) 142.
102. B. A. Palmer, Los Alamos Technical Bulletin, Los Alamos National Laboratory, Los
Alamos, N.M., August 1985.
103. M. L. Parsons and B. A. Palmer, Spectrochim. Acta 43B (1988) 75.
104. J. A. C. Broekaert, F. Leis, and K. Laqua, in: Developments in Atomic Spectrochemical
Analysis (R. M. Barnes, ed.), pp. 84-93, Heyden, London, 1981.
105. P. W. J. M. Boumans and J. J. A. M. Vrakking, Spectrochim. Acta 41B (1986) 1235.
106. H. Bubert and W.-D. Hagenah, Spectrochim. Acta 36B (1981) 489.
107. G. M. Levy, A. Quaglia, R. E. Lazure, and S. W. McGeorge, Spectrochim. Acta 42B
(1987) 341.
108. J. A. C. Broekaert, J. Anal. At. Spectrom. 2 (1987) 537.
109. G. M. Hieftje and K. R. Brushwyler, Appl. Spectrosc. 45 (1991) 582.
174 CHAPTER4
110. G. R. Sims and M. B. Denton, Opt. Eng. 26 (1987) 1009.

111. L. R. P. Butler, K. Kroger, and C. D. West, Spectrochirn. Acta 30B (1975) 489.
112. H. G. C. Human, N. P. Ferreira, R. A. Kruger, and L. R. P. Butler, Analyst (London)
103 (1978) 469.
113. H. Bubert, Spectrochirn. Acta 39B (1984) 1377.
114. B. W. Smith, N. Omenetto, and J. D. Winefordner, Spectrochirn. Acta 39B (1984) 1389.
115. B. W. Smith, J. B. Womack, N. Omenetto, and J. B. Winefordner, Appl. Spectrosc. 43
(1989) 873.
117. T. Masaki, Y. Adachi, and C. Hirose, Appl. Spectrosc. 42 (1988) 51.
118. P. J. Savickas, K. R. Hess, R. K. Marcus, and W. W. Harrison, Anal. Chern. 56 (1984)
819.
119. P. L. Larkins, Anal. Chirn. Acta 132 (1981) 119.
122. H. J. Kim and E. H. Piepmeier, Anal. Chern. 60 (1988) 2040.
123. S. L. Tong and W. W. Harrison, Anal. Chern. 56 (1984) 2028.
125. K. Ohls, J. Flock, and H. Loepp, Fresenius Z. Anal. Chern. 332 (1989) 456.
126. P. R. Banks and M. W. Blades, Spectrochirn. Acta 44B (1989) 1117.
127. W. A. Van Borm and F. C. Adams, Anal. Chirn. Acta 218 (1989) 185.
128. J. B. Ko and K. Laqua, in: 13. Spektrometertagung Düsseldorf (K.-H. Koch and H.
Maßmann, eds.), pp. 51-68, de Gruyter, Berlin, 1981.
129. K. Laqua, in: UI/rnanns Enzyklopädie der technischen Chemie, 4th ed., vol. 5, p. 460, Verlag
Chemie, Weinheim, 1980.
130. J. A. C. Broekaert, T. Bricker, K. R. Brushwyler, and G. M. Hieftje, Spectrochirn. Acta
47B (1992) 131.
5
Glow Discharge
Mass Spectrometry
F. L. King and W. W. Harrison
5. 1. Introduction
Glow discharge mass spectrometry (GDMS) is associated with two c1ose1y

related techniques for the characterization of ion populations in glow dis-
charge plasmas. The materials scientist is familiar with GDMS as a tool for
plasma processing diagnostics, whereas the analytical chemist is familiar
with GDMS as a method for direct solids analysis. The present chapter will
focus on GDMS as a maturing technique in elemental analysis. The reader
interested in GDMS as a plasma diagnostic tool is referred to reviews ofthe
topic by Aita(l) and Coburn.(2)
The relative simplicity of mass spectra compared with optical spectra
makes mass spectrometry an attractive alternative to optical spectrometry
for trace element analysis. By providing isotopic abundance information,
mass spectrometry permits the use of isotope dilution techniques and isotope
tracer studies. The ability to generate a stable analyte ion population directly
from asolid sampie, thereby prec1uding the problems of dissolution, dilution,
and contamination, attendant to techniques requiring solution sampies,
makes the glow discharge an attractive ion source for e1emental mass spec-
F. L. King. Department of Chemistry, West Virginia University, Morgantown, West

Virginia 26506. W. W. Harrison • Department of Chemistry, University of Florida,
Gainesville, Florida 32611.
Glow Discharge Spectroscopies, edited by R. Kenneth Marcus. Plenum Press, New Y ork, 1993.
175
176 CHAPTER 5
trometry. Coupling a glow discharge atomization/ionization source with a

mass spectrometer yields a state-of-the-art technique for the direct elemental
analysis of solid-state materials, GDMS.
The division of the chemical analysis process into three distinct steps,
representation, speciation, and quantitation, (3) permits a parallel division
of the functions performed in GDMS, i.e., atomization/ionization, mass
analysis, and data acquisition/interpretation. In this chapter, we seek to
develop both an understanding of the fundamental principles underlying
these functions and an appreciation of the broad analytical potential
presented by GDMS. The discussion focuses on general principles rather
than any ofthe various turnkey and add-on systems available from commer-
cial vendors.(4) Initially, the principles underlying the operation of various
mass analyzers will be reviewed. The merits of each type of mass analyzer
for use in elemental mass spectrometry will be evaluated. Atomization and
ionization processes in glow discharges will be discussed in the context of
their impact on quantitative interpretation of GDMS data. The reader inter-
ested in a complete description of the atomization and ionization processes
is referred to other chapters of this text where these processes are described
in detail. The configurations and attributes of the various glow discharge
ion source configurations will be described. A sampier of the analytical
results obtained with GDMS will conc1ude this chapter.
5.2. Mass Spectrometry
Traditionally, elemental analysis has been dominated by optical spec-

trometric techniques, such as atomic emission spectrometry and atomic
absorption spectrometry, that probe the transitions between energy levels
characteristic of an analyte element. Mass spectrometry originated as a tool
for the characterization of ions present in gas discharge plasmas. The desire
to extract further information conceming the nature of matter from "positive
rays" emitted by electric discharges led Thompson, Aston, and Bainbridge
to develop a new type of spectrograph. (5) These mass spectrographs separ-
ated the components of these "positive rays" and provided direct informa-
tion regarding the distribution of isotopic constituents in these discharge
plasmas. Unlike the complex optical spectra examined previously, these new
mass spectra were interpreted readily because each elemental isotope yielded
only a single line at a characteristic mass-to-charge ratio (m/z).(5) Charac-
terization of an atom on the basis of its mass-to-charge ratio leads to an
immediate simplification; because there are only a few different isotopes of
any given element, and in general only the singly ionized species will be
detectable under the conditions employed, the resulting mass spectrum will
be less complex and isobaric interferences will be reduced. It is tbis spectral
GLOW DISCHARGE MASS SPECTROMETRY 177
simplicity coupled with the sensitivity and dynamic range of modern ion
detection techniques that makes mass spectrometry a powerful technique for
elemental analysis.
5.2.1. Fundamentals
Although mass spectrometry is fundamentally different from other ana-

lytical spectrometric techniques in that the absorption, or emission, of radia-
tion is not necessarily the physical phenomenon by which the analytical
species are distinguished, * there are certain analogies from which much of
the terminology of mass spectrometry is derived. We begin with a source
of ions, rather than photons, that are characteristic of the analyte. The
composition of the ion beam emitted by the source should be a steady and
reproducible reflection of the original sampie composition. Ion optics serve
to collimate and condition the ion beam optimizing the transmission of the
beam into the mass analyzer. Mass analyzers permit the temporal or spatial
isolation of different m/z ions in the ion beam much the same as a wave-
length-selective device separates photons of different energy in a beam of
light. The beam emerging from the mass analyzer is then of a single m/z
with an intensity that is characteristic of the isotopic composition of the
original sampIe. The beam is detected either by direct measurement of the
ion current or by conversion to e1ectrons or photons that are amplified within
the detector prior to current measurement. A plot of the signal intensity as
a function of transmitted m/z value is the mass spectrum. Mass spectral
resolution, (m/z)/(Am/z), is a measure ofthe ability to separate ions having
small m/z differences just as optical spectral resolution is a measure of
the ability to separate photons having small energy differences. The signal
recorded in mass spectrometry is a direct measure of the ion flux at aselected
m / z; as in optical spectrometry it is a measure of the photon flux at aselected
energy.
5.2.2. Mass Analysis

Mass analyzers can be c1assed loosely into three general categories:
continuous dispersion instruments, sequential filtration instruments, or ion
trapping instruments. At the present time, sector or quadrupole-based instru-
ments are the only commercially available GDMS instruments although the
utility of other mass analyzers is under investigation. A general comparison
of mass analyzers for GDMS appears in Table 5-1.
*Photoselective ionization techniques, such as RIMS, actually employ c1assical spectroscopy in

concert with mass spectrometry to etfect a two-dimensional separation of species.
178 CHAPTER 5
Table 5-1. Comparison of Mass Analyzers for Elemental MS

Mass Analyzer Advantages Disadvantages
Double focussing >Resolution adequate to >Complexity
sector discriminate against some
isobaric interferences
>Commercial systems available >Cost
>Scan speed
Time-of-flight >Mass analyzer simplicity >Complicated ion extraction
>Speed >No results yet
>Good resolution with use of
reflection
Quadrupole mass >Scan speed > Limited resolution
filter
>Peak hopping
>Commercial systems available
Quadrupole ion trap >In situ CID to remove >Problems attendant to
polyatomic isobaric external ion generation
interferences
FT-ICR >High resolution >Cost
>CID possible >Complexity
>Space charge limited
dynamic range
>Problems attendant to
external ion generation
5.2.2.1. Dispersion in Space (Sector MS)
The prototype dispersion instrument was Aston's original mass spectro-

graph, which employed an electromagnetic field to disperse different m/z
ions spatially.(5) The descendants of this instrument are the sector mass
analyzers employed today. Sector instruments employ electrostatic and
magnetic fields to effect ion separation. The mathematics essential to the
explanation of ion trajectories in such fields is relatively straightforward. A
magnetic field acts as a prism, dispersing monoenergetic ions of differing
m/z values across a focal plane (Fig. 5_1).(6) The radius of the curved flight
path of an ion through a magnetic field is given by
r m = (144/B)(mV/z)I/2
where B is the field strength in gauss, m is the atomic mass of the ion, V is
the voltage with which the ion is accelerated into the magnetic field, and z
is the charge on the ion. (6) Such instruments can provide a simultaneous
mass spectrogram of all the species of interest or can scan a mass spectral
region of interest.
Two modes of detection can be employed with sector instruments: (1)
simultaneous detection of a range of m/z ions by a photographic plate or
array detector located in the focal plane or (2) sequential detection by an
Figure 5-1. Magnetic sector mass analyzer.
electronic detector on which different m/z ions are brought into focus by
scanning B or V. Resolution is determined by ion beam focusing in the
detection plane. Two types of focusing are possible with sector instruments:
direction and velocity focusing. In single-focusing instruments, increased
resolution is gained by narrowing the entrance and exit slits.(6) Unfortun-
ately, such instruments depend on the availability of an ion beam having a
very narrow ion energy spread to prevent the occurrence of aberrations
arising from a lack of energy focusing. More complex instruments eliminate
aberrations resulting from wide ion energy spreads by employing double
(velocity and direction) focusing.(6) In such instruments, resolution is deter-
mined largely by the width of the entrance slit, which determines the physical
size of the ion beam image in the detection plane. In both, increased resolu-
tion comes at the expense of ion beam attenuation, resulting in decreased
sensitivity.
The first commercially available GDMS system employed a double-
focusing sector mass analyzer.(7) The practical resolution attainable with
such a mass analyzer is on the order of 4000 while providing suitable sensitiv-
ity to the ppm level. The system was equipped with both a Faraday detector
for matrix and discharge gas ion signals and a Daly-type detector for trace
constituents. An automatie detector switching algorithm permitted a
dynamic range of nine orders of magnitude during a single scan. The
complexity and cost of these instruments limited the consideration of GDMS
to a small segment of the analytieal community whose specialized needs
warranted the purchase of such a system.
5.2.2.2. Dispersion in Time (Time-of-F/ight MS)
Perhaps the simplest form of mass spectrometry is time-of-flight mass

spectrometry (TOFMS). Recalling that kinetic energy, E, is a function of
mass, m, and velocity, v; or charge, z, and acceleration potential, V,
E= mv2 /2 = zV
180 CHAPTER 5
it is obvious that monoenergetic ions of different mjz accelerated to the same

kinetic energy will have different velocities. A TOF instrument operates on
the principle that if all ions exit the extraction lens with equal kinetic energies,
they will travel with different velocities, vx , proportional to the reciprocal
square root of their respective masses, mx :
The time required to traverse the flight distance D and to arrive at the
detector is related to the mj z of the ion (Fig. 5-2) :(8)
t = D(mj2 VZ)\,2
Monitoring the current at the detector yields a time-dependent signal. Con-

version of the time axis in the intensity versus time plot to mjz values yields
a mass spectrum. The mass resolution is determined directly by the temporal
resolution, At j t, which is limited by the initial energy spread of the ions and
the speed of the detection electronics.
The TOF mass analyzer is suited to short pulses of ion signal such as
generated by pulsed laser desorption or pulsed ion beam sputtering. Second-
ary ion mass spectrometry employing TOF mass analysis(9) has developed
into a successful technique for the characterization of a variety of materials.
The potential applicability for TOF with the glow discharge is related to the
use of either a pulsed glow discharge ion source or resonance-enhanced
photoionization of neutrals by a pulsed laser (RIMS employing the GD as
an atom reservoir).
EXTRACTIOO
illS IBECTOO
c , , "z~ t/zz
FL um DISTANI ~I
Figure 5-2. Time-of-flight mass analyzer.
5.2.2.3. Selective m/z Transmission (Quadrupole Mass Filters)
Whereas a sector mass analyzer is analogous to an optical monochrom-

ator, the quadrupole mass analyzer (QMA) is a variable bandpass filtering
ion optic analogous to an optical bandpass filter. These devices are capable
of transmitting ions of all masses simultaneously, or of transmitting ions of
a selected mass window only.(IO) By nature, they are capable of only the
sequential measurement ofvarious isotopes; however, their advantage over
magnetic sector analyzers lies in the ability to scan very rapidly or to "peak-
hop" among aseries of isotopes. (11)
The QMA consists of four cylindrical rods arranged in a square array
to which a combination of dc and rf potentials is applied (Fig. 5_3).<'2) For
optimal approximation of hyperbolic electric fields about each rod, the case
ofthe ideal quadrupole, the ratio ofthe rod radius, r, to the radius enclosed
by the electrodes, ro, should equal a constant value :(13)
r/ro = 1.148
In order to provide a qualitative feel for the operation ofthe quadrupole

mass filter, the treatment of this subject by Miller and Denton(8) will be
followed. When utilized as a mass spectrometer, the quadrupole mass filter
operates as a bandpass filter created by the combination of a high-pass filter
and a low-pass filter (Fig. 5_4).(8) The quadrupole field is established by
the application of potentials, dc and rf, to the rour quadrupole electrodes.
Opposite pairs of electrodes are electrically connected and lie in the xz or yz
planes of the coordinate diagrams as shown in Fig. 5-4. The xz electrode
pair have a positive dc potential applied along with an rf potential, while
Figure 5-3. Quadrupole mass filter. Reprinted with permission from R. E. March and
R. J. Hughes, Quadrupole Storage Muss Spectrometry, p. 43, copyright 1989, Elsevier Science
Publishers.
182 CHAPTER 5
z
Q
VI
!!?
~
VI
z
«
0:
I- L -_ _~==
-+
MASS
B
6 "====
Vi
VI
~
VI
z
«
Cl:
I-
-+
MASS
6 r - - - - , r--------
Vi
VI
~
Vl
z
«
~ ~_ _~~L__ _ __
-+
MASS
Figure 5-4. (A) High-pass mass filter; (B) low-pass mass filter; (C) bandpass mass filter.
Reprinted with permission from P. E. Miller and M. B. Denton, J. ehern. Educ. 63 (1986) 619,
copyright 1986, Division of Chemical Education, American Chemical Society.
the yz e1ectrode pair have a negative dc potential applied along with an rf

potential that is 1800 out of phase with that applied to the xz pair. Ions are
injected into the field along the z axis. In the xz plane, ions of larger masses
are focused along the z axis by the positive dc field, whereas ions of smaller
masses are subject to destabilization by the rf field. In the yz plane, ions of
larger masses are deflected away from the z axis by the negative dc field.
Ions of smaller masses in the yz plane will be stabilized by the ac field. The
net result is the coexistence of a high-pass mass filter in the xz plane, A of
Fig. 5-4, along with a low-pass mass filter in the yz plane, B of Fig. 5-4,
creating a narrow bandpass mass filter, C of Fig. 5-4.
A more rigorous quantitative explanation of quadrupole mass filtering
employs equations of motion for a charged partic1e in an oscillating electric
fie1d.(I3) The electrical potential, <1>, ofthe quadrupole fie1d at any given time,
t, is given by the equation
<I> = [U + V cos(mt)](x2 -l)/21o
where U is the dc potential, V is the rf magnitude, m is the angular frequency

(21Cf) of the rf potential, x and y are the linear displacements along the
respective axes, and ro is the radius encompassed by the four electrodes that
encirc1e the z axis. The partial derivatives of this equation yield the electric
field magnitudes, Ei, along the i axis. Multiplication of the partial derivative
by the partic1e charge yields the force, Ei, applied to the partic1e along the
i axis. From this, equations can be derived that describe the path of any ion
through the field. Parameters a, q, and ifJ
a = 4zU/mm2rö
q = 2z V/mm2r~
ifJ = mt/2
can be substituted yielding a Mathieu differential equation of the form
which is solved for values of a and q yielding u, such that 0 < u < ro. Under
these conditions the ion is never outside of the radius bounded by the quad-
rupole rods along axis x or y and its trajectory is said to be stable. In Fig.
5_5,<'4) these stable ion trajectory conditions are met for those values of a
and q contained under the curve. Imposed upon this stability diagram is
the operational condition that results in the mass filtering ability of the
quadrupole: the ratio between dc and rf potentials is set to a constant value,
a/qoc U/V= constant. The resulting mass scan line intersects the region of
184 CHAPTER 5
R = 100
y-z motion
unstable
a R =1
.1
o 1.0
Figure 5-5. Quadrupole mass filter stability diagram. Reprinted with permission from J. E.
Campana, Int. J. Mass Spectrom. Ion Phys. 33 (1980), 108, copyright 1980, Elsevier Science
Publishers.
a-q space that eorresponds to a stable ion trajeetory, and only ions with
those values of m/z that fall in this region ean sueeessfully traverse the
quadrupole. As the potentials U and V are ramped, the m/z values falling
within this window are ehanged. The width of this window, and thus the
mass resolution, is determined by the slope of the mass sean line.
Now that eommereial vendors have begun marketing quadrupole-based
systems, GDMS should begin to reeeive the same level of aeeeptance and
use as ICPMS has enjoyed. The principallimitation faeed by these systems
is their inability to resolve analyte isotopes from overlapping polyatomie
interferenees. Judieious ehoice of operating eonditions and the implementa-
tion of eollision-indueed dissoeiation ean minimize or remove interferenees
arising from polyatomic ionsY S)
5.2.2.4. Selective m/z Ejection (Quadrupole Ion Traps)
Quadrupole ion traps are similar to quadrupole mass filters in their

operation. Essentially, two opposing rods in the QMA are eonneeted to form
a ring and the remaining two rods are eonverted into hyperbolie end eaps,
(Fig. 5_6).(16) The result is a three-dimensional quadrupole field that is sym-
metrie with respeet to rotation about the center. The end eaps are taken to
be along the former z axis and the former x and y axes now beeome a plane,
r, symmetrie about z. The a and q parameters employed in deseribing the
quadrupole mass filter are modified to fit into the same Mathieu equation(13) :
a = -8zU/mw2r~
q = -4z V/mw 2ro
* __ Electron
source
Figure 5-6. Quadrupole ion trap. Reprinted with permission from R. E. March and R. J.
Hughes, Quadrupole Storage Mass Spectrometry, p.4, copyright 1989, Elsevier Science
Publishers.
The resulting stability diagram is shown in Fig. 5_7.0 6 ) Mass analysis is

achieved by scanning through the regions of ion instability, as opposed to
scanning ion stability with the QMA.(17) Ions become unstable and are
ejected selectively from the trap through orifices in one of the end caps. An
electron multiplier located outside of the trap then detects the ejected ions.
Quadrupole ion traps are smalI, relatively simple, and comparatively in-
expensive. Although resolution of singly charged ions is limited in the trap,
ion traps permit the use of collision-induced dissociation to remove poly-
atomic interferences. The ability to store particular ions for long periods
of time permits the effective integration of low-abundance ions with trap
technology. The limitation at the present time is the efficiency with which
ions can be admitted to the trap from external sourees.
The use of a glow discharge ion source combined with a quadrupole
ion trap mass spectrometer (QITMS) was employed for atmospheric sam-
pling by McLuckey et al. (18) The principal aim was the determination of
airborne trace organic species. Hemberger and colleagues investigated the
utility of a GD ion trap instrument for field elemental analysis.(19) Further
results from investigations of GDMS using QITMS are awaited anxiously.
5.2.2.5. Fourier Transform Ion Cyc!otron Resonance MS
Ion trapping within a cubic cell by a combination of electric and mag-

netic fields is also employed as a mass spectrometric method. In the presence
Figure 5-7. Quadrupole ion trap stability diagram. Reprinted with pennission from R. E.
March and R. J. Hughes, Quadrupole Storage Mass Spectrometry, p. 118, copyright 1989,
Elsevier Science Publishers.
of a strong magnetic field, ions can be trapped in circular orbits with the
application of low voltages. The ions orbit perpendicular to the applied
magnetic field at characteristic cyelotron frequencies, OJ c , that are inversely
proportional to their rn/z values and proportional to the magnetic field
strength, B (Fig. 5-8) YO)
OJ c = zB/rn
Application of an rf-voltage pulse to the transmitter plates of the cubic cell

brings the orbiting ions into coherent motion. The ions are then detected as
image currents generated at the receiver plates of the cello This image current
arises from the field induced as the ions come into elose proximity of the
receiver plate. The image currents rise and fall with the orbit of the ions
yielding a transient that can be Fourier transformed to give the characteristic
frequencies of stored ions and the relative signal contribution at each
frequency.
Workers at IBM and the University of Wisconsin first reported FT-
ICR MS investigations of an rf glow discharge plasma in 1989.(21) Although
Magnetic
field
Trapping plate ~
Collector --cJ ()
I I
I I
I I
. - I o n source
/' '---
/'/' .,---
/' /'
/'
Receiver
"------:--f-----/
~
Transminer plate
Figure 5-8. Ion cyc1otron resonance cell employed in Fourier transforrn mass analysis.
Reprinted with perrnission from H. H. Willard, L. L. Merrit, J. A. Dean, and F. A. Settle,
Instrumental Methods of Analysis, 7th ed., p. 500, copyright 1988, Wadsworth Publishing.
the interest of these investigators was the ion-molecule chemistry occurring

in the deposition plasma, there are compelling motives for the extension of
this work to e1emental analysis. The resolution attainable with FT-ICR MS
can readily solve the problem of isobaric interference in GDMS. As with
QITMS, CID can also be employed to remove polyatomic interferences.
Recently, Barshick and Eyler(22) reported on the use of FTMS coupled with
an external glow discharge source for e1emental analysis. The observed detec-
tion was limited by the transport of ions from the external source into the
ICR cell of the mass spectrometer. Limits to dynamic range due to space
charging effects were minimized by ejection of the major ions prior to data
acquisition.
5.2.3. Mass Spectrometry of Glow Discharge Plasmas
The operating pressure of glow discharge ion sources is greater than

that tolerated by any of these mass analyzers. This necessitates differential
pumping between the ion source region and the mass analyzer region. The
extent of differential pumping will be a function of the type of mass analyzer
employed, the ion trap requiring the least differential pumping, the FT-ICR
MS requiring the most. Typically three vacuum regions are employed: the
glow discharge source at ~l Torr, the intermediate vacuum region at
~IO-4 Torr, and the analyzer region at 10-6 Torr. In order to maintain
analytical signals, ion optics are required to transport ions efficiently through
the differential pumping region. Often, additional ion optics that serve to
188 CHAPTER 5
limit the spread of ion kinetic energy at the mass analyzer entrance are
employed. (23) The mass analyzer provides temporal or spatial distribution of
ions of differing m/z. Except for the case of the FT-ICR MS, mass-selected
ions are detected direct1y with a Faraday cup for major species or with ion
conversion and electron multipliers for trace elements. The double-focusing
instruments use a Daly type of detector where ions are converted to electrons
that are converted to photons and detected with a photomultiplier tube. (24)
This detection scheme is less susceptible to degradation caused by vacuum
impurities or mishaps. Ion counting can be employed to reduce noise and
improve analytical signal. Data acquisition for elemental mass spectrometry
differs from that for organic mass spectrometry in two respects: the widest
possible dynamic range is desired permitting the observation of major and
trace sampIe constituents in the same mass spectrum and sources of back-
ground noise that may be tolerated in organic mass spectrometry but would
obscure the signal for trace sampIe constituents must be reduced
significantly.
5.3. Atomization/lonization and Quantitation
We are treating representation and quantitation together because they

are inextricably interwoven by the differential sensitivity among elements
resulting from the atomization and ionization processes. Most quantitative
protocols in GDMS relY on the use of relative sensitivity factors (RSFs),
sometimes referred to as relative ion yields (RIYs). The RSF of an analyte
element, A, is the ration of its sensitivity (intensity, I, per unit concentration,
C) to the sensitivity of some reference element, R:
These RSFs contain contributions arising from instrumental factors, such

as ion transmission and sensitivity, and glow discharge processes, such as
differential atomization and differential ionization. (25) Whereas the contribu-
tion arising from instrumental factors is generally minimal and uniform for
a given system, the contributions arising from glow discharge processes are
dominant and vary from sampIe to sampIe. Obviously, an understanding of
the atomization and ionization processes aids the analyst in evaluation of
RSF values. Atomization and ionization will be discussed in the context of
their influence on RSFs.
5.3.1. Glow Discharge Processes
As an ion source for elemental mass spectrometry, the glow discharge

is characterized by two attractive attributes, cathodic sputtering and Penning
ionization, that are inherent to its operation. Cathodic sputtering generates

a representative atomic population directly from the solid sampie. Penning
ionization selectively ionizes these sputtered atoms permitting speciation on
the basis oftheir characteristic mass-to-charge rations by mass spectrometry.
5.3.1.1. Atomization
At the heart of the analytical glow discharge lies the process of cathodic
sputtering. Ions accelerated across the cathode dark space impinge on the
surface of the sampie. These projectiles transfer their kinetic energy to the
atoms in the sampie lattice through a collision cascade. If the energy
imparted to an atom in the lattice exceeds its lattice binding energy, the
atom is released. Excess energy above the binding energy can result in the
release of the atom in an excited electronic state or as the ion. The electric
field in the vicinity ofthe cathode leads to the redeposition ofthese secondary
cations onto the cathode surface. Electrons or anions are accelerated into
the negative glow by this same field to sustain the discharge.
The number of target atoms ejected per ineident projectile or sputter
yield, S, is a function of the energy deposited in the collision cascade. Under
the operating conditions of most analytical glow discharges the sputter yield
can be described as a function of the projecti1e kinetic energy, E, projectile
mass, mp, lattice binding energy, Uo, and target atom mass, mt:(26)
S = 3ampmtE/n2(mp + mt)2Uo
For mt/mp = 0.1, a = 0.17 and for mt/mp = 10, a = 1.4
(This equation is valid for analytical glow discharges where E is less than
1000 eV; Eis typically less than half of the discharge operating potential. (27»)
A related value is the sputtering rate, the number of target ions sputtered
per unit time. This value is determined by the discharge operating current
as weIl as the factors mentioned for sputter yields.
From this information it is important to identify those GDMS experi-
mental variables that influence atomization and thereby influence RSFs.
Obviously, the target matrix material is variable from experiment to experi-
ment. For reasons related to ionization, argon is considered to be the best
choice for discharge support gas. Thus, the influence of projectile identity is
normalized for all species. For analytical work, glow discharge operation in
a constant-power mode is preferred and E is considered to be constant
during the time of analysis. The only variable over which the analyst cannot
exercise control is the identity of the target atom. When one element domi-
nates the matrix and all other elements are present in trace amounts, it can
be assumed that sputtering behaves as for a pure material. Thus, there should
190 CHAPTER 5
be little difference in RSFs determined in a similarly pure matrix as reported

by Vieth and Huneke. (28) With alloys containing large percentages of several
elements, it is necessary to match the concentrations of the major constit-
uents closely, as determined by a semiquantitative GDMS analysis before
determining RSF values. The influence of variation in trace constituents on
sputter atomization is minor. It is in ionization where such variations are
no longer subtle.
5.3.1.2. lonization
In general, GDMS does not utilize optical transitions of the analyte

atoms; rather, the mass-to-charge ratio of atoms that have been ionized in
the plasma affords the speciation of analytes. This fact shifts the emphasis
from excitation mechanisms in general to ionization mechanisms specifically.
In some ways this simplifies the relationship between analyte signal and
analyte concentration in the sampIe. But if we assume that the atomization
does not differ strongly between elements in a given matrix, we cannot
assume the same to be true for ionization. RSFs are most likely controlled
by differences in the probability of ionization among the elements.
There exist many processes that contribute to ionization in glow
discharges(26) (Table 5-2). These processes may be characterized as collisions
of the first kind in which kinetic energy is transferred to an atom resulting
in ionization or collisions of the second kind in which potential energy is
transferred to an atom resulting in ionization. (29) The relative importance
of each mechanism will vary with discharge operating conditions, current,
voltage, pressure, and geometry. The Penning process is the dominant ioniza-
tion process occurring in the coaxial cathode-type glow discharge ion sources
operating at 0.4-2.0 Torr, 1-1.5 kV, and 0-10 mA.(30) Langmuir probe
measurements indicate that the electron population in glow discharges con-
sists of three types of electrons in a non-Maxwell-Boltzmann distribution:
fast (E > 25 eV), secondary (E ~ 7 eV), and ultimate (thermal).(31) The most
Table 5-2. Ionization Processes in Glow Discharge

Collisions of the first kind
E1ectron ionization
Collisions of the second kind

Penning ionization Arm+X-.Ar+X+ +e-
Associative ionization Arm+X-.ArX+ +e-
Symmetrie charge exchange A + +A->A+A +
Asymmetrie charge exchange A + +B-.A+B+
energetic of these are localized in the cathode fall region of the glow dis-
charge, whereas ultimate electrons, with energy less than that for optimum
electron ionization cross sections, are the majority species in the negative
glow. In hollow cathode and Grimm-type glow discharges, charge exchange
contributes significantly to sputtered atom ionization. (32) Because Penning
ionization in argon glow discharges uniquely results in the ionization of
atoms in a manner that is relatively uniform among elements and selective
with respect to polyatomic species, ion sources in which this process domi-
nates are preferred.
5.3.2. Quantitation
In the ideal GDMS analysis, the ion signals for all ionized sputtered
species are summed and then the ration of the ion signal for individual
species is calculated. This ratio then corresponds to the concentration of the
species in the bulk. This ion-beam ratio (IBR)(33) method of quantitation
would be an absolute, i.e., standardless, method if there were no difference
in sensitivity among the elements. Of course, sensitivity is not constant
among the elements, and generally varies by less than a factor of ten; there-
fore, the IBR method provides a straightforward means for reliable
semiquantitative analysis. Quantitative analysis is achieved by correcting
IBR values employing RSFs determined for a set of standards having the
same general matrix as the sampIe under investigation. (34)
An alternative method of calibration has been reported recently by
Klingler and Harrison. (35) They employ a source having a sampIe cathode
and a reference cathode. The discharge potential is alternately applied to
these cathodes and mass spectra for both are obtained. The ratio of sample-
to-reference signals for a given isotope yields a direct calibration. This
method offers the potential to eliminate the need for RSF values.
5.4. Glow Discharge Ion Sources
A wide variety of analytical glow discharge geometries have been inves-

tigated as ion sources. In the early 1970s, investigators were interested in
monitoring glow discharge plasmas employed in thin-film deposition. For
applications in microelectronics, high-purity thin films were required. This,
in turn, required that the deposition plasma and sputter target also be of
the highest quality. The group at IBM headed by Coburn used mass spec-
trometry initially to characterize their dc- and rf-powered diode glow
discharge deposition plasmas. (36) They so on realized that the mass spec-
trometric monitoring of the ions produced in these deposition plasmas also
provided a means of determining the elemental composition ofthe sputtering
192 CHAPTER 5
target employed. (37) Their work provided the basis for the development of
GDMS as an attractive alternative to spark source mass spectrometry. Hol-
low cathode glow discharges were coupled with magnetic sector mass ana-
lyzers in the prelirninary investigations of analytical GDMS. (38) In 1978, the
first dedicated elemental analysis system employing a glow discharge
sampled by a QMA was developed at the University of VirginiaY9) The ions
were sampled through an orifice in the base of the sampie cathode. The
resulting beam was found to have a wide energy spread due to the sharp
potential gradient in the near-cathode region. Investigations employing a
variation of the hollow cathode, the hollow cathode plume, exhibited a
similar ion energy behavior. (40) In both instances, energy filtering of the ion
beam was found to offer optirnized performance because of the difference
in kinetic energy between sputtered and discharge gas species. The other
geometries investigated all employ ion sampling through an orifice in the
anode yielding an ion beam characterized by a more narrow ion kinetic
energy distribution. Commercial instruments employ a modified coaxial
cathode geometry, (41) the so-called "pin-type" glow discharge ion source(42)
(Fig. 5-9). This is also the most widely characterized glow discharge ion
source. A Grimm type glow discharge was developed as an ion source for
use with a quadrupole mass filter by Jakubowski et al.(43) A modified version
of the Grimm-type glow discharge, the jet-enhanced sputtering cell, has
recently been adapted to an ICP-MS instrument to perrnit GDMS
determinations.(44) Comparisons of the operating conditions and the ana-
lytical merits ofthe principal glow discharge ion sources are shown in Tables
5-3 and 5-4, respectively.
Whereas different glow discharge ion sources have not exhibited any
significant performance differences, different methods of powering the
sources do show specific performance differences. dc-powered sources are
MHV MACOR
conneclor
probe
lip
Sampie
pin
High-vacuum
feedlhru
Figure 5-9. Coaxial cathode (pin-type) GD ion source.
Table 5-3. Comparison of Glow Discharge Ion Sources
Voltage Current Pressure
Source (V) (mA) (Torr) Cathode
Hollow cathode 200-500 10-100 0.1-1.0 23-mm-deep cylinder with 5-
mm-diameter base
Grimm 500-1000 25-100 1-5 6.5-mm-diameter circle
Jet-enhanced 900 28 2.5 12-mm-diameter circle
Coaxial cathode 800-1500 1-5 0.2-2.0 1.5-2.0-mm-diameter x 4-8-
mm-Iong rod
Table 5-4. Comparison of Glow Discharge Ion Sources

Source Advantages Disadvantages
Hollow cathode >High sputter rate >Charge exchange
mechanisms important
> Intense ion beams >Complicated sampie
geometry
> Amenable to powders
Grimm >Amenable to depth profiling >Flat sampies only
>Easy to use for compacted >Higher discharge gas flow
powders rates
Jet-enhanced >High sputter rate >Flat sampies only
>Easy to use for compacted >Higher discharge gas flow
powders rates
Coaxial cathode >Amenable to various sampie >Powders require conversion
shapes into solid form
> Ionization dominated by
Penning process
the most common. These are suitable for sampies in a conducting matrix.
Operation in a constant-power mode has been observed to yield improved
signal stability as evidenced by studies of run-to-run and sample-to-sample
precision. This results from an inherent compensation for sputter erosion-
induced surface area changes. The use of rf-powered glow discharges permits
the direct analysis of nonconducting sampies that would have to be mixed
with a conductive host matrix prior to analysis with a dc-powered ion
source.(45) Operation of glow discharges with a dc power supply modulated
at less than 100 Hz offers two distinct advantages: higher instantaneous
signal intensities and temporal differentiation of analyte ions from back-
ground gas ions.(46)
5.5. GDMS Applications
Newly developed analytical techniques sometimes struggle to justify

their existence. Applications are posed that could be accomplished much
194 CHAPTER 5
easier by existing methodologies. Not so for GDMS. It arose out of a need

unmet by any available techniques. The driving force in its development was
the limited ability at the time to perform trace elemental analysis directly in
solids, where so many problems and opportunities were present.
In the 1970s, when GDMS was undergoing its metamorphosis from
laboratory curiosity to credible analytical tool, the primary method for
survey trace analysis of solids was spark source mass spectrometry (SSMS).
This technique was limited in scope by its complexity, expense, and un-
reliability. GDMS exhibited all of the strengths of SSMS, but also had
sufficient critical advantages as to suggest it would serve as the eventual
replacement for SSMS, an expectation that has by now generally taken place.
The advantageous features that have led to many GDMS applications
include the following:
• Simplicity-The glow discharge source is a simple two-element

(cathode/anode) device operated at low power and low gas fiows.
• Stability-The source produces a stable supply of low-energy ions
characteristic of the sampIe.
• Sensitivity-The high signal-to-background ratio permits excellent
detection limits, often at the sub-ppb levels.
• Low cost-Actually, commercial mass spectrometers are relatively
expensive, but for those adapting existing units, inexpensive power
supplies are sufficient to drive the source, and the small discharge gas
fiow volumes make for low-cost operation.
• Uniform response-The RSFs do not differ widely across the periodic
table, thus allowing one set of discharge conditions to serve for a
broad range of elements.
• Inert environment-The closed argon source yields a generally inert
plasma that minimizes the formation of polyatomic species, such as
metal oxides.
• Minimal matrix effects-The source can be thought of as a sputter
atomization cell, followed by an ionization source; this pseudo-
tandem configuration negates chemical memory of the sampIe to
minimize matrix-induced differences.
Limitations of GDMS include the potential presence of interferences,

particularly in ultratrace analyses, the inability to handle solutions directly,
and the need for differential pumping to handle the gas load.
Mter a long gestation period, culminating in the first commercial
instrument in the early 1980s, GDMS has spread rapidly, finding broad
application. The major reason for its success has simply been that it solves
problems. GDMS permits the analysis of a broad range of solid sampIe
types, from pure metals to ceramics, from semiconductors to geologicals.
A review of the literature is deceiving in any attempt to determine the

popularity ofGDMS. Much (probably most) ofthe day-by-day applications
are carried out in commercial or industrial laboratories from whence only
technical reports or analytical data sets arise. Some commerciallaboratories
run several GDMS instruments on multiple shifts in order to meet the
demand. GDMS is often the "cash cow" that keeps the business healthy.
We present here only some typical examples of analytical work to which
GDMS is applied. Some ofthe sampies are relatively simple; others present
extremely difficult problems for the technique and the analyst. In all cases,
the need and interest in GDMS is still growing, so that applications that
might be reviewed 5 years from now will show increasingly diverse utility of
this powerful technique.
5.5.1. Solids Elemental Analysis
5.5.1.1. Bulk Metal Analysis
Of all the sampie types, bulk metals are the ones for which GDMS is
almost ideal. Little sampie preparation is required; simply cutting or machin-
ing a pin or disk produces a sampie ready for mounting into the source. By
use of apresputter period, the plasma cleans the sampie surface and readies
the material for analysis. Since reasonably large material consumption (mg)
occurs during GDMS analysis, small sampie inhomogeneities are averaged
out.
Sampie spectra may be obtained in minutes, and a rapid qualitative
analysis can be extracted by examination of the isotopic lines and patterns.
Quantitative analysis follows from line intensity measurements compared
with the matrix element or some selected internal reference element. Software
packages for data treatment can thus very rapidly tell the analyst not only
which elements are present but also their concentration, with linear response
over nine decades.
a. High-Purity Metals. High-purity metals are held to exacting stan-

dards for trace impurities. For example, aluminum used in the electronics
industry must be at least 5-9's purity and 6-9's may be desired. The latter
would require that the entire impurity load be summed to no more than
I ppm. Within the suite of relevant trace elements, some have relatively
relaxed restrictions in the low ppm range, but certain critical elements such
as Th and U (alpha emitters) may be unacceptable at anything above 1 ppb.
To survey the entire periodic table with the detection limits required for such
applications, only GDMS currently meets the need. It is thus not surprising
that this type of sampie comprises a significant fraction of the work load of
GDMS instruments in industry and service laboratories.
196 CHAPTER 5
Vassamillet(7) describes the role ofGDMS in the quality control ofhigh-

purity aluminum production. Because trace impurities can be introduced at
various treatment and melting steps in the production process, multiple
analyses are necessary on a lot-by-lot basis. Before any given shipment of
specified purity aluminum is made to customers, a certificate of analysis is
necessary, requiring an analytical technique that is not only sensitive and
accurate, but also relatively rapid. A VG 9000 GDMS instrument was used
to survey aluminum impurities. After a sputter c1eaning time of 30 min, an
argon discharge was initiated at 1 kV and 2 mA. For each sampIe, 71
elements were monitored and concentrations calculated by use of RSFs
determined from Pechiney reference standards. Most results remained within
a 15% range in the aluminum sampIe, which suffers from inhomogeneity
problems in that the trace elements are not present as a homogeneous solid
solution, but rather occur as particulate precipitates scattered randomly
throughout the sampIe. Even here, however, with the large sampIe consump-
tion of GDMS, reasonable results can be obtained by averaging of sequential
mass spectral runs. In the high-purity aluminum, magnesium (0.86 ppm)
and iron (1.1 ppm) were found as the highest impurities. Gas analyses, such
as C, H, and 0, are difficult because of the background contributions of
each of these species from the glow discharge itself. Gas results vary signifi-
cantly with time as the source is being c1eaned by the discharge sputtering.
After about 45 min, the gas impurities reach aminimum. GDMS analyses
are not rapid, at least not in comparison with solution measurements.
Analysis times of approximately I h, which inc1udes sputter c1eaning of the
sampIe, are indicated as normal. Although only seven or eight sampIes per
day could be analyzed on that basis, it must be recalled that (1) many
elements are analyzed for each sampIe and (2) very little time is required for
sampIe preparation.
Mykytiuk et al.(47) have also analyzed high-purity aluminum with a VG
9000, which they report permits a IOO-fold lower detection limit than with
SSMS and exhibits fewer spectral interferences. By use of cryogenic cooling,
°
background contributions of C, N, and were reduced by three orders of
magnitude, achieving values in the 10-100 ppb range. Typical analyses of
aluminum are shown in Table 5-5, which indicates the reproducibility
obtained when seven pins were prepared and run from the same sampie. In
this paper, the authors also demonstrated the analysis of pure gallium, a
metal characterized by a low melting point (28°C). Without sampie cooling,
the solid metal would quickly become liquid upon initiation of the glow
discharge. By use of the cryogenic cell, however, direct analysis of gallium
in the solid state can be performed. Interestingly enough, sampIe preparation
is carried out in the liquid state by melting the sampie under low heat and
then drawing the molten gallium into micropipettes. After cooling the
pipettes to resolidify the gallium, sampie pins are removed for GDMS
analysis.
Table 5-5. Analysis of Aluminum by GDMS:(47) Values Shown in ppma
Run
Element 2 3 4 5 6 7 Mean %S.D.
B 1.1 1.4 1.1 1.8 1.2 1.0 1.2 1.3 22.0
Mg 0.11 0.11 0.11 0.11 0.11 0.12 0.12 0.113 4.3
Si 4.2 4.1 4.1 4.2 4.3 4.0 4.0 4.13 2.7
P 10.1 9.3 8.5 9.2 10.5 8.1 7.9 9.09 11.0
S 0.033 0.018 0.020 0.035 0.015 0.017 0.042 0.026 41.0
Sc 0.054 0.049 0.059 0.059 0.054 0.054 0.058 0.055 6.6
Ti 0.20 0.11 0.10 0.12 0.12 0.16 0.13 0.134 26.0
V 0.029 0.028 0.026 0.030 0.029 0.029 0.027 0.028 4.9
Cr 0.053 0.053 0.045 0.051 0.050 0.040 0.075 0.052 21.0
Mn 0.033 0.038 0.035 0.038 0.033 0.040 0.038 0.036 7.6
Fe 1.8 1.6 I.7 1.8 1.7 I.7 1.8 I.7 4.4
Ni 0.027 0.029 0.025 0.037 0.024 0.023 0.040 0.029 23.0
Cu 0.77 0.58 0.61 0.67 0.85 0.51 0.71 0.67 17.0
Zn 0.12 0.11 0.10 0.12 0.10 0.12 0.11 0.11 8.1
Ce 0.042 0.039 0.038 0.046 0.053 0.043 0.036 0.042 14.0
Th 0.022 0.018 0.020 0.024 0.022 0.021 0.018 0.021 11.0
U 0.017 0.016 0.015 0.017 0.016 0.014 0.014 0.016 8.2
aSeven pins were cut from the same sampIe. Each result is the mean of live replicate lirings. Preburn 1 h at
5 mA, analyze at 3 mA, 1.1 kV.
Table 5-6 shows a comparison of GDMS and SSMS for two gallium
sampies. Because there are no adequate standards for gallium metal, the
analyst must rely on comparison data from several analytical techniques to
determine the "true" values of the impurities. Sampie 1 contains relatively
low levels of impurities, with the exception of In, Sn, Hg, and Pb. Sampie 2
is a higher-purity sampie in which most "analyses" are simply indicated as
being less than a certain instrumental detection limit. A comparison of the
GDMS estimations relative to those from SSMS shows the lower detection
limits available for the former.
b. Al/oy Metals. Like pure metals, alloys are easily analyzed by GDMS.
Sampie preparation is quite straightforward, and data analysis is com-
plicated only by the need to be aware of potential spectral and polyatomic
interferences. Even for complex alloys, at least one unaffected isotope line
is normally available for each element. Occasional difficulties can be
encountered for monoisotopic elements like aluminum and arsenic. A great
strength of GDMS is that ion yields are relatively unaffected by alloy
composition. (48) Figure 5-10 shows that RSFs collected for seven different
materials, ranging from steels to uranium oxides, are grouped rather tightly
for each element. This may be contrasted with other multielement analytical
techniques (e.g., SIMS, NAA), which exhibit orders ofmagnitude differences
in sensitivity among elements.
198 CHAPTER 5
Table 5-6. Comparative Analysis of High-Purity Gallium by GDMS Versus

SSMS:(47) Values Shown in ppm
Sampie I Sampie 2
GDMS SSMS GDMS SSMS
C 2.0 2.0 0.4 0.2
N 0.2 0.4 0.08 0.2
0 0.6 2.0 0.2 0.2
Na <0.001 Intf." <0.0003 Intf."
Mg 0.002 0.001 <0.0007 <0.002
Al 0.005 0.003 <0.0004 <0.002
Si 0.03 0.02 <0.001 <0.08
S 0.03 0.03 0.002 0.002
CI 0.04 0.03 0.01 0.01
K 0.02 0.05 <0.01 0.005
Ca <0.03 0.01 <0.03 0.002
Fe 0.02 0.04 0.003 <0.008
Cu 3.0 0.9 <0.001 <0.02
Zn 0.9 0.7 <0.001 <0.02
Se <0.02 <0.04 <0.01 <0.04
Ge 0.4 0.2 <0.03 <0.2
Cd 0.3 0.2 <0.002 <0.05
In 10.0 16.0 <0.0003 <0.008
Sn 10.0 30.0 <0.003 <0.05
Hg 4.0 9.0 <0.002 <0.09
Pb 13.0 4.0 <0.002 <0.06
Qlnterference.
An example that illustrates typical alloy analysis is found in the work

by Vieth et al. (49) in which they studied various alloy standards to determine
RSFs and then applied these factors to the analysis of nickel-base alloys.
An interesting aspect of this study for the purposes of this report was a
comparison between two different types of GDMS instruments. The first
commercial instrument was of the so-called "high-resolution" type, which
incorporated a magnetic sector instrument (reverse Nier Johnson geometry)
to permit relatively high ion throughput and the resolution of certain spectral
interferences (e.g., ArO+ on Fe+). Quadrupole-based instruments have been
used for GDMS for many years, (23) but for all their other advantages, such
units cannot resolve same-nominal-mass interferences and exhibit a signifi-
cant decrease in ion throughput at higher masses. Recent access to com-
mercial instruments of both types permitted the comparison shown in Table
5-7, whereby RSFs were determined as an aggregate from five different
matrices. What becomes obvious is that the same sampIes, sputtered and
ionized by essentially identical sources, exhibit generally similar RSFs up to a
point, but beyond about mass 90 the elemental sensitivities of the quadrupole
5 2a 3b 4b Sb 6b 7b Sb 1b 2b 3a 4a 5a 6a
, Y ,
4
CHEMICAl GROUP
m: TI8)~N'
i.~ -wMö-----~~~-:,;}~:-4-----
: :::: =:._'" Ga: J.,.
.2
~ ~~~~~~m
o Tantalum
b. Gallium
CU~ a~6C(;)paS
~_-
[). Zn -_
X Uranlum Oxide
Figure 5-10. Demonstration of the generally uniform relative sensitivity factors characteristic
ofGDMS.(48)
instrument begin to fall off rapidly, presumably due to lower ion transmis-
sion and tuning artifacts. (Note: RSFs calculated by this method are really
"insensitivity factors." That is, larger RSFs reflect lower sensitivity.)
Since RSFs are simply correction factors for the entire instrument sys-
tem (sputtering, ionization, extraction, throughput, and so on), it is feasible
to use the different RSFs generated for the sampIe elements to calculate
analytical results and compare two commercial units. This is shown in Table
5-8, in which RSFs compiled from Ni-based alloys were used to analyze a
CRM 345 standard IN-lOO Ni-base alloy. Several points become clear from
an examination of the data. (1) GDMS is able to provide elemental results
from the major sampIe constituents down to the trace level in one analysis.
(2) The sector instrument versus quadrupole data are in relatively good
agreement; the presence of spectral interferences would make quadrupole
results appear high relative to higher-resolution results, and there is little
appearance of this in the data. (3) Both GDMS instruments yield results
that are reasonably good with respect to the certified value of the alloy
standard. Analyses such as these demonstrate the utility ofGDMS for quali-
tative and quantitative analysis of complex alloys.
5.5.1.2. Semiconductors
A third type of bulk sampIe analysis that is very important in commer-
dal applications is that ofsemiconductors (or semi-insulators). The electrical
properties of these materials are critically dependent on the intrinsic and
doped levels of impurities, making it essential that fabricators know not only
200 CHAPTER 5
Table 5-7. Comparison ofRSF Values Obtained in a Steel

Sampie by Two Commercial GDMS Instruments (49)
RSF (x/Fe)
Element GloQuad VG-9000
Be 2.0 1.1
B 1.5 1.22±O.17
C 5.7 ± 1.4 4.5± 1.1
Mg 1.1 1.3 ±0.3
AI 0.96±0.30 1.39±0.09
Si 2.27±0.65 2.0±0.3
P 5.1±0.7 3.5 ±O.I
S 3.5±0.7 3.3 ±O.1
Ti 0.67±0.11 0.42 ± 0.04
V 0.76±0.06 0.55 ± 0.02
Cr 2.11 ± 0.16 2.2±OA
Mn 1.20±0.22 1.5±0.2
Fe =1.00 =1.00
Co 1.07 ±0.13 1.1 ±OA
Ni 1.50±0.08 1.5±0.1
Cu 5.0±0.7 5.0± 1.2
Zn 4.5±0.6 5.5±0.9
Ga 6.1 4.5±OA
Se 3.0 3.1 ±0.3
Zr 2.2 0.64±0.08
Mo 3.00±O.11 1.3 ±0.2
Sn 7.8 ± 1.9 2A±0.2
Ta 7.0 1.1
W 5.5 1.5±0.1
Pb 9.2±2.1 2.2±0.2
Bi 18 4.3±0.7
what is present but also the concentration levels. The problem for the analyst
is exacerbated by the fact that even extremely low elemental concentrations
of specific elements are known to alter semiconductor properties. Further-
more, the presence of any impurity element can be worrisome enough to
require its complete characterization. The presence of C, 0, and N is often
difficult to measure by elemental techniques (including GDMS), but such
information may be crucial. The net result of all this has been a growing need
for analytical techniques that will survey semiconductors for all elements
and provide at least semiquantitative results. Manufacturers cannot permit
surprises to show up later in a batch of substrate material that has been used
to prepare many expensive semiconductor components.
GDMS has been particularly valuable in supplying rapid, accurate,
and sensitive analyses in this regard. Using GaAs as an example, Evans and
co-workers(50) demonstrated that GDMS could achieve excellent detection
Table 5-8. GDMS Analysis of CRM 345 Standard IN-100 Ni-
Base Alloy:(49) Values in ppm Except Where Indicated as %
Element Certified VG-9000 GloQuad
B 190± 10 230 180
C 0.153 0.21% 0.14%
Mg 5± 1.2 6.1 6.3
AI 5.58% 5.5% 5.4%
Si 520 320
S 18 23
Ca ( <1) :5:0.2 0.9
Ti (4.74%) 5.28% 4.0%
V 1.00% 1.04% 0.85%
Cr 9.95% 10.4% 9.3%
Fe 780 670
Co 14.71% 15.3% 14.9%
Zn <0.5 <0.1 0.49
Ga 8±0.3 8.2 7
As (2) 2.3 4
Se <0.5 2 3
Zr 440 425 410
Mo 3.01% 3.05% 3.3%
Ag <0.2 <2 <0.8
Cd <0.1 :5:0.2 <0.2
In 0.04 0.05
Sn 6± 1.6 5.2 11
Sb <2 2.1 3
Te <0.2 <0.03 <0.3
TI <0.2 :5:0.007 <0.03
Pb 0.2 :5:0.05 0.4
Bi <0.2 :5:0.005 <0.2
limits for many elements, as shown in Table 5-9. However, beyond the 28
elements successfully presented, standing out by their absence are five other
elements, namely H, C, N, 0, and Cl, which may be of great interest to
manufacturers. The authors solved the problem by complementary analyses
using SIMS with Cs bombardment. It was estimated that by the use of
suitable standards, accuracy within 20% was obtained by GDMS and within
a factor of 2 by SIMS.
Mykytiuk et al. (47) have also reported the analysis of GaAs, but they
have taken a different approach to allow the determination of impurity gases
C, N, and 0. After GaAs pins are cut, they are rinsed in 5% bromine in
methanol, then in high-purity methanol, and then dried in an oven. Their
key to obtaining good results for the gases lies in an extended sampie sputter
burn for I h at high-current conditions of 4-5 mA for GDMS. Normal
currents lie usually in the 1-2 mA range. The greater bombardment current
and energy subject the sampie to considerable heating and sputter c1eaning.
202 CHAPTER 5
Table 5-9. Detection Limits for Selected Elements In GaAs

Comparing GDMS and SIMS:(50) Values Shown in ppm
Element GDMS OrSIMS Cs-SIMS
a
H 10,000·
Be 20 2
··
B 2
C 500·
N 100·
a
0 1,000·
F 300 20
Na 30 10
Mg 5 5
AI 5 500
Si 10 40
p 40 40
·
S 200 50
CI 20
Ca 60 6
Ti 0.7 5
Cr 4 I
Mn I 5
Fe 3 10
Cu 10 100
Zn 20 100
Ge 300 100
Br 30 8
Se 400 5
Mo 10 100
Cd 20 6,000
In 200 8
Sn 20 10
Sb 20 20
Te 10 0.5
W 4 5,000
Hg 9 100,000
Au 20 1,000
·Limited by residual atmospheric gases in the instrument.
Table 5-10 shows the time-dependent concentrations of C, N, and 0 during

the sampie treatment step. A question arises as to whether the longer burn
time required to conduct GDMS analyses by this approach is preferable to
the additional time that would be involved in acquiring SIMS data for these
troublesome elements. If no SIMS instrument is available, of course, such
conjecture becomes a moot point. Having confirmational data from either
SIMS or SSMS is always desirable, but falls into the category of high luxury
for most analysts.
Table 5-10. Effect ofPrebum Time on GDMS Analysis ofGaAs:(47)
Values Shown in ppm
Element 5 min 30 min 45 min 60 min
B 0.2 0.08 0.06 0.07
C 0.5 0.03 0.01 0.009
N 0.09 0.02 0.01 0.01
0 0.1 0.09 0.04 0.05
Na 0.02 <0.002 <0.003 <0.003
Mg 0.005 <0.008 <0.003 <0.003
AI 0.05 <0.005 <0.001 <0.001
Si 0.2 0.2 0.004 <0.002
S 0.03 0.01 0.008 0.01
CI 0.003 0.005 0.004 0.003
Fe 0.002 0.003 0.002 0.001
5.5.1.3. Surface Analysis
In one sense, GDMS is always a surface analysis technique, even though

it permits measurement of bulk concentrations. That is, the sputtering pro-
cess central to the glow discharge acts as an atomic mill that regularly erodes
away the surface of the bombarded sampIe. Whatever atoms are on the
surface are sputtered away and measured. Because GDMS consumes signifi-
cant quantities of material (up to milligrams per minute), these sequential
layer analyses combine to yield an averaged composition that is typical of
the bulk concentration. By slowing down the ablation process and limiting
measurements to a brief time span, data indicative of surface concentrations
can be obtained. GDMS and its optical analogue have found considerable
application for the analysis of layered sampIes.
A number of common surface-Iayered materials benefit from GDMS
analysis. For example, metals that are likely to corrode in an oxidizing
atmosphere may be coated with a protective layer 1~ 10 11m in depth. Galvan-
ized steel features a thin layer of Zn on an Fe base material. The emf of Zn
allows it to withstand oxidation in the presence ofwater and oxygen, protect-
ing the Fe, which would readily rust. Typical etch rates of the glow discharge
are 1~5 I1mjmin, permitting mass spectral accumulation for up to several
minutes during the ablation of a given thin layer.
It is not realistic to speak of these analyses as thin-film measurements.
Given its high sputter rate, GDMS lends itself more to the analysis of layers
in the range of micrometers rather than angstroms. In principle, the glow
discharge mills itself through a sampIe, layer by atomic layer. Thus, a layer
comprised of element A should provide a large and steady signal of A + until
that layer is consumed, at which time the A + signal should drop sharply to
204 CHAPTER 5
zero, with a simultaneously abrupt appearance of a signal from the next

layer, comprised of element B. One reasonable approximation of this
behavior is illustrated in Fig. 5-11, in which results from the analysis of a
three-component sandwich sampie are presented. The thin Cr layer (ca.
1 j1m) sustains a Cr ion signal for only a few minutes, at which point it
drops dramatically (note the log scale). The next layer, Cu, is represented
by a signal that essentially rises vertically to a level typical of a matrix signal.
After about 75 min of sputtering, the next interface (Cu/Fe) is reached, at
which point the Fe signal becomes prominent while the Cu intensity drops
off over an approximately 20 min interval.
It is obvious from Fig. 5-11 that the signal responses do not follow an
ideal square wave pattern for their appearance and disappearance. Clearly,
all of the metal sampie layer does not disappear at the prescribed time, but
rather is scattered across a diffuse region that gradually is eroded away.
Note that the initiation of a new layer is quite sharp (Cu, Fe), but the
termination of that layer lingers for many minutes (Cr, Cu). Part of this is
due to the atomic mixing of the sputter process that causes signal from the
depleted layer to extend into the next layer. Another factor is the intrinsic
redeposition that occurs in the collision-rich (l Torr) environment of the
discharge. It has been estimated(52) that up to 67% of the sputtered atom
population is returned to the surface by collisions with argon atoms, to be
subsequently resputtered (perhaps several times) before escaping the surface
region. In this manner, the analytical signal for a given elementallayer will
extend far beyond the nominal demarcation line anticipated.
---Fe
Cr Cu Fe
11 1
~
5.9
miaon
Cu
Cr
10-15 L.-~-:I:-~~-:I:-~~~~~---:L~~
10 20 30 40 50 60 70 60 90 100
Time (min)
Figure 5-11. Thin-layer analysis by GDMS for a three-element sandwich sample.(51)
It is this thin layer cross talk that prevents GDMS from being used for
very thin films. For the micrometer range layers, however, the data obtained
from GDMS are quite valuable and capable of providing quantitative
analysis and depth profile analysis. Because GDMS permits rapid scans, an
entire sweep of the periodic table may be performed in a few tenths of a
second. In this way multiple scans may be accumulated and impurity levels
in the thin layer detected and their concentrations determined. By calibrating
erosion rates using standard thin-Iayer sampies, the thickness of layers may
be determined by examination of signal/time profiles. In this way, thin-Iayer
sampies can be fully characterized. The increasing popularity of GDMS
for thin-Iayer work arises from the general lack of matrix effects exhibited
by this technique, in comparison with the major problems frequently
encountered when using SIMS analysis.
5.5.1.4. Compacted Samp/es
The sampies considered to this point have all arisen from bulk solids,
permitting the ready formation of an electrode for glow discharge analysis.
All solids are not so conveniently prepackaged for the analyst. Many are in
astate that requires compaction to form a suitable sampie electrode (e.g.,
metal filings or powder). If, in addition, these powdered sampies are non-
conductors (e.g., soils, sediments, glas ses) , then an additional step in the
preparation is required to yield a conducting matrix for analysis.
The analysis of metals and alloys calls upon much simpler method-
ologies than in the case of compacted sampies. In addition to the increase
in sampie preparation time, the need to convert to a conducting matrix
introduces the possibility of contamination. Other problems arise because
of the trapping of water vapor and atmospheric gases in the sampie during
the compaction process. Figure 5-12 shows a general schematic of a sampie
+-+-- Die body

+---+-- Plunger
~--+--Sample disc
~==-+---Stainless steel anvil
~--Diebase
Figure 5-12. Cross seetion of a custom die used to fabricate GDMS sampIes from powdered
materials.
206 CHAPTER 5
preparation die. Consider that an ore sampie is to be analyzed by GDMS

using a dc discharge. A 10-20% by weight sampie is prepared by mixing
with a suitable high-purity metal powder, usually silver or copper. (Graphite,
which is used as a matrix for many optical methods, is less satisfactory for
GDMS due to the large amount of hydrocarbon ions produced.) Pressing
the matrix/sample mixture at high pressure by means of a suitable die and
hydraulic press effects the formation of a conductive electrode disk or pellet
for GDMS analysis. If an rf-powered discharge source is available, the
sampie electrode does not have to be conducting, thus eliminating the need
for matrix modification by dilution with a conducting powder. An rf source
will also permit the direct analysis of solid nonconductors, such as glasses
or ceramics.
The glow discharge is confronted with a more difficult sampie matrix
in the case of compacted sampies. Unlike the case of bulk solid metals and
alloys, where the sampie surface is initially c1eaned by the sputter process,
revealing the "true" bulk composition below, a compacted electrode is
comprised of an unending series of surfaces. That is, each micrometer-sized
metal partic1e is coated with a contaminant layer, often the metal oxide,
through which the plasma must etch to reach the metal sampie underneath.
The net result is that the discharge is always being presented a combination
of bulk material and surface "contaminants." In the case of more complex
materials, such as rocks and sediments, the situation is even more difficult.
The elemental analytes are chemically bonded (and tenaciously, at that) to
other elements that may contaminate the discharge and cause it to be more
reactive toward the formation of interferents. For example, examination of
a geological sampie will reveal the trace metals tied up with oxygen and
usually located in a ho stile silicate matrix. When the discharge strikes these
materials, not only are the analytical elements sputtered, but also the matrix
elements, inc1uding atomic oxygen, which is then highly reactive in the
plasma.
In spite of these impediments, GDMS offers some outstanding oppor-
tunities for compacted sampie analysis. Figure 5-13 shows the type of
spectrum that can result from a compacted sampie of a standard reference
sampie, SCo-l (USGS). Only the rare earth region is depicted in the spec-
trum, illustrating the attainable sensitivity and the potential for interferences
in this case. Strong oxides of Ce and La are not totally dissociated in the
plasma, leading to the isobaric interferences shown in Fig. 5-13. Normally,
an uninterfered isotope can be found by which the elemental concentrations
can be determined. Table 5-11 shows some calculated concentrations in the
standard sampie using RSFs previously determined against Ce as an internal
standard. These results demonstrate the still untapped potential of using
GDMS as a qualitative and quantitative analysis tool for compacted
sampies.
Nd 141·Pr =0.38 ppm

Ce 142·Ce = 0.35 ppm
142·Pr = 0.38 ppm
143-Nd =0.17 ppm
144-Nd a 0.34 ppm
~ CeO 145-Nd = 0.17 ppm
'"
r:: Pr 146-Nd • 0.24 ppm
~
.5 147-Sm ,,0.037 ppm
äi 14a-Nd .. 0.080 ppm
r:: Nd 149-Sm s 0.035 ppm
Cl LaO
Ci5 151-Eu = 0.025 ppm
BaO
CD Nd
>
tJ Gd
-ai
a:
150 160 170

m/z
Figure 5-13. Mass spectrum of the rare earth region for an SCo-l (USGS) standard reference
material. Values shown are concentrations in the prepared electrode.
Table 5-11. Rare Earth Concentrations as Determined by GDMS in

USGS Standard Reference SCo-l Values Shown in ppm
Element concentration
Ion peaks Reported Calculated
La-139 29.7 25.6
Ce-140 63.4 (internal standard)
Pr-141 7.62 6.1
Nd-143, 144, 145, 146 27.9 26.6
Sm-147,149 5.07 5.5
Eu-153 1.03 1.3
Gd-160 4.00 4.8
Dy-162,163,164 3.79 3.8
Ho-165 0.83 0.8
Er-166,167,168 2.39 2.7
Yb-171,172,174 2.25 2.1
5.5.2. Solutions
The point has been strongly made that the major advantage of GDMS
is its ability to determine solids direct1y, without sampie dissolution. Thus,
the indusion of solution analysis among GDMS applications seems a con-
tradiction in purpose. Still, there are certain situations in which this approach
may make sense. It should be recognized that the direct analysis of solutions
by GDMS is not normally attempted. Nebulization techniques that work
208 CHAPTER 5
with atmospheric pressure sources such as flames and ICPs cause quenching
of glow discharges. It is necessary to remove the solvent from the sampIe
before introducing it into the plasma. With that caveat, a number of
approaches are possible.
The most direct is simply to evaporate a solution sampIe onto the
surface of a conducting electrode, leaving a residue film of the previously
dissolved analytes. This thin layer then comprises the sampIe, which can be
sputtered in the glow discharge for analysis. In reality, of course, this is
solids analysis, not solutions analysis, as the solution is converted into a
solid residue for conventional GDMS. The opportunities and problems are
more in keeping with Section 5.5.1.3. On the positive side, the sampIe is
effectively concentrated greatly in that all the analytes are presented to the
plasma in the short period of time (seconds) during which the surface layer
is sputtered away. This is also a disadvantage in that little time is available
to average spectra, or even adjust instrument parameters, creating potential
problems of precision.
A somewhat more elegant method of transforming a solution sampIe
into a surface film is by means of electrodepositing selected metals onto a
cathode. In this manner, an analyte contained at a trace level in a large
solution volume may be isolated and concentrated onto a surface for subse-
quent sputter atomization into the glow discharge. This approach offers the
possibility of both selectivity and enhanced detection limits. A third method
serves as a more in situ solution/solid transformation approach. Drawing
upon well-established electrothermal vaporization methodology, microliter
aliquots of solution sampIe are placed on a filament that will also serve as
the GDMS electrode. Currents are passed through the filament to dry the
sampIe to a residue film, then the current is increased to ash away any
organic constituents, if necessary, and finally a large step current is applied
to atomize the residue rapidly into the glow discharge plasma. Again, all the
advantages and disadvantages of microsample techniques are present. Note,
however, that in this approach, the primary atomization step is electrother-
mal, rather than by sputtering. A fourth method oftreating solution sampIes,
and one that gets away from creation of a thin surface residue, is the mixing
of a solution aliquot with powdered matrix material (e.g., silver or graphite).
Drying this mixture under an IR lamp, homogenizing in a Wig-L-Bug
vibrator (Crescent Dental Manufacturing Co.), and pressing of the sampIe
by compaction methods (Section 5.1.4) produce a sampIe electrode contain-
ing the solution residue. This amounts basically to a matrix conversion from
aqueous solution to metal.
Solution analyses by GDMS have been reported since the mid-1970s,(41)
but they have rightly remained on the outer fringe of analytical interest. A
recent application, however, shows that specialized use of GDMS for
solution analysis may be advantageous. lakubowski et al. (53) reported the
detection of sub-picogram levels for Pt and Ir, elements increasingly of
interest in the environment because of their ejection from catalytic converters
in automobiles. Sampies of 1-10 JlI were pipetted onto copper substrates
and the solution dried to a residue at room temperature, producing a 4-mm-
diameter residue spot for a 10-JlI aliquot. Data were acquired by rapid
multiple scans over Pt and Ir mass region or by single ion monitoring of
selected isotopes. Figure 5-14 shows a time trace of the ion signal from a
1-pg Ir sampie. The transient nature of the signal is apparent. The authors
show that satisfactory precision was obtainable as indicated by the error
bars in Fig. 5-15.
Given the effect of solvent introduction on glow discharge operation,
the direct injection of solutions has many obstac1es to overcome, but some
modified efforts in this direction have been reported. Strange and Marcus(54)
have used a partic1e injection system to introduce a solution sampie into the
glow discharge. For the foreseeable future, GDMS is likely to remain asolids
analysis technique, short of some new sampie introduction breakthrough.
5.5.3. Nonelemental GDMS Applications
The versatility of the glow discharge relative to elemental applications

is generally well recognized in the availability of commercial instrumentation
for atomic absorption, atomic emission, and atomic mass spectrometry.
Those outside the elemental community have also taken note of the glow
discharge and found complementary characteristics that are useful in their
field. In general, these emphasize use of the versatile glow discharge plasma
processes more than the atomization process. Two examples will be
presented here for illustration.
2E+M~----------------------------~
o 7 14 21 28 35
Time (8)
-Ir - background .....•...... detecUon IimH
Figure 5-14. Temporal 193Ir ion signal response from 1 pg iridium deposited onto a sampIe
e1ectrode plate. (53)
210 CHAPTER 5
1~~------------------;-:-.:~'
...-
.:
l! 100
§ .:1
.
",:
..
/l!'
//t
0.1 '-<----..------,-----,-----,-----,.1
0.01 0.1 1 10 100 1~
Ir Conten!, ng
Figure 5-15. Linearity of response for I93Ir ion signal by GDMS. (53)
McLuckey et al. (55) took advantage of the inherent ability of the glow
discharge to function on almost any readily ionizable gas. While the inert
gases are normally used as discharge agents, stable plasmas can be obtained
with nitrogen, oxygen, air, and even water vapor. By designing a glow dis-
charge ion source that used ambient air as its discharge gas, the authors
were able to analyze trace impurities present in the sampled air. Figure
A1 mass spectrometer
pump pump
151/s 1000 Vs
Figure 5-16. Atmospheric sampling GDMS ion source.(55)
5-16 shows a sketch of the atmospheric sampling glow discharge source. By

appropriate differential pumping, trace contaminants may be transferred
from the sampled air to the glow discharge cell for ionization by the plasma,
with the subsequent extraction of the ions into the mass spectrometer for
analysis. In this manner a very simple ion source is able to sampie directly
the ambient air contarninants for sensitive mass spectral analysis. Traces of
explosive residues at the ppb level were detected by this arrangement. In this
type of application, it is necessary to observe the sampie in its molecular
form, rather than reducing it to its atornic components. It is the sputter step
in the elemental analysis process that produces major bond breaking. By
introducing the gaseous sampie directly into the low-energy glow discharge
plasma, sufficient ionization and residual molecular ions are produced.
All the applications presented have featured the glow discharge as a
primary ionization source for analyte determination. An increasing use finds
the glow discharge in the role as a detector by ionizing effiuents, normally
organic materials, from separation processes. Figure 5-17 shows one such
use in which the LC effiuent is vaporized and directed into a glow discharge
plasma. The LC tubing itself serves as the discharge cathode around which
the negative glow forms. Sequential chromatographic components pass into
the plasma, where they are ionized by collision with electrons and other
energetic particles and then extracted into the mass spectrometer for sensitive
analysis. The glow discharge serves as a stable, low-maintenance ionizer that
creates efficient ionization of the organic effiuents from the chromatograph.
QUADRUPOLE
MASS ANAL YZER
ION LENSES
ANODE
Figure 5-17. Representation of a glow discharge ionization source used as adetector for
liquid chromatography. (56)
212 CHAPTER 5
This success has led others to use the glow discharge as a primary ionization
source for the analysis of organic sampIes. (56)
5.5.4. Future Directions in GDMS Applications
We have reviewed examples ofthe most common GDMS applications,

and these seem likely to show sustained importance in the future. However,
it might be worthwhile to suggest a few additional areas that could grow
in significance. The dc glow discharge source may find competition from
alternative discharge modes. Most promising at this writing is the rf source,
which has been demonstrated to exhibit the sensitivity and stability needed
for commercial success. At the moment, investigators must build their own
rf systems, limiting the extent of usage. The anticipated appearance of a
commercial rf discharge system and associated direct insertion probe will
stimulate more interest in the analysis of nonconductors such as soils, sedi-
ments, glasses, insulators, and ceramics. Careful shielding of the rf generator
is critical to prevent serious noise problems, and the spectra generated by rf
sources must also prove to have the atomic simplicity characteristic of dc
plasmas. Pulsed glow discharges also offer advantages to the analyst for
time-resolved measurements, although the additional gating electronics
required will be a dis advantage for routine work, not so much in terms of
cost, but in the greater complication of data acquisition.
Other types of spectrometers may find application with GDMS in the
future, particularly if they are able to assist in solving the problem of isobaric
interferences. A magnetic sector instrument (resolution ca. 8000) currently
controls the greatest market share, while a second sector unit has only
recently appeared that claims advantages based on 20,000 resolution. Two
other spectrometer styles that are receiving attention include the FT-ICR
and TOF. Preliminary data(22) from the interfacing of a glow discharge to
an FT-ICR are quite promising based on attainment ofup to 40,000 resolu-
tion, which would permit separation of potentially interfering species that
currently are unresolvable. Finally, recent improvements in TOF spec-
trometer designs suggest that, given the powerful intrinsic advantages of this
technique, it would be worthwhile to couple a pulsed glow discharge to a
state-of-the-art TOF. We can likely expect this in the near future, based on
the growing interest in glow discharges.
References
1. C. R. Aita, J. Vac. Sei. Technol. A 3 (1985) 625.

2. J. W. Coburn, Thin Solid Films 171 (1989) 65.
3. G. D. Patterson, in: Treatise on Analytical Chemistry (I. M. Kolthoff, P. J. Elving, and
E. B. SandelI, eds.), Vol. 10, Wiley-Interscience, New York, 1972.
4. VG 9000, VG MicroTrace, Cheshire, UK. VG Glow Quad, UK. TS SOLA, Turner Scien-
tific, JEOL, Kratos.
5. F. W. Aston, Isotopes, 2nd ed., Longmans, Green, New York, 1924.
6. J. Roboz, Introduction to Mass Spectrometry: Instrumentation and Techniques, Interscience,
New York, 1968.
7. L. F. Vassami11et, J. Anal. At. Spectrom. 4 (1989) 451.
8. P. E. Miller and M. B. Denton, J. Chem. Educ. 63 (1986) 617.
9. T. Niehuis, R. F. ReIler, and A. Benninghoven, J. Vac. Sei. Technol. A 5 (1987) 1243.
10. P. R. Dawson and J. E. Fulford, Int. J. Mass Spectrom. Ion Phys. 42 (1982) 195.
11. R. S. Rouk, Anal. Chem. 58 (1986) 97 A.
12. R. E. March and R. J. Rughes, Quadrupole Storage Mass Spectrometry, p. 43, Wiley, New
York, 1989.
13. P. H. Dawson, in: Quadrupole Mass Spectrometry and Its Applications (P. H. Dawson,
ed.), Elsevier Scientific, New York, 1976.
14. J. E. Campana, Int. J. Mass Spectrom. Ion Phys. 33 (1980) 101.
15. F. L. King, A. L. McCormack, and W. W. Rarrison, J. Anal. At. Spectrom. 3 (1988) 883.
F. L. King and W. W. Rarrison, Int. J. Mass Spectrom. Ion Proc. 89 (1989) 171.
16. J. F. J. Todd, in: Quadrupole Storage Mass Spectrometry (R. E. March and R. J. Rughes,
eds.), Wiley, New York, 1989.
17. G. C. Stafford, P. E. KeIley, J. E. P. Syka, W. E. Reynolds, and J. F. J. Todd, Int. J. Mass
Spectrom. Ion Proc. 60 (1984) 85.
18. S. A. McLuckey, G. L. Glish, and K. G. Asano, Anal. Chim. Acta 225 (1989) 25.
19. P. R. Remberger, J. S. Crain, and J. Williams, Data presented at ASMS Sannibel Island
Conference on Ion Trapping in Mass Spectrometry, 1990.
20. N. M. M. Nibbering, Acc. Chem. Res. 23 (1990) 279.
21. J. L. Shohet, W. L. Phillips, A. R. T. Lefkow, J. W. Taylor, C. Bonharn, and J. T. Brenna,
Plasma Chem. Plasma Process. 9 (1989) 207.
22. C. M. Barshick and J. R. Eyler, J. Am. Soc. Mass Spectrom. 3 (1992) 122.
23. B. L. Bentz, C. G. Bruhn, and W. W. Rarrison, Int. J. Mass Spectrom. Ion Phys. 28 (1978)
409.
24. N. R. Daly, Rev. Sei. Instrum. 31 (1960) 264.
26. B. Chapman, Glow Discharge Processes, p. 181, Wiley-Interscience, New York, 1980.
27. W. D. Davis and T. A. Vanderslice, Phys. Rev. BI (1963) 219.
29. E. W. McDaniel, Collision Phenomena in Ionized Gases, Wiley, New York, 1980.
30. M. K. Levy, D. Serxner, A. D. Angstadt, R. L. Smith, and K. R. Ress, Spectrochim. Acta
46B (1991) 253.
31. F. Roworka and M. Pahl, Z. Naturforsch. 27A (1972) 1425.
32. B. E. Warner, K. B. Persson, and G. J. Collins, J. Appl. Phys. 50 (1979) 5694. E. B. M.
Steers and R. J. Fielding, J. Anal. At. Spectrom. 2 (1987) 239.
33. R. W. Smithwick, J. Am. Soc. Mass. Spectrom. 3 (1992) 79.
34. F. L. King and W. W. Rarrison, Mass Spectrom. Rev. 9 (1990) 285.
35. J. A. Klingler and W. W. Rarrison, Anal. Chem. 63 (1991) 2982.
36. J. W. Coburn, Rev. Sei. Instrum. 41 (1970) 1219.
37. J. W. Coburn and E. Kay, Appl. Phys. Lett. 19 (1971) 350.
38. W. W. Rarrison and C. W. Magee, Anal. Chem. 46 (1974) 461. B. N. Colby and C. A.
Evans, Anal. Chem. 46 (1974) 1236.
39. C. G. Bruhn, B. L. Bentz, and W. W. Rarrison, Anal. Chem. 50 (1978) 373.
40. R. K. Marcus, F. L. King, and W. W. Rarrison, Anal. Chem. 58 (1986) 972.
41. W. A. Mattson, B. L. Bentz, and W. W. Harrison, Anal. Chem. 48 (1976) 489.
214 CHAPTER 5
42. W. W. Harrison, C. M. Barshick, J. A. KlingIer, P. H. Ratliff, and Y. Mei, Anal. Chem.

62 (1990) 943A.
43. N. Jakubowski, D. Stuewer, and G. Toelg, Int. J. Mass Spectrom. Ion Proc. 71 (1986) 183.
44. H. J. Kim, E. H. Piepmeier, G. L. Beek, G. G. Brumbaugh, and o. T. Farmer 111, Anal.
Chem. 62 (1990) 639, 1366.
45. D. L. Donohue and W. W. Harrison, Anal. Chem. 47 (1975) 1528. D. C. Duckworth and
R. K. Marcus, Anal. Chem. 61 (1989) 1879.
46. W. A. Mattson, Ph.D. dissertation, University of Virginia, 1976. J. A. Klingler, P. J.
Savickas, and W. W. Harrison, J. Am. Soc. Mass Spectrom. 1 (1990) 138. J. A. Klingler,
C. M. Barshick, and W. W. Harrison, Anal. Chem. 63 (1991) 2571.
47. A. P. Mykytiuk, P. Semeniuk, and S. Berman, Spectrochim. Acta Rev. 13 (1990) 1.
48. N. E. Sanderson, E. Hall, J. Clark, P. Charlambous, and D. Hall, Mikrochim. Acta I (1987)
275.
49. W. Vieth, A. Raith, and J. C. Huneke, Charles Evans & Associates, unpublished data.
50. Application Note, GDMS-AN-002-01, CharIes Evans & Associates, Redwood City, Calif.,
1988.
51. GDMS Technical Bulletin, APP/GQ/OOl, Fisons/VG Instruments.
52. C. M. Barshick and W. W. Harrison, Mikrochim. Acta 111 (1989) 169.
53. N. Jakubowski, D. Stuewer, and G. Toelg, Spectrochim. Acta 46B (1991) 155.
54. C. M. Strange and R. K. Marcus, Spectrochim. Acta 46B (1991) 517.
55. S. A. McLuckey, G. L. Glish, and K. G. Asano, Anal. Chim. Acta 225 (1989) 25.
56. R. Mason and D. Milton, Int. J. Mass Spectrom. Ion Proc. 91 (1989) 209.
6
Hollow Cathode Discharges
Sergio Caroli and Oreste Senofonte
6.1. Introduction
6.1.1. Historical Background
Pioneering observations on the luminescent phenomena generated in evacu-

ated tubes belong to the history of spectroscopy and date as far back as the
mid-1800sy,2) Within this field of research, the first description of a hollow
cathode discharge (RCD) published in the scientific literature can be found
in the early years of this century, when a German physicist, Friedrich
Paschen, reported on the quite unique features of this radiation source. (3)
At the time he was mainly engaged in the investigation of the spectral series
of R 2 in the IR region and in the distribution of energy in the spectra
emitted by glowing gases. Together with Back in 1913 he discovered the effect
(named for both scientists) that gives rise to the splitting of emission lines
when the source is subjected to a very strong magnetic field. This phenom-
enon is actually a modification of the Zeeman effect that requires less intense
fields.
Thus, interested as Paschen was in the processes occurring when electric
currents are passed in agas at reduced pressure, he arrived almost directly
at the special electrode arrangement to which this chapter is devoted. The
specific reason that prompted him to develop the RCD prototype was the
fact that only a couple of years earlier H. Barteis and co-workers had no ted
Sergio Caroli and Oreste Senofonte • Department of Applied Toxicology, Istituto

Superiore di Sanitä, 00161 Rome, Italy.
215
216 CHAPTER 6
how intense the Fowler lines in the Re spectrum were in the luminescent
layer within the cylindrical cathode of a Geissler tube filled with the said
rare gas operated in the dc mode. Experimental evidence available at that
time showed that a progressive decrease in the Re pressure led to the con-
finement of the glow within the cathode bore and that an enhancement in
current intensity had as a consequence an increase in the radiation output,
while the outer surface of the electrode remained entirely unaffected.
On the basis of these previous findings, Paschen devised a glass tube
whose main axis coincided with that of the cathode. The anode, instead, was
placed perpendicularly to it in a side arm. It may seem surprising that the
cathode, worked out by bending a thin Al foil, was shaped so as to form a
box open at both ends with a pronounced rectangular cross section, in order
that the luminescence in its cavity was ellipsoidal. Paschen noted that at the
relatively low applied direct current of 100 mA the expected lifetime of the
tube did not exceed 70-100 h, but in spite of this, and in obvious conflict
with the long time necessary to photographically record the emitted lines in
higher orders (several hours), he was able to fully reach his goal of elucidat-
ing the hyperfine structure of the spectrum. The RCD was thus expediently
used for the first time in recognition of one of its more striking properties,
i.e., the ability to generate spectra with exceptionally sharp lines, and
Paschen should be given credit for identifying this potential and exploiting
it to the benefit of his studies.
Later on, further applications of this discharge type were reported by
Paschen in investigations of the spectrum emitted by a hollow cathode of
Al. (4) This research should be regarded in its proper light in that what is
quite clear to the experimentalist today was relatively unknown in Paschen's
time. In other words, it became apparent that not only the filler gas, but
also the material forming the negative electrode could issue spectral lines
with the same basic features as the former. A detailed description of the
various zones distinguishable within the cathode cavity was also given,
whereby the presence of the cathode layer of the Crookes dark space and of
the negative glow were highlighted. It was noted in particular that in the
RCD, contrary to what happens with a flat-cathode Re discharge, the more
intense radiation is found at greater distances from the dark space boundary,
with a maximum of intensity along the axis at a given (and relatively low)
value of the gas pressure. The occurrence of Al lines emitted by atoms
ionized once and twice was also stressed as a characteristic of the spectrum.
The outstanding and exhaustive information thus obtained by Paschen was
a further proof of the innovative character of the ReD and to a certain
extent marked a turning point in its history as it was brought to the attention
of a much larger number of researchers. In two subsequent papers, a com-
plete interpretation of the spectra emitted by an Al RCD in an atmosphere
of Re was achieved, with particular regard to the lines of singly ionized Al
atoms. (5,6)
HOLLOW CA THODE DISCHARGES 217
During this same period, other groups started devoting themselves to

this branch of spectroscopy. Substantial contributions in fact were made
by Schüler and his associates for over 35 years. In one of his preliminary
communications, the potential fall in an HCD was thoroughly investigated
employing Al or Fe as the cathode material and Hz as the carrier gas (also He
in the case of AI).(7) The major conc1usions reached at the time established
unequivocally that virtually the entire potential drop resides within the
cathode cavity with minimal anodic fall (which is in no case mandatory to
sustain the discharge ). Simple relations were also established between the
ionization energy of a given gas, the ratio of electron free path to that of
gas atoms, and the potential fall at the electrodes. It is worth mentioning
that the experimental configuration adopted by Schüler for his discharge
tube was based on a co axial mounting of two hollow cylinders, the much
smaller inner one being the cathode and the outer one the anode. Since
this brilliant beginning, Schüler and co-workers successfully achieved in the
subsequent three decades comprehension and applicability of the HCD. (8-12)
His originality was also demonstrated in the use of glow discharge (GD) to
obtain emission spectra from organic substances through electron impact
excitation processes. (13-15)
In addition, among the founders of the technique one should certainly
inc1ude the name of Sawyer. After some time spent at Paschen's laboratory,
he soon developed his own approach to the study ofHCD. Extensive investi-
gations were carried out to assess the discharge behavior of numerous ele-
ments, such as Al, Ca, Cu, Gd, Hg, Mg, Tl, and Zn, and its dependence on
different noble gases (Ar, He, and Ne).(16) The parameters goveming the
excitation of metals in the HCD were thoroughly examined taking into
particular account the ability of cathodic material to sputter. It became
undeniably c1ear that the complexity and variety of possible situations inside
the glow would not always allow results to be predicted. Sawyer is addition-
ally credited with "exporting" the HCD technique from Europe to the
United States while it was still relatively unknown and in doing so set the
premises for its widespread dissemination.
In the 1930s the concoction was thus reaching the right cooking tem-
perature as some preliminary steps were being taken toward quantitative
analysis. In 1934, Konovalov and Frish had already demonstrated the suit-
ability of the HCD for detecting Ar and N Z.(17) It was only 13 years later,
however, when it was possible for McNally and co-workers to establish that
quantitative relationships could be derived between spectral emission and
concentration of analytes present in the cathodic sampleY8) These authors
were able, by means of a water-cooled HCD source, to determine F in metal
oxides in the visible region at concentrations as low as I J1g g-I with a sampIe
weight of only 10 ng.
The facts that ensued after this breakthrough belong more to the recent
chronicle than to history and will therefore be dealt with, to the extent that
218 CHAPTER 6
is necessary to better evaluate the present state of the art, in the following
sections.
6.1.2. Present Impact on the Progress of Spectroscopy
Undoubtedly, the first real success of the HCD should be ascribed to

its use in atomic absorption spectrometry (AAS) as a device capable of
emitting particularly sharp and stable spectral lines.(19,20) The birth and
sudden widespread diffusion in the 1960s ofthis technique were in fact made
possible by the availability of a source with such unique characteristics. One
might wonder why the fortune of HCD should have been bound to the
development of another methodology where the spectral properties of the
discharge simply played an ancillary role and were not directly exploited
for analytical purposes. This was certainly related to the profound crisis
undergone by emission techniques in general in that period, mainly because
there was increasing awareness of their inadequacy in the face of the emerg-
ing analytical challenges. This distrustful attitude has in practice favored
the search for more promising alternatives while discouraging innovative
investigation in the field of atomic emission spectroscopy (AES).
The enormous surge forward given by AAS to modem analytical chem-
istry is, however, out of the scope of this survey, which instead focuses
essentially on more specific applications of HCD in AES. The last decade
witnessed a genuine renewal in this branch ofspectroscopy, mostly thanks to
the maturity reached by the inductively coupled plasma (ICP) methodology,
paired to a sort ofrediscovery ofHCD. The host ofelement-related problems
that today must be faced and that require accurate and precise analyses to
be carried out in an endless variety of matrices has decidedly contributed to
its renaissance, not only because the HCD-AES is an inherently multi-
elemental technique, but also because it offers benefits not easily achieved
through other sources of excitation. (21)
From this standpoint one striking feature is the ability of HCD to be
employed for the analysis of solid sampIes as a direct consequence of its
basic sputtering mechanism. This property makes the discharge at least com-
plementary to what is afforded by the much more popular ICP source, in
its turn particularly suited to assay liquid specimens. This is, however, only
one of the aspects characterizing the wide area of coverage of the HCD, as
gases and solutions (mainly, but not only, in the form of dry residues) fall
within the realm of this atomizationjexcitation system.(22-27) In turn, it is
precise1y the possibility of separating these two fundamental processes typ-
ical of any low-pressure discharge (LPD) that has recently opened a few
major new avenues. Both mechanisms, in fact, display unequaled merits of
their own, exploitable to the benefit of other investigative approaches. The
facets ofthis analytical potential will be treated in detail in Section 6.4; here
HOLLOW CATHODE DISCHARGES 219
it suffices to recall that sputter atomization by means of LPDs is an expedient

means to bring to the vapor state any kind of sampie for subsequent intro-
duction into other spectroscopic sources, (28) whereas an alternative strategy
has led to the excitation within the glow of the ReD of thermally volatilized
materials, as in furnace atomic nonthermal excitation spectroscopy
(FANES)(29) (see Chapter 7).
The renewed interest of the experimentalist in LPDs is moreover well
documented by the increasing frequency with which ad hoc meetings or
special symposia hosted by broader-coverage conferences have been success-
fully organized in the last few years. Among these, mention should be made,
for the time being at least, of the Winter Conference on Plasma Spectro-
chemistry series, of the traditional F ACSS congress, and of the periodic
Anwendertreffen held in Germany. All this rapidly developing background
presumably will open a brilliant future for the RCD source in many respects.
The properties with which this discharge is endowed make this prospect
more than likely and the technical evidence provided in the sections to follow
will give it the necessary credibility.
6.2. Fundamental Aspects
6.2.1. Working Characteristics of a Ho/low Cathode Discharge
Most of the basic characteristics inherent in the functioning mechanism

of LPDs are obviously shared by the RCD and are, therefore, exhaustively
dealt with elsewhere in this volume. Consequently, the present discussion
will focus chiefly on those peculiarities that are unique to the RCD. As
already briefly touched upon, the inner configuration of the cathode can be
thought of as aseries of planar longitudinal narrow cathodic elements placed
in parallel fashion along a main axis so that each lamina has its specular
counterpart with respect to the said axis. The prototypical hollow cathode
set up by Paschen(3) can thus be regarded as two pairs of opposite flat
electrodes forming a box open on both sides. Generally speaking, the macro-
scopic outcome of such an arrangement is that the characteristic lumin-
escence of the GD arising from that side of the plane cathode facing the
anode coalesces, in the case ofthe RCD, with that produced by the opposite
segment, provided that the gap between these two is small enough and the
rare gas pressure is within a suitable interval. In other words, the brightly
emitting zone occupies the space contained within the set of cathodic
laminas. By extrapolation to an infinite number of these laminas, each with
an infinitesimal width and a given length, it is easy to understand that the
overall structure will evolve toward a hollow cylinder, which, if closed at
220 CHAPTER 6
4 2
~----1III--_ _---.J
i
3
Figure 6-1. Schematic diagram of an ReD setup. I, hollow cathode; 2, anode; 3, power
supply; 4, housing.
one end, gives rise to the configuration known as hollow cathode. The basic
scheme of an HCD lamp is given in Fig. 6-1.
The advantages of the cuplike shape of the negative electrode are
remarkable: (1) the previously mentioned formation of the negative glow
takes place exclusively inside the cavity, and Is accompanied by an increase
in current density of no less than three orders of magnitude;(30) (2) closely
related to the former property, the cathodic fall is contained almost entirely
within the bore, where virtually all the relevant processes occur (atomization,
ionization, excitation, emission). All tbis implies that the sampie, be it the
cathode itself or a specimen placed within the hollow, is confined to the
inner volume when it volatilizes and is capable of leaving the cavity only
with difficulty. The practical consequences of such a pattern are very expedi-
ent from many points of view, since consumption of the sampie loaded into
the cathode as weIl as loss of cathode material are minimized.
In contrast to the behavior shown by planar GDs, the mass ablated by
the impinging ions and neutrals undergoes repeated cycles of volatilization
and redeposition before eventually escaping the bore. This characteristic can
be extremely beneficial for a variety of purposes, primarily including
analytical determinations. In this context it is useful to recall what was stated
in a memorable review published in 1984 by Mavrodineanu.(31) According
to this author and in full agreement with present knowledge in the field of
LPDs, the features of such devices can be traced back to:
• The production of narrow spectrallines for atoms, ions, and radicals

of all known chemical species
• The limited extent of self-absorption
• The nonthermal and practically nonselective mechanism ofvolatiliza-
tion induced by the impact of particles from the carrier gas
• The use of a noble gas at low pressure to sustain the discharge, this
giving straight access to the vacuum ultraviolet (VUV) region
• The noticeable stability ofthe process, as the current-voltage charac-
teristics c1early allow one to assume
• The relatively low working temperature even when no cooling of the
system by means of a circulating fluid is provided
• The silent performance of the discharge
The HCD shares these properties to the highest degree, particularly the
subtle natural width of the emitted spectral lines, which is a major conse-
quence of the virtual absence of both the Doppler effect (linked to the low
temperature and pressure of the plasma) and the Stark effect (due to the
very weak electric field existing in the negative glow). Together with the state
of nonlocal thermal equilibrium (non-LTE, or NLTE), the above peculiari-
ties confer to the HCD unique features, as reflected in the nature of its basic
processes.
6.2.2. Plasma Formation
The abundant literature available on the general mechanism of plasma

generation in LPDs provides, on the one hand, evidence of the fact that the
fundamental principles are well understood, while on the other hand, it
testifies to the complexity of the whole phenomenon and its dependence on
a large number of physical and electrical parameters, which to a certain
extent make some minor facets not predictableY9,32-34) This also holds true
for the HCD, where the geometrical confinement of the glow makes matters
less simple. In brief, filling the sealed tube that houses the ReD circuitry with
a noble gas at pressures from tenths to thousands of pascals and applying,a
direct current from a few milliamperes upward, triggers the onset of the
discharge and the attendant luminescence. From the standpoint of
voltage-current characteristics, as a special case of LPDs, the HCD places
itself in the region pertaining to GDs with abnormal cathode fall. (35) In other
words, there is an increase in voltage paralleling that of current intensity,
less pronounced, however, than the one typical of the abnormal planar
model. In this connection it is worth recalling that the three basic quantities
presiding over the HCD operation are gas pressure, voltage, and current
intensity. These are so interrelated that only two can be freely chosen, while
the third is unequivocally fixed by them.
Under the effect of the applied electric field the gas ionizes as the few
electrons and ions always present start heading toward the respective
electrodes and, in so doing, collide with other, mostly neutral, species. In a
cascade process, further charged particles are thus generated until equi-
Iibrium is attained between the two contrasting actions of ionization and
222 CHAPTER 6
recombination. A possibly significant contribution to this overall balance is

given by electrons ejected from the cathode after the impact of energetic
species, as they may play an important role in sustaining the discharge. This
is not so for the thermionic electrons, whose aliquot is normally negligible
unless the temperature of the cathode is allowed to rise, as may be desirable
under circumstances and for preestablished sets of working conditions.
Ultraviolet photons also significantly contribute to the generation of
electrons and are obviously able to travel in any direction in the negative
glow.
Apparently, in an RCD configuration the momentum-gaining particles
and photons have a very small chance of not falling onto the cavity surface
and the electron-production efficiency is significantly higher than for the
planar discharge. A number of physically different zones can be perceived
in the plasma moving away from the cathode toward the axis, the sequence
of which coincides with that displayed by Hat cathode arrangements.
The first of these zones, spreading over the inner surface, is called the
Aston primary (or dark) space. It is fOllowed by a weakly luminous region
and then by a second dark space (named after Crookes or Rittorf). The
bulk ofthe discharge is, however, filled by the intensely luminescent negative
glow. Subsequent to these zones, LPDs are generally characterized by the
presence of the Faraday dark space, the positive column, the thin anodic
dark space, and the anode glow. A more detailed treatment of their function
and relative importance can be found in other chapters of this volume, even
though in this context it is important to stress that some of the mentioned
regions may actually disappear (e.g., the Faraday dark space), due among
other reasons to physical constraints.
What certainly cannot be eliminated (being essential to the discharge
functioning) are the Crookes dark space and the glow region. Both zones
interact in the case of an RCD in a very particular way, as the electrons are
accelerated across the former, where virtually all the cathodic fall concen-
trates. Some of them gain sufficient kinetic energy to travel through the latter
and enter the opposite portion of the dark space, where they experience a
reversal of direction. The same electron can thus accomplish several passages
across the glow and exert an ionization action far larger than that of the
planar system before eventually being quenched. The cathode dark space
acts therefore as an electron gun whose function is to transfer energy to the
equipotential plasma in the glow region. The yield of secondary electrons
genera ted in the dark space increases steadily with the atomic weight of the
inert gas. For He, it is almost negligible and, consequently, a large part of
the primary electrons can cover more than once the whole distance between
the cathode walls. On the contrary, the ionization cross section of Ar makes
this yield of the same order of magnitude as the primary electron current.
Quantitative predictions are rather complex; surely, type and pressure
of filler gas playa decisive role in that they determine the yield of secondary
electrons from the cathode, this quantity also depending on the nature of
the material. Photoelectric emission is estimated to be the main cause of the
HCD phenomenon (36) and increases in the order Xe (9.57), Kr (10.03), Ar
(11.83), Ne (16.85), He (21.22) (values in parentheses are the photon energy
in electron volts at selected wavelengths).(29) For He, photoelectric emission
can reach even 70% of the total. Depending on the type of rare gas and
experimental parameters, a different proportion of metastable atoms is also
expected, whose role in the excitation process may be of primary importance.
Summarizing, there are plenty of events likely to actively participate in the
generation and sustenance of the HCD plasma, as set forth in Table 6-1.
The plasma that ensues from this multiplicity of either competing or
reinforcing processes is highly energetic and provides an ideal environment
for the excitation offoreign species. The remaining paragraphs in this seetion
will elucidate some of the aspects implied in the above capability. Further
and exhaustive information on this subject matter can be found in the
relevant literature. (e.g .• 37)
6.2.3. Energy Distribution

Whatever the HCD variant used, one common feature emerges, namely,
considerable variations of the electron energy distribution from the one
predicted by the Maxwellian model. Experimental observations made by
Table 6-1 Main Ionization Processes in RCD

Ionization type Eventa
Primary
Electron impact x + e- --> X+ + 2e-
X*+e- --> X H + 2e-
Secondary
Penning type G m+e--->G++2e-
G* + E* --> G + E+ + e-
Gm + E -->G + EH + e-
Gm + E* --> G + E+ + e-
Charge transfer G+ + E -->G + E+
G;(fast) + Gb(slow) --> G;(slow) + Gb(fast)
Associative Gm + E --> GE+ + e-
Photoinduced X* + hv --> X+ + e-
Cumulative E+e--->E++2e-
aE. element sputtered [rom the cathode; G, gas atom; X, either E or G; ., excited state; m,
metastable state.
224 CHAPTER 6
various authors(38,39) using probes and electrostatic analyzers lead to a self-

consistent interpretation of the electron energy distribution as expressed by
the equation
(6-1)
where Ee is the electron energy. In the case of He and for settings of 270-400
Pa and 30-60 mA, the functionj*(E) was found to change very little and
to predict that most electrons possess energy below 20 eV.(38-40) The findings
obtained at the time by the same team pointed to the presence of a number
of fast electrons (i.e., with energy >20 eV) higher in the negative glow than
in the vicinity of the anode and, surprisingly enough, even a large number
of slow electrons. The noticeable fraction of electrons possessing energy elose
to the cathodic fall accounts for the previous statement about the similarity
to the planar abnormal discharge.
At lower gas pressures (10-100 Pa), the fast electrons were found to
split into two groups, containing very fast (not undergoing any collision)
and fast electrons (with little residual energy of about 4 eV available for
further ionization and excitation), while the slow electrons were relatively
thermalized at approximately 0.5 eV.(41) It is only the first group that is
responsible for the high current intensities characteristic of an HCD with
electrons repeatedly oscillating from point to point of the inner wall.
All these observations lead to the conelusion that for a given set of
electrical parameters there is a particular press ure of the carrier gas at which
the portion of electrons with higher energy reaches the maximum. Gas
pressure, therefore, exerts a strong influence on the excitation efficiency, the
latter quantity being also a function of the atomic energy level. Typically,
number densities for fast and slow electrons fall in the range of 2-20 x 108
and 2-20 x 109 cm-3, respectively, and their ratio can be as high as 1: 50.(42)
The radial electron number density is not uniform, but shifts from a distribu-
tion with a maximum along the axis for pressures as low as 10-30 Pa to an
annular one with the minimum on the axis for higher pressures
(50-100 Pa).(43) This pattern is also shown in relation to increasing current
intensities.
Positive ions, in turn, float in the almost field-free glow until they enter
the Crookes dark space and are then accelerated toward the cathode surface.
In the planar, abnormal GD, ions are prevented from acquiring the full
energy of the voltage drop by the occurrence of the so-called symmetrical
charge exchange, i.e., the capture of an electron by the fast ion colliding
with an atom of the same species. The net balance will be an ion with thermal
speed, whereby the overallloss of kinetic energy elearly depends on the total
number of collisions, directly related to the gas pressure and the length of
the dark space.
This phenomenon is, however, much less serious for an RCD, where a
considerable number of ions can display energy up to 80% of the entire
cathodic drop. (44) A justification for this difference lies in the very thin dark
space peculiar to this type of discharge, which allows only a limited number
of collisions to occur whatever the pressure iso Moreover, similarly to what
has been observed for electrons, the doughnut configuration is also present
for the ion density, in good conformity with the experimentally ascertained
practical neutrality of the plasma. Besides this, an annular distribution of
brightness of light emitted by the RCD is visible, as would be suggested by
the higher density of energy-rich charged species toward the edge of the inner
space and the consequently more suitable local conditions for excitation and
emission.
6.2.4. $puttering Process
The sea of corpuscles hitting the cathode surface exert a major influence
in that single atoms and clusters are released in a physical state resembling
that of a gaseous phase, albeit the temperature is much lower than that in
true volatilization.(45) The temperature of a working cathode rarely reaches
800-1000 K. Reat production, however, is always concomitant with the
erosion process and may be intentionally augmented through adoption of
suitable working parameters to facilitate the extraction of volatile species
and hence the separation from a refractory matrix.
This kind of vaporization is termed sputtering, as suggested by J. 1.
Thompson in the early 1920s, and is mainly of a physical rather than chemi-
cal nature, since an interaction occurs between the impinging particle and
the lattice structure of the cathodic material without giving rise to significant
formation of volatile compounds which can evaporate on their own. (46) The
ablated species are, as a general rule, neutral as ions are inevitably attracted
back to the surface (the probability of ion ejection can be as low as 1/100).
For impacting ions of 100-300 eV, the kinetic energy of the sputtered
material is relatively small, with part of it immediately diffusing back to the
solid phase. Under such conditions, the released species may travel at a
speed of roughly 3 x 105 cm s-) .
In the pseudovapor formed in this way, molecular aggregates are also
detectable, whose origin can be ascribed to the presence of both exogenous
impurities in the cavity and the sampie loaded on the cathode bottom, if
any. While it is obvious that the crystalline structure is important in the
establishment of the overall sputtering rate, the atomic mass of the gas
employed is of equal worth, as the closer it is to that of the material to be
sputtered, the more efficient is the erosion process. Righ-purity Re is [rom
226 CHAPTER 6
this point of view of poor consequence and the reported use of this gas
together with appreciable ablation may well have been due to traces of other
gases like Ne.
Although target atoms to a certain extent are ejected randomly, it has
been ascertained that the preferential direction of departure is specular to
that of the incident projectile. The duration of the impact may be as short
as 10- 12 s; during this time the colliding partic1e can penetrate the target
surface to a depth of even 0.5 nm. The thickness crossed by the incoming
partic1e is approximately four times the escape depth (the layer from which
the target atoms are re1eased). In spite of the internal collisions suffered,
enough energy is available to eject atoms from the solid phase (normally,
from 12 to 35 eV is necessary for this purpose). The ablation rate can reach
up to 200 monolayers min- I and is greatly affected by the presence of
quenching species like oxygen and water vapor.
So far, numerous published studies have aimed at elucidating the diffu-
sion mechanism of the sputtered sampIe cloud and its density in the cathode
bore.(47-50) Nonetheless, the intricacy of the factors involved in the global
phenomenon allows for few generalizations, such as enhancement of the
vapor density of cathode material with increasing tube current, often linked
by a linear correlation.(5I) Similarly, an increase in gas pressure may cause
the maximum density of sputtered material to shift from the cathode axis to
the walls. If heat is efficiently dissipated and high-vacuum conditions are
attained so as to minimize backscattering phenomena, a general formula for
the sputtering yield r has been found to hold :(34)
r= (knRönomME)j(m - M) (6-2)
where k is a constant, Ro the closest distance of approach under hard-sphere

conditions, no the density of target atoms, E the energy of the impinging
body, and m and M the masses of the impacting and leaving particles,
respectively. On the other hand, more empirical expressions have been
worked out to better account for the experimentally observed behavior in
real-world cases. (52)
Thermal equilibrium of the discharge gas with the incoming pseudo-
vapor is attained after very few collisions and hence can be considered
complete at a distance of only a few tenths of a millimeter from the cathode
surface. It is likely that this process ends when the sputtered material is still
in the transition region at the border between the cathode dark space and
the negative glow. In the long term, the cathode inner cavity will pro-
gressively modify its shape under the effect of sputtering, approaching ever
more that of an hourglass because of both redeposition phenomena and
definitive loss of material through the cathode mouth.
6.2.5. Excitation Mechanism

Since the early work on the RCD, there has been wide consensus that
excitation and emission phenomena normally do not obey classical,
blackbody-based spectral distributions. In connection with this, one aspect
is of great relevance for a correct understanding of the correlations existing
between the dynamic atomic-molecular interactions and spectral emission,
namely, the nonthermal energy distribution with the ensuing different tem-
peratures for the various microscopic states. In more detail, the temperature
of the filler gas atoms can be up to 1200 K, as determined for wide intervals
of current intensity,(53) while several thousand degrees Kelvin marks the
possible electron temperature, with a manifest lack of LTE. The average
path length of electrons in the negative glow is a function, among others, of
carrier gas type and pressure and can range from a few millimeters to several
decades as the gas atomic mass increases.
Considering the multiplicity of parameters that are active in the RCD
phenomenon, the overall excitation process will necessarily be the outcome
of several concomitant actions, a few of which may prevail over others
according to the operational conditions adopted. Collisions of the second
type, in the broadest meaning of the term, are generally acknowledged as
one of the major contributors to this effect, while a less essential role can be
ascribed to collisions of the first kind, i.e., with electrons. As regards the
former, the assumption behind it is that conversion of excitation energy into
kinetic is minimal and that the total spin of colliding atoms is conserved.
On the basis of present knowledge, a number of reactions in the gaseous
phase are deemed to occur, as listed in Table 6-2.
Attention is drawn to the important function that in many instances
metastables from the carrier gas may have in producing excited states. Just
to mention the most frequently employed gases, Ar has important metastable
states at 11.49 and 11.83 eV, Ne at 16.85 eV, and Re at 19.77 and 20.55 eV.
Whether excitation is principally due to electron or metastable-atom colli-
sions is a matter of debate and the response largely depends on the particular
situation at hand. (47,54,55) The nonthermal character of the interactions per se
is a definitive advantage that ultimately leads to considerably favor particular
emission transitions, with the possibility of reaching high spectral levels
Table 6-2 Main Excitation Processes in HCOa

er
x + ---+ X* + e; E+Gm ---+E* +G
X+ + er ---+ x H + e; G+er ---+ Gm+e:
X+hv ---+X· G+ + E + e- ---+ G + E*
aE, element sputtered from the cathode; G, gas atom; X, either E or
G; er, fast electron; e;, slow electron; ., excited state; m, metastable
state.
228 CHAPTER 6
provided suitable parameters and operation modes are chosen. Strong upper-
level first- and second-spectrum emissions are, however, more the rule than
the exception, wherefore nonmetals become wholly accessible to the HCD.
Concerning the radial distribution of radiation emission within the cathode
bore, conclusions reached so far have not been unambiguous, even though
a general correspondence can be foreshadowed with the pattern followed by
electron and ion density touched on in Section 6.2.3. Intensity of the emitted
radiation and electron density are linked by the relationship worked out by
Falk and Lucht :(56)
(6-3)
where dIe is the whole of the radiation emitted by the infinitesimal volume
dV, h Planck's constant, Vo the central frequency of the spectralline, n e the
number of electrons, n the number of atoms in the base state, Qa(Ee) the
cross section of excitation, and viEe ) the velocity of the electrons.
Of all the possible uses of the HCD, analytical applications in AES are
of consolidated importance. Basic to them is the assumption that the vapor
obtained through sputtering from the cathode or sampIe contained therein
and to be excited in the plasma has a stable composition over the entire
discharge duration such as to reflect that of the solid phase if this is multi-
elemental. Normally, this is achieved after a very short predischarge time,
during which the elements with higher sputtering yields are progressive1y
depleted from the cathode surface until an equilibrium is established.
6.3. Main Variants of Hollow Cathode Tubes
6.3.1. Preliminary Comments
The proliferation of HCD lamp configurations for a variety of purposes

has encouraged, from the very beginning, research in this field and has been
limited only by the inventiveness of the experimentalist in facing a given
analytical or technological challenge. In categorizing such devices different
criteria can be followed, each being equally valid in describing the various
models so far developed. Consequent with the multipurpose-oriented
approach chiefly adopted here, it seems more expedient, however, to survey
the principal types of tubes according to their general design and operation
mode rather than to classify them according to their specific uses, since a
given instrumental version is liable to be employed in different sectors.
With regard to the analytical applications (undoubtedly a major portion
of the whole scope), description is focused exclusively on exploitation of the
HOLLOW CA THOOE DISCHARGES 229
ReD in AES, deliberately ignoring, for the reasons given in Section 6.1.2,
its use as a spectralline source for AAS. It goes without saying, at any rate,
that every technieal improvement in stability, emission intensity, and lifetime
of ReD lamps is obviously a potential benefit for this technique.
6.3.2. Selected Models

One fundamental distinction when referring to models is whether the
cathodic block of the lamp has been designed to be cooled by a circulating
fluid, be it plain water, frigorific solution, or liquid nitrogen. The final deci-
sion on the use of a cooled or hot ReD lamp clearly depends on the desired
effect: heat dissipation allows the purest features of the discharge to emerge
in terms of spectralline sharpness and sampling of the cathode (or specimen)
surface through real sputtering, whieh is most frequently the case. (27) This
may not be so when other requirements become a priority, such as the need
for selective volatilization of certain components to separate them from an
analytically disturbing matrix, (57-59) or for a complete and instantaneous
vaporization and atomization of the cavity-hosted sampie to further excite
it in the ReD plasma, as in the F ANES approach. (29,60,61)
A solution intermediate between the cooled and hot ReD is that
reported by Eichhoff(62) with the so-called transitional version. In it, thermal
effects prevail on the cathode bottom and progressively disappear as effi-
ciency ofheat removal increases along the cathode mantle. The sampie vapor
thus experiences transport into a cooler, more ReD-genuine environment
and the spectra emitted are exempt from the thermal influence. Both this
artifact and the F ANES system coincide (in spite of the would-be important
differences) in that production of the sampie cloud and its excitation are
clearly separate (spatially in the former and temporally in the latter), with
the ensuing possibility of optimizing the two processes independently from
one another.
To try to list all of the types of ReD lamps reported to date in the
literature on the basis of their different geometrical arrangements would
be an overwhelming task. Therefore, it seems much more advisable to
spotlight those tubes that deserve particular attention owing to their
innovative characteristics at the time of their introduction, referring the
reader to the appropriate bibliographie sources for a more methodical
coverage.(21,28,31,63,64) Starting with the already quoted prototype by
Paschen, (3) an almost endless series of lamps have been proposed, among
which mention should be made of the one devised by Schüler and Gollnow
in 1935.(10) This displays, in fact, the basie features of a demountable device,
which has since been adopted by others.
Representative of the class of deep-cooled lamps is the twin hollow
cathode set conceived by Török and Zaray,(55) which employs liquid nitrogen
230 CHAPTER 6
,-J=[1=l-
2
~_~\l ~~--1-;:~~nII
I~
IC 2
Figure 6-2. Sketch of the deep-cooled, twin hollow
cathode assembly. I, carrier gas pathway; 2, twin hol-
., 3 low cathode arrangement; 3, liquid nitrogen; 4, Dewar.
4-~,L-f--- Reproduced by permission ofPergamon Press, with fuH
acknowledgment to the authors of Ref. 55.
as a coolant. The system, depicted in Fig. 6-2, was coupled to an

interferometer-spectrometer assembly to thoroughly investigate the excita-
tion process in an HCD in view of its exploitation in trace element analysis.
In turn, the equipment shown in Fig. 6-3 was set up to ascertain whether
the spectrum of singly ionized Cu was convenient for wavelength calibration
in the VUV region.(65) Several Cu(II) lines among the 30 tested were found
to be sufficiently narrow and symmetrical for use as wavelength standards.
In fact, provided that current intensity and He carrier gas press ure were in
the 200-300 mA and 400-800 Pa ranges, respectively, accuracy was better
than ±0.01 pm, whereas no significant shifts were observed with the gas
A
I
5
"
Figure 6-3. Scheme ofthe liquid nitrogen-cooled ReD tube according to Freeman and King.
I, MgF2 window; 2, glass sleeve; 3, liquid nitrogen; 4, expanded polystyrene Dewar; 5, carrier
gas inlet; 6, aluminum anode cylinder; 7, hollow cathode. Reproduced by permission of IOP
Publishing Ltd., with fuH acknowledgment to the authors of Ref. 65.
flow. Another interesting variant ofthe deep-cooled HCD tube was launched
by Appleblad and Schmidt to study the emission spectra of diatomic
molecules .(66) In particular, this apparatus (Fig. 6-4) was devoted to investi-
gation of the CuO excitation under various operating conditions in order to
optimize the emission of the molecular spectrum.
Among the many attempts made to better understand the plasma
behavior in an HCD, it is worth mentioning the stationary hollow cathode
arc (HCA) developed by Bessenrodt-Weberpals et al.,<67) a detailed scheme
of which is given in Fig. 6-5. Attachments like this are becoming more and
more popular for diagnostic development, spacecraft charge control, plasma
wake-field accelerators, and others, and allow for the extraction of high-
intensity, low-emittance, steady-state electron beams. The one described here
achieves discharge stabilization by means of a longitudinal magnetic field
and permits the influence of the anode on the spatial distribution of densities,
temperatures, velocities, and excitation of plasma particles to be ascertained.
For this purpose the authors resorted to powerful investigative techniques
such as Thomson scattering, laser-induced fluorescence, and collective
scattering diagnostics. The hot and dense plasma that forms in such cases
instantaneously expands outside the cathode cavity and hence brings about
a dramatic drop in density and temperature of the charge carriers, this
phenomenon being accompanied by an equally impetuous outward radial
diffusion of particles. Proceeding toward the anode, electron density was
found to decrease by roughly one order of magnitude (from 3 x 10 13 cm- 3
to 0.4 x 10 13 cm- 3). In turn, acceleration of electrons in front of the anode
causes their temperature to considerably increase and consequently thermal-
2 3 4 5 6 7
11 10 9
Figure 6-4. Cross-sectional view of the composite wall HCD tube. 1, window; 2, nylon
insulation; 3, anodic brass cylinder; 4, glass tube; 5, stainless steel cylinder; 6, polystyrene
thermal insulation; 7, copper cooling flange; 8, stainless steel cooling bottle; 9, pump connec-
tion ; 10, hollow cathode; 11, conical copper socket. Reproduced by permission of the University
of Stockholm, Department of Physics, with full acknowledgment to the authors of Ref. 66.
232 CHAPTER 6
2 3
11
Figure 6-5. Hollow eathode are device. 1, hollow eathode; 2, eoils; 3, anode; 4, pumping
system. Reprodueed by permission of the Institute of Eleetrieal and Eleetronies Engineers, with
full aeknowledgment to the authors of Ref. 67.
ization to be attained. The conditions that favor the establishment of the

RCA (e.g., overall voltage 40 V, current intensity 60 A, magnetic field 0.074
T, Ar working gas pressure 1.3 Pa) are quite different from those adopted for
the conventional RCD and open new paths for yet unheard-of technological
applications.
The analytical constraints set by the very limited availability of certain
classes of sampIes for clinical testing, such as spinal cord or renal tubule
fluids, have stimulated the miniaturization of RCD sources, such as the
version described by Williams and co-workers. (68) With the lamp shown in
Fig. 6-6, they were able to quantify Ca, K, Li, and Na at the sub-picogram
level (volumes from microliters to nanoliters were requested). The method
thus seems able to compete with electrothermal AAS, not considering its
multielemental capability.
Atomic fluorescence spectrometry (AFS), once a technique considered
a hybrid of AES and AAS, now enjoys a dignity of its own, profiting from
the RCD properties, such as the brightness of the spectra emitted and the
small width of its lines, which are fundamental requisites for good analytical
performance. This warrants the efforts by many research groups to develop
RCD models apt to best fit the AFS requirements. Of particular value in
this context is the demountable, water-cooled lamp conceived by Rossi and
Omenetto, (69) illustrated in Fig. 6-7. The very favorable fluorescence-to-
scattering ratios and rapidity in interchanging the cathodes and regaining
discharge stability were the major advantages of this device, by means of
which detection limits of between 0.03 and 2 pg mI-I could be reached for
various elements (Ag, Co, Cr, Cu, Fe, Ga, In, Mg, Mn, Ni, Pb, and Tl).
3
I
/ @=
r-r--~----++--5
~~§g +_-++---6 7
11 1"0 9 12
Figure 6-6. Mini-RCD source. 1, cooling water inletjoutlet; 2, auxiliary cathode; 3, anode;
4, hollow cathode housing made of molybdate nylon; 5, carrier gas inlet; 6, quartz window;
7, water-cooled secondary electrode; 8, lamp holder; 9, quartz insulator with a hole; 10, spring-
held hollow cathode; 11, hollow cathode holder; 12, carrier gas outlet. The two components
in the upper part of the figure are seen as a cross section in the lower part. Reproduced by
permission of the Society for Applied Spectroscopy, with full acknowledgment to the authors
ofRef.68.
ONiCKEL
~ GlASS OR Ql./IJ.RTZ
~BRASS
_ STAINLESS STEH
DCOPPER
Figure 6-7. Layout of the water-cooled RCD tube for AFS. 1, glass-to-metal seal; 2, water
cooling; 3, anode; 4, mica shield; 5, O-ring. Reproduced by permission of Pergamon Press,
with fuH acknowledgment to the authors of Ref. 69.
234 CHAPTER 6
Sampling of both positive and negative ions for further studies of gas-
phase reactions is another area of investigation in which the ro1e of RCDs
has recently become ever more important. One of the first effective solutions
to this problem was presented by Roworka et al. (70) who devised the movable
apparatus of Fig. 6-8. The system, able to sampie the entire diameter of the
discharge, was bakeable at least up to 200°C (400°C was still considered
achievable). Residual water may in fact seriously limit the type of processes
that can be investigated. Reactions of positive ions of rare gases with neutral
species, neutral-radical distributions, radial spectral emission of molecular
and atomic states, and measurement of metastable density are among the
applications possible with this device. Similarly, the generation of an atomic
beam for facilitating spectroscopic investigations of Mo, Tb, U, Zr, and
other refractory elements was accomplished by means of the appliance
presented in Fig. 6_9.(71) This attachment was specifically employed to evalu-
ate the characteristics of U atomic beams by laser-induced fiuorescence spec-
troscopy to determine the beam divergence and the most probable axial
velocity.
Metastable population of Li (ls2s2p 4 pO level at 57.4 eV) and Na
(2p 5 3s3p 4 n 7 / 2 level at 33.1 eV) were measured using the RCD tube shown
in Fig. 6-10.(72) The design includes a 5-cm-diameter stainless-steel tube serv-
ing as the discharge anode and container of the metal vapor, and a 1.9-cm-
diameter, 30-cm-long stainless-steel tube functioning as the cathode and
lying inside the former. The assembly is heated to several hundred degrees
centigrade depending on the working conditions adopted. By using a tunable
Figure 6-8. Scheme of the bakeable hollow cathode/mass spectrometer system. I, quartz
window; 2, anode flange; 3, connection to p~p; 4, metal-to-ceramic fitting; 5, cathode; 6,
cathode flange; 7, cathode body; 8, Teflon washer; 9, quadrupole mass spectrometer; 10, mass
spectrometer housing; 11, multiplier; 12, Teflon washer; 13, mass spectrometer flange; 14-16,
18, shift mechanism; 17, hole probe assembly; 19, shutter; 20, gas inlet; 21, anode; 22, connec-
tion to pressure sensor. Reproduced by pennission of Elsevier Science Publishers, with fuH
acknowledgment to the authors of Ref. 70.
Figurs 6-9. Diagram of the hollow

cathode observation chamber appar-
atus. (a), hollow cathode assembly: I,
anode; 2, buffer gas; 3, cathode; 7,
cooling water; (b), observation (b)
, · --5
chamber : 4, laser ; 5, connection to
pump; 6, atomic beam. Reproduced
by permission of the American Insti-
tute of Physics, with full acknowledg-
ment to the authors of Ref. 71.
a b
Figurs 6-10. Sketch ofthe pipelike HeD apparatus. (a): 1, Li heat pipe zone; (b) : 2, anode;
3, cathode (the end view shows the plasma and arrows represent high-energy electrons). Repro-
duced by permission ofthe Optical Society of America, with full acknowledgment to the authors
of Ref. 72.
236 CHAPTER 6
probe laser, densities of 3 x 10 10 and 10 11 atoms cm- 3 were determined for

Li and Na, respective1y.
Approaches of this kind enrich present know1edge of the atomic levels
likely to be used for energy storage for extreme-UV lasers. An exhaustive
survey of the potential possessed by the HCD for the development of a new
dass of metal ion lasers reports the principal transverse type available at the
time, as shown in Fig. 6_11.(73) These devices are comparable in UV output
power to noble gas ion lasers although they require no more than 1/20th of
the normally employed threshold current intensities. Another configuration
was developed by Iijima(74) to improve the oscillation characteristics of the
He-Zn+ ion laser. The structure, shown in Fig. 6-12, permitted high voltage
to be obtained through restriction of the cathode inner wall by an insulator.
The importance of the HCD in this technological area should not be under-
estimated as attested by the multiplicity of configurations continuously being
elaborated.
Further possibilities offered by the HCD extend to deposition of coat-
ings or erosion of surfaces. Figure 6-13 shows, as an example, an assembly
constructed to deposit Allayers on a substrate, elucidating the influence of
cathode separation, bias voltage, and working gas pressure on film thickness
and structure. (75)
This hasty, largely incomplete scan witnesses the ingenuity of experi-
mentalists active in this field of research. Whatever the utilization of the
j(!),
\J~/ ,,
,,
-6
,
'~ ~'
~ ~
a b c
d e
Figure 6-11. Rollow cathode laser types with transverse excitation. (a) Schuebel model; (b)
rectangular slotted RCD; (c) circular slot RCD; (d) waveguide structure: I, envelope; (e)
hollow anode or obstructed glow: 2, discharge. Reproduced by permission of the Institute of
Electrical and Electronics Engineering, with full acknowledgment to the authors of Ref. 73.
a 1 2 3
: :z~~'m**'\*j;!IIj':::~I:' !':'{ll :
6 5 4
b
7 B
9
Figure 6-12. Cylindrical-type hollow cathode discharge after lijima. (a) longitudinal section:
1, anode (I); 2, perforated hollow cathode; 3, anode (II); 4, steatite pipe; 5, macor; 6, macor
space; (b) cross section at A- A': 7, anode; 8, steatite pipe; 9, perforated hollow cathode.
Reproduced by permission of the Japanese Journal of Applied Physics, with full acknowledg-
ment to the authors of Ref. 74.
RCD's various versions, the overall trend is more than ever toward an
increase of research and development.
6.3.3. Boosted Models
The need to enhance the emission output of RCD lamps can be consid-
ered a cornmon feature of all application sectors. From an analytical point
of view this is all the more true, because one of the major drawbacks of the
RCD as an excitation source lies in the fact that, in spite ofthe very favorable
intensity-to-background ratio, the net emission is not sufficient, under nor-
mal working parameters, to detect analytes at ultratrace levels. Therefore,
great effort has been and still is being devoted to the design of discharge
tubes capable of decidedly higher emission power without compromising
the valuable ReD features. Generally, this goal can be achieved either by
expediently modifying the geometrical configuration of the cathode or by
238 CHAPTER 6
o 0
.-
Figure 6-13. Schematic view of an HCD arrangement for ion plating. 1, variable position
insulated support; 2, mask; 3, cathode block; 4, high-voltage feedthrough; 5, insulated support.
Reproduced by permission of Elsevier Sequoia Science Publishers, with full acknowledgment
to the authors of Ref. 75.
enhancing the discharge by superposing an external form of energy compat-

ible with the one inherent to the HCD, without unacceptably altering its
essential positive features.
Examples of the first category include inserting a cone or, better still, a
rod inside the cathode cavity and co axial with it. The sampie to be assayed
is placed on its free end. This version was proposed by Papp and his associ-
ates as illustrated in Fig. 6_14.(76) The advantage ofthis configuration is that
3
9
Figure 6-14. Structure of the three-electrode HCD lamp. 1, hollow cathode; 2, coaxial rod;
3, cooling water; 4, insulator; 5, carrier gas outiet; 6, anode; 7, anode connection; 8, quartz
window; 9, carrier gas inlet; 10, cooling coil. Reproduced by permission ofthe Royal Society
of Chemistry, with full acknowledgment to the authors of Ref. 76.
it increases the sputtering action on the surface of the specimen by staying
in the very core ofthe plasma, and also because the bar, adequately insulated
from the cathode, can be kept at a more negative potential than the latter
and may be varied independently. The net result is a strong enhancement of
the ablation rate and the ensuing reinforced emission.
Further options are the microcavity RCD described in Section 6.3.2,(68)
which has a prominent predecessor in the model described by Czakow. (77)
The desired effect of intensifying the emitted radiation has its rationale in
the relation found by Novoselov and Znamenskii (78) linking in an inverse
fashion the intensity I of a spectralline with the cathode inner diameter:
1= const/d (6-4)
A greater diversification of design is encountered in the case of boosted

versions of the RCD. One of the first approaches simply consisted in
coupling a secondary discharge to the hollow cathode to increase the electron
density within the cavity.(79,80) This effect can be obtained either by a
thermionic source (a heated W filament or plate) or, more satisfactorily and
elegantly, through the lateral insertion into the main bore of an auxiliary
hollow cathode operated in the hot mode.
Superposition of transversal or longitudinal magnetic fields was also
deemed particularly promising and was primarily described by Popovici
and Some~an(81) and Rudnevsky and Maksimov.(82) Under this external
influence, charged particles are greatly deflected from their expected
pathways with the consequence that the density of slow electrons tends to
rise. The process favors an enhancement in the intensity of the spectrallines
with low excitation potential. Quite recently, another step ahead in this
area of investigation has been done by Pavlovic and Dobrosavljevic(83) who
applied a rotating magnetic field to the RCD (Fig. 6-15). Although clearly
2 H20 8
[I ~/
Figure 6-15. Diagram of the ro-
L2J 1
4
tating magnetic field ReD. I,

Pyrex tube; 2, glass joints; 3,
0. . ·.
quartz window; 4, aluminum
anode; 5, graphite cathode; 6,
tungsten electrical connection ; 7,
spherical ground joint; 8, soleno- · . . . . . .. +
5 ....... . 2
ids. Reproduced by permission of ·. . . . . . .
· . . . . . . .
Pergamon Press, with acknowledg-
ment to the authors of Ref. 83. 8 H20
240 CHAPTER 6
5 4
Figure 6-16.
Sectional scheme of the microwave-boosted
ReD. I, lamp body; 2, hollow cathode; 3,
magnetron; 4, quartz window; 5, pump and
2 gas ports.
a function of carrier gas type and discharge settings, the improvement of

emission intensity for analytes like es, Pb, Sb, and Zn did not in the best
case exceed 50-70% of the conventional source.
Another strategy consists in pulsing the discharge, which leads to higher
spectral intensity, lower noise, and often a reduction of linewidth.(84) The
current pulse may be superimposed to the normal dc continuous operation
mode or even replace it. The end application of the pulsed lamp (in AES,
AFS, or AAS) foreordains the optimization ofthe working parameters. Also
of value in this connection is the rf-boosted ReD proposed by Farnsworth
and Walters,(85) where the current pulse is used to sampie the cathodic
material and the rf burst reexcites the atomic vapor thus generated. The
main benefit is due to the noticeable reduction in self-absorption of reson-
anee lines, a phenomenon not at all negligible with a conventional pulse
discharge.
Finally, coupling with microwave (MW) irradiation is particularly
enticing as the increase in the signal-to-background ratio for the analytes
can be of one order of magnitude or more, while emission from the filler
gas is reduced and self-absorption almost vanishes.(86) The positive effect
is thought to be related-especially for Ar-to the larger production of
metastables from the rare gas consequent to the reciprocal reinforcement of
the two congenial types of plasma. A new model of the MW-coupled ReD
tube has been designed, as illustrated in Fig. 6-16, in which the magnetron
is placed within the discharge housing, thus eliminating any need for coaxial
cable and the attendant loss of MW energy due to transmission. (87)
6.4. Recent Applications
6.4.1. SampIe Introduction
In these and the following sections, attention will principally focus on

the developments, trends, and uses disclosed in the scientific literature over
the last decade with the purpose of citing those more representative. From
what has been said so far, it follows that the hollow cathode source, as in
general all GDs, is of considerable importance in the light of volatilization
of solid materials whose vapors can act as atom reservoirs for an array of
other instrumental techniques.(28) Today this is ever more true given the
astonishing development of modern analytical methodologies, which are
plagued by the need for regretfully time-consuming and tedious procedures
to dissolve the sampIes in a reproducible and straightforward way, when
liquid specimens are required.
For solid conducting materials, this is easily achieved with the HCD,
as all that has to be done is to machine the material into a suitable form
(e.g., an entire hollow cathode or a disc to be placed on the bottom of a
supporting electrode) and clean its surface of residual contaminants.
Generally, a short period of presputtering can remove even the slightest
traces of exogenous chemicals. If any powders or ashes are to be assayed,
they should preliminarily be pelletized with pulverized graphite, copper, or
silver at ratios from 1: 3 to 1: 9 in order to make them conductive and,
therefore, liable to attack from the bombarding particles of the noble gas.
The use of these binding agents is also beneficial from the standpoint of
smoothing off inhomogeneities which might seriously affect signal stability
as weIl as of minimizing matrix effects thanks to the presence of a dispersing
medium.
For the above reasons, if solutions are amenable to the technique at
hand it would apparently be preposterous to convert them into dry residues
for the subsequent atomization, were it not for the fact that chemical interfer-
ences from other constituents (in the first place the major and the minor
ones) can be noticeably reduced, while the signal is enhanced owing to the
analyte concentration in the resulting thin film of salts. In consideration of
this very small thickness, emission from the sampIe will be limited in time,
but its transient nature is less than a drawback with multielemental
approaches. Alternatively, solutions can be absorbed on porous supports
(graphite or pellet disks of pressed copper powder) for a longer duration of
the sputtering step.
Obviously, these considerations are particularly valid when the dis-
charge is operated to supply the vaporized sampIe to the very HCD plasma,
in other words, when the source simultaneously performs the two tasks of
vaporizing and exciting the material under test without any practical distinc-
tion between the two processes, or when the two functions are temporally
separated as in the F ANES system. (29) Beyond AES, however, the techniques
that at the moment greatly benefit from the availability of HCD devices
are AAS, AFS, plasma etching and deposition, mass spectrometry, laser
resonance ionization, and laser technology at large. (88-93)
242 CHAPTER 6
6.4.2. Analysis 01 Solids
One of the best explored application areas of the HCD, the determina-
tion of trace, minor, and major elements in massive or pelletized specimens
still records numerous and often innovative uses.
Resorting to the procedure of machining massive cathodes from the
specimens to be tested, brass and stainless steel sampies were analyzed for
their Cu and Zn, and Cr, Fe, and Ni contents, respectively, with accuracy
comparable to that ofICP-AES.(94) Analogously, the quantification of Au
in Au alloys containing from 79.91 to 97.87% of the element was carried
out employing the internal standardization method. (95) Both accuracy and
precision were about 0.03% (w /w), thus comparing favorably with fire
assays.
A comparison of spectral patterns obtained from dc and high-frequency
(27.2 MHz) HCD source was carried out for solid copper, brass, and steel
using Ar as the carrier gas. (96) Results showed that the electronic energy
distributions are similar for Ar II ions, while they may be quite different for
cathode elements. Spectral lines from Cd, Cu, and Zn showed increases in
emission intensity of more than one order of magnitude for the high-
frequency HCD. Mixture offluoride salts (NaF, PbF2, LaF3, ZrF4, Na3AIF6,
and K 2NbF 7) with Si02 were analyzed in a hot HCD tube through quant-
ification of the InF bands. (97) The method was able to measure F in the
1-104 J.l g g-I concentration range, the lowest amount detectable being 10 ng.
After introduction of metal chips of Pb, Sb, and Zn or metal salt
of Cs and Tl into the cathode cavity, the behaviour of these elements in
the rotating magnetic field HCD was reported by Pavlovic and
Dobrosavljevic.(83) The MW-HCD was also employed to analyze Cr, Cu,
Mn, Mo, and Ni in stainless steel with considerable improvement over the
conventional dc mode. (98,99) The briquetting procedure, in turn, was used to
discharge nonconductive powders containing trace elements like Ag, Al,
Cr, Ga, Mn, Ni, Pb, Sn, Ti, and Zn with detection limits between land
10 J.lg g-I (100) Cu, Fe, Mg, Mn, Ti, and Zn could be quantified in Al alloys
with the MW-HCD operation mode at concentrations from 0.01 to 6%.(101)
Finally, the differences in sputtering of solid sampIes between the conven-
tional HCD and the MW-boosted version were examined in view of the
benefits of the latter in the analysis of trace elements. (102)
On the basis of these examples, one can certainly infer how broadly the
HCD is being exploited in this specific field.
6.4.3. Analysis 01 Liquids
Apart from a few attempts to introduce solutions as such into the

discharge by spraying them, (24) the most widely adopted mode resorts to the
sputtering of the thin layer of salts left by the liquid sampIe after drying
within the cathode bore.
In Section 6.3.2 we discussed the ability of a mini-RCD source to quan-
tify Ca, K, Li, and Na.(68) Absolute detection limits of 0.70, 0.38, 0.88, and
0.72 pg, respectively, could be obtained. Several rare earth elements (Er, Ro,
Nd, Tm, and Y) were determined after drying of solutions within a Cu
hollow cathode operated in the hot mode. (103) Absolute detection limits
ranged from 0.02 (Y) to 0.25 (Nd) Jlg, with an improvement of about one
order of magnitude with respect to those attained in graphite cathodes. (104)
Given the occurrence of serious spectral interferences in steel matrices,
the determination of P in this material has to be performed after dissolution
of the solid and drying of the solution within thecathode cavity.(105) Using
Re as the carrier gas in the MW-coupled discharge, amounts of Pas low as
20 Jlg liter -I could be detected. The ability ofthe RCD as such or combined
with an MW field to analyze trace elements in solutions obtained after
digestion of biological materials was extensively verified by Caroli and his
team.(l06-110)
Cr, Mn, Ni, and Pb were analyzed in solution dry residues using both
a conventional dc and an MW-boosted RCD lamp.(99) The data obtained
with the two operation modes using Ar or Re as the filler gas point to
a significantly better detection power and sensitivity for the MW-RCD.
Desiccated solutions of fluoride salts were analyzed with an RCD setup
using graphite supporting cathodes.(97) Quantities from 10 to 1000 Jlg g-I F
were amenable to determination.
An elegant way to circumvent the problems posed by liquid sampies in
an RCD was proposed by Foss et al.,(27) who held the solutions at liquid
nitrogen temperature within the cathode and thus could sputter them in the
solid form. Seventy elements, variously grouped, could be analyzed in vol-
umes from 20 to 50 JlI extracting the ions from a miniaturized hollow
cathode and sending them to a double-focusing mass spectrometer. Detec-
tion power ranged from submicrogram to milligrams per liter.
A number of elements, such as Ag, AI, Au, B, Cd, Co, Cr, Cu, Fe, Mn,
Mo, Ni, Pb, Se, Sn, W, and Zn, were analyzed by means ofthe RCD-excited
ICP-AFS.(I1I) To achieve acceptable compromise conditions, rf power,
carrier-gas flowrate, and nebulization type were thoroughly examined and
optimized. The use of an ultrasonic nebulizer afforded detection limits in the
low microgram per liter range, with the exception of refractory elements,
which showed significantly higher levels (up to 200 Jlg liter- 1 for W). Further
improvements in ICP-AFS were gained by using an RCD pulsed at peak
current higher than I A.(1l2) Detection limits for Ag, Al, Cu, Cr, Mo, Sr,
and Zn were in the interval from 0.02 Jlg liter- I for Cu to 5.0 ng liter- 1 for
Mo. Molecular species were also investigated by probing the vapors sput-
tered from the Cu-Pb cathode in the GD source using an N 2 pump dye laser
to excite the fluorescence.(1l3)
Again, the suitability of boosted RCD sources for use in ICP-AFS was
scrutinized and compared with that of the ICP source. (114) The test was
244 CHAPTER 6
performed for CU, Ni, and Pb. It was concluded that the HCD is less suitable
than the ICP torch in terms of detection ability, although other advantages
are obvious, e.g., low cost of the device. In an attempt to overcome the
rather poor detection power of HCD-excited ICP-AFS (one to two orders
of magnitude lower than in the case of ICP-AES), some technical
improvements were made such as the use of ultrasonic nebulization and of
potentiated HCD tubes.(1I5) The potential of this analytical approach was
tested for numerous trace elements in water, blood, and fuel oils.
Although not as straightforward as direct nebulization, the procedure
generally adopted to analyze liquid specimens is reasonably reliable and
rapid, and the preconcentration undergone by analytes on the cathode sur-
face during the desiccation step enhances the intensity of the emission signal
resulting in better detection power.
6.4.4. Analysis 01 Gases

Although inherently suited to quantify components in a gaseous mix-
ture, very few such instances have been reported. This is all the more surpris-
ing if one considers that most studies have been, and are being, conducted
with gases. Nor should it be overlooked that often analysis of gases with
HCD means their quantification in asolid host matrix, as was reported for
oxygen in iron(1I6) or for nitrogen in refractory alloys.(l17) One significant
example of application of this kind of analysis, however, can be found in
the monitoring of artificial atmospheres in spacecraft. (22) Again not really a
determination of gases, the system devised by Iida(118) permitted the analysis
of Al alloys for the content of Cu, Fe, Mg, Mn, and Ni (besides Al) after
either single or multiple pulse laser vaporization of the solid sampie. Detec-
tion power was in the range of a few micrograms per gram.
Perhaps the most significant example in this field relates to the qualita-
tive and quantitative assay of Br, Cl, and F in a He gas stream.(119)
The system, based on the use of a microcavity HCD, permitted the determin-
ation of amounts in the low nanogram range and was seen as a more than
likely candidate for an element-specific detector in vacuum-outlet gas
chromatography.
No further significant analytical applications of this kind have been
reported recently.
6.4.5. Special Uses
6.4.5.1. Optoga/vanic Spectroscopy
An exhaustive study of the optogalvanic effect in the HCD plasma was

made by van Veldhuizen et al. (120) The time-dependence of the effect was
analyzed in terms of ambipolar diffusion. Further, the relative enhancement
of optogalvanic signal related to the presence of the 235U species was
described for the U atomic lines at 436.3 and 437.2 nm, the cause of which
was ascribed to saturation and radiation trapping effectsY2l)
Of the two possible concomitant processes leading to the optogalvanic
effect (increased cross section for electron impact ionization and increased
electron temperature), a study conducted in an Ne ReD yielded results
favoring temperature coupling as the dominant event under the chosen
experimental conditions. (122) Agreement with the predictions of a mathemati-
cal model was more than satisfactory.
Two-step optogalvanic spectroscopy was applied to the measurement
of the radial distribution of the electric field in the cathode fall region of
an Ne ReD. Observations of the linear Stark effect of the nd' (n =
10-12)-3p'[lj2]1 transitions were made. The depth of the cathode fall in
the ReD is one-fourth to one-fifth (ca. 0.8 mm) that of a flat cathode GD
under similar experimental conditions (3-4 mm), whereas the electric field at
the cathode surface is approximately 5.2 and 3-4 kV jcm in the two cases. (123)
A rather complicated relationship could be arrived at through laser
optogalvanic measurements linking the spatial distribution of the electric
field in an ReD with the spatial variation of the spectral width of the
Ar 7d[5j2]3-4p[3j2]2 transition.(124) No inverted population between the
levels 2P 1,2,3,4,5 and 1PI was detected in an Ne ReD by using the two-step
optogalvanic method, in full agreement with the observed very large
enhancement of the optogalvanic signalsY 25 ) Further optogalvanic observa-
tions revealed a linear decrease in electric field with radial distance from the
cathode surface in the case of a Kr ReD laser through study of the Stark
shift of the Kr 8d[3j2]~-5p[3j2h atomic line.(126)
The inherent ability of the ReD to lend itself to such investigations
favors both the comprehension of its basic working mechanism and the
development of other technological applications, such as those dealt with in
the next subsection.
6.4.5.2. Laser Techn%gv
A fascinating area of investigation, laser development through the use

of the ReD process has gained tremendous momentum over the last decade.
The problems of efficiency, instability, cataphoresis, and sputtering
underIying the development of a proper discharge tube for cw laser operation
were faced by Apai et al.(127) They designed a dc Re-Kr+ laser affording an
output power of 5 mW at 469.4 nm which could adequately control the
sputtering rate by increasing the total gas pressure, with a considerably
lünger lifetime of the discharge tube.
246 CHAPTER 6
To overcome the drawbacks inherent in the excessive length of ßutelike

or hybrid-sized HCD tubes for laser operation, a compact (46 cm long)
device of cylindrical structure was fabricated for white light laser operation,
whose performance is equivalent to that of the other systems mentioned
above. (128) Transverse and longitudinal HCDs for laser operation were com-
pared with regard to the 469.4 nm Kr transition. The first type affords better
axial uniformity, considered to be superior for laser media excitation.(129)
Longitudinal HCD is affected by plasma inhomogeneity seriously affecting
the laser tube operation. (130)
A new model of HCD He-Zn laser at 492.4 nm was developed with a
grid as the auxiliary electrode to inßuence the gain of the laser line through
variations in the charge ßow distribution.(I3I) The Cu HCD He-Ne laser at
632.8 nm (3s 2-2p4 transition) produces a higher efficiency than an Al one.(132)
This could be ascribed to the fact that the upper laser and 4S0 He metastable
levels are more populated in the former case. Cd and Zn neutral atom
densities were measured by Arai et al.(133) in the HCD He-Cd+ and He-Zn +
laser discharge by monitoring the decays of the Cd(II) 636.0-nm and Zn(II)
610.2-nm line intensities. Density values of 6.2 x 10 14 and 7.1 x 10 14 cm- 3
were obtained, respectively. Although these values are lower by a factor of
0.3 than expected, they still are higher by one order of magnitude than those
found in the positive column-type laser discharges.
A number of elements (Mo, Sm, Ta, V, Y, Zr) were studied after
machining them in the form of massive hollow cathodes to demonstrate the
suitability ofthe HCD for very high resolution laser saturation spectroscopy
and consequently to ascertain the isotope shifts and hyperfine structureY34)
As some amount of ions in the interelectrode space is necessary before
every pumping pulse for regular and stable pulse laser operation, a high-
voltage electrodeless discharge was combined with a pulsed HCD Cu+ laser,
resulting in a roughly 30% enhancement in laser output power and lower
likelihood of arc formation.(135) For the first time four laser transitions
[Te(II) at 629.4, 564.9, and 548.8 nm and Ge(II) at 589.3 nm] were reported
in HCD-operated laser systems. (136) With these elements it was possible to
verify that laser action can take place via charge exchange, Penning ioniza-
tion, and radiation cascade processes. A lifetime of ca. 360 h was achieved
in a dc He-Kr ion HCD laserwith an output power of5 mWat 469.4 nm.(137)
Reportedly, laser light noise may be a major problem in writing applica-
tions and means to abate it are constantly being sought. (138) An He-Zn
HCD laser tube, used to conduct an investigation of this issue, revealed that
ßuctuations in He + are responsible for the instability of the power output
from which noise originates. The addition oftiny amounts ofNe suppressed
noise down to less than 0.2% (rms).
Additional information on the use of HCD systems in laser operation
can be found in an excellent Hungarian reviewY39)
6.4.5.3. Deposition and Etching
The ability ofthe HCD to both erode and coat surfaces by appropriately
guiding the sputtering mechanism and the chemical reactions occurring in
the plasma can be expediently used to achieve results in the treatment of
surfaces that could not otherwise be obtained.
How deposition processes can be advantageously combined in a single
HCD chamber was brilliantly dealt with by Horwitz et al. (140) The system
was reported to behave excellently in mixed metal-polymer deposition and
Si02 etching on commercial complementary metal oxide semiconductor
wafers, especially with regard to process kinetics and uniformity. High depo-
sition rates together with obtaining abrasion-resistant films were the points
of merit in a new procedure developed by Horwitz and McKenzie(\41) to
deposit amorphous silicon with an rf-operated HCD.
The effect of the metal cathode on methane dissociation was elucidated
in order to generate precursors for the polymerization of methane. (142) Mixed
with Ar, methane could be polymerized at a deposition rate of 70 nm S-I in
a 19% Pt-81% W cathode and 42 nm S-I for a W cathode, both rates being
much higher than more conventional procedures based on rf. The same
authors had previously obtained very encouraging results in similar
undertakings using He, which suppressed the phenomenon of carbon
formation. (143)
Ion beatn sputter deposition of metallic films was achieved by means of
a superdense HCD plasma.(I44) Moreover, a triode device made up of a
magnetron cathode coupled with an HCD source as a second cathode to
enhance the plasma density was devised. (145) The coupling efficiency was
:::;60%. Compared with conventional magnetrons, the combined system
allows for a tenfold increase in discharge current and deposition rate.
Although still in their early stage, such applications are definitely pro-
mising and open entirely new avenues to GDs in general.
6.4.5.4. Beam Generation

HCD properties allow for the production of ionic and atomic beams
that can be extracted for many other scopes. The combination of this source
with mass spectrometry is the foeus of the studies carried out by Harrison
and his team(50,146) and is certainly among the most exciting promises of this
technique. A small hole on the bottom of the cathode aets as a minieleetrode
cavity giving rise to an inereased HCD effect with ejection of a "plurne,"
whose ions are then assayed on the basis of their masses.
A neutral beam injector ion source consisting of a combination of HCD
and ion accelerator system was deseribed by Aston. (147) In it, the cathode
works at high stagnation pressure and plasma density, which ensures field-
enhanced thermionic emission. The ion aeeelerator system amplifies the
248 CHAPTER 6
extracted ion current density by a factor of four through a set of multiple

hole and slit apertures.
The suitability of the He HCD as a neutral beam injector was further
investigated using a BaO-impregnated porous W tube in an attempt to over-
come the drawbacks involved in directly heated pure W or oxide-coated
filaments. (148) The combination of Mo buttons set in front of the orifice and
antiparallel magnetic fields was found effective in reducing the minimal gas
flowrate down to ca. 320 Pa liter S-1
The production of a U atomic beam by Babin and Gagne(71) was men-
tioned in Section 6.3.2. The two-dimensional emittance and brightness fora
given high ion current of an HCD ion source were quantified by Saad and
Holk. (149) The use of the HCD as an electron beam is also compatible with
the technique and has been exhaustively described in the literature.(150,151)
6.4.5.5. Plasma Diagnostics
Not to be forgotten are the impressive number of recent contributions

attempting to shed further light on unclear aspects of the HCD processes.
The use of a steady-state low-pressure HCD arc in Ar for diagnostic
studies was described in Seetion 6.3.2. (67) In continuation of previous works,
Kulakowska and Zyrnicki (152) studied the chemiluminescent reaction
SnCl4+ [0 2+ Ar(He)]* and also compared the characteristics ofthis process
with those ofexcitation in an ReD. The 0-0 and 0-1 bands ofthe 1n _X11;+
system and the 3norX 11;+ 0-0 subbands ofthe InCI spectrum were subjected
to further examination by Borkowska-Burnecka and Zyrnicki.(153) Rota-
tional constants of the upper states could thus be recalculated. Similarly, the
rotational constants of 115In79 Br and 115In81 Br were calculated by a high-
resolution study of the A3no-X11;+ and B3nl-XI1;+ subsystems. As already
reported, the two-component electron population was investigated by Rersh-
covitch et al. (154)
The occurrence of plasma satellites was detected by an assembly of two
pulse-operated hollow cathodes and its dependence on voltage modulation
was clarified.(155) Preliminary observations by Hildebrandt('56) evidenced the
advantages offered by the HCD in the study of such plasma satellites. The
temporal dependence ofthese results on the increase ofthe forbidden compo-
nent of He (I) at 447 nm showed that their intensities can exceed that of the
said forbidden component while, for a short period, the increase in the near-
satellite intensity is inversely proportional to that of the far-satellite.
Reflection of 8-mm microwaves was used to determine both the radius
and speed of the cylindric ionization front in an HCD under the expansion
conditions that occur when applying high-voltage pulses.('57) In an attempt
to model the radial dependence of electron concentration in an He HCD it
was found that it becomes nonuniform at pressures higher than ca.
130 Pa. (158) A simple one-dimensional equation was deduced. Later, it was
pointed out that for electrons with energy of hundreds of electron volts,
their distribution relaxes faster than the angular one.
Spectral lines of weak intensity emitted by Ne atoms and ions were
examined by a microcomputer-controlled photon-counter system, which
proved to be particularly advantageousY 59) By measuring variations in
plasma conductivity, internal elemental processes in the plasma could be
ascertained for Al, Cd, Cr, Cu, and Fe.(160) Along the same line, the utility
of the RCD for high-resolution, Doppler-free saturation experiments was
further demonstrated by ascertaining the hyperfine structure of V atoms
and the relevant constants for the five 3d 34seF)4p6D o levels and the four
e
3d4 D)4s6D metastable levels. (161)
The absolute values ofthe electric field, measured by Stark optogalvanic
spectroscopy, were found not to decrease linearly from the cathode surface
in an Ar RCD, but to follow a radial pattern mirroring the linear enhance-
ment of space charge density along the current flOW.(162) Further experiments
revealed that this behavior is described by a function of the Q(r) = -Ar + B
type, where Q is the space charge distribution and r the radial distance from
the cathode axis in the zone between the cathode surface and the boundary
with the negative gIOW.(163) Finally, the same team ascertained the pressure
broadening and frequency shift ofthe Ar 7d[5j2]3-4p[3j2]2 transition.(I64)
Row discharge temperature and hydrogen density can influence the
intensity of the Rj emission was elucidated by Miderski and Gellene by
monitoring the 3;Al --+ 2p 2A 2 transition. (165) The effect was found to be
accurately described by assuming that the primary process leading to Rj is
the Rtj e- dissociative recombination.
A good agreement between experimental findings and theoretical pre-
dictions was obtained in the Stark width measurements of the 589.6 nm
neutral Na hyperfine lines in the presence of charged perturbers. (166) Another
team strove to elucidate the processes involved in high-current RCD devices
for metal ion lasers by examining the case of an Ne-Cu discharge. A set of
expressions was derived that accounted for the dependence of current density
on the spatially averaged densities of noble gas ions and metal ions and
atoms.
In aseries of precise measurements of saturation spectroscopy in the
RCD, the isotope shifts and hyperfine splitting were determined for the
270.24-and 270.59-nm transitions of Pt(I).(168) Notwithstanding relatively
large specific mass variations, the isotope-shift ratios turned out to be tran-
sition independent.
The incorporation of a confined positive column in front of the hollow
cathode could feed additional electrons into the RCD plasma without
increasing sputtering or broadening self-absorption.(169) The net enhance-
ment ofthe hyperfine doublet ofthe Cu(I) 324.8 nm transition was measured
250 CHAPTER 6
as a function of current intensity and gas type and pressure. One of the more
annoying problems with spacecraft operation is reported to be electrical
charging, especially for satellites in geosynchronous orbits, with spacecraft
frame potential of thousands of volts negative with respect to the ambient
space plasma.(170) To compensate for this effect, an HCD neutral plasma
source was designed to serve as a plasma emitter, capable of operating with
either Ar or Kr and to supply ion currents of up to 325 J1.A and electron
currents of 0.02 to 6.0 A. Another exotic use of the HCD consisted in the
precise determination of the radiance output of a sealed Pt/Cr-Ne tube to
calibrate the faint object spectrograph on the Hubble space telescope.(171)
The applicability of fluorescence techniques with a repetitively pulsed
tunable dye laser to investigate electron densities and neutral-partic1e tem-
peratures was thoroughly examined. (172) A power-broadening relationship
was developed to accommodate rather large saturation parameters. To
study the mechanism and kinetics of HCD plasma chemical reactions,
the decomposition of TiC4 under conditions of non-LTE was also
c1arifiedY 73 ) While the interaction of Ar metastables with TiC4 molecules
appears to play only a minor role, the overall reaction is assumed to be
TiCl4 + ne-* -+ Ti* + Ch + 2CI + ne-(n = 4-6).
The remarkable improvement in the signal-to-background ratio in the
MW-boosted HCD lamp, especially when Ar is used as the carrier gas, was
examined in the spectral range 200-400 nm.(174) Moreover, Pataky-Szab6
and her associates gained further data on the functioning of the three-
electrode HCD source. (175) The crater that formed on top of the central
rod turned out to be quite independent of the physical and chemical
inhomogeneities.
A number of other contributions in this field deserve mentioning: for
example, those studies devoted to the mechanism ofTh atomization,(176) the
measurement of the density of U ions, (177) the effects of a magnetic field
parallel to the cathode axis on the radial emission profiles(l78) (also investiga-
ted for commercial AAS lamps(179», the achievement of optical phase conju-
gation in an HCD tube through degenerate four-wave mixing,(180) the
performance of the Cu-Ne HCD in the current density interval of 10-
100 mA cm -2, (181) the formation of metastable Ar atoms in an Ne HCD
arc,<'82) the ablation pattern of Al and Cd under normal dc and MW-coupled
HCD discharge conditions, (183) the generation of Cu vapor(184) and its emis-
sion and absorption characteristics in a pulsed HCD, (185) the lifetime of some
P(II) levels,<'86) the appearance ofincoherent Hanle signals caused by mutual
orthogonalization of the entrance slit height, (187) the determination of the
excitation temperature of Ar(l), Ar(II), Cr(l) , Cr(II), Cu(l) , Si(I), V(I),
and V(II) species(188) and of Al(I), Mg(I), Mg(II), Pb(l), Ti(I), and Ti(II)
species,(189) and the study of HCD tubes operating at both dc and high-
frequency currents.(190)
For those interested in obtaining further information, updated reviews
on the various analytical, technological, and diagnostic uses of the HCD are
available. (e.g., 21,31,37,88)
6.5. Future Trends
Although divination is an unreliable art, the only chance of succeeding

lies in the accurate and sound analysis of past and present events. Trying to
apply this general rule to the disclosure of the most probable pathways the
HCD will take, one must scrutinize foregone facts and current tendencies.
Of the many possible ways to achieve this, the approach adopted herein
consists in searching the available literature on this subject (more than 1300
documents) from the early years up to 1980 and comparing the data thus
obtained with the data for the last decade, in order to highlight any signifi-
cant changes.
The issues being considered (selected on the basis of their role as trend
markers) are the topic ofpublished document, the nationality ofthe authors,
the medium used to report the results, and finally the language in which the
document is written. The information obtained is set forth in Tables 6-3
through 6-6. With regard to subject matter, the highest effort has been
devoted to classifying each document as systematically as possible according
to its predominant content. Despite this effort a certain amount of arbitrari-
ness in the categorization process cannot be avoided, although this would
not affect the overall breakdown and the hierarchy that descends from it.
The data allow some main conclusions to be drawn, namely, that works
centered on the investigation of fundamental aspects of the HCD are still
about a third of the total, and more than half the volume of the literature
up to 1980, but still clearly highlighting the need for a better understanding
Table 6-3. Percentage Distribution of Documents on HCD with

Reference to the Topic
Fraction %
Topic Up to 1980 From 1980
Analysis 24.6 11.1
Basic mechanism 65.1 33.7
Beam generation 0.1 4.2
Depositionjetching 0.1 13.0
Instrumentation 9.5 8.3
Laser technology 0.1 22.3
OptogaIvanic spectroscopy 0 5.8
Review 0.5 1.6
252 CHAPTER 6
Table 6-4. Percentage Distribution of Documents on

HCD with Reference to the NationaJity of the Authors
Fraction %
Country Up to 1980 From 1980
Australia 0.7 3.0
Canada 0.7 4.1
China 0 6.2
Europe 45.5 38.7
Japan 5.3 18.1
Soviet Union 33.2 13.2
United States 14.3 14.7
Others 0.3 2.0
Table 6-5. Percentage Distribution of Documents on

HCD with Reference to the Language Used
Fraction %
Language Up to 1980 From 1980
Chinese 0 5.1
English 39.5 65.4
French 5.6 1.5
German 14.3 3.3
Japanese 2.3 8.2
Russian 34.8 13.4
Others 3.5 3.1
of some facets of the discharge processes. At the same time, the percentage
of analytical papers has more than halved, reflecting the difficulty in coping
with the present requirements of trace determinations. While studies report-
ing on the development of instrumentation do not show appreciable changes,
other previously unrecorded areas have become dramatically important, as
is the case for laser research and coating and etching technology. This trend
clearly points to the ongoing diversification of HCD applications.
With regard to nationality of the authors, there is a marked decline in
the number of contributions from Europe and the USSR (once clearly the
major sources of activities in the field) and an extraordinary increase in the
number of Japanese and Chinese teams. The leading role of Europe and the
USSR seems still indisputable, accounting for more than half of the total
publications. It should be noted that the most active countries in Europe
have been Germany (both West and East), France, and Italy and that most
of the above decrease can be ascribed to the presently reduced amount of
research with HCDs by the first two.
Table 6-6. Percentage Distribution of Documents on HCD with
Reference to Publication Type
Fraction %
Publication
typeQ Up to 1980 From 1980
Journals
Canada 0.7 1.3
China 0 5.7
Europe 34.3 18.6
Japan 3.1 6.4
USA 17.4 31.2
USSR 32.6 12.5
Conference proceedings 5.5 8.2
Technical reports 4.3 1.8
Dissertations 1.0 0.9
Patents
Europe 0.2 2.7
Japan 0 9.1
USA 0.9 1.6
"Journals also incIude chapters in books or monographs.
Quite understandably, such changes are reflected in the languages used

with the additional point that the enhanced diffusion of English in the
scientific literature is now documented by approximately two-thirds of publi-
cations being written in this language.
An analogous pattern is shown in the breakdown of diffusion media,
as journals printed in North America are now in the lead, replacing Europe
as the most productive region. Among journals preferred for disseminating
results, until1980 an apparent predominance was shown for Spectrochimica
Acta (7.8%), followed by Applied Spectroscopy (2.7%) and Analytical
Chemistry (1.6%) (values in parentheses refer to the frequency ofappearance
in journals only). An appreciable number of papers on HCD also used to
be published in Fresenius' Zeitschrift für physikalische Chemie. Over the last
decade, the situation has moved toward a better balance, with Spectro-
chimica Acta still prevailing (4.5%), but with a smaller gap to Applied
Spectroscopy (3.6%) and Analytical Chemistry (1.7%). A significant number
of papers on HCD have appeared in a periodical first released in the mid-
1980s, the Journal of Analytical Atomic Spectrometry, totaling 1.7% ofpubli-
cations on RCD. In both periods (before and after 1980) the preferred media
for publishing studies on HCD have been two European and two North
American journals. Another phenomenon that should not be overlooked is
certainly the quickly increasing numberofpatents pivoting on HCD develop-
ments and applications. The economic implications of this spectroscopic
branch seem to be weIl understood by Japan, followed by Europe and the
USA.
254 CHAPTER 6
Therefore, a crystal ball is not needed to forecast that the above-outlined

trends will be all the more valid in the future and that technological exploita-
tion of HCD in laser development and surface treatment procedures will
play a major role. Regarding analytical impact, current premises do justify
the belief in an imminent adequacy of the source to the challenging require-
ments of trace determinations as posed by today's problems. Whether these
avenues will be explored and accomplished is a question for the year 2000.
Acknowledgment
The expertise and patience of Dr. Isabel A. Robinson in revising the

English style of the various drafts of this chapter are gratefully
acknowledged.
References
1. J. Plücker, Continued observation on the electrical discharge through rarefied gas tubes,
Ann. Phys. 104 (1858) 113.
2. J. Plücker and J. W. Hittorf, On the spectra of ignited gases and vapors, with especial
regard to the different spectra of the same elementary gaseous substance, Phi/os. Trans.
115(IV) (1865) 1.
3. F. Paschen, Bohrs HeliUllllinien, Ann. Phys. 50 (IV) (1916) 901.
4. F. Paschen, Die Funkenspektren des Aluminiums. I Teil, Ann. Phys. 71 (IV)(1923) 142
5. F. Paschen, Die Funkenspektren des Aluminiums. 11 Teil, Ann. Phys. 71(IV) (1923) 538.
6. R. A. Sawyer and F. Paschen, Das erste Funkenspektrum des Aluminiums AI 11, Ann.
Phys. 84 (IV) (1927) 1.
7. H Schüler, Ueber Potentialgeralle an Elektroden in Gasentladungsröhren, Phys. Z. 22
(1961) 264.
8. H Schüler, Ueber eine neue Lichtquelle und ihre AnwendungsmögJichkeiten, Z. Phys. 35
(1926) 323.
9. H. Schüler and J. E. Keystone, Hyperfeinstrukturen und Kernmomente des Quecksilbers,
z. Phys. 72 (1931) 423.
10. H. Schüler and H. Gollnow, Ueber eine lichtstarke Glimmentladungsröhre zur spektro-
skopischen Untersuchung geringer Substanzmengen, Z. Phys. 93 (1935) 611
11. H. Schüler, Possibility of applying the hollow cathode discharge to spectroanalytical
investigations, Proceedings of the I Co/loquium Spectroscopicum Internationale, Strasbourg,
1950, pp. 169-171.
12. H. Schüler and A. Michel, Ueber zwei neue Hohlkathodenentladungsröhren, Spectrochim.
Acta 5 (1952) 322.
13. H. Schüler, Ueber die Emissionsspektroskopie organischer Substanzen mit Hilfe der Elek-
tronenstossanregung in der Glimmentladung. I, Spectrochim. Acta 4 (1950) 85.
14. H. Schüler and L. Reinebeck, Ueber die Emissionsspektroskopie organischer Substanzen
mit Hilfe der Elektronenstossanregung in der Glimmentladung. 11, Spectrochim. Acta 6
(1954) 288.
15. H. Schüler and L. Reinebeck, Ueber eine Methode der Variation der Anregungsbedin-
gungen organischer Substanzen, Z. Naturforsch. 5a (1950) 657.
16. R. A. Sawyer, Excitation processes in the hollow discharge, Phys. Rev 36 (1930) 44.
17. V. A. Konova10v and E. S. Frish, Illumination of the mixture of argon and nitrogen, Zh.
Tekh. Fiz. 4 (1934) 523.
18. J. R. McNally, Jr., G. R. Harrison, and E. Rowe, A hollow cathode source applicable to
spectrographic analysis for the halogens and gases, J. Opt. Soc. Am. 37 (1947) 93.
19. A. Walsh, The application ofatomic absorption spectra to chemical analysis, Spectrochim.
Acta 7 (I955) 108.
20. A. Wa1sh, Atomic absorption spectroscopy and its app1ications-Old and new, Pure Appl.
Chem.49 (1977) 1621.
21. S. Caro1i, Low-pressure discharges: Fundamental and applicative aspects, J. Anal. At.
Spectrom. 2 (1987) 661.
22. E. L. Grove and W. A. Loseke, Hollow cathode excitation of air-type atmospheres, Can.
J. Spectrosc. 18 (1973) 33.
23. J. A. C. Broekaert, Emission spectrographic determination of all rare earths in solutions
by hollow cathode excitation, Bull. Soc. Chim. Belg. 85 (1976) 261.
24. P. Zanzucchi, A study of direct solution analysis with a new hollow cathode discharge
system, Dissertation, University of Illinois, Urbana, University Microfi1ms Ltd., Ann
Arbor, Mich., 1967.
25. S. Caro1i, A. Alimonti, P. Delle Femmine, and S. K. Shuk1a, Determination of gallium
in tumor-affected tissues by means of spectroscopic techniques, Anal. Chim. Acta 136
(1982) 225.
26. S. Caro1i, O. Senofonte, and P. Delle Femmine, Determination oftrace elements in bio10gi-
ca1 materials using a hollow-cathode discharge : Comparative study of matrix effects,
Analyst (London) 108 (1983) 196.
27. J. O. Foss, H. J. Svec, and R. J. Conzemius, The determination of trace elements in
aqueous media without preconcentration using a cryogenic hollow-cathode ion source,
Anal. Chim. Acta 147 (1983) 151.
28. S. Caroli, Low-pressure discharges, in: Sampie Introduction in Atomic Spectroscopy
(J. Sneddon, ed.), pp. 225-253, Elsevier, Amsterdam, 1990.
29. H. Falk, Hollow cathode discharge within a graphite furnace: Furnace atomic non-
thermal excitation spectrometry (FANES), in: Improved Hollow Cathode Lamps for
Atomic Spectroscopy (S. Caro1i, ed.), pp. 74-118, EHis Horwood Ltd., Chichester, UK,
1985.
30. A. Günther-Schulze, Glow discharge in a hollow cathode, Z. Techn. Phys. (1930) 49.
31. R. Mavrodineanu, Hollow cathode discharges: Analytical applications, J. Res. Nat!. Bur.
Stand. 89 (1984) 143.
32. G. Francis, The glow discharge at low pressure, in: Handbuch der Physik (S. Flügge, ed.),
Vol. 22, pp. 53-208, Springer-Verlag, Berlin, 1956.
33. R. Papoular, The Glow Discharge, Electrical Phenomena in Gases, pp. 123-140, I1iffe
Books, London, 1965.
34. M. Kaminsky, Atomic and Ionic Impact Phenomena on Metal Surfaces, Springer-Verlag,
Berlin, and Academic Press, New York, 1965.
35. F. M. Penning, Electrical Discharges in Gases, Philips Technical Library, Eindhoven,
Servire B.V., Katwijk, The Netherlands, 1957.
36. P. F. Little and A. V. Engel, The hollow cathode effect and the theory of glow discharges,
Proc. R. Soc. London Sero A 224 (1954) 209.
37. M. E. Pillow, A critical review of spectral and related physical properties of the hollow
cathode discharge, Spectrochim. Acta 36B (1981) 821.
38. V. S. Borodin and Yu. M. Kagan, The investigation of a hollow cathode discharge, Zh.
Tekhn. Fiz. 36 (1966) 181.
39. H. Falk, Einige theoretische Ueberlegungen zum Vergleich der physikalischen Grenzen
thermischer und nicht-thermischer spectroskopischer Strahlungsquellen, Spectrochim.
Acta 32B (1977) 437.
256 CHAPTER 6
40. V. S. Borodin, Yu. M. Kagan, and L. I. Lyagushchenko, Investigation of a hollow cathode

discharge. 11. Zh. Tekh. Fiz. 11 (1967) 887.
41. F. Howorka and M. Pahl, Experimental determination of internal and external parameters
of the negative glow plasma of a cylindrical hollow cathode discharge in argon, Z. Natur-
forsch. A 27 (1972) 1425.
42. P. A. Büger and W. Fink, Analysis of solutions in a hollow cathode, Fresenius Z. Anal.
Chem. 244 (1969) 121.
43. C. Howard, M. E. Pillow, E. B. M. Steers, and D. W. Ward, Intensities of some spectral
lines from hollow-cathode lamps, Analyst (London) 108 (1983) 145.
44. F. Howorka, W. Lindinger, and M. Pahl, Ion sampling from the negative glow plasma
in a cylindrical hollow cathode, Int. J. Mass Spectrom. Ion Phys. 12 (1973) 67.
45. H. Patterson and D. H. TomIin, Experiments by radioactive tracer methods on sputtering
by rare gas ions, Proc. R. Soc. London Sero A 265 (1962) 474.
46. R. E. Honing, Sputtering of surfaces by positive ion beams of low energy, J. Appl. Phys.
29 (1958) 549.
47. K. B. MitchelI, Spectroscopic studies of ionization in a hollow cathode discharge, J. Opt.
Soc. Am. 51, (1961) 846.
48. T. Musha, Cathodic sputtering in a hollow cathode discharge, J. Phys. Soc. (Jpn.) 17
(1962) 1440.
49. A. D. White, New hollow cathode glow discharge, J. Appl. Phys. 30 (1959) 711.
50. E. H. Daughtrey, D. L. Donohue, P. J. Slevin, and W. W. Harrison, Surface sputter
effects in a hollow cathode discharge, Anal. Chem. 47 (1975) 683.
51. B. E. Warner, Investigation of the hollow cathode discharge at high current density,
Dissertation, University of Colorado, Boulder, 1979.
52. G. Knerr, J. Maierhofer, and A. Reis, Application of high-current hollow cathode for
quantitative analysis of conductors and glasses, Fresenius' Z. Anal. Chem. 229 (1967) 241.
53. H. Falk, The limiting factors for intensity and line profile of radiation sources for atomic
absorption spectrometry, Prog. Anal. At. Spectrosc. 5 (1982) 205.
54. J. G. Hirschberg, E. Hinnov, and F. W. Hofmann, Spectroscopic investigations of a
weakly ionized plasma in a helium hollow cathode discharge, MATI (Plasma Physics
Laboratory, Princeton University), 236, 1963.
55. T. Török and G. Zaray, Versuche mit einem tiefgekühlten Zwillingshohlkathoden Inter-
ferometer-Spektrometer. I. Spectrochim. Acta 30B (1975) 157.
56. H. Falk and H. Lucht, Investigation of excitation processes in a hollow-cathode discharge
by time-resolved measurements in the vacuum UV, J. Quant. Spectrosc. Radiat. Transfer
16 (1976) 909.
57. T. Lee, S. Katz, and S. A. MacIntyre, The spectrographic determination of uranium 235.
V. Routine application of a multiple hollow grating spectrograph, Appl. Spectrosc. 16
(1962) 92.
58. I. A. Berezin and K. V. Aleksandrovich, Determination of sulfur, chlorine and fiuorine
in beryllium oxide by a spectrographic method, Zh. Anal. Khim. 16 (1961) 613.
59. M. P. Chaika, Analysis of low-volatile oxides for halogens, Opt. Spektrosk. 2 (1957) 421.
60. H. Falk, E. Hoffmann, and C. Lüdke, F ANES (furnace atomic nonthermal excitation
spectrometry). A new emission technique with high detection power, Spectrochim. Acta
36B (1981) 767.
61. H. Falk, E. Hoffmann, and Ch. Lüdke, A comparison of furnace atomic nonthermal
excitation spectrometry (FANES) with other atomic spectroscopic technique, Spectro-
chim. Acta 39B (1984) 283.
62. H. J. Eichhoff and R. Voigt, Use of a metallic hollow cathode for spectrochemical analyses,
in: Proceedings of the IX Colloquium Spectroscopicum Internationale Lyon, 1961, Vol. 3,
pp. 309-317 (1962).
63. S. Caroli (ed.), Improved Hollow Cathode Lampsjor Atomic Spectroscopy, Ellis Horwood
Ltd., Chichester, UK, 1985.
64. S. Caroli, Hollow eathode lamps as excitation sourees for analytieal atomie speetroseopy,
Fresenius' Z. Anal. Chem. 324 (1986) 442.
65. G. H. C. Freeman and W. H. King, Cu II speetrallines and their suitability as wavelength
standards in the vaeuum ultraviolet, J. Phys. E 10 (1977) 984.
66. O. Appelblad and K. Sehmidt, A liquid nitrogen eooled eomposite wall hollow eathode,
USIP Report (University of Stoekholm), 85-07, 1985.
67. M. Bessenrodt-Weberpals, A. Broekhaus, P. Jaunernik, H. Kempkens, C. Nieswand, and
J. Uhlenbusch, Diagnosties of a steady-state low-pressure hollow cathode are in argon,
IEEE Trans. Plasma Sei. PS-14 (1986) 492.
68. Y. R. Jong, R. L. Davis, J. C. Williams, and J. C. WilIiams, Jr., Evaluation of the mini-
hollow eathode emission source for the analysis of mierosamples, Appl. Spectrosc. 42
(1988) 1379.
69. G. Rossi and N. Omenetto, Application of a demountable water-eooled hollow eathode
lamp to atomie fluoreseenee speetrometry, Talanta 16 (1969) 263.
70. F. Howorka, A. Seherleitner, V. Gieseke, and 1. Kuen, Bakeable hollow eathode for the
study of ion-moleeule reaetions in diseharge in gaseous mixtures, Int. J. Mass Spectrom.
Ion Phys. 32 (1980) 321.
71. F. Babin and J.-M. Gagne, Charaeterization of an atom beam produced with the help of
a hollow eathode discharge, Rev. Sei. Instrum. 57 (1986) 1536.
72. D. E. Holmgren, R. W. Falcone, D. J. Walker, and S. E. Harris, Measurement oflithium
and sodium metastable quartet atoms in a hollow cathode diseharge, Opt. Lett. 9 (1984)
85.
73. D. C. Gerstenberger, R. Solanki, and G. J. Collins, Hollow eathode metal ion lasers,
IEEE J. Quantum Electron. 8 (1980) 820.
74. T. Iijima, He-Zn + ion laser using hollow eathode discharge with high-voltage operation,
Jpn. J. Appl. Phys. 21 (1982) 1732.
75. N. A. G. Ahmed and D. G. Teer, Charaeterization of aluminium coatings deposited in a
hollow cathode discharge, Thin Solid Films 80 (1981) 49.
76. L. Papp, Development of hollow cathode radiation sources. Part 2. Study of the effeet of
a cylinder placed in the cathode eavity on the emitted light intensity, J. Anal. At. Spectrom.
2 (1987) 407.
77. J. Czakow, The microcavity hollow cathode and its analytieal potential, in: Improved
Hollow Cathode Lamps jor Atomic Spectroscopy. (S. Caroli, ed.), Ellis Horwood Ud.,
Chiehester, UK, 1985, pp. 35-51.
78. V. A. Novose1ov and V. B. Znamenskii, Correlations between the intensity of speetral
lines, discharge parameters in a hollow eathode and its diameter, Spektrosk. Tr. Sib.
Soveshch.,4th, 1965, pp. 273-278.
79. Z. Szilvassy, Principles and use of a boosted hollow cathode discharge source for atomic
spectroscopy, in: Improved Hollow Cathode Lamps jor Atomic Spectroscopy. (S. Caroli,
ed.), Ellis Horwood Ltd., Chiehester, UK, 1985, pp. 178-202.
80. Z. Szilvassy, A. Buzäsi, and E. Häzi, Effeet of eathode material and sampIe composition
on the plasma eharaeteristics ofhollow cathode discharge, Acta Chim. Hung. 126 (1989)
353.
81. C. Popovici and M. Some§an, On the emission speetrum of the negative glow plasma of
a hollow cathode diseharge in magnetic field, Appl. Phys. Lett. 8 (1966) 103.
82. N. K. Rudnevsky and D. E. Maksimov, Hollow cathode discharge in a magnetie field,
in: lmproved Hollow Cathode Lampsjor Atomic Spectroscopy. (S. Caroli, ed.), Ellis Hor-
wood Ltd., Chichester, UK, 1985, pp. 148-177.
83. B. Pavlovic and J. Dobrosavljevic, Influenee of a rotating magnetic field on a hollow
258 CHAPTER 6
84. R. B. Djulgerova, The pulsed hollow cathode discharge. New spectroanalytical possibilit-
ies, in: Improved Hollow Cathode Lamps for Atomic Spectroscopy (S. Caroli, ed.), Ellis
Horwood Ud., Chichester, UK, 1985, pp. 55-73.
85. P. B. Farnsworth and J. P. Walters, The radiofrequency-boosted, pulsed hollow cathode
lamp, in: Improved Hollow Cathode Lampsfor Atomic Spectroscopy (S. Caroli, ed.), Ellis
Horwood Ud., Chichester, UK, 1985, pp. 119-147.
86. S. Caroli, A. Alimonti, and F. Petrucci, Analytical capabilities of the microwave-coupled
hollow cathode discharge, Anal. Chim. Acta 136 (1982) 269.
87. R. Tomellini, M. Cilia, O. Senofonte, G. Guantera, M. G. DeI Monte Tamba, and S.
Caroli, A newly devised microwave-boosted low pressure source in atomic emission spec-
trometry. Berichte des 3. Anwendertreffen Analytische Glimmentladungs-Spektroskopie,
Jülich (FRG), April, 1990.
88. S. Caroli, The hollow cathode emission source: A survey of the past and a look into the
future, Prog. Anal. At. Spectrosc. 6 (1983) 253.
89. Atomsource Direct Solids Atomizer, Varian Report, 1988.
90. W. Pekruhn, L. K. Thomas, 1. Broser, A. Schroder, and U. Wenning, Chromiumjsilicon
monoxide on copper solar selective absorbers, Sol. Energy Mater. 12 (1985) 199.
91. N. Jakubowski, D. Stuewer, and W. Vieth, Performance of a glow discharge mass spectro-
metry for simultaneous multielement analysis of steel, Anal. Chem. 59 (1987) 1825.
92. W. A. Mattson, B. L. Bentz, and W. W. Harrison, Coaxial cathode ion source for solids
mass spectrometry, Anal. Chem. 48 (1976) 489.
93. K. R6zsa, Hollow cathode discharges for gas lasers, Z. Naturforsch. 35a (1980) 649.
94. R. De Marco, D. Kew, and J. V. Sullivan, Determination of major constituents in metal
sampIes by emission spectrometry using a demountable hollow cathode source and internal
standardization, Spectrochim. Acta 41B (1986) 591.
95. R. De Marco, D. J. Kew, C. Chadjilazarod, D. W. Owen, and J. V. Sullivan, Precision
and accuracy of quantitative emission spectrometry with particular reference to gold
alloys, Anal. Chim. Acta 94 (1987) 189.
96. J. Borkowska-Burnecka and W. Zyrnicki, Comparison of spectra excited in HF and dc
hollow cathode discharges, Spectrosc. LeU. 20 (1987) 795.
97. J. Borkowska-Burnecka and W. Zyrnicki, Fluorine determination in a hollow cathode
discharge by monofluoride emission spectra, Anal. LeU. 18 (1985) 1539.
98. A. Buzasi Györfine, S. Caroli, A. AJimonti, and O. Senofonte, Comparative study of the
hollow cathode and glow discharge radiation sources for spectrographic determination of
alloying elements and impurities in steel, Acta Chim. Hung. 113 (1983) 295.
99. S. Caroli, F. Petrucci, and A. Alimonti, The microwave-coupled hollow cathode discharge
and its analytical potential for the determination of trace elements in steel, Can. J. Spec-
trosc. 28 (1983) 156.
100. S. Caroli, O. Senofonte, N. Violante, F. Petrucci, and A. Alimonti, An investigation of
the power of detection of the hollow cathode sou~ce in emission spectroscopy, Spectro-
chim. Acta 39B (1984) 1425.
101. S. Caroli, F. Petrucci, A. Alimonti, and G. Zaray, Analysis of minorelements in metals
by microwave-coupled hollow cathode discharge, Spectrosc. Lett. 18 (1985) 609.
102. O. Senofonte, N. Violante, O. Falasca, and S. Caroli, Solid sampIe investigation by means
of a novel version ofthe microwave-coupled hollow cathode discharge (MW-HCD), Acta
Chim. Hung. 126 (1989) 317.
103. J. Mierzwa and W. Zyrnicki, Determination of Nd, Ho, Er, Tm, and Y in solutions by
hollow cathode discharge with copper cathodes, Anal. Lett. 21 (1988) 115.
104. J. A. C. Broekaert, The investigation of two sampIe preparation techniques applied to
the determination of rare earths in solutions with the aid of hollow cathode excitation,
Spectrochim. Acta 35B (1980) 225.
105. S. Caroli, O. Falasca, O. Senofonte, and N. Violante, The hollow cathode emission source
and its analytical potential for the determination of major, minor and trace elements. II.
Phosphorus, Can. J. Spectrosc. 30 (1985) 79.
106. S. Caroli, Development of a hollow cathode method for the spectroanalytical determina-
tion of trace elements in biological materials, Ann. Ist. Super. Sanita 19 (1983) 495.
107. S. Caroli, KatOdporiasztasos sugärforräsok jelenlegi a1kalmazasai es värhatö fejlesztesi
iranyzatai az emissziös spektrosköpiaban, Kern. Közl. 62 (1984) 57.
108. A. Alimonti, S. Caroli, F. Petrucci, and C. Alvarez Herrero, Determination of arsenie by
hollow-cathode emission spectrometry, Anal. Chirn. Acta 156 (1984) 121.
109. S. Caroli, A. Alimonti, and K. Zimmer, Applicability of a hollow cathode emission source
for determining trace elements in electrically non-conducting powders, Spectrochim. Acta
38B (1983) 626.
110. S. Caroli, A. Alimonti, P. Delle Femmine, and S. K. Shukla, Determination of gallium
in tumor-affected tissues by means of spectroscopic techniques, Anal. Chim. Acta 136
(1982) 225.
111. E. B. M. Jansen and D. R. Demers, Hollow-cathode lamp-excited inductively coupled
plasma atomic-fluorescence spectrometry: Performance under compromise conditions for
simultaneous multi-element analysis, Analyst (London) 110 (1985) 541.
112. W. R. Masamba, B. W. Smith, R. J. Krupa, and J. D. Winefordner, Atomic and ionic
fluorescence in an inductively coupled plasma using hollow cathode lamps pulsed at high
currents as excitation sources, Appl. Spectrosc. 42 (1988) 872.
113. B. M. Patel and J. D. Winefordner, Laser-excited fluorescence of diatomic molecules of
copper and lead in glow discharge sputtering, Appl. Spectrosc. 40 (1986) 667.
114. S. Greenfield, T. M. Durrani, and J. F. Tyson, A comparison ofboosted-discharge hollow
cathode lamps and an inductively coupled plasma (lCP) as excitation sources in ICP
atomic fluorescence spectrometry, Spectrochirn. Acta 45B (1990) 341.
115. D. R. Demers and C. D. Allemand, Atomic fluorescence spectrometry with an inductively
coupled plasma as atomization cell and pulsed hollow cathode lamp for excitation, Anal.
Chern. 53 (1981) 1915.
116. B. Rosen, New developments in the application of a hollow cathode discharge tube
designed for the quantitative determination of oxygen in metals, Appl. Spectrosc. 5 (1951)
20.
117. N. A. Zakorina, I. S. Lindstrem, and A. A. Petrov, Spectral-isotopic determination of
nitrogen in refractory alloys by using a discharge with a hot hollow cathode, Zavod. Lab.
50 (1984) 30.
118. Y. Iida, Laser vaporization of solid sampies into a hollow-cathode discharge for atomic
emission spectrometry, Spectrochim. Acta 45B (1990) 427.
119. L. Puig and R. Sacks, Hollow cathode plasma emission determination of F, CI and Br in
gas streams, Appl. Spectrosc. 43 (1989) 801.
120. E. M. van Veldhuizen, F. J. de Hoog, and D. C. Schram, Optogalvanic effects in a hollow
cathode glow discharge plasma, J. Appl. Phys. 56 (1984) 2047.
121. B. M. Suri, R. Kapoor, G. D. Saksena, and P. R. K. Rao, Relative enhancement of
optogalvanic signal of less abundant isotope in uranium hollow cathode discharge, Opt.
Commun. 49 (1984) 29.
122. R. A. Keller, B. E. Warner, E. P. Zalewski, P. Dyer, R. Engleman, Jr., and B. A. Palmer,
The mechanism of the optogalvanic effect in a hollow-cathode discharge, J. Phys. 44
(1983) C7/23.
123. N. Ami, A. Wada, Y. Adachi, and C. Hirose, Optogalvanic measurement of the electric
field inside the cathode fall region of neon hollow cathode discharge, Appl. Spectrosc. 43
(1989) 245.
124. C. Hirose, T. Masaki, A. Wada, and Y. Adachi, Radial profile of the spectral width of
the Ar 7d[5/2]3--4p[3/2]2 line inside the cathode fall region of Ar hollow cathode dis-
charges, Appl. Spectrosc. 43 (1989) 87.
125. T. Caesar and J. L. Heully, Experimental evidence of non-inverted population in a neon
hollow cathode, Opt. Commun. 45 (1983) 258.
260 CHAPTER 6
126. S. Fujimaki, Y. Adachi, and C. Hirose, Optogalvanic measurement of the cathode-fall

region of Kr hollow cathode discharge, Appl. Speetrose. 41 (1987) 567.
127. P. Apai, M. Janossy, I. Pälmai, K. Rozsa, and G. Rubin, D.c. hollow cathode He-Kr
discharge, SPIE 473 (1984) 198.
128. I. Ebina, W. Sasaki, and T. Ohta, A 46 cm-discharge-Iength white light He-Cd II laser
with coaxial type hollow cathode, Trans. IECE Jpn. 69 (1986) 367.
129. J. Mizeraczyk, J. Wasilewski, J. Konieczka, W. Urbanik, M. Grozeva, and J. Pavlik,
Comparison of He-Kr+ laser oscilJations in transverse and longitudinal hollow-cathode
discharges, Proceedings ofthe International Conferenee on Lasers, 1981, pp. 877-881.
130. J. Mizeraczyck, J. Konieczka, J. Wasilewski, and K. Rozsa, High-voltage hollow-cathode
He-Cd+ laser, Proceedings ofthe International Conferenee on Lasers, 1980, pp. 177-181.
131. M. Cilea, C. P. Cristescu, I. M. Popescu, and A. M. Preda, Hollow cathode He-Zn laser
with an additional command electrode, Rev. Roum. Phys. 27 (1982) 357.
132. S. S. Cartaleva, S. V. Gateva, and V. J. Stefanov, Investigation of spontaneous lines from
upper and lower levels of He-Ne laser line 632.8 nm excited in hollow cathode, SPIE 473
(1984) 192.
133. T. Arai, K. Nihira, T. Iijima, and T. Goto, Excitation mechanism in the hollow cathode
He-Zn+ laser, Ik. Kog. Doi. Ken. Hok. B-9 (1985) 67.
134. D. S. Gough and P. Hannaford, Very high resolution laser saturation spectroscopy in
hollow-cathode and glow discharges, Opt. Commun. 55 (1985) 91.
135. A. Baczynski, M. Dzwonkowski, and P. Targowski, The influence of ignition method on
laser output of pulse hollow cathode copper ion laser, Opt. Appl. XIII (1983) 231.
136. J.-b. Liu, Investigation ofhollow cathode Ge II and Te 11 lasers, Appl. Phys. B 32 (1983)
211.
137. M. Janossy, K. Rozsa, P. Apai, and L. CsilJag, D.c. hollow cathode He-Kr ion laser,
SPIE 473 (1984) 177.
138. T. Iijima, Noise measurement in a He-Zn hollow cathode laser tube, Physiea 115C (1983)
257.
139. K. Rozsa, Gaz- es femgözlezerek celjara alkalmazott üreges katOdu kisülesek, Magy. Fiz.
Foly. XXXIV (1986).
140. C. M. Horwitz, S. Boronkay, M. Gross, and F. Davies, Hollow cathode etching and
deposition, J. Vae. Sei. Teehnol. A6 (1988) 1837.
141. C. M. Horwitz and D. R. McKenzie, High-rate hollow-cathode amorphous silicon deposi-
tion, Appl. Surr Sei. 22/23 (1985) 925.
142. P. Meubus, H. Lange, and G. Jean, Methane polymerization with a hollow cathode:
Influence of the cathode metal, Plasma Chem. Plasma Process. 9 (1989) 527.
143. P. Meubus and G. Jean, Methane polymerization using a hollow cathode, Prepr. Pap.
Am. Chem. Soe. Div. Fuel Chem. 32 (1987) 260.
144. H. Nagasaka, K. Yanagida, T. Tanabe, M. Takeuchi, and H. Mase, Compact ion source
using superdense hollow cathode discharge and its application to thin film formation, Pur.
Kot. 5 (1985) 155.
145. J. J. Cuomo and S. M. Rossnagel, Hollow-cathode enhanced magnetron sputtering, J.
Vae. Sei. Teehnol. A4 (1986) 393.
146. W. W. Harrison and B. L. Bentz, Glow-discharge mass spectrometry, Prog. Anal. Spee-
trose. 11 (1988) 53.
147. G. Aston, Hollow cathode startup using a microplasma discharge, Rev. Sei. Instrum. 52
(1981) 1259.
148. S. Tanaka, M. Akiba, Y. Arakawa, H. Horiike, and J. Sakuraba, Reduction of gas flow
into a hollow cathode ion source for a neutral beam injector, Rev. Sei. Instrum. 53 (1982)
1038.
149. H. M. Saad and O. Holk, Investigation ofbeam emittance ofa hollow cathode ion source,
Egypt. J. Phys. 15 (1984) 45.
150. R. J. Barker, S. A. Goldstein, and R. E. Lee, Computer simulation of intense electron
beam generation in a hollow cathode diode, NRL Memorandum Report 4279, 1980.
151. J. J. Rocca, J. Meyer, and G. J. Collins, Hollow cathode electron gun for the excitation
of cw lasers, Phys. Let!. 87A (1982) 237.
152. B. Kulakowska and W. Zyrnicki, Spectroscopic studies of low pressure plasma. Tin and
its compounds, Phys. Ser. 33 (1986) 424.
153. J. Borkowska-Burnecka and W. Zyrnicki, High resolution study of the A31Co-X1r+ and
B31Cl-X1I;+ subsystems ofthe 115In79Br and 115In81 Br molecules, Phys. Ser. 35 (1987) 141.
154. A. I. Hershcovitch, Y. J. Kovarik, and K. Prelec, Observation of a two-component elec-
tron population in a hollow cathode discharge, J. Appl. Phys. 67 (1990) 671.
155. J. Hildebrandt, Yoltage modulation in a pulsed hollow-cathode discharge and its relation
to the occurrence of plasma satellites, Phys. Let!. 95A (1983) 365.
156. J. Hildebrandt, Temporally and spatially resolved plasma satellites in a hollow-cathode
source, J. Phys. B 16 (1983) 149.
157. J. Hildebrandt, Microwave diagnostic ofthe pulsed generation in the hollow-cathode glow
discharge, Z. Naturforsch. 38a (1983) 1088.
158. Yu. M. Kagan, Rate of ionisation and density of electrons in a hollow cathode, J. Phys.
D 18 (1985) 1113.
159. M. Kimura, Time-resolved spectral measurement in hollow cathode discharge tube by
photon-counting method, Jpn. J. Appl. Phys. 23 (1984) 105.
160. Z. Szilvassy, A. Buzasi, and E. Hazi, Effect of cathode material and sampie composition
on the plasma characteristics of hollow cathode discharge, Acta Chim. Hung. 126 (1989)
353.
161. D. S. Gough, P. Hannaford, M. Lowe, and A. P. Willis, Hyperfine structures in 51y using
saturation spectroscopy in a hollow-cathode discharge, J. Phys. B 18 (1985) 3895.
162. T. Masaki, A. Wada, Y. Adachi, and C. Hirose, Space charge distribution in the cathode
fall region of an Ar hollow cathode discharge, Appl. Spectrosc. 42 (1988) 49.
163. T. Masaki, A. Wada, Y. Adachi, and C. Hirose, Pressure dependence of space charge
distribution in the cathode fall of Ar hollow cathode discharges, Appl. Spectrosc. 42 (1988)
51.
164. T. Masaki, Y. Adachi, and C. Hirose, Application of optogalvanic spectroscopy to the
measurement of pressure broadening and shift in the negative glow region of an Ar hollow
cathode discharge, Appl. Spectrosc. 42 (1988) 54.
165. C. A. Miderski and G. I. Gellene, Experimental evidence for the formation of H 3* by
H+ 5/e- dissociative recombination, J. Chem. Phys. 88 (1988) 5331.
166. F. Moreno, J. M. Alvarez, J.C. Amare, and E. Bernabeu, Stark effect of atomic sodium
measured in a hollow cathode plasma by Doppler-free spectroscopy, J. Appl. Phys. 56
(1984) 1939.
167. H. Koch and H. J. Eichler, Particle densities in high current hollow discharges, J. Appl.
Phys. 54 (1983) 4939.
168. R. D. LaBeIle, W. M. Fairbank, Jr., R. Engleman, Jr., and R. A. Keller, Isotope shift
and hyperfine structure of Pt I transitions in a hollow-cathode discharge, J. Opt. Soc. Am.
B 6 (1989) 137.
169. H. A. Phillips, H. L. Lancaster, M. Bonner Denton, K. R6zsa, and P. Apai, Self-
absorption in copper hollow cathode discharges: Effects on spectralline shape and absorp-
tion sensitivity, Appl. Spectrosc. 42 (1988) 572.
170. W. D. Deininger, G. Aston, and L. C. Pless, Hollow cathode plasma source for active
spacecraft charge control, Rev. Sei. lnstrum. 58 (1987) 1053.
171. J. Z. Klose and J. M. Bridges, Radiance of a Pt/Cr-Ne hollow cathode spectral line
source, Appl. Opt. 26 (1987) 5202.
172. P. ChaII, E. K. Souw, and J. Uhlenbusch, Laser diagnostics of a low-pressure hollow-
cathode are, J. Quant. Spectrosc. Radiat. Transfer 34 (1985) 309.
262 CHAPTER 6
173. J. F. Behnke, H. Lange, and H.-E. Wagner, Dissociation of titanium tetrachloride to

titanium in hollow cathode discharges. Part I. Determination of the dissociation degree,
Pol. J. Chern. 56 (1982) 1175.
174. S. Caroli, O. Senofonte, N. Violante, and R. Astrologo, Analytical capabilities of a
microwave-coupled hollow-cathode discharge, J. Anal. At. Spectrorn. 3 (1988) 887.
175. M. Pataky-Szab6, L. Papp, and B. Derecskei, Investigations of processes in the three-
electrode hollow cathode emission source, Acta Chim. Hung. 126 (1989) 359.
176. P. Pianarosa, Y. Demers, and J. M. Gagne, Atomization of thorium in a hollow-cathode
type discharge, Spectrochirn. Acta 39B (1984) 761.
177. P. Pianarosa, P. Bouchard, J. P. Saint-Dizier, and J. M. Gagne, Density ofuranium ions
in the 41°9 / 2 ground state in a hollow-cathode type discharge, Appl. Opt. 22 (1983) 1568.
178. R. Simonneau and R. Sacks, Modulation of commercial hollow cathode lamps by a
magnetic field in a magnetron configuration, Appl. Spectrosc. 42 (1988) 1032.
179. S. Tanaka, M. Akiba, H. Horiike, Y. Okumura, and Y. Ohara, Effect ofmagnetic field
on the characteristics of hollow cathode ion source, Rev. Sei. Instrum. 54 (1983) 1104.
180. W. G. Tong and D. A. Chen, Doppler-free spectroscopy based on phase conjugation by
degenerate four-wave mixing in hollow cathode discharge, Appl. Spectrosc. 41 (1987) 586.
181. E. M. van Veldhuizen and F. J. de Hoog, Analysis of a Cu-Ne hollow cathode glow
discharge at intermediate currents, J. Phys. D 17 (1984) 953.
182. M. J. Verheijen, H. C. W. Beijerinck, P. W. E. Berkers, D. C. Schram, and N. F. Verster,
A hollow cathode arc in neon: Simultaneous laser probing and molecular-beam sampling
ofmetastable atoms as a plasma diagnostic, J. Appl. Phys. 56 (1984) 3141.
183. N. Violante, O. Senofonte, A. Marconi, O. Falasca, and S. Caroli, An investigation of
the sputtering process in the microwave-coupled hollow cathode discharge, Can. J. Spec-
trosc. 33 (1988) 49.
184. W. Winiarczyk and L. Krause, Production of copper vapour in a pulsed hollow cathode
discharge, J. Quant. Spectrosc. Radiat. Transfer 33 (1985) 581.
185. W. Winiarczyk and L. Krause, Emission and absorption studies of copper vapor ejected
from a pulsed hOllOW cathode discharge, J. Quant. Spectrosc. Radiat. Transfer 34 (1985)
163.
186. D. Z. Zhechev and I. T. Koleva, Self-alignment and radiative lifetime measurements of
some PIllevels in a hollow-cathode discharge, Phys. Ser. 34 (1986) 221.
187. D. Z. Zhechev, Incoherent signal at Hanle experiments in a hollow cathode discharge,
Spectrosc. Lett. 20 (1987) 111.
188. W. Zymicki and A. B. Basily, Atom and ion temperatures in a hollow cathode discharge,
Spectrosc. Lett. 19 (1985) 713.
189. W. Zymicki, Excitation temperatures of atoms and ions in a hollow cathode discharge,
Spectrochim. Acta 40B (1985) 995.
190. W. Zymicki, Z. Tomasik, and I. Nowicka, Spectroscopic measurement of plasma tempera-
tures in a hollow cathode discharge, Spectrosc. Lett. 17 (1984) 207.
7
Analysis of Nonconducting
Sampie Types
Michael R. Winchester, Douglas C. Duckworth,
and R. Kenneth Marcus
7. 1. Introduction
New developments in the field of specialized materials continually present

new challenges to the analytical chemist. The improved performance of novel
glasses and advanced ceramics demands improved analytical techniques
because trace contaminants in such materials critically affect their perform-
ance. Advances in the material sciences are thus dependent on improved
analytical chemistry and more sensitive methods for the analysis of noncon-
ducting materials. Broekaert et al. have reviewed state-of-the-art techniques
for the analysis of advanced ceramics, comparing and contrasting the per-
formance of eachYl
Traditional methods for the elemental analysis of bulk materials, such
as atomic absorption and inductively coupled plasma mass spectrometry
and atomic emission spectroscopy, are typically solution-based techniques.
These are tried-and-proven successful analytical techniques; however, they
require dissolution procedures that can be difficult and time-consuming.
Dilution of the analyte and concomitant contamination, resulting in lowered
sensitivities, quantitative errors, and spectral interferences, are obviously a
concern.
Michael R. Winchester. Inorganic Chemistry Research Division, National Institute
for Standards and Technology, Gaithersburg, Maryland 20899. Douglas C.
Duckworth • Analytical Chemistry Division, Oak Ridge National Laboratory, Oak Ridge,
Tennessee 37831. R. Kenneth Marcus • Department ofChemistry, Howard L. Hunter
Chemical Laboratories, Clemson University, Clemson, South Carolina 29634
263
264 CHAPTER 7
Direet solids teehniques for noneondueting material analysis are quite

powerful in some respeets and do not require dissolution procedures; how-
ever, they are not free of analytically undesirable effects. Techniques such
as secondary ion mass spectrometry provide very low deteetion limits and
lateral spatial resolution, though matrix and ehemical sensitivities in the ion
formation process can be severe. X-ray and electron spectroscopies offer
lateral spatial resolution but typically lack the sensitivity for trace compo-
nent detection. Erratic vaporization and ionization processes in spark
sources require long integration times, although relative and absolute sensi-
tivities are very good.
Much research effort has been devoted to the analysis of nonconducting
species by glow discharge techniques. The methods to be presented in this
chapter inc1ude mixing the nonconducting material with a condueting host
matrix (much like spark source methodology) and direct analysis of solids
by radio frequency-powered glow diseharges. Both teehniques are solids
techniques and though dilution of the analyte is necessary in the first, none
is necessary in the latter. As mentioned in the preceding chapters, glow
discharges provide a steady-state source of excitation and ionization that is
relatively free of matrix effects due to separate atomization and excita-
tion/ionization steps. Also, the simplicity of the glow discharge makes it an
attraetive and a relatively inexpensive souree for nonconducting materials
analysis. In light of the specialized analytical techniques that are necessary
for the deve10pment of materials with improved performance, the glow dis-
charge may be an attractive alternative to more traditional approaches.
7.2. Application of the Direct Current Glow Discharge
7.2.1. Introduction
In the direct current (de)-powered glow discharge, the solid sampie

serves as the eathode, and so must be eleetrieally conduetive. In order to
atomize noneonductive solids in the dc glow discharge, the sampie must first
be rendered eonductive, generally by compaction into a suitable host matrix
material. This methodology is similar to the briquetting technique used with
are and spark discharge sourees and was first introduced for use with the dc
glow discharge by Dogan et al. in 1972Y)
The advantages that characterize the application of the dc glow dis-
charge to the direct analysis of nonconductive solids are basically those that
characterize the glow discharge itself. In particular, in comparison with
solution-based analytical methodologies, the need for dissolution of the
sampIe is avoided, often resulting in a savings of time, effort, and expense.
This is particularly evident in the ease of noneonductive materials that are
ANAL YSIS OF NONCONDUCTING SAMPLE TYPES 265
usually difficult to dissolve. Also, since the cathodic sputtering step is non-
thermal in nature, sampie atomization rates for various species are quite
uniform, varying typically by less than an order of magnitude. Because of
this fact, and aided by the process of steady-state sputtering (see Chapter
2), the gas-phase concentrations of sputtered species in the negative glow
region of the discharge are thought to be directly representative of the bulk
solid concentrations. Direct solids techniques that rely on thernlal atomiza-
ti on mechanisms are often characterized by atomization rates that vary by
several orders of magnitude from element to element, making representative
gas-phase concentrations an impossibility. Uniformity of sputter rates also
results in the fact that analyte speciation effects usually prove to be
insignificant.
An additional advantage of the glow discharge is the relative absence
of matrix effects. This matrix insensitivity results from the fact that the
atomization, excitation, and ionization steps are segregated in time and
space. This characteristic of the glow discharge is particularly relevant in
comparison with x-ray spectroscopies, which are inherently limited by severe
matrix effects. Another advantage is that the discharge is operated in a low-
pressure, inert atmosphere, which effectively inhibits efficient formation of
interfering molecular species often found in atmospheric flames and plasmas.
Furthermore, relatively low degrees of collisional and Doppler broadenings
result in narrow optical transitions, allowing optical isotopic analysis of
constituents with suitably large isotopic shifts. The narrow atomic transitions
in the glow discharge also lead to enhanced sensitivity and linear dynamic
range in atomic absorption analysis because of the elose match of line profiles
in the glow discharge atomizer and the hollow cathode light source.
There are, however, some disadvantages inherent to the sampie prepara-
tion process. For example, mixing with an appropriate host matrix material
is equivalent to analyte dilution and loss of sensitivity. Furthermore, prob-
lems of contamination, loss of analyte, and inhomogeneity mayaiso result.
Despite these pitfalls, the application of the dc glow discharge to the direct
solids analysis of electrically nonconductive sampies has been shown by a
number of researchers to be an attractive alternative to competing methods
of analysis. In the following sections, the factors and analytical practices
important to this methodology are discussed in detail. Also, example applica-
tions are present. It is hoped that the discussion will provide the reader with
an understanding of the advantages and limitations of the dc glow discharge
for such applications.
7.2.2. Sampie Preparation
Conventional dc glow discharge devices can be applied to the direct

analysis of electrically nonconductive solids if the sampie is first rendered
266 CHAPTER 7
conductive by compaction into a suitable conductive host matrix material. (2)

This involves intimately mixing the sampie with a conductive host matrix
material, both in powder form, and pressing a portion of the resulting mix-
ture to form an overall conductive solid sampie that can serve as the cathode
in the glow discharge. In order to perform this type of sampie preparation,
and obtain acceptable analytical data, a number of factors that affect ana-
lytical performance must be considered.
Handling of sampie materials prior to mixing is an important considera-
tion. This is because atmospheric gases trapped inside the interstices of the
sampie mixture during pressing can degrade the analytical characteristics of
the glow discharge. Additionally, water, which may exist as adsorbed or
hydrate waters on the surface of the sampie particles can be particularly
damaging. Partic1e size control may be somewhat effective at reducing inter-
stitial trapped contaminants. Heating procedures are often utilized to reduce
the amounts of water and other atmospheric gases on the surfaces of the
sampie particles prior to pressing. However, if strongly hygroscopic
materials and volatile analyte species are simultaneously present in the
sampie, it may not be possible to remove the majority of the chemisorbed
water without volatilizing analyte-containing compounds. Thus, heating
procedures may be inappropriate. All sam pie materials should be stored in
a desiccator prior to use. Likewise, pressed sampies should be isolated from
the atmosphere if immediate analysis is not anticipated.
Winchester et al.,(3) investigating the implementation of glow discharge
devices for the direct analysis of automotive catalysts, have reported the
observation of deleterious effects of water trapped in compacted sampies.
Automotive catalysts are composed principally of r-alumina particles into
which noble metals are chemisorbed. In this work, chemisorbed Pt and
Rh were determined by means of atomic absorption spectrophotometry.
Alumina is a highly hygroscopic material. Overnight drying of the sampies
at 110°C proved unsuccessful at removing a sufficient amount of the water
present in the sampies, but more rigorous drying procedures were not
attempted in order to avoid altering the chemical composition of the sampies.
Therefore, significant amounts of water were incorporated into the
compacted sampies. Conveniently, sputtering of a sampie in a glow dis-
charge results in sampie heating due to the transfer of kinetic energy from
the impinging ions to surface atoms. In this way, water may be removed
from the compacted sampie through vaporization. Unfortunately, if the rate
of vaporization is too fast, damage to the sampie may result. In this particu-
lar case, for example, as the compacted catalyst sampies were heated by this
mechanism, the trapped water expanded too rapidly as it was vaporized,
resulting in sampie cracking. In some instances, the sampies were even
rendered unusable by this process. At the least, the precision of analysis was
worsened, and presputter routines designed to slowly eliminate trapped
water, as weIl as any other adsorbed gases, were undertaken. Even with this
difficulty, however, the glow discharge analysis of these catalyst materials
was successful. As will be discussed in a subsequent section, the presence
of water in the glow discharge environment is harmful for other reasons
as weIl.
Mixing of sampie powders with host matrix materials is usually accom-
plished by means of a mechanical shaker or grinder. These vibrating mills
usually produce changes in the particle sizes through attrition. This is not
normally seen as detrimental and may in fact be viewed as advantageous.
Mixing times of approximately one-half to several minutes are necessary to
produce mixtures of sufficient homogeneity. Lomdahl et al. utilized the glass
mixing vessel illustrated in Fig. 7-1 to homogenize sampie powders.(4) The
vessel is rotated about a horizontal axis at 20 rpm, such that each rotation
results in an approximately 50% division and recombination. After a 30-min
mixing period, mixture homogeneities superior to that obtainable in vibrat-
ing mills are claimed. Additionally, litde or no particle attrition occurs. Such
a procedure would be advantageous if studies of the effects of particle size
distributions on analytical results are to be performed.
Once the samplejhost mixture has been satisfactorily homogenized, an
appropriate portion is compacted in a die press assembly to form the final
Figure 7-1. Glass mixing vessel (sampIe divider) for mixing powders.(4)
268 CHAPTER 7
electrically conductive solid sampie used in the analysis. Figure7-2 illustrates

a typical die press assembly used to produce sampie cathodes. Normally, a
pressure of at least several hundred psi is necessary to produce cathodes of
suitable strength. Pressed sampies may take the form of disks,(4-10) pins,(1l,12)
or hollow cathodesY3,14) The pressed cathode may constitute the entire
cathode, or only part of the cathode. A multiple-section hollow cathode
assembly is illustrated in Fig. 7-3a.(\3) With this assembly, the compacted
sampie mixture may comprise the entire inner surface of the hollow.
However, sputtering in the closed-end hollow cathode has been shown to be
concentrated toward the closed endY5-17) Therefore, pressing an entire
hollow cathode may be considered unnecessary. The multiple-section
assembly offers the analyst the flexibility to utilize the compacted sampie
mixture as only the base of the hollow cathode if so desired. The remainder
of the cathode inner surface may consist of a blank material that does not
contribute analyte species to the negative glow. As shown in Fig. 7-3b, a
compacted pellet of sampie mixture may alternatively be placed into a base
plate, which acts as a sampie holder and forms the closed end of a hollow
cathode assembly.(14)
Compacted sampies may not be capable of withstanding a great deal
of mechanical stress, such that a specialized sampie holder may be required.
This is usually the case in the application of the Grimm-type or similar glow
discharge lamps, in which the sampie is mounted on the outside of the
atomizer by pressing it firmly against an O-ring by means of a compression
bolt assembly and thereby forming a vacuum seal. For example, using a
sputtering cell similar in construction to the Grimm emission source,
Winchester and Marcus employed a brass sampie holder with a 1.25-cm-
diameter, O.l-cm-deep, recess in one face.(5) This sampie holder is diagram-
matically illustrated in Fig. 7-4. The inside walls of the recess are tapered
D Brass
[] Stainless Steel
Body ____-------r-Sample Mixture
Base Flate fer
Figure 7-2. Typical die press assembly used for pressing powder sam pie mixtures.
A
A a b
-+---8
C
E
o
Figure 7-3. Hollow cathode assemblies used for the analysis of compacted powder sampie
mixtures. In panel a, the compacted sampie mixture may constitute any or all of the inner
surface sections,(13) while in panel b, it constitutes only the end.(14) A, compacted sampie; B,
copper or graphite supporting cylinder; C, sealing O-ring; D, supporting plate; E, copper pellet
incorporating sampie disko
COlllpacted Sample
---~~~
~ ~ ~/
1. 25 CD! Diameter Recess
................
Water In.~~=::;~ _water Out
l e i================::::::Jlj
Figure 7-4. Brass, water-cooled, sampie holder for the analysis of powder mixtures compacted
into the shape of disks.
such that the diameter of the recess decreases slightly with increasing depth.
In this way, compacted sampie disks of 1.25-cm diameter and approximately
I-mm thickness can be press-fit into the recess against a clean, flat surface
until flush with the holder surface. The holder containing the sampie may
270 CHAPTER 7
then be mounted for analysis. Only the holder is in physical contact with
the O-ring and the compression bolt, such that no stress is placed on the
sampIe. Additionally, since only the compacted sampIe is exposed to the
discharge environment, only the sampIe is sputtered. The conductive holder
serves to couple the electrical power to the sampIe and is water-cooled in
an attempt to maintain ambient sampIe temperatures. Various solid-state
processes, such as diffusion, which can be detrimental from an analytical
standpoint, are known to occur in compacted sampIes (and alloys to a lesser
extent) at rates that are a function of sampIe temperature. These types of
processes will be discussed at length in a subsequent section. An additional
advantage of such a holder is that relatively little sampIe is needed for an
analysis, with each disk containing only about 200 mg of actual sampIe
material.
7.2.3. Analytical SampIe Composition
The composition of the sampIe is an important consideration for the

successful implementation of the dc glow discharge to the analysis of electri-
cally nonconductive solids. Compacted sampIe composition direct1y affects
the performance of the methodology, principally by determining the sputter
rate and structural integrity of the compacted sampIe and the amounts of
water and trapped gases in the compacted solid. These unwanted constitu-
ents are harmful to glow discharge analysis in terms of their effects both on
the structural integrity of the compacted sampIe and on gas-phase processes
within the discharge. SampIe composition factors to consider include the
choice of host matrix material, the relative percentages of sampIe and host
in the samplejhost mixture, and the particle sizes ofthe sampIe constituents.
Each of these considerations is discussed in the sections that follow.
7.2.3.1. Choice of Host Matrix Material
The choice of conductive host matrix material is paramount to the

success of the analysis. Boumans has shown that the sputter rate of an
alloy is determined by the sputter rates and concentrations of the individual
components comprising the alloy. (18) This should also be true of compacted
samples.1t follows that for good sensitivity, host materials with high sputter
rates are desirable. Marcus and Harrison compared the use of graphite,
silver, and copper as conductive host matrix materials for the emission spec-
trometric analysis offlint clay added in equal mass portions with the host (see
Table 7_1).0 9 ) Emission intensities for the analyte elements were significantly
higher when the clay sampIe was mixed with copper than when the sampIe
was mixed with either silver or graphite, which provided similar inten-
sities. The reduction in copper sputtering rate is similar to that found by
Table 7-1. Comparison ofEmission Intensities (Counts) and Net Sputter Weight Losses for
I : I Mixtures of NBS 97 Flint Clay in Graphite, Silver, and Copper Matrices(19)
Relative intensities in:
Element % in sam pie disk Graphite Silver Copper
AI 10.25 2507 3183 6354
Si 10.28 641 810 1133
Fe 0.16 265 288 325
Mg 0.05 454 534 796
30-min weight loss (mg)

Sampie and matrix 1.71 1.62 2.21
Pure matrix 2.31 18.52 9.02
Winchester and Marcus who observed that 1:4 (w:w) sampie mixtures of
bauxite (NBS SRM 69a) in copper sputter only 65% as fast as pressed pure
copper powder under similar discharge conditions. (20)
Silver is intrinsically a rapid sputterer, as shown in Table 7-1. However,
the presence of the clay in the mixture degraded the sputter rate by a factor
of 11.4. This effect was attributed to the large clay j silver volume ratio in the
mixture, owing to the relatively large density of silver, which caused the
conductance of the mixture to be seriously depleted relative to the conduct-
ance of the silver host itself. The large suppression of sputter rate with the
silver host as compared with the copper host has also been observed by EI
Alfy et al. (7) The low sputter rate of the graphitejclay mixture was attributed
to the intrinsically low sputter rate of graphite.
It is also desirable that the host matrix material not interfere with the
determination of the analytes of interest. In the implementation of optical
spectroscopies, such an interference would most likely take the form of
overlapping absorption or emission lines, but because of the inherently
narrow linewidths of atomic transitions in the glow discharge, such spectro-
scopic interferences are rare. Isobaric interferences in low- to medium-resolu-
tion mass spectrometry, on the other hand, are not rare, and so the choice
of host matrix material is somewhat more important in such applications.
Therefore, it may be necessary to judiciously choose a host matrix material
that does not interfere isobarically. Other desirable characteristics of the
host include good mechanical strength upon pressing, high thermal conduc-
tivity, low cost, and availability in high purity and a variety of particle sizes.
Copper,(4,6,8-1O.\4) silver,(8) graphite,(1l,12) tantalum,czI,22) iron,(7) nickel,(7) and
a 4: 1 (w:w) mixture of copper and graphite(13) have all been used success-
fully. It is interesting to note that Ta possesses the added advantage of
behaving as a getter, meaning that it tends to form strong oxide bonds,
etfectively "sucking" oxygen from the discharge environment. Thus, if it is
272 CHAPTER 7
used as the host matrix material, isobaric molecular oxide interferences in

glow discharge mass spectrometry experiments may be substantially
reduced. (22)
7.2.3.2. Relative Percentages of SampIe and Hast in Mixture
When choosing an appropriate samplejhost composition, a trade-off

between sputtering rate and analyte concentration must be considered. As
the percentage ofnonconductive sampie material in the samplejhost mixture
is increased, the concentrations of analyte species in the compacted sampie
also increase. However, sputter rates are seen to decrease with increasing
percentages of nonconductive materials. (22) The reasons for this are some-
what complex and not well understood. Because of the larger impedance
associated with the incorporation of the nonconductive material, the dis-
charge voltage is larger at constant current and pressure, meaning that ions
of the support gas impinge on the surface of the cathode with a higher
average kinetic energy. This should result in enhanced sputtering, but due
to the usually stronger molecular bonding in the nonconductive sampie
(typically oxides, carbides, and the like) as compared with the metallic bond
strengths of the conductive host matrix material, lower sputter rates are
expected.
The choice of sampie composition may be complicated if high sputter
rates are desired and the sputtering source utilized is characterized by spa-
tially nonuniform sputter patterns across the surface of the sampie, such as
with the so-called "jet-enhanced" discharges (see Chapter 3). The non-
uniform sputtering results in spatially nonuniform heating of the sampie
surface, which imposes considerable stress on the compacted sampie,
particularly at the rapid heating rates induced by high discharge powers.
Because of the presence of temperature gradients across the sampie surface,
the compacted sampie may fracture, producing an erratic discharge. The
thermal conductivity and mechanical strength of compacted sampies usually
decrease with increasing nonconductive material content, meaning that
higher percentage compacted sampies cannot necessarily be sputtered at
discharge powers as large as lower percentage compacted sampies. The
effects of sampie heating may be somewhat alleviated by use of sampie
cooling. Even using water cooling, however, near-surface sampie tempera-
tures of 800-900 K have been measured.(23) As a result, the choice of non-
conductive material percentage must incorporate considerations of this
effect.
Discharge current and pressure (voltage) influence not only sampie
sputter and heating rates, but also excitation and ionization conditions in
the discharge. Therefore, in the case of spatially nonuniform sputtering, the
most effective method of optimization involves monitoring the analytical
signal of interest while varying discharge current and pressure, for a range
of sampie makeups. Although somewhat time-consuming, this methodology
is the only way to deduce the parameters necessary for optimum analytical
operation. In those systems where spatially nonuniform sputtering is not
problematic, the nonconductive percentage may generally be optimized
independent of other discharge conditions. In such cases, maximum ana-
lytical signals are usually obtained at nonconductive material percentages of
20% and most researchers agree on this.(7,13,16,17) However, other percentages
may be used successfully. In cases involving spatially nonuniform sputtering,
lower percentages may be analytically optimum. For example, 10% sampIes
were found to be optimal in one study involving the analysis of alumina-
based automotive catalyst materials. (3)
7.2.3.3. Effect of Particle Size
Another important factor influencing the success of the technique is the

range of particle sizes present in the sampie mixture. In order to obtain a
discharge whose gas-phase concentrations are representative of the bulk
sampie, the particle sizes present in the pressed sampie must be small com-
pared with the rate of removal of surface layers. Furthermore, discharge
stability appears to be inversely related to particle size. Owing to these facts,
small particle sizes are preferable in terms of accuracy and precision. It
should be kept in mind, however, that the use of smaller particles increases
the specific analyte surface area, allowing for more adsorbed or chemisorbed
water in the final pressed solid. Nevertheless, homogeneity and discharge
stability are usually the dominant concerns, and many researchers agree that
particle sizes should be kept below approximately 30-40 J1.m in diameter in
order to obtain acceptable accuracy and precision of analysis. (4,7,13) However,
so long as thorough mixing procedures are adopted and sputter rates are
relatively high, successful analyses may be performed with little or no atten-
tion paid to particle size. (5)
7.2.3.4. Effect of Trapped Gases and Water
Pressing of powders inevitably results in the incorporation of trapped

gases, including water vapor, into the pressed sampie. The presence of such
unwanted constituents can effectively degrade the analytical characteristics
of the glow discharge environment. Completely ridding the sampie of these
types of components is not possible. In fact, even the sputtering of alloys,
rather than compacted solids, does not totally eradicate the problems
associated with these constituents, because the materials commonly used
to construct glow discharge sources possess some propensity to maintain
adsorbed water and gases on their surfaces. These gases will outgas during
274 CHAPTER 7
discharge operation at a rate determined by the temperature of the glow

discharge source and the nature of the material from which it is
constructed. (24)
Nitrogen is a common contaminant in compacted sampIes and in the
glow discharge source since it composes 78% of the Earth's atmosphere.
The effects of molecular nitrogen on argon and neon discharges have been
studied.(25) In these studies, N 2 was found to effectively quench argon meta-
stable atoms through collisions in which the nitrogen molecule is excited
into the C 3Pu level. The C 3Pu level of N 2 lies 11.05 eV above the ground
state while the argon metastable levels reside 11.55 and 11.72 eV above the
ground state. The resulting small energy defect of 0.50 or 0.67 eV accounts
for the successful quenching of argon metastables by N 2. Since argon ions
are most rapidly produced in the discharge by electron impact ionization of
argon metastables, quenching is observed to lower the discharge current
under constant voltage conditions. Sputter rates and emission intensities of
sputtered species also decrease compared with pure argon conditions, as a
result of the lower current density at the cathode surface. Many other gases
have been found to have similar quenching effects on the metastables of
argon, krypton, and xenon. (26) In fact, molecular oxygen has been found to
quench argon metastable atoms at a rate that is a factor of seven faster than
molecular nitrogen. Other common gases on the list of perpetrators inc1ude
H 2, CO, NO, N 20, CO2, HCI, CCI4, H 20, and CH4.
Because of the fact that waters of hydration may often accompany
sampIe components and the fact that water vapor may be the atmospheric
constituent most prevalently adsorbed onto the inside surfaces of the
discharge source, water may be the most analytically damaging of the
commonly encountered contaminants. As noted above, water is known to
collisionally quench metastable species. Through various excitation and ion-
ization processes, water vapor yields several contaminant ion species in the
discharge, inc1uding H+, Ht, 0+, OH+, H 2 0+, H 30+, and H 30+· nH20,
where n = 1 to 5.(11) Mass spectra are often hampered by isobaric interfer-
ences from such species. Water, however, is particularly detrimental for
another reason as weIl. Stern and Caswell approximated the mobilities of
H+ and Ht, the primary dissociated ionic species of water, as being between
20 and 40 times that of Ar +. (27) These ions may thus carry a proportionately
large percentage ofthe total discharge current and produce very little sputter-
ing. This is because the ratio of the masses of the ions to the atoms present
in the typical sampIe is such that energy transfer cross sections are quite low,
resulting in very low sputter yields. Gough et al. observed that absorbance
values of sputtered species, which should refiect relative sputter rates, were
indeed lower when "moist" argon with only 17 ppm water was used as
opposed to "dry" argon with only 3 ppm water.(24) Similar correlations
ANAL YSIS OF NONCONOUCTING SAMPLE TYPES 275
between absorbance values (sputter rates) and approximate hydration water
concentrations of compacted sampIes have been observed by Winchester and
Marcus. (20)
Although sputtering and other discharge processes may be altered by
the presence of contaminants, analyses should still be relatively successful,
provided the concentrations of the contaminant species remain constant.
Unfortunately, this is never the case, and so methodologies for reducing the
effects of contaminants are essential to the successful analysis. Proper clean-
ing and outgassing of the glow discharge source prior to analysis can be
quite effective. As mentioned previously, control of the particle sizes of
sampIe components, heating of sampIe materials before mixing, and storage
of sampIes in a desiccator can help prevent the trapping of gases, especially
water, into the compacted sampIe. As a further precaution, a mild vacuum
may be applied to the sampIe mixture during pressingY) Additionally, the
compacted sampIe may be outgassed under vacuum andj or mild heat prior
to analysis. Perhaps the most effective means of reducing the effects of
contaminants, however, is the use of a suitable presputtering period prior to
data acquisition. Loving and Harrison demonstrated mass spectrometrically
the erratic behavior of glow discharge-produced ions immediately after dis-
charge initiation as a result of changing concentrations of water vapor. (11)
After the discharge had time to "dry out," ion signals became temporally
stable, intense, and reproducible.
Presputter procedures are generally undertaken simply by allowing the
analytical signals to stabilize at the discharge conditions desired for analysis
before acquiring data. However, the use of a sputtering cell with spatially
nonuniform sputtering may necessitate more complicated procedures. If
large discharge powers are desired, such that sputter-induced heating is fast,
the compacted sampIe may not be able to withstand the mechanical stress
caused by the rapidly expanding trapped gases. An erratic discharge is usu-
ally the result. For this reason, presputter methods in such cases should
involve attaining the desired discharge conditions through several steps,
each successive step employing a somewhat higher discharge power than
the preceding step. This methodology has been demonstrated to be quite
appropriate, though possibly time-consuming. (3)
7.2.4. Atomic Transport in the Compacted Solid
Atomic transport processes have been suggested to occur in compacted

sampIes under glow discharge sputtering. These are important to be aware
of, since their presence contributes to the approach to steady-state sputtering
conditions. At first glance, it would appear that these processes are so severe
276 CHAPTER 7
that the acquisition of accurate quantitative data would be impossible. How-

ever, numerous experiments have demonstrated the feasibility, and appropri-
ateness, of acquiring such data with the glow discharge. The following
discussion is offered simply to provide insight into the processes possibly
occurring in the sampIe during glow discharge operation.
The approach to a steady state of sputtering necessarily results in the
formation of a near-surface altered layer that possesses a composition some-
what different than the bulk material deeper in the sampIe. A number of
factors contribute to the formation of the altered layer and thus the develop-
ment of and approach to the steady-state sputtering condition. Differing
sputter rates of the various components at the surface, as weIl as solid-state
diffusion and chemical reactions occurring in the near-surface layers, will
affect sputter-equilibration times.
The role of surface temperature in determining the composition and
thickness of the altered layer has been observed in a number of studies
involving alloys. Swartzfager et al., sputtering a 50: 50 Cu: Ni alloy with 2.0-
keV Ne + (SIMS-type sputtering), found an exponential increase in the
altered layer thickness with sampIe temperature, particularly above 200 K. (28)
Similar results have been reported by Lam et al.(29) In another study, the
surface of a Cu-Ni alloy was found by Auger electron spectrometry to be
enriched in Ni after sputtering at near room temperature, while sputtering
at 250°C effectively enriched the surface with Cu atoms. (30) The Cu enhance-
ment was attributed to solid-state diffusion of the Cu from the bulk to the
surface, accelerated by the higher temperature. Anderson reported that the
relative sputter rates of Ag and Cu in Ag:Cu eutectic (60:40) targe!s bom-
barded by 100-eV Ar+ are significantly different at low sampIe temperatures
than they are at high sampIe temperatures.(31) Since sputter yields of pure
metals are thought to be relatively independent of temperature, (30) this implies
that the differing surface compositions as a function of temperature are the
result of the presence of temperature-dependent solid-phase processes.
Solid-state diffusion and chemical phenomena are likely present in com-
pacted sampIes. However, solid-state processes may be somewhat different
in compacted sampIes than they are in alloys and pure metals, because of
the presence of the individual particles that compose the sampIe. Mai and
Scholze have developed a model of the formation of the altered layer in
compacted sampIes (see Fig. 7_5).(23,32) Similar to alloy sputtering, bulk
diffusion (within individual partic1es), surface diffusion, and radiation-
enhanced diffusion are present. Contact boundaries (the boundaries between
individual partic1es) and dislocations within individual particles formed by
the compaction process provide regions of enhanced solid-phase diffusivity
relative to the more dense regions of a bulk solid. Therefore, "short-circuit"
diffusion along contact boundaries or dislocations will dominate in such
sampIes. In fact, diffusion of large groups of atoms mayaIso be possible
:t.
~.cu ~
.....
~''''I ~
;;;
~:;:::::;:::::;::::l' contlct
bound.,les
~
cont.ct grain l'\J"'6.. . plrtlll tri nl port

bound.ry lo,med phenomenon ~o
by complctlon
- - - _. modilied ne .. ~
surl.ce region
~
:::l
~
~
~
.....
",
'"'4
~
g,lln Zl
boundary
_ dlslocltlon
IIne
•
(@
Figure 7-5. Schematic representation ofpartial transport phenomena supposed to occur during glow discharge excitation.
I, sputtering; 2, redeposition; 3, bulk diffusion; 4, grain boundary diffusion; 5, contact grain boundary diffusion; 6, I\)
diffusion along dislocations; 7, surface diffusion; 8, radiation-enhanced diffusion; 9, solid-state reactions. (32)
~
278 CHAPTER 7
along boundaries. It has been estimated that for a wide range of sampie
temperatures the rates of surface and "short-circuit" diffusions may be quite
similar, while bulk diffusion may be orders of magnitude slower. The pres-
ence of the boundaries also results in enhanced chemical reaction rates in
these areas as weIl. The rates of any such reactions will of course also depend
on temperature and the reactive characteristics of the components.
Small-diameter component particles were shown previously to be an
important factor in terms of accuracy and precision of analysis. Particle size
also partially determines the rates of the various solid-state processes being
discussed here. For example, as the particle sizes of the components of a
compacted sampie decrease, the density of contact boundaries increases.
This could result in a further enhancement of the rate of short-circuit diffu-
sion along contact boundaries, as weIl as the overall rates of chemical
reactions. Simultaneously, small particles may decrease these rates by
increasing the thermal conductivity of the sampie and thereby effectively
decreasing sampie surface temperatures.
Although the precise effect of particle size on the rates of solid-state
processes is not known, the rates ofthese processes will determine the forma-
tion of the altered layer and the time necessary to reach the condition of
steady-state sputtering, presputter times necessary for the establishment of
equilibrium sputtering conditions may in fact be shortened by the use of
highly homogenized sampies of small particle size. Mai and Scholze observed
this type of effect by emission spectroscopy for sampies of ZnFe204 and
ZnO/Fe203 mixed in Cu powder.(23) As illustrated in Fig. 7-6, emission
intensities from the coarse mixtures needed nearly 9 min for stabilization,
while those from the highly homogenized mixtures required only about
4 min. Additionally, the behavior ofthe two coarse sampie mixtures contain-
ing identical overall stoichiometries but different speciations was significantly
different before the establishment of equilibrium, while the behavior of the
two isostoichiometric homogenized sampie mixtures was quite comparable
both before and after the establishment of equilibrium. These data are strong
support for the importance of solid-state processes in establishing steady-
state sputter conditions. Furthermore, the data indicate the need for homo-
genized sampie mixtures of small particle sizes in terms of reducing necessary
presputter times. Thus, from all viewpoints, smallparticle sizes may be
considered advantageous to the analysis.
The necessary presputter time required for attainment of steady-state
sputtering has also been observed to depend on the discharge power used.(23)
As illustrated in Fig. 7-7, the emission signals of W from compacted WC
(both Wand C are electrically conductive) and from a mixture of W in Cu
powder (50% w/w) do not coincide even after 30 min of sputtering at 14 W.
Sputtering at 30 W attains equilibrium only after approximately 11 min,
while sputtering at 100 W produces the altered layer and steady-state sput-
~~------,--------r------~--
o ~
~.nm
~
Figure 7-6. Temporal behavior of emission intensities of Zn (crosses) and Fe (solid dots) in
different compounds compacted into Cu host matrix for (a) coarse m;xture and (b) sufficiently
homogenized mixture. (23)
tering in less than 3 min. The use of large discharge powers during pre-
sputtering, however, may not necessarily be advantageous. This is due to
the fact that the altered layer composition and thickness may be quite differ-
ent from one set of discharge conditions to the next, owing to the differing
current densities and heating rates. The most expedient procedure may be
to presputter the sampie at whatever discharge conditions are necessary for
analysis. Additionally, as mentioned previously, the use of stepped pre-
sputtering routines may be needed in some cases.
In the experience of the authors, the sputtering rates of even compacted
sampies are so high as to make the influence of any local atomic transport
within the sampie negligible in the attainment of steady-state sputtering
conditions. Studies of the sputtering of various inorganic compounds of
the same metal ions show little influence on sputtering characteristics, (5,19)
suggesting minimal amounts of "chemistry" within the sampie matrix. In
280 CHAPTER 7
o
Ip""oowl
• mon «
w-cu
lAU
-I
o
4 \
IP"30W I
-~"""'-:=s:,= r » ' 7"f
I •·
Y ". A• A
w
Ip",~w t
-I
10 20 . . 30
time, mit)
Figure 7-7. Temporal behavior ofW emission intensity for WC single phase (solid dots) and
W jC mixture (crosses) compacted into Cu host matrix for three different discharge powers.(23)
terms of plasma stabilization and the like, the effects of outgassing and
thermal breakdown seem to be far more important than transport
phenomena.
7.2.5. Methods for Improved Quantitation
Qualitative, semiquantitative, and quantitative analytical data may be

obtained from the glow discharge through optical and mass spectroscopies.
In the simplest experiments, data are collected as raw signals using dc glow
discharges that are continuously operational, Le., nonpulsed. Quantitative
analyses done in this way rely on the reproducibility of the glow discharge
environment and conditions, and the lack of systematic errors, to obtain
acceptable accuracy. Owing to the fact that glow discharge devices are in
fact quite reproducible and appear to suffer from few systematic errors,
analyses can often be performed in this way with accuracies of better than
5% relative. However, electrically nonconductive sampie materials com-
pacted into conductive host matrices introduce other variables that may be
somewhat irreproducible. These include the amount of trapped water vapor
and atmospheric gases contained within the compacted sampie, the sampie
structural integrity, and the like. Therefore, experimental conditions may

be somewhat less reproducible when analysis of electrically nonconductive
materials is to be compared with that of alloys. As a result, the accuracies
of such simple analyses may be affected.
The analytical performance ofthe glow discharge source for the analysis
of electrically nonconductive materials may be improved by the use of inter-
nal standardization, as weH as by the use of a pulsed discharge. In fact, those
methodologies may in some cases be essential to the successful implementa-
tion of the discharge. Therefore, each is discussed in some detail in the
following sections.
7.2.5.1. Internal Standardization
Determinations of analytes in nonconductive sampIe types compacted

into host matrix materials have been done most often by comparison with
suitably prepared external standards (i.e., through calibration curves).
Because of the relative lack of matrix and speciation effects in the glow
discharge, matrix matching of external standards to sampIes is not a neces-
sity, although it may be advantageous. However, the reproducibility and
stability of analytical signals obtained from the glow discharge are affected
by a number of parameters. Some of these have already been discussed, such
as the presence of contaminant species. This is most likely a problem if the
discharge source must be vented to atmosphere to change sampIes, such as
with Grimm-type sources. The placement of the sampIe into the discharge
source mayaIso affect the analytical signal attained. This may be particularly
true with the use of sampIe holders, where the efficiency of sampIe cooling
could be changed by irreproducible placement of the sampIe in the holder.
If the assumption is made that factors affecting a given analyte will
similarly affect every analyte in the sampIe, then utilization of internal
standardization is suggested. This approach assumes that the ratios of the
concentrations of analyte species in the vapor phase are the same as those
in the solid sampIe. This is likely a valid assumption, provided that a steady
state has been attained and the only physical sputtering is involved. This
is also true of chemical sputtering, which is species specific. Nevertheless,
McDonald has presented evidence to support the validity of this assumption
for five types of alloyed sampIes analyzed by atomic absorption spectro-
photometry, as shown in Table 7_2.(8) Although the ratio ofNi to Cu concen-
trations (Re) in the alloy sampIes varied over four orders of magnitude, the
ratio oftheir absorbances (RA) in the negative glow were observed to behave
similarly, such that Re/RA remained relatively constant.
Because of the relative absence· of matrix effects in the glow discharge,
practically any atom present in the sampIe, including the host matrix
material, a constituent in the original sampIe material, or a deliberately
282 CHAPTER 7
Table 7-2. Proportionality of Concentration Ratios to Absorb-

anee Ratios for Solid Alloy Samples(8)a
Sampleb R e = CNi/Ccu RA = ANi/A cu Re/RA
SS406 5.28 0.851 6.2
H1l68 3.96 0.572 6.9
SAC1576 0.625 0.0952 6.6
C1100 0.000771 0.000116 6.6
Cll15 0.000841 0.000136 6.2
·Cu measured at 324.7 nm and Ni measured at 352.5 nm.
·See Ref. 8 for sampie descriptions.
added component, may be employed as the internal standard. This is in

contrast to the use of internal standardization in other spectrochemical
sources prone to gas-phase matrix effects, where the internal standard must
be quite similar to the analyte chemically and spectroscopically in order to
adequately compensate for interferences that might be present. However,
even in the glow discharge several characteristics of the internal standard
are desirable. In particular, the added internal standard should be chemically
inert, nonhygroscopic, available in high purity, and easily handled. For use
of a constituent other than the host matrix material, it should be an element
with reasonable sensitivity in whatever analytical methodology is to be
employed, such that a small but easily measured quantity can be added
reproducibly to the sampie mixture. In this context, use of the host matrix
material requires the use of spectroscopic transitions with low sensitivity for
optical spectrometries. For mass spectrometry, a minor isotope of the host
matrix may be used.
According to synthesis of variance, (33) the use of internal standardiza-
tion necessarily reduces analytical precision by virtue of the variance associ-
ated with the added measurement of the reference signal from the internal
standard. Limits of detection may be worsened by this phenomenon, but the
improvement in accuracy of the analysis usually makes internal standardiza-
tion quite beneficial.
Because only a specific volume of the discharge is monitored analytically
during analysis, accuracy of determinations involving internal standard-
ization may be affected by changing concentration gradients of sputtered
species in the negative glow if various species cannot be expected to behave
similarly. (34) Analyte species of widely differing masses are known to diffuse
at differing rates in the gas phase, producing concentration gradients. (34)
Other studies have revealed the presence of significant concentration gradi-
ents of sputtered species in the discharge, (22,35,36) due not only to differing
diffusion rates, but also to spatially differing atomization(20) and
ionization(36) rates. Provided that concentration gradients remain constant
throughout the analysis, no problems should result. Nevertheless, the
presence of gradients should be kept in mind in case of any problems that
might ensue.
Use of the host matrix material as the internal standard has met with
mixed success. Determination of Ni (in sampies of nickel salts compacted in
Cu) using Cu as the internal standard was quite successful in one atomic
absorption study.(8) In another study involving the analysis of automotive
catalysts, the host matrix material, again Cu, was used successfully in AAS. (3)
However, Lomdahl et al. reported an unsuccessful attempt to determine
nickel and iron, as oxides, compacted in Cu, utilizing the Cu as the internal
standard for emission spectroscopy.(4) The reason for this apparent discrep-
ancy is unknown. However, the authors suggested that the high concentra-
tion of oxygen in the discharge environment may have contributed to
environmental instabilities.
Use of an additive (spike) as the internal standard, on the other hand,
has been shown to be quite appropriate in several investigations. (3,8,13) In
particular, McDonald reported accurate determinations of Cu (in Cu slag
sampies incorporated into Ag) using Ni as the internal standard, even though
sputter rates for the compressed sampies varied by a factor offive at constant
discharge current and pressure,(8) Ni was added to the sampie mixtures to
yield a final concentration of 1%. For five sampies, relative quantitative
errors were less than 4.5%. Without the use of the internal standard, errors
of up to 500% would have been expected, based on the widely varying
sputter rates. Additionally, Marcus and co-workers have demonstrated the
successful use of Ba as an added internal standard in the determination of
Pt and Rh in y-alumina-based automotive catalyst materials by AAS.(3)
Average accuracy was improved by a factor of 6 for Pt and a factor of 9 far
Rh by using the internal standard. Ba was added as Ba2C0 3 to obtain a final
Ba concentration in the compacted solid of 0.27%. It is interesting to note
that the Ba(II) 455.4 nm line was utilized to measure the reference absorb-
ance signal. Although likely to be affected by changing ionization rates
in the negative glow, this ionic transition was still useful for the internal
standardization. Quantitation in this study used both calibration plots and
relative sensitivity factors. These quantitation methodologies will be outlined
further in a subsequent section.
In short, although glow discharge analyses of compacted sampies can be
performed quite successfully with direct external standardization (calibration
curves), it would appear that internal standardization might be a generally
more attractive means of improving analytical determinations. Certainly,
this form of quantitation should be investigated further.
7.2.5.2. Pu/sed G/ow Discharges
An electrically nonconductive material placed directly in contact with

an electrically conductive cathode being sputtered in a dc glow discharge
284 CHAPTER 7
will eventually sputter. However, the onset of sputtering of the nonconduc-

tive sampie will begin only after the material has been reduced to the metallic
state, (37) possibly by thermal decomposition or random bombardment with
energetic particles near the cathode, or by redeposition. If the discharge is
pulsed rapidly (>ca. 1kHz), the nonconductive sampie may be sputtered
directly immediately after discharge initiation, in the same way that alloys
may be sputtered in the dc glow discharge. (38) High-frequency sputtering of
tbis type will be discussed in detail in the second half of this chapter.
Even if electrically nonconductive sampies compacted into host matrix
materials are to be sputtered in the dc glow discharge, use of a pulsed
(power modulated) power supply also holds analytical benefits. Pulsed glow
discharge operation has proven to be quite beneficial in the analyses of
electrically nonconductive materials compacted into host matrix materials. (3)
For example, high discharge powers which result in arcing in the continuous
dc mode may be applied while avoiding arcing. This is possible because the
pulse length is kept short enough that any electric arc does not have sufficient
time to be established before the end of the pulse. Duty cycles may be kept
low to maintain a low average power, such that the problems associated
with excessive sampie heating may be a10ided. Shorter presputter times may
result from the higher discharge powers, because of the more rapid but still
gentle release of trapped gases from the compacted solid.
The use of the large discharge powers during the pulse, coupled with
synchronous detection, enables higher sensitivities and signal-to-noise ratios
to be obtained relative to those attainable in the continuous dc mode,
because of higher sputter rates during the pulse and a more energetic gas-
phase environment. (39) Higher sputter rates are produced by virtue of the
fact that sputter rates are determined primarily by discharge power.(40) The
sensitivity improvement associated with enhanced sputtering is especially
advantageous when determining analytes in sampies with intrinsically low
sputter rates (e.g. compacted sampies). The more energetic gas-phase
environment is a result of increased electron number density and electron
temperature during the high-power pulse, promoting more efficient electron
impact collisional processes in the gas phase.
When using the pulsed glow discharge, accuracy and precision can
possibly be degraded if pulse lengths utilized are too short. As species are
sputtered from the cathode, different analytes may diffuse into the sampling
volume of the negative glow at differing rates, because of dissimilar masses
and gas-kinetic cross sections. If the pulse length is short relative to the time
necessary to establish an equilibrium concentration of the various analytes in
the sampled volume, then the gas-phase concentrations of sputtered species
sampled will not necessarily accurately reflect their concentrations in the
original bulk solid, (39) even in the presence of steady-state sputtering. Pro-
vided that the behavior is reproducible, no inaccuracies should ensue with
short pulse durations. However, reduced accuracy cannot necessarily be
mIed out, and worsened precision is likely to result. This fact should be
considered when choosing a pulse duration.
Pulse power, duration, and repetition frequency are mutually dependent
in terms of their effects on discharge behavior, and optimization of these
pulse parameters may be somewhat time-consuming. Typical pulse param-
eters might include pulse powers in hundreds of watts, pulse durations on
the order of milliseconds,· and pulse repetition frequencies less than 100 Hz.
7.2.6. Illustrative Analytical Applications

Employing the sampling methodologies described in preceding sections,
the dc glow discharge source may be applied to the analysis of electricaHy
nonconductive materials. The remainder of this section focuses on practical
applications of dc glow discharge devices for the analyses of nonconductive
sampie types. Since atomic absorption, emission, and mass spectrometries
are the most commonly utilized detection methodologies, one example of
each is presented. These examples are given to illustrate the range of sampie
types amenable to dc glow discharge analyses, as weH as the sampie prepara-
tion considerations, quantitation methods, and representative figures of
merit. The reader is referred to the original publications for further details.
7.2.6.1. Atomic Absorption Spectrophotometry: Analysis of

Automotive Catalyst Materials
Winchester and colleagues have described the determination of Pt
and Rh in automotive catalyst materials using atomic absorption
spectrophotometry. (3) Such sampies consist of the noble metals chemisorbed
onto y-alumina particles. The Atomsource glow discharge atomizer (Analyte
Corp., Medford, Oreg.) was used to atomize the compacted solid sampies.
The Atomsource is described in some detail in Chapter 3. Of interest here
is the ability of the Atomsource to operate in a pulsed mode. As discussed
in a preceding section, such atomizer operation holds significant analytical
advantages. The Atomsource's mate spectrometer, the Analyte 16, atomic
absorption spectrophotometer, was used to detect the sputtered atoms. Inter-
nal standardization, in conjunction with calibration curve standardization,
and the use of relative sensitivity factors were employed as possible means
of quantitation.
Compacted calibration standards consisting of Ba2C03 , Pt02' H20,
Rh20 3, and diluent in a Cu host matrix (total nonconductive content = 5%)
were utilized. The raw support material onto which active platinum and
rhodium were deposited in the actual catalysts, consisting primarily of
286 CHAPTER 7
y-alumina, was employed as the diluent. Since alumina is considerably hygro-

scopic, it was dried overnight at approximately 110°C. Ba2C03 was added
to the sampies, Ba being used as an internal standard to compensate for
sample-to-sample sputter rate differences. The Cu host was also investigated
as a possible internal standard. SampIes and standards were presputtered
before analysis in order to allow water and trapped gases to escape slowly
without rupturing the sampIe disks and to allow absorbance values to stabi-
lize. Analyses were performed under discharge conditions of 10 Torr argon
and 300 mA for platinum and 10 Torr argon and 255 mA for rhodium,
barium, and copper. During both presputtering and data collection the
Atomsource was operated in a 7.75-Hz pulsed mode, with either a 26- or
52-msec pulse length, under current regulation.
The results of the analyses employing calibration curves are presented
in Table 7-3. Those data in the "corrected by ... " rows consist of the ratio
of analyte absorbance to internal standard absorbance plotted against their
relative concentrations. The data in the "raw absorbance" rows are simply
the analyte absorbance versus concentration. As is illustrated by the R 2
(goodness offit) values, linearity was acceptable in all cases. The determined
Pt and Rh concentrations are presented, along with "accepted" values. The
first three sampIes were produced gravimetrically from the metal oxides, in
order to determine the accuracy of the methodology in the absence of
certified standards, since no such standards are commercially available.
Therefore, the accepted values for these sampies are simply calculated from
the amounts of the constituents known to have been added to the sampie
mixtures. The fourth sampIe was an actual catalyst obtained from a commer-
cial producer. The accepted values listed for this sampIe were determined by
competing methods of analysis as indicated in Table 7-3. Of particular inter-
est, the value for the platinum in SampIe 4 was determined by the stannous
chloride method (differential spectrophotometry), which is currently consid-
ered to be the method of choice.
As indicated in Table 7-3, the glow discharge analyses using raw
(unweighted) absorbances were somewhat successful, with average relative
errors of 13.5% for Pt and 8.3% for Rh. This degree of accuracy may be
acceptable for less stringent analyses. The employment of the internal
standard elements greatly improved the performance of the method. Using
the Cu internal reference, the average relative errors were 2.59% for Pt and
3.27% for Rh. Using the Ba reference, the average relative errors were 1.91%
for Pt and 1.85% for Rh.
Relative sensitivity factors (RSFs) were also used in this study as a
possible means of quantitation. An RSF is an empirically determined quan-
tifier, based on the assumption that variations in analytical signal intensity
caused by variations in glow discharge conditions are not analyte specific
(see Section 7.2.5.1). An RSF is defined mathematically as the second
:t.
...~
~
Ci;
0
Table 7-3. Pt and Rh Determinations via Calibration PlotS(3) ."
:c:
0
Sampie No. :c:C')
Equation R2 2 3a 4 0
Platinum
~
c::
C')
Raw absorbances (1.11 x 1O-5 )x - 7.07 x 10- 5 0.998 0.578 b (±0.040%) 2.124 (±0.078%) 1.305% 0.463 (±0.049%)
:::!
Corrected by (13.0)x + 3.58 x 10- 4 0.997 0.470 (±0.023%) 1.938 (±0.140%) 1.164% 0.418 (±0.024%) :c:
CO)
Cu(I) 324.7 nm
Corrected by (9.93 x 1O-2 )x + 4.75 x 10-5 0.992 0.487 (±0.040%) 1.873 (±0.163%) 1.197% 0.432 (±0.052%) ~
Ba(II) 455.4 nm
Accepted value 0.477% 1.877% 1.149% 0.437%C (0.450%)d ...,~
Rhodium 0000f
"'<
Raw absorbances (1.91 x 1O-4 )x + 8.49 x 10- 5 0.998 0.467 (±0.036%) 0.186 (±0.003%) 0.400% 0.0860 (±0.0023%) ~
11)
Corrected by (225)x + 1.30 x 10- 3 0.995 0.449 (±0.043%) 0.164 (±0.007%) 0.364% 0.0798 (±0.0068%)
Cu(I) 324.7 nm
Corrected by (1.70)x + 2.50 x 10- 4 0.996 0.431 (±0.012%) 0.167 (±0.014%) 0.373% 0.0838 (±0.0025%)
Ba(II) 455.4 nm
Accepted value 0.427% 0.163% 0.366% 0.0856%e
aOnly one sampIe disk analyzed.
bAll values represent concentrations in original solid sampIe. Values in parentheses are one standard deviation.
CValue obtained by stannous chloride method (current method of choice).
dValue obtained by ICP-AES.
eAverage ofvalues obtained by flame AAS, DCP-AES, and ICP-AES.
~
00
"""J
288 CHAPTER 7
bracketed term in the equation
C~ = [Cf(A~/Ai)][(C~/CD(Ai!A~)] (7-1)
where Ca and C represent the concentrations of the analyte and internal

standard, A a and Ai their respective absorbances, and the subscripts rand s
represent the values obtained for reference and analytical sampIes, respec-
tively. Therefore, RSF values can be determined from a single external stan-
dard, or multiple standards, containing known concentrations of analyte
and internal standard. The analyte concentration in the unknown sampIe
may then be determined direct1y using the RSF, the absorbances of the
analyte and internal standard, and the known concentration of internal
standard in the unknown. Results using RSFs in this study proved to be
quite similar to those obtained with the calibration plots, and so will not be
presented here. The advantage ofusing RSFs as opposed to calibration plots
is that a single standard could conceivably be utilized for the analysis of a
large number of sampIes of widely varying composition.
In conclusion, the glow discharge methodology was shown to compete
quite weIl with more established methods of analysis, and both the Cu host
and the added Ba are suitable as internal standards. In general, analytical
precisions were found to be better than 7.5% (relative) when internal stan-
dardization was employed. No limits of detection were reported in the paper,
but were in the single-digit ppm range for Rh and in the double-digit ppm
range for pt.(20)
7.2.6.2. Atomic Emission Spectrometry: Analysis of

Geological Sampies
Marcus and Harrison have described the analysis of geological materials

by means of glow discharge atomic emission spectrometry.(19) In that work,
the hollow cathode plume (HCP) discharge source(41) was used to atomize
compacted solids and excite the sputtered analyte atoms. The hollow cathode
plume, depicted in Fig. 7-8, is a modified hollow cathode source incorporat-
ing agas ftow through the center and an exit orifice in the base, which
comprises the compacted sampIe disko Under certain conditions of pressure,
current, and orifice diameter, two hollow cathode discharges exist, one in
the main body and the second inside the orifice, which acts as a microcavity
hollow cathode under such conditions. Microcavity hollow cathode dis-
charge devices are known to produce intense analyte emission. (42) The gas
ftowing through the plume device causes the negative glow region of the
dual discharge to extend out of the orifice as depicted in Fig. 7-8. Intense
analyte emission from this extended glow region was monitored.
PLASIIA
PLUIIE
HIGH
ENERGY
REGION
SAIIPLE
DISK
t
DISK
HOLDER
dlacherge
gaa
Figure 7-8. Schematic representation of the hollow cathode plume source.(41)
Graphite, silver, and copper were investigated as possible host matrix

materials for geological specimens. These investigations showed that, on the
basis of analyte emission intensities (see Table 7-1), Cu was the best choice.
Studies of the emission intensities of various analytes (from a flint clay
sampIe compacted into Cu) as a function of the percentage of sampIe in the
sample/host mixture are shown in Table 7-4. These results show that 10-25%
sampIe mixtures may be considered analytically optimal. The authors con-
cluded that 10% sampIe mixtures could be considered better, however, on
the basis of the smaller required sampie sizes, the consequentiallowering of
the amounts of trapped gases and water incorporated into the compacted
sampies, and the normalization of the sampie matrix material's effect on
sputtering as a result of the larger host fraction.
Table 7-4. Elemental Emission Intensities as a Function of Flint Clay Concentration in a

Copper Host Matrix(19)
Relative intensities from single scan
Element 5% 10% 25% 50% 75%
Cu 1220730 1124420 711744 242160 22720
AI 7204 11984 11964 6710 1435
Si 1356 2245 2119 1121 522
Fe 277 321 337 284 250
Mg 803 1007 1026 600 297
Na 482 538 435 269 390
290 CHAPTER 7
As was noted in earlier discussions, presputtering is often necessary with

compacted sampIes in order to allow trapped gases and water to escape
slowly from the sampIes and to allow analytical signals to stabilize. The
authors found this to be the case in these studies as well. Figure 7-9 illustrates
the temporal response of AI(l) 309.2-nm emission upon initiation of the
discharge. A 5-min presputter (preburn) at a lower current (50 mA, 2 Torr
argon) was followed by an increase in current to optimal analytical condi-
tions (150 mA, 2 Torr argon). With this routine, about 10 min was required
to attain temporally stable emission.
In order to estimate the feasibility of using direct, single-standard analy-
sis of geological sampIes with this technique, the authors employed NBS 97a
flint c1ay as the single external standard (Le., no calibration plots were
generated) for the analysis of NBS 98 plastic c1ay. Their results are given in
Table 7-5. The calculated concentrations of the plastic c1ay constituents
demonstrate reasonable agreement with the assayed values. The deviations
ranged from 3.4% for Al to 14% for silicon, with an average of 8.1%. As
pointed out by the authors, the determinations were quite successful even in
the cases of Fe and Mg, which were present at a factor of three difference
in concentration between the two c1ays. Reproducibilities were reported to
be in the range of 5% relative. Detection limits for five of the elements are
also given in Table 7-5. These values were reported to be approximately
twice those found for the same elements by the HCP atomic emission tech-
nique applied to conductive alloys, the 10wer sputter rate of the compacted
geologicals accounting for the lower sensitivity.
PRE-
BURN
t
10 15 20
,
25
L
30
TIME (min)
Figure 7-9. Temporal behavior of AI(I) 309.2-nm emission from a 10% flint c1ay jeu matrix
sampie (AI concentration 2.05%): presputtered (prebumed) at 50 mA, final current of 150 mA;
2 Torr.(J9)
Table 7-5. Quantitative Analysis of NBS 98 Plastic Clay Using NBS 97a Flint Clay as an
External Standard(l9)
Plastic clay
Flint clay
Element Iplas/Iflint Calcd Assay (assay) LOD
Al 261516/296188 a 1.8% 1.74% 2.05% 5 ppm
Si 111966/87698 2.61% 2.28% 2.04% IOppm
Fe 13715/4823 903 ppm 945 ppm 317 ppm 2ppm
Na 21549/10793 54.6 ppm 60 ppm 27 ppm 0.1 ppm
Mg 86077/32678 241 ppm 256 ppm 92 ppm 0.2 ppm
Cr 3818/4267 18.7 ppm 21 ppm 21 ppm
a Average of three 3-s integrations.
7.2.6.3. Atomic Mass Spectrometry: Analysis of Nuclear Materials
Robinson and Hall have described the analysis ofuranium oxide (U3Üs)
reactor fuels by glow discharge mass spectrometry (GDMS).(43) They
employed the VG9000 glow discharge mass spectrometer (VG Isotopes,
Limited, Cheshire, UK) for these analyses. The VG9000 is a high-resolution
magnetic sector instrument, somewhat unusual for "atomic" mass spectro-
meters, which usually employ quadrupole mass filters. The instrument uses
pin cathodes, typically 15 mm long by 1.5 mm wide, mounted in a removable
sampie holder. The U3Üs sampies were mixed with graphite (1: 3 by weight)
and compacted into the shape of pin cathodes.
The graphite matrix was used as an internal reference element, and ion
beam ratios, analyte versus carbon, were utilized as the means of quanti-
tation. Thus, no external standards were employed, and "no standards"
analysis was achieved. The results of the analysis for 30 elements in four
standard reference U3Üs nuclear fuels, along with certified values, are
presented in Table 7-6. As is evident, the glow discharge analyses were quite
successful. Analytical precisions, as indicated in Table 7-6, were generally in
the range of 5 to 10% relative. Although no limits of detection were reported
in tbis study, they are normally in the sub ppm range for dc GDMS applied
to e1ectrically nonconductive sampies.
7.3. Application of rf-Powered Glow Discharges
7.3.1. Introduction
As discussed in the previous section, the mixing of nonconductors with

a conducting host matrix is a viable method for nonconductor analysis.
However, this method of nonconductor analysis may be undesirable due to
10ss of homogeneity (or inhomogeneity), entrained gases, sampie contami-
I\,)
Tab/. 7-6. Comparison of VG9000 Results with Certified Values for Analysis of NBL98 Series Standard U 3Os(43)G
NBL98/7 (ppm) NBL98/5 (ppm) NBL98/3 (ppm) NBL98/1 (ppm)
Element Cert VG 10- Cert VG 10- Cert VG 10- Cert VG 10-
Li 0 <0.04 0.01 1.2 1.6 0.22 5.0 8.4 1.0 26.2 30.3 1.0
Be 0 <0.04 0.01 0.8 1.0 0.16 5.3 6.1 0.8 25.7 23.8 4.0
B 0.1 0.81 0.51 0.4 0.3 0.08 1.2 1.8 0.3 5.5 6.8 0.3
Na 4.0 9.3 4.8 16.0 15.7 2.7 88.0 55.0 2.0 455.0 230.0 12.6
Mg 1.0 <0.4 4.0 2.5 0.45 17.0 13.5 2.2 91.0 42.0 3.9
AI 5.0 0.8 0.05 25.0 20.0 4.9 115.0 120.0 7.7 522.0 429.0 13.4
Si 2.0 10.0 10.2 1.9 65.0 410.0 5.3 315.0
P 3.7 0.7 0.04 23.0 7.2 0.82 99.0 39.5 2.1 505.0 175.0 9.8
K 2.3 4.8 2.1 28.0 22.2 2.6 138.0 102.0 7.3 725.0 582.0 40.0
Ca 4.5 9.6 5.0 19.0 40.0 5.1 100.0 167.0 19.0
Ti 0.3 0.44 2.1 3.6 0.6 11.0 14.0 0.5 50.0 78.0 2.3
V 0 <0.05 0.01 10.0 12.0 1.4 50.0 55.0 3.3 250.0 328.0 11.7
Cr 2.0 <1.4 9.0 6.7 1.3 22.0 16.1 3.4 101.0 43.0 4.2
Mn 0.8 2.9 1.3 0.11 10.6 7.5 0.9 49.0 28.0 0.5
Fe 13.0 9.3 0.9 32.0 26.6 4.0 110.0 87.3 6.4 515.0 394.0 17.0
CO)
~
'b
!il
::Q
.....
•
...~
i:i
Ci;
Table 7-6. Continued ()
NBL98/7 (ppm) NBL98/5 (ppm) NBL98/3 (ppm) NBL98/1 (ppm) ""Ci:

Element Cert VG 10" Cert VG 10" Cert VG 10" Cert VG 10" ~
()
Ni 2.0 1.0 0.2 5.6 3.5 0.66 22.0 12.8 1.7 103.0 61.0 2.1 ~
Co 0.06 <0.04 0.01 1.0 0.84 0.11 5.0 3.5 0.1 25.0 22.0 0.6 ~
Cu 0.4 <0.06 2.4 0.38 0.10 10.0 2.6 0.1 51.0 9.6 0.4 ;::!
:c:
C)
Zn 1.5 <0.2 19.0 3.1 0.77 96.0 24.8 3.5 480.0 96.0 2.70
Sr 0 2.6 1.9 0.21 10.0 9.0 0.9 55.0 52.0 2.0 t!
Mo <0.1 <0.28 0.07 2.0 2.3 0.62 10.0 9.5 2.7 51.0 45.0 0.8
Ag 0.1 <0.08 0.3 0.25 0.1 0.8 0.43 0.1 6.0 2.7 0.5
...~
Cd 0.3 <0.33 0.6 1.9 < 1.0 5.6 1.6 0.65 '"
In 0.2 <0.04 0.3 1.4 0.65 0.13 8.0 1.45 0.7 ~
Sn <1.0 <0.5 2.5 10.0 5.7 1.5 50 18.3 0.2
l2
Sb 0 0.14 1.0 1.0 5.0 1.8 0.3 25.0 6.4 0.5
Ba 0 5.2 1.6 2.0 1.6 0.91 10.0 8.2 0.5 50.0 28.8 4.2
W 0.1 <0.13 0.04 2.0 1.8 0.38 9.9 7.1 1.0 48.0 42.5 0.54
Pb 0.8 <0.08 2.5 0.53 0.16 9.0 3.2 0.4 46.0 18.0 1.0
Bi <0.2 <0.04 0.01 1.0 0.7 0.26 7.0 3.2 0.7 46.0 10.0 0.6
"NBL vaIues: provisionaI certificate concentrations by weight. VG9000 vaIues: ion beam ratio concentrations.
~
294 CHAPTER 7
nation, and dilution. As described briefly in Chapter 2, rf-powered glow

discharges allow one to avoid most ofthese sampie preparation-related prob-
lems because no mixing with a conducting host is required. The sampie need
only be shaped, or pressed in the case of a powder, into the appropriate form.
An attractive benefit of the rf-powered glow discharge is its applicability to
most sampie types. Because the fundamental dc bias may be sustained on
both conducting and nonconducting surfaces, both conductors and non-
conductors can be analyzed without altering the experimental arrangement.
Generally, any glow discharge source designed for rf operation is amenable
to dc operation, although the reverse is not usually true because of funda-
mental differences in the discharge operation. The potential of such a source
is clear, yet its analytical applications have been surprisingly limited. Before
looking at the analytical applications of the rf-powered glow discharge, a
brief review of its history, theory, and other fundamental considerations is
merited.
7.3.2. Historical
Jackson has speculated that rf-powered plasma sputtering may have

been observed as early as 1920 during investigations of rf plasmas.(44)
Robertson and Clapp(45) were perhaps the first to record (1933) the observa-
tion that material was removed from the walls of a glass tube when a high-
frequency discharge was initiated between two external electrodes. Hay(46)
observed that material removal, which he attributed to sputtering, occurred
only at high frequencies, although the reason for this was not understood. In
1948, Lodge and Stewart(47) compared the nature of dc and high-frequency
discharges along with the spatially dependent removal of material from the
walls of the discharge and attributed the rem oval to sputtering induced by
a high negative wall charge and a high potential field at the surface of the
insulator. Levitskii(48) investigated ion energies, performed probe measure-
ments, and sputter studies of high-frequency discharges maintained between
internal metal electrodes. In 1955, Wehner(49) proposed a rationale for
sputtering nonconducting materials, which Wehner and co-workers(50)
demonstrated in 1962 with the application of an external high-frequency
potential to a glass wall of a thermionically supported sputtering system.
The discharge served to clean the glass, and therefore, it was proposed that
because of the high sputtering rates observed, the method could be used for
relative yield measurements, etch studies, and sputtering of insulating layers.
Indeed, in subsequent studies by Davidse and Maissel,(5I,52) the technique of
depositing insulating thin films was developed into a practical method involv-
ing a relatively simple two-electrode geometry, the grounded diode
arrangement.
7.3.3. Theory 01 Operation
A thorough diseussion of the theory of rf glow diseharges has been

presented elsewhere, (53) and will therefore be presented here in limited detail.
The eonsideration of a de potential applied to a sampIe (eleetrode) made of
a noneondueting material provides a simplistie explanation of rf plasma
operation. Obviously, the eurrent neeessary to sustain a glow diseharge ean-
not flow through the target. The resulting response of the insulator surfaee
to the applied potential is analogous to the eharging of a eapaeitor as demon-
strated in Fig. 7-10. When a high negative voltage (- Vs) is applied to the
insulator, the surfaee potential drops initially to - Vs and deeays with time
to more positive potentials beeause of ion neutralization reaetions at the
eathode surfaee. (53) A short-lived diseharge will exist until a minimum thresh-
old voltage is reaehed and the plasma is extinguished. Thus, the solution to
the problem of sputtering insulators is to provide a souree of eleetrons for
ion neutralization while maintaining a eontrolled negative potential at the
eathode. As previously mentioned, Wehner(49) eorreetly proposed that the
application of a high-frequeney potential to a metal base at the baek of an
insulator would eause the positive eharges, which aeeumulate on the surfaee
of the insulator, to be neutralized with plasma eleetrons during the positive
portion of eaeh eyde, maintaining a negative potential on the surfaee of the
noneonduetor. Based on the rate ofpositive eharge accumulation in redueed
pressure environments, a high-voltage pulse of 1 MHz or greater will sustain
a pseudoeontinuous diseharge. (50)
The key fundamental phenomenon occurring from the applieation of a
high-frequeney ae potential is the "self-biasing" of the eleetrodes. It is this
self-biasing that supports eathodic sputtering in the plasma. In order to
demonstrate this phenomenon, eonsider the 2-kV peak-to-peak square-wave
potential (Va) illustrated in Fig. 7-11 and the resultant potential on the
eathode surfaee (Vb). As the potential is applied during the first half-eyde,
the surfaee eharges to -I kV followed by a deeay to approximately -0.7 kV
as the surfaee is bombarded by positive ions. As the seeond half-eyde begins,
VOLTAGE 0 I.,.----====""""""----;~
-Vs
Figure 7-10. Temporal response of an insulating material to an applied negative voltage

pulse. (114)
296 CHAPTER 7
--~~-+--~~-+--~.t
Va
(kV)
-1
o 't 2't 3't
Vb t
(kV)
dc
offs et
Figure 7-11. Electrode response to an applied square-wave potentiaL(1l4)
the applied + 1 kV (+2 kV, relative) potential results in a + 1.3 kV potential

on the surface. During this half-cycle, electrons are accelerated to the
insulator surface and charging results analogous to that of the prior half-
cycle. However, because of the greater mobility of the electrons relative to
the heavier positive ions, the surface potential decays toward zero at a faster
rate than during the previous half-cYcle, thus reaching a value of +0.5 kV.
As the second full cYcle is initiated and the polarity of the electrode is
switched, the resulting potential is -1.5 kV ( + 0.5-2 kV). After several such
cYcles, the waveform of Vb will reach a constant dc offset. This dc offset is
the self-bias potential that maintains a time-averaged cathode and sustains
the sputtering ion current. The dc offset potential is approximate1y one-half
the applied peak-to-peak voltage for analytical applications, depending on
discharge parameters and source geometry. The insulating target material is
alternately bombarded by high-energy ions and low-energy electrons; how-
ever, the discharge is, for all intents and purposes, continuous, and can be
considered a dc discharge with a superimposed ac potential.
7.3.4. Fundamental Considerations

Anyone wishing to use an rf-powered glow discharge as an analytical
source will soon come to appreciate the wealth of prior investigations that
arose from the industrial application of these sources in deposition and
etching processes. The electronics industry was the first to apply rf-powered
glow discharge sputtering sources to the development of thin films. Since

then, the rf-powered glow discharge has proven itself useful in the
preparation of various thin films inc1uding semiconductors, (54-57)
dielectrics, (58-61) and superconducting films. (62,63) Radio-frequency-powered
discharges have also been used for generating polymer films by plasma-
induced polymerization. (64-68)
With the development of deposition and etching processes arose the
need for a better understanding of these discharges both in terms of moni-
toring the fabrication processes as weIl as obtaining a more fundamental
understanding of the collisional processes occurring in the negative glow.
Successful methods of monitoring these deposition and etching systems
inc1ude optical emission,(69-72) fluorescence,(73.74) and luminescence from the
growing film. (75) Various researchers have also successfully demonstrated
mass spectrometric monitoring of the deposition process by sampling per-
tinent discharge ions.(76,77)
Various diagnostic techniques have been utilized in studying the pro-
cesses occurring in the plasma. Gottscho and Miller(78) have reviewed the
optical diagnostic techniques that have been applied to glow discharge
plasmas. (Techniques for plasma diagnostics are described in Chapter 11.)
Diagnostic techniques applied to rf-powered glow discharges inc1ude
emission, (79-81) atomic absorption, (82) laser-induced fluorescence,<83,84) and
actinometry. (81) Electrostatic probes(85-87) and mass spectrometers(82,88-93)
have been used to study both the energy of charged species in the plasma
and factors affecting their formation. The various techniques utilized in
applications of monitoring and diagnostics suggest a range of useful ana-
lytical applications of the rf-powered glow discharge.
Such fundamental studies of rf deposition processes have generated a
wealth of information and have suggested typical operating considerations
for analytical devices. While a detailed discussion of such findings is beyond
the scope of this work, a presentation of the most fundamental and relevant
considerations necessary to the successful application of rf glow discharges
to analytical techniques is merited. The critical parameters in rf glow dis-
charges--discharge geometry, frequency, and electrical coupling-have been
the focus of much of the diagnostic investigations to date.(79,89-91,94-96) The
discussion that follows merely highlights these critical parameters. For a
more detailed consideration, the reader is referred to Chapman(53) and other
references inc1uded therein.
7.3.4.1. Source Geometry
The geometry of the rf glow discharge source is critical in terms of

isolating the bombarding species to the sampIe and in determining the energy
of the bombarding ions as related to the dc bias. In highly asymmetric rf
298 CHAPTER 7
discharges (those with large difference in electrode surface areas), the ion
bombardment is generally isolated to the smaller electrode and the negative
bias approaches half of the rf peak-to-peak voltage. As presented by Kohler
et al.(89) in a thorough investigation ofrfplasma potentials, Fig. 7.12 demon-
strates the effect of geometry and mode of rf coupling, either directly or
capacitively coupled, on the dc bias voltage (Vdc ) the excitation electrode
voltage (VI> dashed curve), and the plasma potential [Vp(t), solid curve], for
de Coupled Capaeitively Coupled
~V(t) F~V(t)
V(t)
~
~ V~f" ~ o 1---_,--+,---',c-
\\
- - -t--- -,---..,-
," \
",
,,
\ I \
\ \ I \
Asymmetrie \ \ I
.,
\ I
(small electrode '. " V(t) \ I
powered) ',,'
V(t)
+
Symmetrie
,,
,,
,
,
'.
,
\ I
I
I
I
.'
\ , I
'-'
r!l
V(t)
~ \ , I
Asymmetrie ,,
\
I 0
(large electrode
powered) , I
I
"
Figure 7-12. Plasma potentials [Vp(t), solid curves], excitation electrode voltages [V(t),
dashed curves], and dc bias voltages (Vdc ) as a function of discharge coupling and geometry. (89)
a sinusoidal waveform (assuming a purely capacitive sheath). As illustrated
for the capacitively coupled excitation electrode, decreasing the cathode-to-
anode ration causes the dc bias voltage potential on the target electrode
to become more negative with the plasma potential following suite. This
distribution of voltage between the electrodes is caused by differences in the
capacitances of the dark space sheaths of the two electrodes. At sufficiently
high frequencies, a few megahertz, the sheaths can be considered to behave
capacitively(97) because the rf frequency is below the electron resonant
frequency and above the resonant frequency for most ions. Kohler et al.
conduded that the capacitive sheath model was in good agreement with
experimental results in most applications. (89) The relative capacitances of
the dark spaces are proportional to the relative electrode surface area and
determine the distribution of the dc bias developed across the electrodes.
The larger electrode has a larger capacitance and thus a smaller capacitive
reactance. Therefore, the impedance of its sheath should be smaller than the
impedance of the sheath of the smaller electrode, and its bias potential is
doser to zero than that of the smaller electrode. Koenig and Maissel(94)
developed (to a first approximation) the relationship VJ/V2 = (A 2 /A 1)4,
where VJ/ V2 and Az/AI are the ratios of the dc potential difference between
the glow space and the electrodes and the electrode surface areas, respec-
tively. More recent studies indicate that this relationship may be doser to
the first power than the fourth in cases of higher power densities and for
cathode-to-anode area ratios less than 0.6.(91,98) In analytical applications,
sputtering of the chamber or resputtering of the deposited films is obviously
undesirable and thus the cathode-to-anode surface area ratio should be as
small as practical.
It is not the absolute relative sizes of the electrodes that are critical, but
rather it is the area of the electrode surfaces in contact with the discharge
species that determines the relative electrode areas. In some instances the
nature of the discharge gas can affect the symmetry of the discharge. For
example, Roth(79) observed spatially resolved emission profiles between the
electrodes of a dosely spaced parallel plate rf discharge positioned in a larger
vacuum chamber. Emission profiles of nonreactive gases indicate that the
symmetry of the discharge can be dramatically affected by the nature of
the gas without physical changes in the areas of the parallel plate electrodes.
This is explained in terms of the degree to which the plasma diffuses into
the chamber. As the plasma diffuses from the grounded electrode (anode)
into the chamber, which is also at ground, the discharge becomes asymmetric
with respect to the powered electrode because of the differences in surface
areas. The diffusion is dependent on both the pressure and the electron
temperature of the gas. In molecular gases the electron temperature is lower
than that of atomic gases because of vibrational and rotational energy losses
not available in the latter. Thus, diffusion in the atomic discharges results
300 CHAPTER 7
in increased asymmetry of the discharge as demonstrated by a change in the

spatially resolved emission profiles of the discharge gases at each electrode.
This study demonstrates the need for asymmetry between the electrodes.
Generally, analytieal applieations of rf glow discharges use argon or other
atomic gases and are asymmetrie in design.
7.3.4.2. rf Power Coup/ing
The electrical coupling of the rf power to the target electrode is also

critical. Without capacitive coupling in the circuitry, any dc potential differ-
ence that would otherwise exist between the electrodes cannot be maintained
and the dc bias potential would be divided equally between the electrodes
via the external circuitry. Note that in Fig. 7-12 the bias potential is zero for
every geometry where the rf power is directly coupled. With direct coupling,
little sputtering of the target electrode is likely because of the absence of a
steady potential bias between the electrodes. Koenig and Maissel(94)
described an equivalent electrical circuit of an rf glow discharge for sputter
applications where the effects of geometry and other parameters such as
pressure and magnetic fields can be evaluated. As adapted from their circuit,
an equivalent circuit for analytical applications is shown in Fig. 7-13. The
blocking capacitor serves to maintain the dc bias voltage (capacitive
Blocking Capacitor
RF Electrode
Target Capacitance
Cathode
} Sheath
} Bulk Plasma
} Anode
Sheath
Figure 7-73. Equivalent circuit ror an rf glow discharge plasma.

coupling). When analyzing insulating materials, an external capacitor is not

necessary because of the capacitance of the nonconducting surface. In this
case, the sampIe provides the required target capacitance. For conductors,
the matching network provides the capacitance required to sustaln the bias
potential. The magnitude ofthe dc bias is largely responsible for the energies
of incident ions on the sampIe surface and subsequent sputter rates. The
division of the bias voltage between electrodes determines the degree of
sputtering at each electrode. Various methods, such as substrate tuning, (99)
have been developed to control the division of the dc bias between the
electrodes. Division of the bias potential between electrodes promotes higher
film quality in deposition systems as a result of resputtering of impurities
from the surface of the growing films by energetic ion bombardment.(IOO.IOI)
However, for analytical purposes, sputtering from surfaces other than the
target is undesirable because of resulting spectral interferences. Resputtering
would also degrade any depth profiling analyses. Therefore, discharge sys-
tems that maximize the dc bias at the target surface are desirable for most
analytical purposes. (Jackson has reviewed rf sputtering with respect to
the equipment and process mechanisms related to deposition and etching
systems.(44» Also illustrated in Fig. 7-13 are the electrical properties ofthe
bulk plasma and the anode and cathode sheaths. The glow region of the
plasma is resistive in nature while the sheaths have both resistive and capaci-
tive properties. Resistances in the sheath are typically very large at low
pressures and therefore the rf discharge is primarily capacitive in nature.
Only at higher pressures, and at lower operating frequencies do the sheaths
exhibit resistive behavior. The diodes in the circuit represent the high electron
mobility.
7.3.4.3. Operating Frequency
The operating frequency of the discharge generator is another param-

eter that should be considered. Typically, a 13.56-MHz potential is chosen
for rf glow discharge applications simply because this frequency has been
designated by the Federal Communications Commission as a "free fre-
quency." This frequency is generally suitable for all analytical applications,
with detrimental effects occurring only at lower frequencies where neutraliza-
tion of ions on the sampIe surface is less efficient, resulting in a lower dc
bias. For optimum results, low megahertz frequencies are required. The
effect of frequency has been investigated primarily in terms of the energy of
ions bombarding the substrates in deposition systems. Such studies have
provided important information for the mass spectrometrist while little
emphasis has been placed on optical phenomena occurring as a function of
frequency.
302 CHAPTER 7
Analytically, phenomena occurring at lower frequencies have implica-

tions for mass spectrometric analyses in terms of ion energy distributions
observed in the discharge. If the frequency chosen for the discharge is too
low, rfmodulation may occur. Radio frequency modulation is an ion transit
time-related broadening ofthe ion energy distribution. Kohler et al.(9O) have
investigated the frequency dependence of ions bombarding grounded sur-
faces in a low-pressure planar rf argon discharge. In order to minimize the
spread of the ion energy distribution, the frequency chosen should be high
enough that ion transit across the plasma sheath between the bulk plasma
and the grounded surface occurs over several cyc1es of the applied potential.
The energy of the ions arriving at a grounded surface is determined primarily
by the plasma potential, Vp(t), and collisional processes occurring in the
sheath. If the frequency is greater than the ion resonant frequency the energy
of the ions will be determined by the time-averaged plasma potential, Vp •
This is demonstrated in Fig. 7-14 where the ion energy distributions are
illustrated for both lOO-kHz and l3.56-MHz plasmas. The ion energy distri-
bution at l3.56 MHz is relatively monoenergetic as determined by Vp
because the ion transit time across the sheath occurs over several rf periods.
In comparison, the lOO-kHz ion energies are a function of the temporal
100 kHz 13.56 MHz

ii
g
..
>
::
:s.
....
~
.c
:J
.
U
+N
c(
-50
J.
0 50
::1. 100 0
~
20 40
Ion Energy (with respect to ground) in eV
Figure 7-14. Ion energy distribution of discharge species sampled through the grounded
anode of a planar diode rf g10w discharge with l00-kHz and 13.56-MHz excitation potentials. (90)
v
n Plasma
Potential
Vph)
n----------- 50 Ion
Energy
ot-+-4\--,-'\--4-foo.-... --=====::::~O (eV)
/Voltage on
, Powered
Electrode
V(t)
v
Figure 7-15. Illustrative figure ofthe relationship between the lOO-kHz energy distribution
in Fig. 7-14 and the plasma potential, Vp (t).(90)
plasma potential Vp(t) because ion transit times are less than one rf period.
Figure 7-15 illustrates the ion energy distribution dependence on Vp(t) at
the IOO-kHz excitation frequency. At low frequencies, the resistive nature of
the plasma sheath is demonstrated as ion energies are dictated by the positive
excursions of the sinusoidal excitation electrode potential V(t).
Even at higher excitation frequencies transit effects may still be
observed. Figure 7-16 shows the ion energy distributions of ions traversing
the ground sheath at the substrate plane in a 13.56-MHz discharge.(91) The
discharge was confined by a cylinder which increased the plasma density.
Increasing the plasma density decreased the sheath thickness. Subsequently,
the high mass Eu+ required several rf periods to traverse the sheath while
the much lighter Ht ion traversed the sheath within one rf period or less as
evidenced by the relatively large spread in the energy distributions.
7.3.5. Analytical Applications

Sputter deposition and etching of nonconducting materials, as weIl as
the fundamental studies mentioned, suggest the potential of the rf-powered
glow discharge for the analysis of nonconducting materials. Radio-
frequency-powered glow discharges are useful as both primary and second-
304 CHAPTER 7
H 0+
2
-----'
o 20 40 180
Figure 7-16. Energy distribution of H;, HzO+, and Eu+ at the substrate plane in a confined
discharge (13.56 MHz, 100 W, 75 mTorr).(91)
ary sources of excitation and ionization. This section first describes the
analytical applications of rf-powered sources as secondary excitationj
ionization sources and concludes in describing their use as primary atomiza-
tionjexcitationjionization sources.
7.3.5.1. rf Discharges as Secondary Excitationjlonization Sources
A variety of applications have utilized rf glow discharges as secondary

excitation or ionization sources. As mentioned in Chapter 4, secondary
discharges within hollow cathode lamps can boost the emission intensity
by enhancing electron impact in the gas phase while minimizing resonance
broadening by controlling the sputtering current. Unfortunately, any
attempts to increase excitation intensity by increasing the current or decreas-
ing the pressure of the discharge gas in the hollow cathode discharge are
accompanied by an increase in the vapor pressure of the analyte, and sub-
sequently resonance broadening. Secondary sources of excitation allow
increased excitation without increasing the vapor pressure of the analyte,
minimizing self-absorption. Direct current,(I02-I06) and microwave(107-109)_
boosted discharges have demonstrated increased emission intensity of
spectral lines and lower background levels relative to nonboosted lamps.
Likewise, Walters and Human demonstrated that an inductively coupled, rf-
boosted glow discharge lamp resulted in higher emission intensities and more
linear calibration curves over a wider dynamic range as a result of decreased
self-absorption. (110)
The technique of sputtered neutral mass spectrometry (SNMS) devel-
oped by Oeschner(lll) incorporates an inductively coupled high-frequency
(27 MHz) discharge as either the primary sputterjionization or secondary
ionization source. In the direct bombardment mode a small dc potential (a
few hundred electron volts), which is applied to the cathode surface, extracts
ions formed in the plasma to the surface. Neutrals sputtered from the surface
are then ionized in the rf plasma, and sampled mass spectrometrically. As
in the glow discharge, separation of the atomization step from the ionization
process results in minimized matrix effects and improved relative elemental
sensitivities.
7.3.5.2. The rf G/ow Discharge as a Primary Spectrochemica/ Source
The rf discharge has served as the primary atomizationjexcitationj

ionization source for various analytical applications. The sputtering process
that occurs in the rf discharge is analogous to that of the dc discharge. High-
energy ions, under the infiuence of the bias potential, bombard the sampIe
(cathode) surface. Sputtered neutral atoms diffuse into the negative glow
where subsequent excitation and ionization occur. Analytical applications
of the rf-powered glow discharge source have been limited primarily to mass
spectrometry with a few applications in atomic absorption, emission, and
laser-induced fiuorescence spectroscopies.
SampIe preparation is typically similar for any of the rf glow discharge
techniques. Because of the nature of the rf discharge (i.e., its ability to sputter
nonconductors), sampIe preparation is quite simple. For solid sampIes, no
grinding or mixing of the sampIes is necessary, and sampIes need only con-
form to the appropriate shape and dimensions of the holder. Powdered
sampIes can often be pressed in an appropriate die assembly without the
need for binders which would dilute and possibly contaminate the sampIe.
In the case of solids, no additional gases are entrained in the preparation
step; thus, gas analysis of nonconducting materials may be possible. As will
be presented, depth profiling of elemental compositions in nonconducting
solids is possible because sampIe homogeneity is preserved during sampIe
preparation.
Source designs for analytical applications vary but obviously share com-
mon features. The discharge is maintained between two electrodes, either
external or internal with respect to the vacuum system employed. Donohue
and Harrison(112) employed an rf-powered hollow cathode for the analysis
of conductors, nonconductors, and solution residues. Their rf-powered glow
discharge cavity ion source is shown in Fig. 7-17. The hollow cathode, B,
served as the sampIe electrode and was also used for analyzing deposited
solutions. The excitation potential (30 kV, 500 kHz) was supplied from a
spark source generator to electrode C, externally positioned around the
discharge tube as shown. The resulting plasma could be sampled optically
through the quartz window (D) or mass spectrometrically through a small
hole in the base of the hollow electrode. Typical operating pressure for this
306 CHAPTER 7
Figure 7-17. Radio-frequency-powered bolJow catbode source. A, gas inlet; B, holJow elec-
trode; C, rf counter electrode; D, quartz optical window; E, beat sink; F, liquid N 2 cold finger;
G, electrode block; H, insulators.(112)
source was approximately 300 mTorr. Acquired mass spectra of Zn, Sn, and
glass (Pyrex) electrodes, as weIl as an analysis of elements deposited on glass
from solution, demonstrate the analytical capabilities of this source.
The grounded, diode rf discharge source design is most common. In
this configuration the sampie is mounted on the powered electrode while the
vacuum chamber typically serves as the opposite (grounded) electrode for
experimental simplicity. This geometry also serves to minimize the cathode-
to-anode ratio, which, as previously discussed, distributes the dc bias on the
smaller electrode and thus isolates sputtering to the sampIe surface. A practi-
cal concern in such a design is proper shielding of the sampie holder in order
to restrict sputtering to the sampIe surface. A typical rf source design is
illustrated in Fig. 7-18. The sampie is mounted on the rf excitation electrode,
,....----- -....--Grounded Electrode
Sampie
'iijiijiii~-- Electrode
.~
RF Powered
-4-- - Grounded Shield
Figure 7-18. Typical rf glow discbarge source design.

which may take many forms. Typically, the sampie is deposited on, press-
fit into, or otherwise mounted onto the rf electrode. The grounded shield is
positioned within the cathode dark space (typically about 0.5-1.0 mm spac-
ing) and prevents a discharge from forming on the rf backing electrode,
thereby limiting sputtering to the sampie surface.
Coburn et aIY13) utilized a six-position, rotary, water-cooled sampie
holder for depth profile analyses. In this configuration, shown in Fig. 7-19,
up to six sampIes could be mounted and analyzed separately. Only one
sampIe is exposed to the discharge at a time while the others are protected
within a grounded shield. Such a design maintains vacuum integrity between
analyses and increases sampIe throughput. The exposed sampIe holder is
1"--20cm~1
rf
IlrrL
Grounded
+ Six Position
Rotary
Water-Cooled
To Gas
Admission
System and - .
Shielding Sampie
Sublimation
Holder
~ Pump
To
+- Diffusion
Pump
Figure 7-19. Experimental design incorporating a six-position, rotary, water-cooled rf

discharge source. (113)
308 CHAPTER 7
approximately 5 cm in diameter, though sampie sizes were somewhat

reduced in this application.
Duckworth and Marcus developed an rf glow discharge source (Fig.
7-20) that is more amenable to typical sampie sizes found in bulk analytical
applications. (114) In this design sampie sizes are reduced to approximately
0.5 inch in diameter. Again the grounded steel sleeve is positioned to restrict
sputtering to the sampie surface and the source is water-cooled in order to
minimize volatilization of atoms and/or changes in relative sputter yields.
Sampies are machined to the appropriate height (-0.5 inch) for analysis.
Conducting sampie mounts of various heights are also used to position
nonconducting or small sampIes relative to the grounded steel sleeve. A
Macor insulating sleeve is used to isolate the rf-driven sampIe holder from
the conflat flange (2.75 inch)-mounted source housing. This source is ana-
lytically sound, but sampie throughput is low because of loss of vacuum
integrity upon changing sampIes and the time required to remove and
reassemble the conflat flange source mounting hardware.
A direct insertion probe (Fig. 7-21) has been designed that increases
sampIe throughput, maintains vacuum integrity, and is more amenable to
small sampie sizes.(1I5) The probe accommodates both pins and disk sampIes
of up to 3/16 inch in diameter. In this design the rf is supplied through a
high-voltage feedthrough that is vacuum-welded to the end of the probe.
The sampIes are press-fit or silver-painted onto a copper sampIe support.
CATHODE
STEELBODY
SLEEVE
/
/FLANGE
RF
POWER
Figure 7-20. Flange-mounted (2~ inch) rf-powered g10w discharge source for analytical
applications. (114)
1/2" Tubing Eleclrlcal Mocor Steel

Feedlhrough Sieeve Sieeve
Glass Tublng Sampie Sampie

Holder
Figure 7-27. Radio-frequency-powered glow discharge direct insertion probe.(IIS)
This support is further insulated from ground by a Macor cap and a final
stainless steel cap that restricts sputtering to the sampie surface. Shielding
on the coaxial cable is pulled back to allow the insulated (Teflon) rflead to fit
through the half-inch stainless-steel probe body. Coaxial shielding integrity is
maintained by the probe body over the length ofthe probe to reduce radiated
power losses. The probe is inserted through a ball valve assembly, thus
maintaining vacuum integrity and allowing rapid sampie interchange. The
addition of an expandable bafHe to the ball valve assembly allows the axial
positioning of the discharge relative to the sampling orifice and subsequent
optimization of the analyte signal. As will be discussed in Section 7.3.5.3,
Duckworth and Marcus(1l4) have observed a strong spatial dependence of
ion densities in rf GDMS. Movable source assemblies, which allow the
sampling position to be optimized, were critical for optimum analytical
performance.
The limiting feature of the direct insertion probe is that solids must be
machined into a compatible size (e.g., <3/16 inch in diameter) and shape.
This is often difficult and time-consuming. Nonconducting sampies are often
nonmachinable and must be ground and pressed into asolid of appropriate
dimensions. Such sampie preparation is restrictive and in the latter case will
result in a loss of depth profiling capabilities. The latest design in rf discharge
sources incorporates an external sampie mount that circumvents such
310 CHAPTER 7
Fused Silica
Windows
Stainless ....~r===j~=::~~:j:===ll
Steel Body
Vacuum Port '--r===dF-f-- Negative Glow

~~~~----J
Disk
Female Coax
Connector
Glass Insulator
Copper Conductor
Male Coax
Connector
,1 inCh.. ,
RG-213/U Coax Cable

(to matching network)
Figure 7-22. Diagrammatic representation of the rf glow discharge atomization/ excitation

source.(116)
problems.(116.1I7) The source, shown in Fig. 7-22, is amenable to various

sampie sizes-the only requirement is that the sampie surface be flat enough
to form a vacuum seal on an O-ring (typical for many nonconductors such
as glasses and ceramics). Powder sampies mayaiso be analyzed with this
source by compacting them into a brass disk that is pressed against the
O-ring (Fig. 7-3). The sampie is pressed against a lO-mm-diameter Teflon
O-ring on an orilke disk by a brass torque bolt, forming a vacuum seal. A
Macor spacer electrically isolates the sampie from the torque bolt. A copper
conductor, insulated in glass tubing, couples the rf power from the RN
coaxial connector to the sampie. In atomic emission applications, the source
is equipped with two optieal ports, one on and one perpendieular to the
sampie axis. With these optieal eonfigurations, the authors were able to
eompare analyte emission intensities, signal-to-noise ratios, and emission
speetra from both optieal geometries. (116) These eomparisons will be summa-
rized in Seetion 7.3.5.4. A similar souree, shown in Fig. 7-23, has been
designed for mass speetrometry. (117) This souree is eomparable in design in
all respeets exeept for sampIe size and an ability to position the sampie
relative to the mass speetrometer sampling orifke. A torque serew, aeting in
eoneert with the rf feedthrough bolt, serves to move the sampIe (diseharge)
with respeet to the sampling orifiee in order to optimize the mass speetro-
metrie sampling ofthe plasma ions. Beeause ofthe sampIe positioner, sampIe
sizes were restrieted to dimensions varying from 1.3 to 2.2 em, though larger
versions of the same design are feasible.
Initial mass speetrometric studies using the external sampIe mount
design indieate that the total preparation and plasma stabilization times for
analysis are on the order of a few minutes, eonsisting of 1 min of pumpdown
time and 1-2 min presputtering. Figure 7-24 indicates the temporal stability
of the 63CU + ion signal. In less than lOs the ion eurrent stabilizes and remains
stable throughout 18 min of monitoring. In order to minimize pumping
VACUUM
SOURCE LlNE
HOUS ING /
P05' \ " ,. - -. :. - ,
SAMPLE \
Figure 7-23. Diagrammatic representation of the extemal sampIe mount geometry for rf
glow discharge mass spectrometry. (117)
312 CHAPTER 7
10
Time (mln.)
Figure 7-24. Temporal stability of 63CU+ intensity utilizing the external sampie mount,
rf-powered glow discharge (13.56 MHz, 15 W, 0.11 Torr).(I17)
times, the discharge volume is reduced as much as possible while maintaining

asymmetry. Large vacuum lines were utilized to increase the conductance
and thus minimize initial rough pumping time. Short pumpdown and pre-
sputter times, as weIl as source amenability to sampIe shape and size, should
allow for convenient routine analyses.
It must be stressed that in any rf-powered glow discharge source design,
radiated noise must be minimized. Complete coaxial shielding of all rf
circuitry and proper grounding are critical requirements. Obviously, any
break in coaxial shielding will result in rf power losses. Such noise is known
to interfere with associated electronic instrumentation. For further consid-
eration of proper grounding and shielding, the reader is referred to
Morrison. (118)
7.3.5.3. rf G/ow Discharge Mass Spectrometry
Mass spectrometric sampling of ions formed within the discharge com-

prises most of the analytical applications of rf glow discharge to date. Most
of the rf GDMS has been limited to quadrupole instruments although some
attempts to interface rf glow discharges to high-resolution mass spec-
trometers have been made with limited success. Ion energies in the rf glow
discharge are typically 20-30 eV with energy spreads of a few electron volts.
These energies, as weIl as the lower operating pressures relative to dc sources,
are quite amenable to quadrupole mass spectrometry. Coburn(93) has
reviewed quadrupole mass spectrometric sampling of ions formed in rf glow

discharges and conc1uded that the most promising application of the tech-
nique appears to be elemental analysis of solids.
Ionization in the rf-powered glow discharges is known to be quite
efficient, attributed in part to the oscillating nature of the rf discharge.
Electron impact is thought to be more efficient at high frequencies and other
mechanisms such as "multipacting," "electric field amplification," and "surf
riding" have been proposed as means of increased ionization in high-
frequency discharges. (53) As in the case of dc glow discharges, Penning ioniza-
tion is thought to be prominent in rf discharges. Eckstein et al. (82) have
demonstrated quite c1early that Penning ionization was the predominant
mechanism for ions sampled in their discharge. As shown in Fig. 7-25, they
found a linear correlation between the observed ion signal and the product
of Cu neutral and Ne metastable atom densities as determined by atomic
absorption spectrophotometry. The dependence on metastable concen-
trations indicates that inelastic collisions with the neon metastables is the
primary source of analyte ionization.
Some studies suggest that Penning ionization may not be the dominant
ionization mechanism in other rf discharges. In a similar set of experiments
TO· 3 L..----L.--'-...&....I..........L.LL..._..L..-L-L................._ ................"-'-.........u.---L--L......................
10.4 10-3 10-2 10- 1

Cu· Nem (arbitrary units)
Figure 7-25. First-order relationship between Cu + intensity and Cu· Nem products
(20 mTorr Ne).(82)
314 CHAPTER 7
by Duckworth and Marcus, no correlation was observed between the analyte

ion signal and discharge gas metastable atom densities.(119) The lack of
correlation has been attributed to differences in the experimental parameters
including source design, discharge parameters, and, perhaps most impor-
tantly, the sampling position in the discharge. Demonstrated in Fig. 7-26 is
a diagrammatic representation of fundamental sampling considerations
within an rf discharge as a function of pressure. (114) The authors observed the
greatest analyte signals when sampling from the cathode dark space-negative
glow interface, shown in the middle panel at a typical operating power and
pressure (25 Wand 0.2 Torr). As pressure is decreased (0.1 Torr) and mean
electron free paths are increased, the interface expands and only discharge
gas ions, predominantly formed by electron impact and charge exchange,
are detected. At higher pressure (0.35 Torr), analyte intensities are increased
though not as high as when sampling from the interface. Such spatial phe-
nomena have not been mentioned by other authors and it is believed that
ions sampled by Eckstein et al. (82) were extracted from the negative glow
region where electron energies are lowered because of multiple collisions.
An increased contribution by electron impact ionization at the negative
glow-cathode dark space interface may explain the lack of correlation
between the mass spectrometric and atomic absorption measurements. It
should also be no ted that average electron energies in the rfplasma (4-6 eV)
are much higher than those in dc plasmas ('" 1 eV) because of the rf oscillat-
ing fields superimposed on the dc bias potential. (120) Increased contributions
to ionization from high-energy electrons would not be surprising, particu-
lady at the interface of the cathode fall region.
Mass spectral characteristics of rf-powered glow discharges are very
similar to those of dc glow discharges, with some exceptions. In Fig. 7-27 a
mass spectrum of a stock copper sampie is shown. Typical matrix and dis-
charge gas species are observed. The most notable difference between the dc
and rf spectra is the increased intensity of hydrides and molecular species in
the rf mass spectrum. Water and contaminant species in the discharge are
efficiently ionized by the enhanced electron impact in the rf plasma and
promote the formation of hydride species. Most notable are the argon
hydride species, which are typically at a factor of ten less for dc-powered
25 watts, 0.35 torr 25 watts, 0.2 torr 25 watts, 0.1 torr

Figure 7-26. Illustration of the influence of discharge pressure and power on the cathode
dark space/negative glow interface location. (114)
x10
x10
A~ ArW
/
Zn+
e'o 70 a'o 90 ,
100
,
120
,
110 130
Mass Number
Figure 7-27. Radio frequency glow discharge mass spectrum of a stock copper alloy
(0.20 Torr, 10 W).(114)
discharges. Copper dimers as well as argides, thought to be formed through

associative collisions, (53) are quite prominent in the rf discharge. This is likely
the result of the lower operating pressures in the rf glow discharge, which
should reduce the efficiency of collisional dissociation within the plasma.
King et al. have demonstrated that such polyatomic species are reduced at
higher discharge pressures because of the increased collision frequency within
the plasma,021)
Isobaric interferences in the rf glow discharge are similar to those of its
dc counterpart; however, the intensities of polyatomic species may be
greater. The preferential removal of polyatomic isobaric interferences,
generated in the glow discharge, by collision-induced dissociation (CID) in
the collision cell of a multiquadrupole mass spectrometer is discussed in
Chapter 5. Duckworth and Marcus have demonstrated the utility of
performing CID on polyatomic species in the collision cell of a double-
quadrupole mass spectrometer.(122) Polyatomic species are preferentially
lost relative to the analyte as are atomic species which are lost through
symmetrie and asymmetrie charge exchange. Though effective in the prefer-
ential removal of interfering species, analyte intensities are reduced by
approximately 80% because of scattering within the collision cello
The interfacing of an rf glow discharge with a high-resolution mass
spectrometer is another alternative to the quadrupole mass filters typically
employed. However, because of the required high accelerating potentials of
sector mass spectrometry and the required electronic configurations of the
rf discharges, attempts have met with only limited successY23) Efforts typi-
316 CHAPTER 7
cally have employed an isolation transformer to float the rf components to

the mass spectrometer's accelerating potential (-8 kV). Since probe inlets
and vacuum chambers are at ground potential, this high reference voltage
makes complete shielding of the rf cables and source particularly difficult.
Consequently, appropriate shielding requirements are not met and rf noise
adversely affects detectors, Hall probes, and other electronic systems. Recent
modifications in electrical design have allowed complete shielding of the rf
componentsY24) Radio frequency power is efficiently coupled to the sampIe
through a direct insertion probe similar to that of Fig. 7-21 and is isolated
from the high voltage power supply. Likewise, the high voltage is coupled
to the sampIe and source housing while isolated from the rf-generating elec-
tronics. A typical mass spectrum of Sn in NIST SRM 1103 brass is shown
FuUscala deIIacIlon = 5.DOE-13 Amps
I
112
I
114
I
116
LLl I
118
m/z
I
120
A
I
122
1I
124
~ 115Sn
(31.7 ppm)
11
I Il
\
I I
I I
I
I I
II U' Lj I
11 11
1 I
I
I
: tJ U lJ I
11 11
I I
B
L----.JU
I I I
.11
v
I I I I
112 114 116 11. 120 122 124
m/z
Figure 7-28_ Radio frequency glow discharge mass spectrum ofNIST SRM 1103 brass. (A)
xl, (B) xIOOY24)
in Fig. 7-28a. The Sn isotopes (0.88% by weight) are c1early detected. This
spectrum was acquired utilizing an rf-powered source coupled to the VG9000
double-focusing glow discharge mass spectrometer (VG Isotopes, Ltd.,
Cheshire, UK).(124) The rf glow discharge provides a very stable ion signal
in thisapplication. The electrical design limits the noise level to only a few
counts per second. As noted in Fig. 7-28b, 115Sn, present at 3l.7 ppm, is
detected quite easily under moderate mass resolution conditions
(M / ~M - 1200). Limits of detection under these conditions are calculated
to be -lI ppb (20').
Comparisons of the analytical characteristics of rf and dc glow
discharges as ion sources are difficult because of the large parametric
differences-pressure and power regimes in particular. However, the
optimum performances of each are similar. Ion currents for the rf sources
are comparable to dc discharge ion currents. Ion currents are typically about
10- 11 for both conducting and nonconducting matrices. (11 4) Conducting
sampies which tend to form oxides, such as aluminum, can be quite difficult
to analyze with a dc source because of nonconductive coatings that develop
on the surface of the sampie. Such dielectric layers are efficiently sputtered
and dissociated in an rf glow discharge. Duckworth and Marcus have
reported increases in ion currents of a factor of 50 over dc glow discharges
for Al matrix samplesY25) In the same study, relative sensitivity factors for
rf and dc glow discharges were investigated for a variety of aluminum
sampies, and the relative ion yields were found to be similar. It has also been
suggested that matrix effects may be small even between conducting and
nonconducting sampies. (114)
In the course of the development of rf GDMS, much of the analytical
studies and development have been performed on conducting materials such
as copper, brass, or steel. This is primarily related to the ease of sampie
preparation, high sputter yields for some metals, and the need to compare
sources with similar materials (i.e., dc versus rf). However, spectra such as
that shown in Fig. 7-29 demonstrate the analytical potential of rf-powered
glow discharge sources in the analysis of nonconducting materials. (114) Figure
7-29 is a mass spectrum of a mixture oftransition metal oxides (As, Cr, Ga,
Ni) pressed, without binder, into a pellet form. Important features to note
are that the oxides are efficiently sputtered, dissociated, and ionized in what
is reported to be a very stable plasma. Analyte ion currents, approximately
10- 11 A, are comparable to that observed for conducting matrices. The
spectrum indicates that dissociation of oxides is rather efficient in this plasma
(0.3 Torr, 25 W). No enhancement in oxygen species such as ArO+ and
ArOH+ (as a result of the oxide dissociation) is evident, although hydrides
such as AsH+ are observed. Such hydrides result from water species which
are commonly entrained in the preparation of powder sampies.
All oxide species may not be dissociated equally. In fact, Coburn and
colleagues(126) found that in certain cases the observed atomic signal may be
318 CHAPTER 7
C,
14.5%
\ NI
.
11.1%
IU,.
\
xl0
10
~
55
J
10 15 7.
1~ /::...
75
I. ! 2.
IA~
3. 4.
A
I.
.Y.r
Mass Number
I. 7. I.
J.
I. 100 11.
Figure 7-29. Radio frequency gJow discharge mass spectrum of a mixture of transition meta!
(Cr, Ni, Ga, As) oxides (0.30 Torr, 25 W).(114)
less than molecular ion signals. Figure 7-30 indicates the relative amounts
of molecular-to-atomic ions as a function of M-O bond energies where 1] is
the fraction of molecular species reaching the sampling orifice. The dissoci-
ation efficiency of metal oxides was found to be related to the metal-oxygen
bond strength to a large extent, although contributions from other factors
are evident. Of note is the arsenic oxide fraction, approximately 0.5, indica-
ted here. Arsenic appears to be one of the more difficult oxides to dissociate
although its binding energy is not as high as some of the more easily dissoci-
ated oxides indicated. In fact, arsenic oxide is present at a similar fraction
in the spectrum shown in Fig. 7-29. One factor proposed that may account
for its unlikely tenacity is a mass-dependent bias in the degree of dissociation
during the sputtering event. As argon ions impinge on oxygen species at the
surface of the sampie, the energy of the sputtering atom is transferred to the
oxygen atom. The ability of oxygen atoms to elastically exchange the kinetic
energy to the metal atoms is a function of the mass of the metal, with more
efficient energy exchanges occurring with atoms of like mass. Subsequently,
oxygen atoms bonded to light metal atoms are likely to be sputtered as
molecules while oxygen atoms bonded to higher-mass metals would leave
the surface as atoms. This phenomenon is also illustrated qualitatively in
Fig. 7-30 where oxides with lower-mass ions have higher 1] values (i.e., above
the dashed line) and oxides with higher-mass cations have lower 1] values
1.0 I I I
0.7
As
•
/
/ :Pr
0.4 -
V
Ti
•
I
-
Ta
• I-Nb
0
0.2
/
.
• ::!;
~ +
::!;
/H_O
/
• 0.1
'"
0.07
..
Ni
/:sn
~e/
0.04 -
Zn
-
Eu
• /
0.02
/
0.01
/ Pb
•
2 3 4 5 6 7 B 9
M - 0 Bond Energy (eV)
Figure 7-30. The dependence of 1) on the M-O bond energy (60 mTorr Ar, 13.56 MHz,
100 W).(126)
(i.e., below the dashed line). The authors also concluded that oxygen-bearing
contaminant species tend to increase the amount of molecular species
observed because of chemisorption of the contaminant onto the sampIe
surface. Good vacuum practice tended to reduce the relative amounts of
molecular species.
Other exemplary applications include various vitrified glass matrices.
An rf glow discharge mass spectrum of NIST SRM 1412 multicomponent
glass is shown in Fig. 7-31a. These ions were generated from the flange-
mounted source shown in Fig. 7-20 and were analyzed by a quadrupole mass
spectrometer.(127) The sampIe is a silicon-based vitrified glass. The low signal
intensity for silicon is related primarily to its low concentration (19.8%),
even though it is the "matrix" element. Higher-mass analyte ions, indicated
in the expansions, represent various minor elements present at concentra-
tions of 3-4% atomic. These species are readily atomized, including barium,
320 CHAPTER 7
Ar+ Sr Cd Ba Pb
J"OOI
(X100) (X1000) (X1000)
A~
2+
Ar
(X1)
20 40 60 80 100 120 140 160 180 200

m/z
b
2+
Ar Si Ar+
Fe +
Ar 2
U
Na
B
Sr
(X1000) (X1000)
AI
Mg
zn! 1'~~iA(
(X11JO) JJ~
(X10) ~~=~~~~::::::~~::!:::=;(====;=(~~:;=~~~=!~==
20 50 60 70 80 90
m/z
Figure 7-31. Radio frequency glow discharge mass spectrum of (a) high- and (h) low-mass
elements in NIST SRM 1412 multicomponent glass.(J27)
which tends to form strong metal oxides. Various low-mass and transition
elements (1-4% atomic) of interest are seen in Fig. 7-31b. Few molecular
interferences are noted at masses of interest. Argides of low-mass elements,
such as the labeled silicon and aluminum argide species, can present prob-
lems in the transition element region. Such interferences often do not pose
a problem in that other isotopes of the element of interest may be chosen,
but care must be exercised.
Radio frequency glow discharges also find applications in the area of
depth profiling. Lower operating pressures of rf-powered glow discharges
should reduce redeposition of sputtered materials and thus should increase
depth resolution relative to dc glow discharges. Coburn et al. demonstrated
and characterized the use of rf glow discharges for the determination of
elemental composition profiles in solids. (113,128,129) The sampies were mounted
on target holders that were much larger in diameter (typically 10-50 times
larger) in order to provide a uniform ion current density across the sampie
surface. Uniform sputtering ion current densities help to maintain uniform
etching rates across the sampie. Using relatively small sampies also tends to
normalize discharge characteristics between sampies, as weIl as to minimize
sampie heating that could lead to vaporization and unacceptably high etch
rates. One disadvantage of the technique is that the holder is also sputtered
during the analysis, and all constituents of the holder material contributed
to the observed ion signal. In such applications care must be exercised in
choosing an appropriate backing electrode. Analyses indicated a lack of
matrix effects on ion signals and the possibility of performing standardless
quantitations. In characterizing the technique, the authors found that sensi-
tivity increased with pressure from 10 to 160 mTorr, and was attributed to
increased ionization efficiency resulting from an increased metastable density
and collision frequency. Sensitivity was found to decrease with increasing
power, although the reason for this was not clear. As expected, depth resolu-
tion was found to decrease with increasing pressure because of collisional
redeposition. Obviously, a trade-offbetween sensitivity and resolution exists
in terms of optimum discharge pressure. Depth resolution was also found to
decrease with increasing power as weIl. Parametric control of the rf discharge
allowed sputter etch rates to be varied from 1 A/min for low-power He
discharges to 1000 A/min for high-power Ar discharges. Single ppm detec-
tion limits were observed in the course of these investigations.
An example of the depth profiling potential of rf-powered glow dis-
charge mass spectrometric analysis of thin films is shown in Fig. 7_32.(129)
The sampie, which is illustrated in the inset, consists of a 1000-A copper
layer on a fused quartz substrate. A 3000-A nickel film was evaporated on
the copper film. Resolution, as defined by [t(O.1) - t(0.9)]/t(0.5), is 3.3%
for Ni+ and 2.95% for Cu+ where t(0.5) is the time where the ion signal is
0.5 times the steady-state ion current; 1(0.1) and 1(0.9) are defined similarly.
322 CHAPTER 7
•0 0
Area 1.1 cm 2
., 0
...c
'V
0
•
~'" 2.
() I
300M Ni
0
. °
~ :t
0
100011. ~u
o· 0
I Si0 2 Substrate I o.
o.
°
0
!
•
J.-"...1...-%°
Q
20 40 60 100 120 140

80 160 180
t (minI
Figure 7-32. Radio frequency glow discharge mass spectrometry-determined depth profile of
the thin-film structure shown (13.56 MHz, 50 W, 30 mTorr Ar).(129)
Differences in the resolution reflect the differences in relative ion yields.

(Figure 7-32 also indicates a higher sputter yield for copper, in that sputter-
ing times are not directly proportional to the thickness of the layers.) The
authors present possible mechanisms that influence interface broadening
including depth-dependent mechanisms such as nonuniform sputter etching
and depth-independent mechanisms such as surface mixing and diffusion.
Depth resolution is proposed to be determined by a 30-A depth-independent
interface broadening plus a broadening that varies between 2 and 8% of the
total depth of analysis.
7.3.5.4. rf G/ow Discharge Optica/ Spectroscopies
Analytical applications of rf-powered glow discharge are only beginning

to attract the interest of optical spectroscopy practitioners. Publications of
analytical applications of the rf glow discharge source for spectroscopy have
been limited; however, the recent renaissance of rf GDMS is beginning to
spawn interests in optical applications.
Winchester and Marcus demonstrated the potential applicability of rf
glow discharge atomic emission spectroscopy. (130) Initial work demonstrated
the potential of the rf glow discharge source for atomization and subsequent
excitation of both electrically conducting and nonconducting materials.
These studies were performed using the six-way cross mounted direct inser-
tion probe ofFig. 7-21. Subsequent studies(1l6) employed the external sampie
mount shown in Fig. 7-22. As mentioned in that section (7.3.5.2), this source
incorporates two optical ports, one viewing down the sampie axis
(glow+cathode) and one perpendicular to the sampie axis (glow only).
This geometry allowed the comparison of the contribution of the cathode
blackbody radiation and spurious electrieal breakdowns to the observed
signal. It was found that the "glow only" and the "glow + eathode" optical
geometries had signal-to-noise ratios that were statistically equivalent, while
analyte emission intensities were a factor of 2-5 higher for the latter config-
uration. Optimization of souree parameters, discharge pressure and power,
demonstrated a trade-off between the atomization rate and the extent of
excitation. Atomization rates increase with increasing discharge power at
constant pressure and increasing pressure at constant discharge power.
Excitation temperatures (though not at Ioeal thermodynamic equilibrium
eonditions) tend to increase with increasing power but decrease with inereas-
ing pressure. Conditions of high pressure (6-10 Torr) and high power
resulted in strong analyte emission intensities. These conditions are radically
different from those in the rf GDMS case and point to another differenee
between the rf and de, where AES and MS conditions are more comparable.
Exemplary emission spectra of both nonconducting and condueting sampies
were presented. The emission speetrum of Macor, a noncondueting glass
ceramic, is shown in Fig. 7-33. Diseharge pressures were decreased in order
to minimize the sputter rate and therefore the subsequent rapid release of
trapped water which contributed a large amount of band structure to the
spectrum. At these compromised discharge eonditions the diseharge emission
is still atomic in nature, even though oxides are being sputtered. Several
emission lines of Al and Si, sputtered from the Macor, were identified.
Excited argon emission lines are also observed in the 350-500 nm range.
Al (r) 396.2
Al (r)
Al Ir) 309.3
"'" '~
250 300 350 400 450 500 550 600
Wavelength (nm)
Figure 7-33. Radio frequency glow discharge emission spectrum ofMacor collected with the
"glow + cathode" optical geometry (13.56 MHz, 2 Torr Ar, 60 W).(1l6)
324 CHAPTER 7
Page et al.(13I) used an rf sputter source to produce atoms and clusters

for study by absorption, emission, and laser-induced fluorescence spec-
troscopies. In comparison with more exotic and expensive sources such as
lasers, the rf discharge serves as an economical and efficient source of atoms
and small clusters. The rf source used was of a typical deposition (diode)
design, such as that of Fig. 7-18, and is mounted via a 6-inch conflat flange
to a six-way cross. The source is water-cooled and utilizes 3-inch-diameter
targets. The advantages of such a source are its simplicity, low cost, produc-
tion of large fluxes of both atomic and small cluster species (Le., dimers),
and low blackbody radiation because of the near room temperature of the
target and the discharge chamber. The atom number density of an iron target
was found by atomic absorption to be 2 x 1012 cm- 3 , with a translational
temperature of 800 K. Copper dimer number densities were estimated to
be 10 10-10 11 cm- 3 for copper targets with a 500 K rotational temperature
calculated. As an example of the source utility for the study of small cluster
species, an LIF spectrum of the (0,0) band of the CU2 C ~ X transition is
shown in Fig. 7-34. This is the first rotationally resolved spectrum of this
transition, with the Q branch head shown at ~21842 cm- I and the R branch
at ~21845 cm-I. High signal levels allowed the delayed detection (1 flS) of
the Astate while ignoring interference from the short-lived B state. With
rotational temperatures only slightly above room temperature, rotationally
10,0)
CU2 C-X LlF Spectrum
( 1,11
~~~~~d
21760 21780 21800 21820 21840 21860
II(cm- 1) -
Figure 7-34. LIF spectrum of the CU2 C ..... X transition, obtained from the sputtering of a
copper target in an rf-powered glow discharge.(131)
resolved spectra of small clusters should be possible for other metals. As in
this case, assignments of the excited-state symmetry and rotational constants
may be determined.
7.4. Conclusions
While bulk metal and semiconductor analyses have been, and will
continue to be, the major applications of glow discharge spectroscopies, the
needs presented in those areas where oxidejrefractory materials are the raw
and finished products are so great that future investigations into glow dis-
charge sampling methodologies are guaranteed. The two basic approaches
outlined in this chapter demonstrate the basic characteristics and great
potential that glow discharge spectroscopies hold for this very important set
of sampie matrices.
EarIy in the development of glow discharge atomic emission spectros-
copy, the application of the sampie briquetting technique showed promise.
The combination of analytical methodology development and fundamental
plasma studies have allowed for the more straightforward and precise appli-
cation of the compaction technique. Problematic inclusion of water and
gases in the compaction process has been greatly minimized by more careful
sampie handling and the use of conducting host matrices which have getter-
ing properties. The ability to easily add an internal standard element, or in
the best case, an isotopic spike in mass spectrometry applications is an
important asset to this method. The compaction technique will see increased
usage as it is applicable to any GD system without modification of the base
instrument or source, permitting optimum system ftexibility.
The inherent capability of radio-frequency-powered glow discharges to
operate with either conductive or insulating sampie types is probably the
single most important development in this field over the last decade. Sputter
atomization of bulk nonconductors such as glasses and ceramics is direct
and straightforward. Analysis of powder sampIes still requires compaction
to form the appropriate sampIe shape. While many powders may be com-
pressed to sufficient strength, addition of some amount of metal binder may
be necessary from a practical standpoint. Conversion of an existing GD
system to rf operation may require some capital investment, but it is expected
that the long term benefits for total solids sampIe coverage may be well
worthwhile.
The analysis of nonconductive sampIe types will always pose unique
challenges not faced in traditional GD applications. It is these applications,
however, that may generate the most new interest in the use of GD
techniques.
326 CHAPTER 7
References
1. J. A. C. 8roekaert, T. Graule, H. Jenett, G. Tölg, and P. Tschopel, Fresenius Z. Anal.

Chem. 332 (1989) 825.
2. M. Dogan, K. Laqua, and H. Massmann, Spectrochim. Acta 278 (1972) 65.
3. M. R. Winchester, S. M. Hayes, and R. K. Marcus, Spectrochim. Acta 468 (1991) 615.
4. G. S. Lomdahl, R. McPherson, and J. V. Sullivan, Anal. Chim. Acta 148 (1983) 171.
6. G. S. Lomdahl and J. V. Sullivan, Spectrochim. Acta 398 (1984) 1395.
7. S. EI Alfy, K. Laqua, and H. Massmann, Z. Anal. Chem. 263 (1973) 1.
9. H. 8ubert, W. -D. Hagenah, and K. Laqua, Spectrochim. Acta 348 (1979) 19.
10. H. Bubert and W. -D. Hagenah, Spectrochim. Acta 368 (1981) 489.
11. T. J. Loving and W. W. Harrison, Anal. Chem. 55 (1983) 1526.
12. R. B. Keefe, Ph.D. dissertation, University of Virginia, Chariottesville, August, 1983.
13. S. Caroli, A. Alimonti, and K. Zimmer, Spectrochim. Acta 388 (1983) 625.
14. S. Caroli, O. Senofonte, N. Violante, F. Petrucci, and A. Alimonti, Spectrochim. Acta
398 (1984) 1425.
15. S. Caroli, A. Alimonti, P. D. Femmine, and S. K. Shukla, Anal. Chim. Acta 136 (1982)
225.
16. S. Caroli, O. Senofonte, A. Alimonti, and K. Zimmer, Spectrosc. Leu. 14 (1981) 575.
17. P. J. Sievin and W. W. Harrison, Appl. Spectrosc. Rev. 10 (1975) 201.
18. P. W. J. M. 80umans, Anal. Chem. 44 (1972) 1219.
19. R. K. Marcus and W. W. Harrison, Anal. Chem. 59 (1987) 2369.
20. M. R. Winchester and R. K. Marcus, unpublished work.
21. F. L. King, A. L. McCormack, and W. W. Harrison, J. Anal. At. Spectrom. 3 (1988) 883.
22. W. J. Kolar, M. S. thesis, Clemson University, Clemson, S. C., May, 1989.
23. H. Mai and H. Scholze, Spectrochim. Acta 418 (1986) 797.
24. D. S. Gough, P. Hannaford, and R. M. Lowe, Anal. Chem. 61 (1989) 1652.
26. J. E. Ve1azco, J. H. Kolts, and D. W. Setser, J. Chem. Phys. 69 (1978) 4357.
27. E. Stern and H. L. Caswell, J. Vac. Sei. Technol. 4 (1967) 128.
28. D. G. Swartzfager, S. 8. Ziemecki, and M. J. Kelley, J. Vac. Sei. Technol. 19 (1981) 185.
29. N. Q. Lam, H. A. Hoff, H. Wiedersich, and L. E. Rehn, Surf Sei. 149 (1985) 517.
30. H. Shimizu, M. Ono, and K. Nakayama, J. Appl. Phys. 46 (1975) 460.
31. G. S. Anderson, J. Appl. Phys. 40 (1969) 2884.
32. H. Mai and H. Scholze, Spectrochim. Acta 428 (1987) 1187.
33. J. Workman, Jr., and H. Mark, Spectroscopy 3(3) (1988) 40.
35. N. P. Ferreira and H. G. C. Human, Spectrochim. Acta 368 (1981) 215.
37. H. Schuler and H. Gollnow, Z. Phys. 93 (1935) 611.
38. A. I. Drobyshev and Y. I. Turkin, Spectrochim. Acta 368 (1981) 1153.
39. R. 8. Djulgerova, in: Improved Hollow Cathode Lampsjor Atomic Spectroscopy (S. Caroli,
ed.), Ellis Horwood Ltd., Chichester, England, 1987.
40. D. Fang and R. K. Marcus, Spectrochim. Acta 438 (1988) 1451.
41. R. K. Marcus and W. W. Harrison, Anal. Chem. 58 (1986) 797.
42. J. Czakow, in: Improved Hollow Cathode Lampsjor Atomic Spectroscopy (S. Caroli, ed.),
Ellis Horwood Limited, Chichester, England, 1987.
43. K. Robinson and E. F. H. Hall, J. Met. 39 (1987) 14.
44. G. N. Jackson, Thin Solid Films 5 (1970) 209.
45. J. K. Robertson and C. W. Clapp, Nature 132 (1933) 479.
46. R. H. Hay, Can. J. Res. AI6 (1938) 191.
47. J. 1. Lodge and R. W. Stewart, Can. J. Res. A26, (1948) 205.
48. S. M. Levitskii, Sov. Phys. Tech. Phys. 27 (1957) 913.
49. G. K. Wehner, Adv. Electron. Electron Phys. 7 (1955) 239.
50. G. S. Anderson, W. N. Mayer, and G. K. Wehner, J. Appl. Phys. 33 (1962) 2991.
51. P. D. Davidse and L. 1. Maisse1, 3rd International Vacuum Congress, StuUgart, 1965,651.
52. P. D. Davidse and L. 1. MaisseI, J. Appl. Phys. 37 (1966) 574.
53. B. N. Chapman, Glow Discharge Processes, Wiley~Interscience, New York, 1980.
54. J. E. Greene and J. M. Whelan, J. Appl. Phys. 44 (1973) 2509.
55. S. B. Hyder, J. Vac. Sei. Technol. 8 (1971) 228.
56. J. L. Vossen and E. S. Poliniak, Thin Solid Films 13 (1972) 281.
57. K. Wasa and S. Hayakawa, Jpn. J. Appl. Phys. 12 (1973) 408.
58. R. M. Goldstein and S. C. Wigginton, Thin Solid Films 3 (1969) R41.
59. J. G. Titchmarsh and P. A. B. Toombs, J. Vac. Sei. Technol. 7 (1970) 103.
60. S. Iida and S. Kataoka, Appl. Phys. Lett. 18 (1971) 391.
61. 1. H. PraU, Thin Solid Films 3 (1969) R23.
62. R. Frerichs, J. Appl. Phys. 33 (1962) 1898.
63. R. Frerichs and C. J. Kircher, J. Appl. Phys. 34 (1963) 3541.
64. P. K. Tien, G. Smolinsky, and R. J. Martin, Appl. Opt. 11 (1972) 637.
65. M. J. Vasile and G. Smolinsky, Int. J. Mass Spectrom. Ion Phys. 12 (1973) 133.
66. G. Smolinsky and M. J. Vasile, Int. J. Mass Spectrom. Ion Phys. 12 (1973) 147.
67. M. J. Vasile and G. Smolinsky, Int. J. Mass Spectrom. Ion Phys. 13 (1974) 381.
68. G. Smolinsky and M. J. Vasile, Int. J. Mass Spectrom. Ion Phys. 16 (1975) 137.
69. H. Biederman and L. Martinu, Acta Phys. Slovaca 35 (1985) 207.
70. N. Mutsukura, K. Kobayashi, and Y. Machi, J. Appl. Phys. 66 (1989) 4688.
71. H. Chatham and P. K. Bhat, Mat. Res. Soc. Symp. Proc. 118 (1988) 31.
72. J. W. Coburn and M. Chen, J. Appl. Phys. 51 (1980) 3134.
73. H. Ratinen, Appl. Phys. Lett. 21 (1972) 473.
74. H. Ratinen, J. Appl. Phys. 44 (1973) 2730.
75. H. Ratinen, J. Appl. Phys. 44 (1973) 3817.
76. T. C. Tisone, B. F. T. Bolker, and T. S. Latos, J. Vac. Sci. Technol. 17 (1980) 415.
77. C. R. Aita, T. A. Myers, and W. J. LaRocca, J. Vac. Sei. Technol. 18 (1981) 324.
78. R. A. GoUscho and T. A. Miller, Pure Appl. Chem. 56 (1984) 189.
79. R. M. Roth, Mat. Res. Soc. Symp. Proc. 98 (1987) 209.
80. C. R. Aita and M. E. Marhic, J. Vac. Sei. Technol. A 1 (1983) 69.
81. A. D. Kuypers, A. Koch, and H. J. Hopman, J. Vac. Sei. Technol. A 8 (1990) 3736.
82. E. W. Eckstein, J. W. Coburn, and E. Kay, Int. J. Mass Spectrom. Ion Phys. 17 (1975)
129.
83. R. A. Gottscho and M. L. Mandich, J. Vac. Sei. Technol. A 3 (1985) 617.
84. V. M. Donnelly, D. L. Flamm, and G. Collins, J. Vac. Sei. Technol. 21 (1982) 817.
85. M. B. Hopkins, C. A. Anderson, and W. G. Graham, Europhys. Lett. 8 (1989) 141.
86. T. 1. Cox, V. G. I. Deshmukh, D. A. O. Hope, A. J. Hydes, N. St. J. Braithwaite, and
N. M. P. Benjamin, J. Phys. D 20 (1987) 820.
87. D. N. Ruzic and J. L. Wilson, J. Vac. Sei. Technol. A 8 (1990) 3746.
89. K. Kohler, J. W. Coburn, D. E. Horne, E. Kay, and J. H. Keller, J. Appl. Phys. 57 (1985)
59.
90. K. Kohler, D. E. Horne, and J. W. Cobum, J. Appl. Phys. 58 (1985) 3350.
91. J. W. Coburn and E. Kay, J. Appl. Phys. 43 (1972) 4965.
92. F. Shinoki and A. Itoh, Jpn. J. Appl. Phys. Suppl. 2, Pt. 1 (1974).
328 CHAPTER 7
93. J. W. Coburn, Thin Solid Films 171 (1989) 65.

94. H. R. Koenig and L. I. Maissei, IBM J. Res. Dev. 14 (1970) 168.
95. R. A. Gottscho, G. R. Scheller, D. Stoneback, and T. Intrator, J. Appl. Phys. 66 (1989)
492.
96. J. W. Coburn and K. Kohler, Proc. Electrochem. Soc. 87 (1986) 13 (Proc. Symp. Plasma
Process., 6th, 1986).
97. R. W. Gould, Phys. Lett. 11 (1964) 236.
98. C. M. Horwitz, J. Vac. Sci. Technol. A 1 (1983) 60.
99. J. S. Logan, IBM J. Res. Dev. 14 (1970) 172.
100. L. I. Maissei and P. M. Schaible, J. Appl. Phys. 36 (1965) 237.
101. J. E. Greene, CRC Crit. Rev. Solid State Mater. Sei. 11 (1983) 47.
102. J. V. Sullivan and A. Walsh, Spectrochim. Acta 21 (1965) 721.
103. R. M. Lowe, Spectrochim. Acta 31B (1976) 257.
104. D. S. Gough and J. V. Sullivan, Analyst 103 (1978) 887.
105. J. V. Su1livan and J. C. Van Loon, Anal. Chim. Acta 102 (1978) 25.
106. G. S. Lomdahl and J. V. Sullivan, Spectrochim. Acta 39B (1984) 1395.
107. F. Leis, J. A. C. Broekaert, and K. Laqua, Spectrochim. Acta 42B (1987) 1169.
108. N. Violante, O. Senofonte, A. Marconi, O. Falasca, and S. Caroli, Can. J. Spectrosc. 33
(1988) 49.
109. F. Leis, J. A. C. Broekaert, and E. B. M. Steers, Spectrochim. Acta 46B (1991) 243.
110. P. E. Walters and H. G. C. Human, Spectrochim. Acta 36B (1981) 585.
111. J. F. Geiger, M. Lopnarski, H. Oeschner, and H. Paulus, Mikrochim. Acta (Wien) I
(1987) 497.
112. D. L. Donohue and W. W. Harrison, Anal. Chem. 47 (1975) 1528.
113. J. W. Coburn, E. Taglauer, and E. Kay, J. Appl. Phys. 45 (1974) 1779.
115. D. C. Duckworth and R. K. Marcus, 16th Annu. Meet. Fed. Anal. Chem. Appl. Spectrosc.
Soc. 1989, Paper No. 138.
116. M. R. Winchester, C. Lazik, and R. K. Marcus, Spectrochim. Acta 46B (1991) 483.
117. P. R. Cable and R. K. Marcus, 1991 European Winter Conference on Plasma
Spectrochemistry.
118. R. Morrison, Grounding and Shielding Techniques in Instrumentation, Wiley-Interscience,
New York, 1986. .
119. D. C. Duckworth, P. R. Cable, and R. K. Marcus, 38th ASMS Conference on Mass
Spectrometry and Allied Topics, 1990.
120. D. Fang and R. K. Marcus, 17th Annu. Meet. Fed. Anal. Chem. Appl. Spectrosc. Soc.
1989, Paper No. 339.
121. F. L. King, A. L. McCormack, and W. W. Harrison, J. Anal. At. Spectrom. 3 (1988) 883.
122. D. C. Duckworth and R. K. Marcus, Appl. Spectrosc. 44 (1990) 649.
123. N. Ketchell, Ph.D. dissertation, University of Manchester, 1989.
124. D. L. Donohue, W. H. Christie, D. C. Duckworth, D. H. Smith, and R. K. Marcus, 39th
ASMS Conference on Mass Spectrometry and Allied Topics, 1991.
125. R. K. Marcus and D. C. Duckworth, 1989 Pittsburgh Conference and Exposition, 1989,
Paper No. 657.
126. J. W. Coburn, E. Taglauer, and E. Kay, Jpn. J. Appl. Phys. Suppl. 2, Pt. 1 (1974) 501.
127. D. C. Duckworth and R. K. Marcus, 1989 Pittsburgh Conference and Exposition (1989),
Paper No. 548.
129. J. W. Coburn, E. W. Eckstein, and E. Kay, J. Appl. Phys. 46 (1975) 2828.
131. R. H. Page, C. S. Gudeman, and M. V. Mitchell, Chem. Phys. 140 (1990) 65.
8
Thin Film Analysis
Hubert Hocquaux
8.1. Historicallntroduction
8.1.1. Bulk Analysis
Grimm (1968) was the first to demonstrate the principle of using a glow
discharge lamp for the analysis of Hat sampies. (I) Since the Grimm lamp
appeared, low-pressure gas discharges have found many applications.
Several authors have investigated the potentialities of such a discharge. (2-9)
Principal applications are in the bulk analysis of metals or of nonconducting
materials pressed into pellets with a conducting binder. Several other
configurations have been described for bulk analysis, mostly to improve the
sensitivity, e.g., hollow cathodes,(IO·II) boosted lamps,(12,I3) and magnetic
field-enhanced glow discharges.(14)
8.1.2. Surface Analysis
Berneron and co-workers (1971) introduced the use of the glow dis-
charge lamp for depth profilingY S) They plotted the variation of the light
intensity of aluminum and mangane se versus sputtering time in order to
study defects on the surface oflow-carbon steels. In 1972, Belle and Johnson
published some practical applications of the Grimm lamp for depth
profiling. (16) They measured the composition of various elements in steel
within the depth range ofO.I-40 pm. In the same year, Greene and Whelan
proposed the use of glow discharge optical spectrometry as a method for
Hubert Hocquaux _ IRSID UNIEUX, BP 50, 42702 Firminy, Cedex, France.
329
330 CHAPTER 8
analyzing thin filmsY7) Concentration profiles of boron and phosphorus

implanted into silicon are described in several publications of Greene et
a1YS,19) The lamp had a flat cathode but the geometry differed from that of
the original Grimm tube. The stability of the discharge was poor during the
first 50 sand the authors estimated the depth resolution with such a system
to be approximately 80 nm. Schneider and Shuman described in 1974 one
of the first examples of surface analysis by GD-OES,(20) but the recording
of intensities on photographic medium resulted in a rather poor depth
resolution.
Following these pioneering studies, Bemeron and Moreau carried out
extensive work exploring the use of GD-OES in the determination of gases
in metals and depth profiling analysis.(21,23) As a result ofthis work, a proto-
type was made commercially available by RSV in 1978. This system was
bought by Unirec and designed specially for depth profiling analysis.
Between 1978 and 1985, GD-OES emerged as a technique among the most
widely used for surface analysis, first in France and thereafter predominately
in Germany and Japan. Today, systems for depth profiling are available
from several manufacturers of OES equipment.
8.2. Principle of Surface Analysis by Glow

Discharge Sputtering
The glow discharge lamp operates by removing material from the

sampIe surface by cathodic sputtering in an atmosphere of a carrier gas at
low pressure, generally argon. The material is removed layer by layer, more
or less parallel to the original surface. Therefore, this technique is suitable
for depth profiling analysis. The ablation can be easily controlled by adjust-
ment of operating parameters: buming voltage, current, and gas pressure.
The sputtering process produces a crater the size of which is determined by
the anode geometry. The sputtering rate is defined as the sampie erosion per
unit time.
Atoms released by the impact of positive ions and neutral atoms are
raised to an excited state by further collisions, producing a cathodic glow.
The resulting plasma emits radiation consisting of characteristic lines of the
elements eroded from the sampie and the lines of carrier gas. The emission
from the plasma is analyzed by an optical spectrometer. By integrating the
element-specific signals at a high rate, information on the in-depth variation
of elemental composition is obtained. (23) The high sputtering rate, combined
with very high speed intensity measurement of analytical signals, led to a
very attractive technique capable of in-depth profiling of surface layers with
thicknesses ranging from a few nanometers to several tens of micrometers.
THIN FILM ANAL YSIS 331
8.3. Equipment
8.3.1. Glow Discharge Source
Various modes of glow discharge operation have been used as radiation

sources in atomic spectroscopy. For analytical applications the lamp consists
of a low-pressure electrical discharge in a noble gas at pressures from 10 to
1000 Pa. The analytical sampie acts as the cathode. For surface analysis only
plane cathodes have been considered. The principle of glow discharge lamps
and their electrical characteristics are described in several publications(8,24--26)
and summarized in previous chapters.
For surface analysis the original Grimm Lamp(l) is the most frequently
used (Fig. 8-1). This tube consists of a particular shape discharge chamber.
The sampie to be analyzed acts as the cathode: it seals and separates the
chamber from the external atmosphere. In order to eliminate the heat gener-
ated, the cathode must be cooled. The anode is tubular in shape and its
frontal distance to the sampie is 0.2 mm. Because of the shape of the tube,
only the negative glow is present. A first pump maintains the gas pressure
in the lamp. In order to avoid a second discharge being created between the
sampie and the face of the anode tube, an additional pump is used to evacu-
ate tbis interspace. A further development is the glow discharge lamp with
floating anode, which has been introduced by Siemens (Fig. 8_2).(27) This
system is now used by the firm Leco. Anode, floating anode, and cathode
are separated from each other by an isolator. The distance from the cathode
K- t - + - E
o
Figure 8-1. Schematic diagram of a Grimm

B
glow discharge lamp. A, anode; B, cathode; C,
anodejcathode insulator; D, MgF2 window;
E, sampIe; F, cooling; G, argon inlet; H, PI
vacuum pump; I, P2 vacuum pump; J, O-ring;
K, anodic chamber. H c
332 CHAPTER 8
Wind ••
Anode
Teflu
f leating PUIiP
1.L.L.J:...L.L.L.LLL;~-----1~~
hode
- ,,,",'
L.I-I-/____~I Cathode
Sampl e
Figure 8-2. Schematic diagram of the glow discharge lamp introduced by Siemens.
to the anode is longer and the positive column is present contributing to the
radiation output.
8.3.2. Spectrometer
The glow discharge lamp is placed in front of the slit of an optical
spectrometer. A few studies have been performed with a monochromator or
a fast sequential spectrometer, (28) but for surface analysis it is necessary to
use a multichannel spectrometer. The sinoptic diagram of the more sophis-
ticated commercial system is presented in Fig. 8-3. The light emitted by the
plasma enters the spectrometer via a window and a primary slit. The optical
components are made with magnesium fluoride (MgF2) in order to extend
the useful spectral range to 110 nm. The light is spectrally resolved by a
diffraction grating and led through a secondary slit to a photomultiplier
where it is transformed into an electric signal. The principal characteristics
of the first prototype used by IRSID are summarized in Table 8-1. Some
complementary information on the theory of optical spectrometers can be
found in various papers.(22,29,30)
Most frequently a wavelength range from 110 nm to 460 nm is chosen
by a suitable selection of the individual optical components. For the
detection of spectral lines with wavelengths lower than 150 nm, special
photomultipliers are needed. But another solution is commonly used, con-
sisting of covering the envelope of the photomultipliers with a fine coat of
sodium salicylate. Thus, the analysis of H, N, 0, Cl, and C is possible in the
UV region. Some studies have been made in the visible spectral region for
......-t;-- sampie
glo. discharge
VACUUM
PO LV
CHROMATOR
Figure 8-3. Schematic diagram of the IRSID surface analysis system.
Table 8-1. Optical System of the RSV Analymat 2504
Paschen-Runge configuration
F ocal length: 2 m
Jobin-Yvon holographie concave grating with a surface covered with MgF 2
Reciprocal dispersion: 0.22 nm/mm
the determination of these elements. (31) It is expected that, with neon or

mixtures of argon and helium as carrier gas, spectrallines of high excitation
energy (F, Cl, 0, N) can be excited, which are not possible in an Ar atmos-
phere, but at present better detection limits are obtained in the UV region
(Cl, 0, H, N). In Table 8-2, a selection of analyticallines usable for surface
analysis is presented. As can be seen, there are a number of unusuallines that
may not be incorporated in commercially available multichannel emission
spectrometers. An interesting optical device to increase the spectral range
has been proposed recently by Jobin-Yvon (Longjumeau, France) (Fig.
8-4). In the near future, such developments will be necessary to improve the
334 CHAPTER 8
TBble 8-2. Analytical Lines for Surface Analysis

Element Ä Element Ä Element Ä
Ag 338.28 Cd 228.80 H 121.56

346.62 486.13
AI 237.84 656.30
256.80 Ce 413.76
394.40 Hf 286.63
396.15 Cl 118.87
133.57 Hg 253.65
Ar 137.72 134.72
157.49 479.50 I 145.79
183.04
As 189.04 Co 340.51
200.33 345.35 In 410.17
Au 242.79 Cr 267.71 Ir 203.35
298.92 322.07
B 208.95 425.43
249.60 K 404.72
Cu 219.22 766.49
Ba 23Q.42 327.39
455.40 La 408.67
Fe 259.90
Be 313.04 271.40 Li 323.26
332.12 273.95 610.36
371.99 670.77
Bi 306.77
Ga 403.14 Mg 277.66
Br 148.84 280.27
Gd 376.83 383.82
C 156.14
165.70 Ge 303.90 Mn 257.61
403.14
Ca 393.36 403.44
performance of GD-OES for research purposes. The systems for routine

control are less sophisticated.
8.3.3. Electronics and Computer
Various systems are proposed by the manufacturers of optical spectro-

meters for GD-OES. The equipment shown in Fig. 8-3 is not up-to-date,
but for the moment it exhibits the best performance among the commercially
available systems. It was first introduced by Siemens. In the near future
other systems with microcomputers will become commercially available.
Some design criteria should be mentioned:
Table 8-2. (Continued)
Element A. Element A. Element A.
Mo 317.03 Pd 360.95 Te 200.20
386.41 238.57
Pr 433.39
N 149.26 Ti 337.20
174.20 Pt 265.94 365.35
411.00
Rb 420.18 U 385.95
Na 330.23
588.99 Rh 437.48 V 311.07
589.59 318.39
S 180.73 411.17
Nb 316.34 437.90
410.09 Sb 206.83
416.46 W 400.87
Sc 424.68 429.40
Nd 430.35
Se 196.09 Y 371.10
Ni 225.38 377.43
341.47 Si 251.61
349.29 288.15 Zn 213.85
330.29
0 130.21 Sn 189.98 334.50
777.19 242.16 377.21
317.50 481.05
P 177.49 326.20
178.28 Zr 339.19
185.94 Sr 407.77 360.11
253.56
Ta 301.70
Pb 220.35 362.66
405.77
1. The signal from each photomultiplier is amplified (linear or logarith-

mic amplifier). A fast scanning frequency ofthe multiplexer is neces-
sary to reach a good depth resolution (30,000 s -I for the Siemens
system in Fig. 8-3).
2. The use of high-resolution color display permits areal-time monitor-
ing of the light intensity for each selected element so that the sputter-
ing process can be interrupted at any time and the bottom of the
crater examined by a complementary method (e.g., SEM, x-ray
diffraction, ESCA).
3. The number of selected lines and the scanning frequency of the multi-
plexer are often very important, therefore the capacity of storage of
the computer has to be large.
336 CHAPTER 8
SDS
~
salPLE
Figure 8-4. Optical system of the lY 50 GDS.
4. Complex and specific software is needed for checking, modifying,

smoothing the element profiles, and if possible quantifying the
results.
8.4. Analytical Results
In a glow discharge lamp the surface analysis of the sampIe is possible

because of the fact that the material is removed layer by layer in a cathodic
sputtering process. The light intensity of each line is simultaneously recorded.
Thus, in the strict sense of the word, the result is not a surface analysis
because the sampIe is continuously consumed. In fact, GD-OES is a depth
profiling analysis. When the sputtering process is interrupted at the end of
the analysis the first recorded result is the change with time of the intensities
of analytical lines of selected elements. This qualitative analysis is often
sufficient, especially when comparisons are made. To illustrate the pos si-
bilities of GD-OES for qualitative analysis, an example is given in Fig. 8-5
and the information that is extractable from such arecord is summarized
below. The sampIe was exposed for a corrosion test in a crevice with caustic
species concentrated from river water pollution: liquid sampling from the
crevice during the test indicated that the solution contained a large amount
of sodium.
GD-OES showed the presence of a surface layer made of interpenetrated
oxides and deposits:
::i
~ ______ calcium
z:-
'0;
c
!!
.5
.fo
;.3
silici
o 10
Abrasion time (min)
niCkel/
~\
~
Cr/Ni ratio "-

\
~ 'FelNi I1I.tio sulfur--..
~
oxygen/
o 10 15
Abrasion time (min)
t L er cnricbmUII
L.-=-: Fe <nrichmcn1
Loxidelay.,.
Figure 8-5. Qualitative surface analysis by GD-DES: study of corrosion products on InconeL
1. The presence of metallic oxides is demonstrated by the presence of

a layer with well-defined levels of the main metallic elements of the
base material (profiles ofNi, Cr, and Fe) associated with the presence
of oxygen. The abrasion of this layer takes about 1.5 min and corre-
sponds to a thickness on the order of a few micrometers. Nickel is the
main metallic component of the oxide but a chromium enrichment is
338 CHAPTER 8
located at the metall oxide interface and an iron enrichment occurs

at the top of the layer (see Fe/Ni and CrjNi plots in Fig. 8-5).
2. The deposits are rich in calcium, and contain some magnesium.
3. A small amount of sulfur is present in the surface layers. It is mainly
located in the near-surface deposit but there is also a small peak
dose to the metalj oxide interface; in this interface region oxygen is
also present.
4. Hydrogen and carbon peaks are also present: their peaks are dose
to the metalj oxide interface but they are not exactly corre1ated.
Oxygen could be associated with some of the above species.
5. Although the sensitivity of GD-OES for sodium is quite poor, this
element appears to be present and perhaps corre1ated with carbon,
hydrogen and/or sulfur (the profile of sodium is not shown in Fig.
8-5).
If we compare these results with those of other sampies exposed to

corrosion, it is possible to draw numerous metallurgical conc1usions. If one
needs more quantitative results, some modifications of the curves are
necessary:
Time has to be correlated to depth.

Intensities of analyticallines have to be converted into concentrations.
8.5. Typical Operating Parameters for Surface Analysis
If the material, the anode, and the carrier gas are constant, the glow
discharge conditions are described by three important parameters: current
intensity, buming voltage, and gas pressure. These factors are not indepen-
dent; if two of them are selected, the third one is then automatically deter-
mined. The operating parameters frequently used for bulk analysis are
summarized in Table 8-3.
For surface analysis it is necessary to select softer conditions to improve
depth resolution and to avoid local melting. In most cases the current
intensity is lower than 100 mA and the buming voltage is chosen between
400 and 800 V. Constant-VOltage operation is seldom employed because of
Tab/e 8-3. Operating Conditions for Bulk Analysis

Current intensity: 80-200 mA
Burning voltage : 800-1600 V
Gas Pressure : a few millibars
Constant-voltage or constant-power modes
insufficient control of the discharge in the beginning. Such a problem is
mainly encountered when the oxide layer at the surface of the material is an
insulator (e.g., Ti, Zr, Al oxides). The discharge is then not stable because
charge accumulation occurs near the surface. A few studies have been per-
formed with a constant-voltage mode in the case of materials with a low
melting point such as Sn, Sb, and Zn.
If the thickness of the layer is an important quantity, another system
can be interesting. It consists of a mass flowmeter and a PID regulator
(proportional integral derivative pressure controller). With this apparatus it
is possible to maintain constant both the intensity of current and the burning
voltage by monitoring continuously the flow of gas. Thus, in a constant-
current mode the voltage does not change even if the electrical characteristics
of the sputtered layers are completely different. Currently it is not possible
to regulate the voltage during the very few first seconds (5 s) of the discharge.
An example of such a mode of operation is illustrated in Figs. 8-6 and 8-7.
In this case the principal advantages of the system concern the profile of
secondary elements like 0, C, H. For example, without regulation of the
voltage the profile of oxygen changes rapidly at the interface. In fact, this
artifact is only due to the variation of the voltage. With the correction system
the profile of oxygen does not vary in the interfacial zone.
~ ,------------------------------------------,
~
..........................................
-
=!
~ ..-
./0"\.
~
H! \ .. ' ....... .
FE
/0.: 00 \
..
>-
=: .,. . 0:
"
0 \
'"......
/
..
z "'! /~. 0 \
on . , 0 \
...=
:!!: ~ ".
.
o o ' .,
o "
....
~ o .
~.
............. .
_________o~.~ ____ 1G. ___________ _
-- ~.. • ••
.............
c
._ .. __ ._.-- ._-._ .. ..::-:.'":::.

~.-
\.
-- --:/-
..
~~~~.~.~~~~~~=r~r=F9~~,-~F9=T=r9F~
0.00 46.15 92.31 138.46 184.62 230.77 300.00
SPUTTERING TIME (sec)
Figure 8-6. Analysis of a zinc coating on asteeI sheet with the constant-current mode. 1=
75 mA, Vg ~ 1200 V in steel.
340 CHAPTER 8
gr--------------------------------------------.
~
•.........
~! \
= ..
c! ,. ~ZN
'."
>- "
...
...
~"'!o
~:e
... FE
... t
z:
::; .\. "y/
,/
L. " ".
n:ji. :\.,
1'•
"..... ,'" ".
.
11. _ ........
. " •• .1 ....... J.!...~ "_" VG
/ •• c. ...... "'-
g _ '-_/"/ _ _ _ _ _ o~·~~··:. :.:~~~~:.~,.:~~~.;;;~~:.~:;:.~~;
.,; 0.00 46.15 92.31 138.46 184.62 230.77 300.00
Figure 8-7. Analysis of a zinc coating on a steel sheet with the constant-current mode and
regulation of the burning voltage by the PID controller. 1= 75 mA, Vg = 1200 V in steel and
in the zinc coating.
8.6. Shape 0' the erater

The size ofthe crater is determined by the anode geometry. In Fig. 8-8,
three types of craters are shown (two circular anodes and an oval one). In
the first publications, (21,23,32) the authors mentioned that the bottom of the
crater was not plane and that metal deposits were observed at the periphery.
The influence of the type of lamp has been discussed. (32) In fact, these results
are not complete because they were obtained with an anode the diameter of
which was 8 mm and with a high lamp power. More recent studies have
demonstrated that these conditions are not weIl suited for surface analysis.(33)
These conclusions have been drawn by a group ofFrench laboratories work-
ing together in a Groupement po ur l'Avancement des Methodes Spectro-
scopiques (GAMS) commission. In fact, the shape of the crater changes
with the diameter of the anode, and with the operating parameters as can
be seen in Fig. 8-9. Bouchacourt(34) tried to draw for low-carbon steel some
areas on the diagram 1= f (Vg ) where typical shapes are found (Fig. 8-10).
When the conditions are mild, near the "normal" glow discharge region, the
bottom of the crater is not completely sputtered. With high intensities of
current the results are similar to those of earlier publications. A flat bottom
Figure8-8. Different types of craters as a function of anode diameter (rp). (Left) rp = 0.8 cm;
(center) ljJ = 0.4 cm; (right) L = 0.8 cm, 1= 0.2 cm.
Figure 8-9. Influence of the anode diameter and of the operating parameters on the shape
ofthe crater (iron matrix). (I) Diameter ofthe anode rp = 0.8 cm, Vg = 1000 V, 1= 75 mA. (2)
rp = 0.8 cm, Vg = 1200 V, 1= 100 mA. (3) rp = 0.4 cm, Vg = 600 V, 1= 25 mA.
is obtained with a current below 40 mA and a voltage that is high enough

to have an "abnormal" discharge. These examples demonstrate that the
conditions are to be optimized for the material and the type of anode. As
seen in Figs. 8-9 and 8-10, a roughening of the surface takes place during
sputtering. The roughness of the bottom increases linearly with time. This
is one of the factors limiting the depth resolution. It is clear that the choice
of the operating parameters for depth analysis is a compromise between a
good depth resolution and a high intensity of spectrallines.
342 CHAPTER 8
I
NO DISCHARGE NORMAL GLOWI AJlNORMAJ. GLOW DISCIlARGE
, DISCIlARGE /
• I ~\
I I"
50 >< \ I>< I \
I I J \ CRATER PROFILES
\
I
\
\ I \
\
\
I \\ (brudth x4. dtpth x300)
40 >< \ ~\I \,
\
\
\h \
\
Xl)< ><I' .:>'~
30 \ 1\ ,
\ \ I \ \
, \ f \ "
\ \ I' "
\ \ I" ' ... ,
\ ~ 16 Ä " .6. Ä' ......
\\ . . .
\I " ... " ........
J",.6.
\
'(b...... " .............. 3PCl.D.ln.. 1
',:
\\ +
\ ............ ... ................
\ + '" ... .............. ... ... 1Er .. 2pm. "ln""
\', 1......... . . ..................
....... 1 . . . . . . . . . . _ . . -...... -.......... 1,..••• ln·1
t ................... --...... -- ___ .. _ .. _.O • .sp,..1l1n·'
t ... ---------------O.2.5pD..llln"'1
o 200 ~oo 600 800 1000 1200
VOLTAB! (V)
Figure 8-10. Current/Voltage dependence on crater shapes and sputtering rates. Pure iron;
diameter of the anode IP = 0.4 cm. Reproduced with permission from Ref. 34.
8.7. Sampie Geometry
The shape of the lamp dictates the geometry of the sampies. With the
original Grimm lamp, and without special sampie holders, the material must
be flat and its dimensions must be larger than the seal, generally 20-mm
diameter for an anode of 8-mm diameter. In more recent studies, several
authors have used smaller anodes (4-mm and even 2-mm diameter) but with
poorer light intensity. It is possible to use smaller specimens by mounting
them in a conducting material, such as Ag or Cu, but the sputtered area
must be larger than the diameter of the anode. The choice of the anode
diameter is a compromise between lateral resolution and sensitivity. The
analyses of circular sampies have been performed in IRSID in France with
some special tools.(35,37) In Figs. 8-11 and 8-12, two examples of sampie
holders for the analysis of tubes or wires are given. In Fig. 8-13, depth
profiling results for circular sampies are illustrated. The tests have been
carried out on drawn wires of small diameter (cfJ = 0.8 mm) in stainless steel
from industrial production. The in-depth profiles of the elements Fe, B, Ca,
Cr, S, and C were recorded. Particular attention was given to boron and
calcium, which take part in the composition of wire-drawing antifriction
grease and remain as a contamination at trace levels. In contrast, with chemi-
calor electrochemical cleaning, ion erosion in a glow discharge does not
cause a "preferential cleaning" of the pore sites. SEM observations of the
Figure 8-11. Special holder for the analysis of wires or tubes (diameter> 2.2 mm).
Figure 8-12. Special holder for the analysis of wires (diameter <2.2 mm).
crater at different depths show the presence of pollutants in the surface

defects until their total elimination. The ion bombardment of these curved
surfaces leads to the formation of a Bat part on the upper generating lines
of the wires; because of this phenomenon the in-depth resolution is not as
344 CHAPTER 8
Fe __- - - - - - - - - - - ,
~
Ci)
Z
W
I-
Z
I-
J:
(!)
~
SPUTTERING TIME (min)

Figure 8-73. In-depth profile on the surface of a stainless-steel wire (tP = 0.8 cm). Reproduced
from Ref. 36.
good as for flat sampIes and the interface between two layers not so weIl
defined. The results described in different publications(35,37) have shown the
complementary nature of the SEM and GD-OES methods for the charac-
terization of industrial surfaces. The results can be easily used for control
purposes or for perfecting a manufacturing process.
8.8. Quantitative Surface Analysis with GD-DES
As has been demonstrated in the previous sections, applications of the

GD-OES technique for surface analysis are numerous. But the results are
most frequently presented as a relative intensity versus a sputtering time. To
quantitatively determine the distribution of an element as a function of the
distance from the surface, two problems have to be considered:
1. Sputtering rate of the target under argon ions accelerated in the

cathode dark space
2. Relationship between concentration and light intensity
If the sampIe is homogeneous, quantitative analysis is very easy. The
sputtered depth and the chemical content can be determined with reference
to the bulk composition. Standardization of light intensities can be made
with the help of reference sampIes. In a glow discharge there are few matrix
effects because of the fact that the analysis is conducted under equilibrium
conditions in a quasi-stationary state. This is true in depth profiling analysis
when rather small changes of composition have to be investigated. Some
applications of direct quantitative analysis have been described in the
literature. (23,32.35) An example of calibration with the help of bulk analysis is
given in Fig. 8-14. During cold-rolling a superficial enrichment of chromium
appears by diffusion from the bulk to the oxide layer. Behind the peak of
chromium adepleted zone can be observed. It is important to measure
quantitatively this phenomenon to know the corrosion properties of the
sheet.
If the sampIe matrix changes, the buming voltage and the subsequent
sputtering rate are no longer constant and the problem of quantification
is more complicated, as was pointed out very early by Jäger for bulk
analysis. (38,39) In the field of surface analysis, many authors have now
presented quantitative results with either a practical method or a more theor-
etical approach.
25
20
..
.
0-
~ 15
0-
8
,.
,.'"co
...'""" 10
~
0_5 0.1 0.15 UD 0.25 UD 0.35 040 0.45 0.50 ~ 1.00 1.50 ;,00 ~
DEPTH (~.)
Figure 8-14. Study of chromium-depleted zones on a ferritic steel after cold-rolling.

346 CHAPTER 8
8.8.1. Factors Affecting Sputtering Rate
8.8.1.1. Practical Considerations
a. Measurement of Eroded Depth. In the case of a homogeneous

sampie, calibration of the depth is most frequently obtained by weighing
after sputtering for a definite time or by measuring the crater depth with a
roughness meter or an optical method. To know the sputter behavior of a
layer the same approaches can be used especially in the case of thick films.
Some experimental results are given in the literature.(8,23,40,41) For thin layers
the measurement is more complicated. In the case of oxides a different
approach has been described. (42) Determination of the sputter rates is
obtained by comparison with those of high-grade metals under the same
excitation conditions. The sputter rate of an oxide can be calculated from
the ratio of the measured to the calculated intensity of the oxide when the
sputter rate of the corresponding metal is known. This method is used to
estimate the sputtering rate of oxides and nitrides on steels. In another study,
GD-OES was used to analyze the thin oxide film grown in dry air on the
surface of a chromium-molybdenum low-alloy steel.(42) The thickness
measured through calibration by x-ray interferometry was found to be about
2.5-6.5 nm. In Table 8-4, some experimental results obtained at IRSID are
presented.
b. Influence ofthe Homogeneity ofthe Sampie. It has also been demon-

strated experimentally that the in-depth resolution is limited by the forma-
tion of structures at the bottom of the crater. (32,33,43) These phenomena are
the result of many factors such as the discharge conditions and nature of
the carrier gas. But the most important parameters are the chemical and
metallurgical properties of the sampie. An example is given in Fig. 8-15 to
illustrate the infiuence of the size of grains in the crystal structure of the
layer, in this case a zinc coating on a steel sheet. Sometimes the interpretation
of results is further complicated by these phenomena and by the roughness
Table 8-4. Experimental Sputtering Rate for Some Materials

Sputtering rate
Material (Il m jmin) Material Sputtering rate
Carbon steel 1.7 Brass 6.2
Stainless steel 2.0 Zirealoy 1.7
Nickel alloy 2.3 Vanadium 1.2
Titanium sheet 0.8 Molybdenum 1.2
Aluminum sheet 1.4 Zine 18
"Burning voltage, 800 V; current intensity, 50 mA.
VZZZZZZJ ezaVZJ VZZZZZZI rzm

10 ~m
Figure 8-15. SEM observation of the bottom of a erater for a zine eoating.
created by the sputtering process especially near the interfacial region. The
depth resolution is limited by the fact that signals originating from different
layers may appear simultaneously. These problems do not occur only in
GD-OES but in all methods using an ionic sputtering of the sampies. Some
authors have found that the gradually deteriorating depth resolution can be
described by a dynamic instrument function.(40)
Preferential sputtering can occur when the eroded layer contains inclu-
sions. The physical and chemical properties of these particles are generally
very different from those of the matrix sampie. The sputtering process is
changed and cones are formed on the sampie surface. Such observations
have been made by Jäger and Blum(44) in the case of gold and brass sampies.
Fig. 8-16 gives another example of these problems. It can be seen that the
formation of cones during the sputtering of this zinc coating is caused by
the presence of inclusions of oxides containing silicon and aluminum. These
particles are insulating and more resistant to sputtering than the matrix. The
348 CHAPTER 8
10 11m
Figure 8-16. Formation of cones during the sputtering process (SEM analysis of a zinc
coating).
rem oval of the top of the cones may be achieved by sputtering, but some
data from electron micrographs indicate that the top is sometimes broken
off as soon as the cone reaches a certain height. These phenomena, too, are
not only observed in GD-OES but also with other methods using ionic
sputtering. Greene has reported similar results in another type of
discharge. (45)
c. Influence 01 the Temperature 01 the Sampie. The influence of this

parameter has been discussed for low-melting-point matrices such as zinc,
tin, and lead.(46) In fact, it is weIl known that the sputtering rate increases
rapidly when the temperature reaches approximately 100°C under the
melting point. Different cooling systems have been studied to reduce this
phenomenon. At the moment the best results are obtained with a direct
circulation of water on the back of the sampIe. In Fig. 8-17, the effect of the
application of a cooling system during the analysis of galvanized steel sheets
a ,. h .111,
b
~ ......... ~
/ \
\ F.
\
\..
..... --'-10
.... ,......
......
SPUTTERIIS TlIE (111') SPUlTE.IU TIIE (_I')
Figure 8-17. Analysis ofthe surface ofa galvanized steel sheet. (a) Without a cooling system;
(b) with a cooling system. Vg , burning voltage. Reproduced from Ref. 46.
is illustrated. With this system the results are considerably improved in the
case of thin sheets.
8.8.1.2. Study of the Sputtering Rate Variation with Respect to

Discharge Parameters
The variations of the buming voltage investigated at IRSID for several

materials concem the range from 400 to 1600 V. An example is given in
Fig. 8-18. In this case the current was fixed at 25 mA and the diameter of
the anode was 4 mm. The erosion speed variation (which is proportional to
the sputtering yield), at a constant current, is a linear function of Vg • Several
authors have investigated this dependence and have obtained the same
results.(8,34,47) The curves in Fig. 8-18 prove the existence of an energetic
erosion threshold, an important parameter that surely govems the sputtering
in the field of low energies.
The results presented in Fig. 8-19 were also obtained at IRSID. The
erosion speed variation as a function of I is linear in the range considered
(25 to 200 mA). The working voltage was kept constant (by pressure regula-
tion) at 1000 V at each point.
8.8.1.3. Experimental Results
As discussed previously, the sputtering rate of a given material is often

easy to measure. To quantify the results, it is necessary to know the sputter-
ing yield, which is the average number of eroded atoms per incident ion. In
the field of glow discharge in-depth analysis, there are few purely theoretical
approaches to predict the sputtering yield of a target. In fact, authors
350 CHAPTER 8
11
'2
:€E 11
.=.
>
11
ft
511 1111 1111 Vg (volt)
Figure 8-18. Variation of the sputtering rate as a function of the cathodic fall Vg • Diameter
of the anode, 0.4 cm; intensity of current, 0.025 A. Reproduced from Ref. 51.
•
0
•
0
•
•
'2
:€E
.=. 0
> • ..1,101
0 Oh,.
•
11 111 151
I (ma)
Figure8-19. Variation ofthe sputtering rate as a function ofthe current intensity. Constant-
vo!tage mode (Vg = 1000 V); diameter of the anode, 0.8 cm. Reproduced from Ref. 51.
commonly obtain from the measured eroded mass an equation leading to

experimental sputtering yield, (8,40,47-49) as was discussed in Chapter 2.
The sputtering yield S is given by(8)
S = 1O-6qNejMt+- (8-1)
where q is the sputtering rate (pg/s-I), N the Avogadro number (mole-I),
e the electronic charge (C), M the atomic weight, and i+ the ionic current
(A). The ionic current is related to the total current by
i+ = i/I +Y (8-2)
where y is the average number of secondary electrons emitted by the target

under the impact of a colliding ion. Many authors, especially Boumans, (8)
have not found precise values for y. Thus, in the literature, sputtering yields
are reported in terms of S/I + Y or with a value of y equal to 0.1 for all of
the energies and all materials under argon bombardment. In more recent
work on glow discharge lamps,(50,51) the following empirical relationship has
been used to express the secondary electron emission yield y:(52)
y = 0.032 (0.78Ei - 2qJ) (8-3)
where qJ is the work function of the solid.

The results of these calculations are shown in Table 8-5 for 11 elements.
With these values the experimental sputtering yield can be calculated.
The example of results given in Table 8-5 is obtained with the following
operating parameters: diameter of the anode, 0.4 cm; current intensity,
0.025 A; buming voltage, 1000 V.
8.8.1.4. Theoretical Predictions-Modeling

The approach described below is a summary of the model developed
by L. Ohannessian at IRSID.(51)
a. Prinäple. This target study can be divided into two steps: (1) ener-
getic distribution of colliding ions, atoms, and neutrals and (2) physical
mechanisms of the collision cascades leading to the sputtering. The authors
have assumed a linear electric field in the dark space, computed the length
of this space, and then the distribution of ions and neutrals at the target
Table 8-5. Calculated Value ofthe Secondary Electron Emission Yield for Different Materials
and Experimental Sputtering Yield
Target
AI Ti V Fe Ni Cu Mo Ag Sn Au Pb
r 0.127 0.164 0.152 0.096 0.074 0.111 0.12 0.118 0.11 0.053 0.148
Sexp 0.45 0.27 0.36 0.54 0.74 1.54 0.32 2.49 1.09 1.53 2.44
"From Ref. 51
352 CHAPTER 8
surface. The Sigmund theory offers a basic formalized system for the sputter-
ing phenomena resulting from this bombardment.
b. Calculation 0/ the Length 0/ Cathode Dark Space and the Energetic

Distribution 0/ Ions and Neutral Atoms. The ca1culation of the length of the
cathode dark space L remains within the hypothesis that the electric field
undergoes a linear decrease from the cathode to the border which separates
the dark space from the plasma.
The following equation has been used by Ohannessian:
with L in cm, R the anode radius in cm, I the current in A, Vg the buming
voltage in V, P the pressure in Torr, and T the temperature of carrier gas
in K. For a voltage of 1000 V, apressure of 6 Torr, and a current density
ofO.2 A/cm2 , L is equal to 0.03 cm. Allis et al.(54) mention a value ofthe same
degree of magnitude in the same field of the selected operating parameters.
After having characterized the cathode dark space (area of argon accel-
eration) by its length L, the distribution of ions and neutral atoms can be
ca1culated. In a pioneering work, Davis and Vanderslice(55) measured the
energy distributions of ions bombarding the cathode in an abnormal glow
discharge. Chouan and Collobert(56) offered an exact ca1culation ofthe distri-
bution function in the case of a continuous diode discharge. Ohannessian
has developed her ca1culation from the work of Abril et al. (57) These authors
take the ca1culation back up by considering all of the particles: ions and
neutral atoms. Abril et al. set the following hypothesis :
1. All of the ions come from the region situated between the cathode
fall area and the area of emission of negative glow.
2. The predominant collisions in the dark space are charge transfer
collisions where a fast ion leaves its charge to a rest neutral in order
to provide a rapid neutral and rest ion.
The complete ca1culation is described by Ohannessian.(51) It is necessary

to point out that the distribution of neutral atoms, taking into account the
energy losses through successive impacts in the dark space, is different from
that ca1culated by Abril et al. The amount of neutrals with low energy (some
eV) is substantial; it is assumed that all the mobile neutral atoms will reach
the cathode after a direct path, which is not true for those created too far
from the cathode.
c. Sputtering Yield. 1. Theoretical expression of sputtering yield. From
Sigmund's(53) theoretical expression established in 1969, Matsunami and
Yamamura(58) derived an initial empirical formulation ofthe sputtering yield
of monoatomic targets that takes the sputtering threshold energy into
account. At the same time Yamamura et al.(59) established, by the compila-
tion of experimental results, the following formulation:
where S( W) is the sputtering yield, a *, Q, and Wth are empirical parameters,

Us is the sublimation energy in eV, Sn(e) and Se(e) are reduced stopping
cross sections, either elastic or inelastic, respectively (derived from LSS
theory(60), and e is the reduced energy.
e = 0.03255M2WjZ 1Z2(Zf/ 3 + Z~/3)1/2(Ml + M 2) (8-6)
K= SNjSn=8.478Z1Z 2MJ/(Zf/ 3 + Z~/3)1/2(Ml + M 2) (8-7)
Wth is the threshold voltage:
Therefore, this is the sputtering yield expression that has been used:
It is based on the principle of a colliding mechanism, resulting from

works by Sigmund.
In addition, it takes the sputtering threshold energy into account.
Finally, it results from the recent compilation of experimental studies.
2. Comparison of experimental and predicted results. Energy distribu-

tions of colliding argon species at the target were ca1culated-both ions and
neutral atoms. The formulation by Matsunami and Yamamura(58) is inserted
into each pro gram of distribution ca1culation. Thus, a mean value of the
sputtering yield is obtained under the impact of ions on the one hand and
fast neutrals on the other. The mean value of the number of atoms eroded
because of colliding ions St is finally obtained by taking the contribution of
neutrals into account (the latter being in the region of 50% of the sputtering
yield).
354 CHAPTER 8
Table 8-6. Total Theoretical Sputtering Yie1d for Different Materials

Material
AI Ti V Fe Ni Cu Mo Ag Sn Au
Sion 0.23 0.11 0.15 0.28 0.28 0.53 0.12 0.78 0.28 0.46
Sneutral 0.27 0.12 0.16 0.39 0.38 0.85 0.14 1.52 0.53 0.82
Stotal 0.5 0.23 0.31 0.67 0.66 1.38 0.26 2.3 0.81 1.28
For T = 380K, Vg = 1000V, P = 6 Torr, and J t = 0.2 A/cm2, we come

across the results shown in Table 8-6. The comparison between theoretical
and experimental results is summed up in Table 8-7 and is represented in
Fig.8-20.
Tbe experimental value of S is obtained by integrating the results of
the secondary electron emission yield. It is possible to introduce different
temperature values of the carrier gas in the ca1culation of the theoretical
sputtering yield. It is recalled, in fact, that this parameter is not known
precisely for given discharge conditions (as it is not measured). Thus, for a
set of operating conditions where the voltage and current are known, and
where the pressure of the gas is determined theoretically, we observe the
influence of the temperature on the correlation of experimental and theoreti-
cal results (Table 8-7 and Fig. 8-20). The best correlation is obtained at
445 K (Fig. 8-21). Fig. 8-22 is a different illustration ofthe theoretical varia-
tions of the sputtering yield as a function of the carrier gas temperature.
The importance of this experimental parameter on the sputtering rate is
proven here.
3. Conclusion. From the original work of Ohannessian the results
obtained are summed up by putting them into their analytical context:
I. Tbe role played by the fast neutrals formed by charge exchange in

the dark space and their contribution to the sputtering phenomenon
is proven. The number of neutral atoms striking the target is much
higher than that of ions (probably ten times more), but their energy
is much lower (lower than the threshold energy of sputtering for a
Table 8-7. Comparison between Experimental and Theoretical Results of Sputtering Yield
for Different Values of Temperature
AI Ti V Fe Ni Cu Mo Ag Sn Au
Scxp: 0.45 0.27 0.36 0.54 0.74 1.54 0.32 2.49 1.09 1.53
S'heory 320K 0.41 0.19 0.26 0.55 0.54 1.16 0.22 1.95 0.69 1.05
380K 0.5 0.23 0.31 0.67 0.66 1.38 0.26 2.3 0.81 1.28
445K 0.59 0.28 0.37 0.78 0.76 1.58 0.31 2.63 0.92 1.43
512K 0.69 0.32 0.43 0.91 0.87 1.82 0.37 2.95 1.05 1.6
320 K
512K
.....
oe
...
I-
z:
...""
::E
...
~
><
CI>
1
S THEORETICAL
Figure 8-20. Investigation ofthe correlation between theoretical and experimental sputtering
yields for different values of gas temperature. Reproduced from Ref. 51.
large fraction of atoms). Thus, their contribution to the total sputter-

ing yield is an average of ~50%.
2. A bibliographical investigation of the secondary electron emission
phenomenon carried out elsewhere(51) enables calculation of its yield
and the contribution of electrons to the total current. Using these
data the experimental yield of the sputtering is obtained.
3. A good correlation exists between the experimental and theoretical
sputtering yields for pure materials investigated under given dis-
charge conditions. Therefore, it is possible to employ the realistic
precalculations of the sputtering yield of monoatomic solids.
8.8.2. Concentration Calibration
For a full quantitative analysis, emission intensities have to be converted

into concentrations. In the case of a homogeneous material the conversion
is very easy but in general the problem is more complicated. Fig. 8-23 shows
the spectral intensity-time profile of a multilayer coating prepared by chemi-
cal vapor deposition on a carbon steel. The outer layer consists of vanadium
carbide and the inner layer, chromium carbide. The concentrations of carbon
calculated and measured by other methods are 17 and 8% in the vanadium
356 CHAPTER 8
....
C
...
I-
Z
.........
:.
..,...
S TH EORETICAL
Figure 8-21. Correlation between Sexperimental and Stheoretical. T= 445 K; Vg = 1000 V; J t =

0.2 Ajcm2 • Reproduced from Ref. 51.
and chromium carbide, respective1y. It is clear from this example that content
is not proportional to profile intensity. If the apparent thickness of the
two layers are compared with measurements by SEM, it appears that the
horizontal axis of the spectral intensity-time profile is not proportional to
depth. The same results have been obtained in Japan for two-Iayer Zn-Fe
electroplating. (61)
8.8.2.1. Practical Approaches
It has been demonstrated that the main parameters influencing the

intensities in the constant-current mode are the burning voltage and the
sputtering rate. If the intensities in Fig. 8-23 are corrected point by point
with these two parameters, the carbon content in the different carbides is
quite good. This type of correction is very simple and gives a first estimation
of the concentration of the main constituents. But it is too empirical for
general application.
Different authors have proposed methods to quantify surface pollutants
or enrichments by integrating the spectral intensity. Good correlations have
/
S total
/ " Cu
Au
,/
1.5
D.5
300 400 500 T (I I
Figure 8-22. Theoretical variations of sputtering yield as a function of carrier gas tempera-
ture. Vg = 1000 V; J, = 0.2 A/cm2 • Reproduced from Ref. 51.
( Vanadium carbide )( Chromium carbide Sleel

)(
Fe
." \ ......
"'- ....,..
'-.-.-.-
SPUTTERING TIME
Figure 8-23. Qualitative analysis of a multilayer coating. V g , voltage.
358 CHAPTER 8
been obtained for example in the case of carbon content in surface or enrich-
ment of manganese after annealing. The calibration is realized by chemical
analysis: carbon in surface by the Ford method or combustion in high-
purity oxygen(62) and Mn by ICP-OES.(63) If the sampIe matrix changes,
the associated variations complicate the problem of quantification and less
empirical methods must be used.
8.8.2.2. Theoretical Approach
In recent years, many authors, especially in Japan, have proposed

more or less sophisticated methods for the quantification of in-depth
profiles.(40,61,64-66) These methods are based on the assumption that the emis-
sion yield, Rom (emission intensityjsputtering rate), of a spectralline m of
the element n is independent of the matrix. Several investigations(61,67) have
demonstrated that Rom is practically independent of the matrix composition,
but changes with voltage and current. The method described below has been
developed in IRSID (Institut de la Recherche Siderurgique, France).(43,49,67)
The intensity JA(A) of the spectralline of an element A located at the
surface of the sampIe eroded by cathode sputtering is given by:
(8-10)
where CA is the atomic concentration of element A; K A is the collection

efficiency of the line A(A), which is an instrumental factor depending on the
spectrometer operating parameters; R A(A)is the emission yield of the element
A for the line A, which is independent of the matrix(49) and represents the
number of excited atoms per eroded atoms; r~ is the total sputtering yield
of the matrix (average number of eroded atoms per incident ion); and I is
the current intensity of the lamp.
Equation (8-10) is correct if the sputtering steady state is reached. If
the composition of the matrix changes during analysis (e.g., multilayer coat-
ings) a modification of the electrical properties occurs and the buming volt-
age changes (in constant-current mode). As the emission yield depends on
the voltage and the current, this means that the operating parameters have
to be chosen in such a way that both voltage and current remain essentially
constant during the sputtering time. This can be achieved in constant-current
mode by selecting a buming voltage between 500 and 700 V. Thus, the
discharge conditions will be approximately the same for all the matrices and
RA(A) will be considered as a constant (KAR A = K A ).
An extensive investigation of sputtering yield has been made by

Ohannessian.(SO,SI) At constant current, variations with the voltage and the
composition of the matrix have been indicated. Thus, r is an important
parameter that determines the shape of the profiles. For iron and zinc the
following expression has been established :(67)
rT/v I - Vo = rI/v 2 - Vo (8-11)
where rT (rn is the total sputtering yield for the voltage VI (V2 ) and Vo is
the threshold voltage (Vo ::::; 300 V). If the sputtering yield of the matrix is
approximately constant throughout the analysis, Eq. (8-10) is used direct1y.
An example of this case is the surface analysis of a sampie treated by ion
implantation. The spectral intensity is proportional to the concentration in
depth. If the total quantity of the implanted element is known, it is very easy
to find the concentration at each depth. An example of such an application
is presented in Fig. 8-24 (implantation of phosphorus in an iron matrix).
Ifthe sputtering yield changes with time, Eq. (8-10) can be written for
each element A, B, C ... of the matrix. Thus, it is possible to estimate the
atornic fraction of the elements at any erosion depth and for the area of the
specimen submitted to the discharge, i.e., a few square millimeters, through
the expression:
(8-12)
Practically, the ratio of the coefficient K for each element with respect
to that of one major element is used:
(8-13)
31 P in Fe
e
-= 5
o 20 60 100 140 180 220 2&0 300 340 Deplh( n m)

Figure 8-24. Concentration profile of implanted phosphorus in iron. 31p in Fe: 125 keV,
5 x 10 14 P/cm 2 • Reproduced from Ref. 43.
360 CHAPTER 8
It is sufficient to know the ratios K B / K A , K B / K c , .... They may be deter-

mined with the aid of an analysis with the same operating parameters on
binary system B-A, B-C, ... of known composition or any kind of system
containing B-A, B-C, and other elements.
a. Application to the Study 0/ a Zinc Coating. This example has been

chosen to illustrate the different shapes of the zinc profiles as a function of
the discharge conditions. The sampie is a zinc-electroplated steel sheet. Fig-
ures 8-25, 8-26, and 8-27 illustrate the different types of profiles obtained for
zinc as a function of the operating conditions (burning voltage and current
intensity).
In the first example (Fig. 8-25) the discharge parameters are approxi-
mately equivalent in the coating and in the steel (Vg ~ 500 V), so that the
variations of the emission yield are minimal. The zinc profile decreases
almost regularly from the coating to the substrate with a little "shoulder"
in the interface region. The operating parameters are:
• 1= 50 mA (constant-current mode)
• Diameter of the anode = 7 mm
• Vg varies from 455 V in the zinc coating to 535 V in the steel
The light intensity of the zinc line (JZO) decreases rapidly when the
signal ofiron (]Fe) appears. The shoulder in the interface region is observed
c::o
~
FE
...
c::o ZN
:; ""
~
....
>-
....... ...
e;; c::o
z "'I
!!:
....
:z:
CD
::::;
...
c::o
"'!
V,
o 30 60 90 120 150 180 210 2'0 270 300

Figure 8-25. Analysis of a zine eoating in a eonstant-eurrent mode. 1= 50 mA; Vg ~ 500 V
in the steel. Reproduced from Ref. 43.
~r-----------------------------------------,
=
'"
"'"
...
~
ZN
>-
!::
... :ä=
CI>
z
0-
FE
:!
0-
:z: VG
~
.....
=
"'!
N
o 48 64 80 96 112 128 144 160

SPUTTERING TIME (SEC)
Figure 8-26. Analysis of a zine eoating in a eonstant-eurrent mode. 1= 50 mA; Vg ~ 1000 V

in the steel. Reprodueed from Ref. 43.
=
e
~
;;; ZN
FE
~
>-
0-
;;; =
...
z
0-
'"
"'"
!:
0-
:z:
=
CD
::;
~
Va
=
~
la
30 60 90 120 150 180 210 240 270 300

SPUTTERING TIME lsec)
Figure 8-27. Analysis of a zine eoating in a eonstant-voltage mode. I g , eurrent intensity;
Vg ~ 750 V. Reproduced from Ref. 43.
362 CHAPTER 8
for a rather long time of erosion. The zinc content is calculated from Eq.
(8-13) at each depth taking into account only zinc and iron in the composi-
tion of the coating. In Fig.8-28, it is observed that the concentration profile
of zinc decreases regularly. Because of the roughness of the bottom of the
crater, the concentration profile spreads largely.
If land K Zn are considered as constant, the variation of r T as a function
of the depth may be described by:
(a is determined with a pure zinc sampie) and corrected from the variation
of the voltage with the aid of Eq. (8-11) (Vo ~ 300 V). Thus, the variation
of rT is calculated for a constant voltage (curve rJ). It can be seen (Fig.
8-28) that the sputtering yield is then divided by 2 when the zinc concentra-
tion changes from 100% to 80%. Ifthe zinc amount is then divided by 2, r T
is divided by 3.8. Blaise has also found that the sputtering yield varies very
quickly for two-phase systems. (68)
The important slope of the zinc profile is the result of the change in the
sputtering yield which changes with the zinc content. In the interface region
(at constant current) the voltage increases between the coating and the steel,
eIn ."
. 144 112
SPUTTERING TIME (secl
240 211
Figure 8-28. Variation of the sputtering yield as a function of zine eontent. yT, sputtering
yield;y;,sputtering yield calculated for a eonstant voltage; CZn , zine eontent; /zn, intensity
of the zine line. Reproduced from Ref. 43.
thus the variation is less important and it can be observed that a smaH
shoulder appears on the profile ofzinc (Jzn) as so on as the voltage increases.
Ifthe constant-current mode is used (with a higher voltage), the varia-
tion of this last parameter increases between the coating and the substrate.
Thus, yT decreases more slowly and the shoulder in the interface region is
more important. The shape of the profile JZn looks like the curve recorded
for a two-Iayer coating (Fig. 8-26). FinaHy, if a constant-voltage mode is
selected (Vg = 750V in Fig. 8-27), the current intensity decreases in the inter-
mediate zone between zinc and iron. yT and JZn thus decrease more rapidly
in this region and the shoulder observed in the first example disappears.
It is necessary to mention that, if the discharge parameters are very
different for the two materials, the variations of the emission yield can also
lead to another modification of the shape of the profile.
b. Study 0/ the Enrichments in a Passive Layer. Quantitative analysis

in the case of very thin films is much more difficult because of the changes
in burning voltage at the beginning of the discharge. If the emission yield is
considered as a constant during the analysis, the atomic fractions of the
different elements in the matrix can be calculated from expression (8-13).
Thus, it is possible to compare the evolution of the ratio of the element (A)
to another one (B) between the substrate (S) and the superficiallayer (L):
(8-14)
This determination can be calculated without calculating the ratio KAI K B •

In fact, it is possible to write:
(8-15)
Such an approach has been applied to the determination of the enrichment

of chromium and silicon in passive layers on stainless steels. The results
obtained by this method and by the determination of the atomic fractions
are very similar. The differences between the two modes of quantitation are
less than 10%. (69)
As discussed earlier, the conversion of erosion times into depths is easy
for a thick layer or a homogeneous material but more difficult in the interface
region and for thin films where the burning voltage is not very weH defined.
To establish a general method of quantitative in-depth profiling (concentra-
tions and depths), more fundamental work remains to be done. Specific
approaches are now frequently proposed.(40,41,43,49,64-67,70)
364 CHAPTER 8
8.9. Main Applicafions in Surface Analysis
Compared with the main techniques for surface analysis and in-depth
profiling, GD-OES is fast and relatively easy to use. This is why these systems
are used not only for research and development, but also for quality control.
The primary applications are in the metallurgical, automotive, and nuc1ear
industry fields. The list of applications is very long, so we give only a few of
the technically most important works:
• Surface states after pickling, degreasing, and annealing(21,23,31,35,71)

• Oxide scales on hot-rolled steel(23,32,40)
• . Iayers, d epIeted ch
PassIve '
romlUm zones, and ano d'IC OXl'des(4271-74)
'
• Corrosion products(75)
• Transfer layers(76)
• Chromate films(77,78)
• Zinc and zinc alloy coatings on steel(31,40,61,65-67,79,80)
• PVD and CVD coatings(35,76)
• Ionic implantation profiles(35,40,76)
• Surface analysis of titanium(70)
It is not possible to describe all of them so we have chosen to summarize

in the following section some examples in the field of coatings.
8.9.1. Analysis of Coatings and Surface Treatment by GD-OES
In this type of application, the thicknesses mostly concemed vary from

1 to 150 pm: glow discharge spectrometry is then practically the only method
suitable for the study and control of the layers formed. GD-OES allows the
continuous analysis of the first 150 pm within a reasonable time; there is no
loss of information as may happen with the techniques that carry out analy-
ses of a very superficial layer altemated with ion or chemical abrasions
(discontinuous profiles). As for thicker layers (up to 300 pm), the sampie
may be recentered after elimination of the deposits on the edge of the crater
which cause anode-cathode short circuits.
8.9.1.1. Hot Nitriding Treatment on Tool Steels
The quantitative evaluation of the concentrations of the modifying ele-

ments has been made by means of bulk test sampies such as oxides, nitrides,
and carbides. The sputtering rate and the voltage being practically identical
in iron and in the iron nitrides analyzed, this estimation is fairly good in the
absence of a compact oxide layer. In the case of nitriding in "tenifer," the
values are certainly much less precise because of the high oxygen content in
11
N (\)
,I
.i
ÜC··.
....... ~ '\
\'\ " (2)
3 @: \Ci)\
....... \,.::..~~:_ .
•••••••••••••• •.~
•••:-:-:-
•••:7..,.::-::
•••::7 =...=.. =
•..:-:-:-
•.• ...=... 7:7:
••••
I+---~~---r----,-----r---~----,-+
o 2 4
SPUTTERING TIME (Illn)
Figure 8-29. Nitrogen concentration profiles obtained with various tool steel surface treat-
ments. (I) "Tenifer" (molten salts); (2) gaseous nitridation; (3) ionic nitridation (E); (4) ionic
nitridation (r') Reproduced from Ref. 81.
the combination layer. In all cases the quantitative estimation is very good
for the zone of diffusion (evolution of the profile of one element in a
homogeneous material). The tracings of concentration profiles of nitrogen,
oxygen, and carbon are shown in Figs.8-29, 8-30, and 8-31, respectively.
a. Nitrogen Concentration Profile. (Fig. 8-29). Taking into account the

calibrations carried out, the maximum nitrogen content in the case of "ten-
ifer" and of gaseous nitriding seems to exceed 10 %, which is quite reasonable
o ('f, )
20
16
12
CD
Figure 8-30. Oxygen concentration pro- 4

files obtained with various tool steel surface
treatments. (I) "Tenifer" (molten salts);
(2) gaseous nitridation; (3) ionic nitrida-
tion (E); (4) ionic nitridation (r') Repro- 2 3 4
duced from Ref. 81. SPUTTERING TIME (Mln)
366 CHAPTER 8
I
\
\
C (\1 \.
\
\.
'. @
'.'.
0,5 '.'.
'.'.
'-.
' .......... -._.-.-
\
,, , ---- ,-
@.............. ~ :::~.:::::::.:-..~. ~ ::.:. ~ ~.:. ................. '

O~----~--~----~----~----~----~+
o 2 4 5 6
SPUTTERING TIME (min 1
Figure 8-31. Carbon concentration profiles obtained with various tool steel surface treat-
ments. (I) "Tenifer" (molten salts); (2) gaseous nitridation; (3) ionic nitridation (c); (4) ionic
nitridation (r'). Reproduced from Ref. 81.
for obtaining [; nitrides. On the contrary, in the case of the sequenced treat-
ment of ionic nitriding, the carbonitride [; contains about 8 % nitrogen,
which is logical taking into account the presence of carbon. The level
observed in that case in the underlayers, is about 5% nitrogen, corresponds
in all probability to the r' nitride formed during the first sequence of the
treatment.
b. Oxygen Concentration Profile. (Fig. 8-30). If one makes exception

of a very slight superficial contamination, the two ionic nitriding treatments
are characterized by the absence of oxygen. In the case of gaseous nitriding
and above all the "tenifer," we observe a very large superficial enrichment
in oxygen which relates to an approximate thickness of 8 pm for the "ten-
ifer" and of 5 to 6 pm for the gaseous nitriding. In parallel with this phenom-
enon, and over the same thicknesses starting from the surface, an increase
is noted of the content in oxidizable alloy elements such as chromium and
vanadium.
c. Carbon Concentration Profile. (Fig. 8-31). lonic nitriding with a

treatment sequenced to obtain an [; configuration presents a carbon profile
that rises up to more than 2% on the surface, which corresponds in the iron-
nitrogen-carbon diagram to the composition of the carbonitride [; with 8%
nitrogen. lonic nitriding with the production of r' nitride, if superficial
contamination is excepted, brings about a large decarburization over a depth
estimated at more than 50 pm.
Nitriding in a "tenifer" salt bath could eventually bring about a super-
ficial enrichment in carbon, but the profile of this element is very strongly
disturbed on the surface by the presence of oxygen (see Fig.8-30). Gaseous
nitriding under an ammonia atmosphere also brings about a decarburization
over a depth estimated at 40 J.1m, and the carbon profile on the surface is
disturbed by the presence of oxygen (see Fig.8-30).
These investigations have made it possible to explain differences of
behavior due to friction and wear in a steel Z38 CDV05 treated by the three
methods. (81)
8.9.2. lonic Implantation

The superficiallayer concemed in this treatment is much thinner than
in the previous example: most of the time it is less than 150 nm with the
maximum concentration of the implanted element ranging from 30 to 60 nm.
During the last few years, many studies have demonstrated the improvement
of the resistance to wear by ionic implantation of several materials. Different
ion species have been tried (e.g., C, B, N, P, V, Ti, Co, Sn, Ni, Cr) with
greater or lesser success. Nevertheless, it is nitrogen that is most often used.
Two remarkable facts have been observed:
1. A notable improvement of the resistance to wear

2. A persistence of this effect on depths of wear very much deeper than
the implanted zones (up to 100 times more)
GD-OES is perfect1y suitable for the study and the control ofthe process
(Fig.8-32). It makes it possible to track down a large number of elements:
• Nitrogen (quantity implanted and depth involved)

• Oxygen and carbon: checking of the superficial pollution layer which
may prevent the penetration of the nitrogen into the material.
• Pollutants (Ca, Si, Al, B) brought by the initial preparation of the
surface (e.g., machining, polishing under water). Control of these
elements, for which the sensitivity is high with GD-OES, makes it
possible to complete the interpretation of the phenomena: informa-
tion on the porosity of the superficiallayer and showing up as "ther-
mal" effects or as erosion by sputtering by comparing the composition
of the surface before and after treatments.
On a more fundamental plane, one can see that the results show an
excellent correlation with those obtained by the nuclear method using the
368 CHAPTER 8

Figure 8-32. lonic implantation of nitrogen in steel. 15N: 40 keV, 5 x 10 17 N/cm2 .
resonant reaction 15N(P, ay) 12C (for the analysis ofN) which is nondestruc-
tive but much more difficult to use (Fig.8-33).
8.10 Limits and Advantages of GD-DES
8. 10. 1. Limits
The main negative properties of GD-OES are the following:
1. As for all the existing methods using ionic sputtering for in-depth
proilling (e.g., AES, ESCA, SIMS), direct measurement of the
eroded depth is not possible (see Section 8.8.1). The roughness of
the surface increases during the course of sputtering and the shape
of the bottom of the crater depends on the operating parameters and
on the structure of the metals. This fact limits the depth resolution
in the interface region especially if the studied layer is thick.
2. GD-OES has no lateral resolution. The eroded area commonly has
a diameter of 0.4 or 0.8 cm.
3. The method does not give any direct information on chemical bonds.
But the ratio between the intensity of different elements permits
a b
'0000
5Y. /.":::\\
\
.
....
souo :
. I"':.
.
• ...•
....
...
.... ~ to 000
3
/ ....\
":
\\
e:;
!:! ...
D f \~,
!!
.:::= l
/
2
\ \1"
,:.l..\
,
.1000 { \
~ { \.~~
t
~,
\"~
..
r ~.
uo ,so
SPUTTUIIB "aE (S .. , EUI'Y (I .. ,
Figure 8-33. Comparison between GD-OES and nuc1ear method-ionic implantation of N.

15N/cm2: I = 10 16 ; 2= 5 X 10 16 ; 3= 10 17 ; 4=2 X 10 17 ; 5=4 X 10 17 ; 6=6 X 10 17 • (a) GD-OES; (b)
nuc1ear method. Reproduced from Ref. 76.
drawing rapidly some conclusions on these bonds. Quantitative ana-

lysis may be difficult because the analysis is made at nonequilibrium
conditions.
4. At the moment the sampie must be electrically conductive or at
least semiconductive. In arecent work, Duckworth and Marcus(82)
introduced a new discharge tube using radio frequencies for the
analysis of nonconductive materials.
8.10.2. Advantages
The main properties of GD-OES are as follows:
1. Almost all elements can be determined (0, N, and H in particular).

But with argon as a carrier gas the sensitivity is poor for some of
them, e.g., F, Cl, and 1.
2. High vacuum is not required, so the technique is fast and rather easy
to use.
3. Quantitative analysis in a homogeneous matrix is very easy because
there are few matrix effects in the plasma and because calibration
curves are commonly linear.
370 CHAPTER 8
4. The initial and running costs are low relative to other surface
methods.
5. The main advantage for industrial products is the ability to draw
profiles from a few nanometers to several tens of micrometers in a
short time.
8. 11. Future Trends
Considerable progress has been made in recent years with the develop-
ment of commercial spectrometers with a larger spectral range and suitable
computers. GD-OES is now used either in research centers or direct1y in the
plants for quality control (especially in Japan). In many aspects,
improvements are possible. First it is necessary to point to the fact that the
lamp in all commercial spectrometers is very similar to the Grimm tube. In
the near future, new sources will appear, e.g., regulation ofthe flow of carrier
gas by a mass-flowmeter (l and Vg constant) and the use of microwaves or
rf lamps to increase sensitivity or to analyze nonconductive materials. The
field of GD-OES will become larger in materials study. Because the speci-
ficity of the technique is the glow discharge tube, improvements in the lamp
will be the most important changes. Other carrier gases may be used to
improve the sensitivity. Progress is also needed in optical systems to increase
the spectral range with the use of combined spectrometers (monochromator
and polychromator). Two types of systems will be developed: a sophisticated
spectrometer for research and a simplified apparatus for quality contral of
specific surfaces (e.g., coatings).
Compared with other techniques, the matrix effects in GD-OES are
relatively simple, and quantitative in-depth profiling is thus likely to develop
in the near future. Different mathematical models will emerge either for
specific applications or for a more general approach. The number of
GD-OES systems in the production plants will increase especially in the
metallurgical and automotive industries.
References
1 W. Grimm, Spectrochim. Acta 23B (1968) 443.

2 M. Dogan, K. Laqua, and H. Massman, Spectrochim. Acta 26B (1971) 631; 27B (1972)
65.
3. H. de La Follie, 8th Spektrometertagung, Linz, 1970.
4. W. Grimm, 8th Spektrometertagung, Linz, 1970.
5. H. Jäger and R. P. Butler, 16th eSI, Heidelberg, 1971, Preprints, Vol. 2, p. 204, Adam
Hilger, London, 1971.
6. M. E. Ropert, 16th CSI, Heidelberg, 1971, Preprints, Vol. 2, p. 214, Adam Hilger, London,
1971.
7. M. Dogan, Ph.D. thesis, University of Munster, 1970.
9. R. Berneron, J. Cretin, and J. P. Moreau, Report IRSID RI 320, 1972.
10. M. E. Pillow, Spectrochim. Acta 36B (1981) 821.
11. A. I. Drobyshev and Y. I. Turkin, Spectrochim. Acta 36B (1981) 1153.
12. D. S. Gough and J. V. Sullivan, Analyst 103 (1978) 888.
13. N. P. Ferreira, J. A. Strauss, and H. G. Human, Spectrochim. Acta 38B (1983) 899.
14. R. A. Kruger, B. M. Bombelka, and K. Laqua, Spectrochim. Acta 35B (1980) 581.
15. M. Ancey, R. Berneron, and P. Pamiere, R.C. Met. Phys. (1971) 582.
16. C. J. Belle and J. D. Johnson, Appl. Spectrosc. 27 (1973) 118.
17. J. E. Greene and J. M. Whelan, J. Appl. Phys. 44 (1973) 2509.
18. J. E. Greene, F. Sequeda Osorio, and B. G. Streetman, Appl. Phys. Let!. 25 (1974) 435.
19. J. E. Greene, F. Sequeda Osorio, and B. R. Natarajan, J. Vac. Sci. Technol. 12 (1975) 366.
20. H. Schneider and H. Shuman, Kernforschungszentrurn Karlsruhe, Report KFK 2009, 1974.
21. R. Berneron and J. P. Moreau, 18th CSI, Grenoble, 1975, Preprints, Vol. 1, p. 263, GAMS,
Paris, 1975.
22. R. Berneron and J. P. Moreau, CECA EUR 6279 FR, 1978.
23. R. Berneron, Spectrochim. Acta 33B (1978) 665.
24. M. Laporte, Decharge electrique dans les gaz, Arrnand Colin, Paris, 1948.
25. F. M. Penning, Electrical Discharges in Gases, p. 41, Philips Technical Library, Eindhoven,
1957.
26. J. A. C. Broekaert, J. Anal. At. Spectrom. 2 (1987) 537
27. R. Plesch, Siemens Analysentechnische Mitteilungen N2, p. 292.
28. A. Bengtson, L.Danielson, and A. Nordgren, 23rd CSI, Arnsterdam, 1983.
29. D. A. Skoog and D. D. West, Principles 0/ Instrumental Analysis, Saunders College, Phila-
delphia, 1980.
30. K. Laqua, W. D. Hagenah, and H. Waechter, Z. Anal. Chem. 225 (1967) 142.
31. A. Bengtson and L. Danielson, 6th International Conference on Thin Films, Stockholm,
1984.
32. K. Laqua, A. Quentmeier, and D. Demeny, Proceedings of 10th Jubilee National Confer-
ence on Atomic Spectroscopy, Veliko Turnovo, 1982.
33. J. Pontet and H. Hocquaux, Actes des journees de la Commission Decharge lurninescente
(now 1988), GAMS, Paris, 1988.
34. M. Bouchacourt, Actes des journees de la Commission Decharge Lurninescente, GAMS,
Paris, 1988.
35. H. Hocquaux, Met. Corros. Ind. 58 (1983) 693.
36. H. Hocquaux, J. P. Colin, et al., Siemens Colloquiurn "Caracterisations de surface par
SDL," Paris, 1984, unpublished work.
37. H. Hocquaux and P. Hunault, 1989 Pittsburgh Conference, Paper No. 686, unpublished
work.
40. A. Bengtson and M. Lundholm, J. Anal. At. Spectrom. 3 (1988) 879.
41. K. H. Koch, M. Kretschmer, and D. Grunenberg, Mikrochim. Acta (Wien) 11 (1983) 225.
42. A. M. Brass, L. Nevot, M. Aucouturier, and R. Berneron, Corros. Sci. 24(1) (1984) 49.
43. J. C. Charbonnier, H. Hocquaux, J. P. Moreau, and D. Nizery, Mesucora 88, GAMS,
Paris, 1988.
44. H. Jäger and F. Blurn, Spectrochim. Acta 29B (1974) 73.
45. J. E. Greene, J. Vac. Sei. Technol. 15 (1978) 1718.
372 CHAPTER 8
46. R. Berneron, B. Chetreff, J. P. Colin, P. de Ge1is, H. Hocquaux, F. Hoffert, and S. Mathieu,

24th CSI, Garmisch Partenkirchen, 1985, unpublished work.
47. J. P. Moreau, CNAM thesis, Paris, 1975.
48. N. Laegreid and G. V. Wehner, J. Appl. Phys. 32 (1961) 365.
49. J. Pons-Corbeau, Surf Interface Anal. 7 (1985) 169.
50. H. Hocquaux, L. Ohannessian, and J. Tousset, 24th CSI, Garmisch Partenkirchen, 1985,
unpublished work.
51. L. Ohannessian, Ph.D. thesis, Lyon, 1986.
52. R. A. Bargolia, E. V. Alonso, J. Ferron, and A. Oliva Florio, Surf Sei. 90 (1979) 240.
53. P. Sigmund, Phys. Rev. 184 (1969) 183.
54. W. P. Allis, G. Fournier, and D. Pigache, S. Phys. (Paris) 38 (1977) 915.
55. W. D. Davis and T. A. Vanderslice, Phys. Rev. 131 (1963) 219.
56. Y. Chouan and D. CoJlobert, S. Phys. (Paris) 43 (1977) 915.
57. I. Abril, A. Gras-Marti, and J. A. VaJles-Abarca, Phys. Rev. 28 (1983) 3677.
58. N. Matsunami and Y. Yamamura, At. Data Nucl. Data Tables 31 (1984) 2.
59. Y. Yamamura, N. Matsunami, and N. Itoh, Radiat. Eff. 71 (1983) 65.
60. J. Lindhard, N. S. Scharff, and H. E. Schiott, Mat. Fys. Medd. Dan. Vid. Sellsk. 33 (1963)
14.
61. K. Takimoto, K. Nishizaka, K. Suzuki, and T. Ohtsubo, Nippon Steel Tech. Rep. 33 (1987)
28.
62. W. Kreschmer, K. H. Koch, and D. Grunenberg, Stahl Eisen 104 (1984) 193.
63. P. Netter, Report IRSID RI 1004, 1985.
64. K. H. Koch, D. Sommer, and D. Grunenberg, Mikrochim. Acta Suppl. 11 (1985) 137.
65. Y. Ohashi and Y. Konushi, Tetsu To Hagane 69(12) (1983) 1052.
66. K. Suzuki, K. Nishizaka, and T. Ohtsubo, Tetsu To Hagane 70(4) (1984) 295.
67. J. Pons-Corbeau, J. P. Cazet, J. P. Moreau, R. Berneron, and J. C. Charbonnier, Surf
Interface Anal. 9 (1986) 21.
68. G. Blaise, Colloquium on Sputtering, Les Ares, France, 1985.
69. J. Pons-Corbeau, Report IRSID SCA 85/205, unpublished work.
70. K. Suzuki, T. Ohtsubo, and T. Watanabe, Nippon Steel Tech. Rep. 33 (1987) 36.
71. R. Berneron and H. Hocquaux, Methodes Usuelles de Caracterisation des Surfaces, p.91,
EyroJles, Paris, 1988.
72. R. Berneron, J. C. Charbonnier, R. Namdar-Irani, and J. Manenc, Corros. Sei. 20 (1980)
899 .
. 73. B. Alexandre, R. Berneron, J. C. Charbonnier, and J. Manenc, Surf Technol.9 (1979) 317.
74. F. Gaillard, M. Romand, H. Hocquaux, and J. S. Salomon, Surf Interface Anal. 10 (1987)
163.
75. M. Dernat, H. Hocquaux, and L. Ohannessian, L'eau, l'industrie, les nuisances 86 (1984)
44.
76. H. Hocquaux and R. Leveque, Galvano-Organo:Traitements de Surface 750 (1984) 837.
77. K. Suzuki, S. Yamazaki, and T. Ohtsubo, Tetsu To Hagane 3 (1987) 565.
78. A. Quentmeier, H. Bubert, R. P. H. Garten, H. J. Heinen, H. Puderbach, and S. Storp,
Mikrochim. Acta Suppl. 11 (1985) 89.
79. M. Iwai, M. Terada, H. Sakai, and S. Nomura, Tetsu To Hagane 11 (1986) 1759.
80. Y. Matsumoto, N. Fujino, and S. Tsuchiya, Trans. ISIJ Jpn. I (1987) 891.
81. R. Leveque, H. Michel, and M. Gantois, International CoJloquium on Tool Steels, Saint-
Etienne, France, 1977, unpublished work.
9
Discharges within Graphite
Furnace Atomizers
James M. Harn/y, David L. Styris, and Philip G. Rigby
9.1. Introduction
In recent years, the concept of using electrical discharges within graphite

furnace atomizers as atomizationjexcitation sources for atomic emission
spectrometry has been established and evaluated.<1-31) The excitation in these
sources is principally dependent on the discharge and is not a direct result
of the electrothermal heating of the furnace. Consequently, this unique
concept has been designated fumace atomization nonthermal excitation
spectrometry (FANES). This can lead to some confusion since "non-
thermal" also implies that the line radiance and profile associated with the
discharge do not conform to Maxwell-Boltzmann statistics. (32) While it is
certain that the discharge is necessary for excitation, there is considerable
uncertainty whether all the discharges described in this chapter can be gen-
erally categorized as nonthermal. We shall nevertheless continue to use
FANES to describe the concept, bearing in mind the above ambiguity.
To date, direct current (dc) and radio frequency (rf) discharges have
been examined in this source. In both cases the analyte is thermally vaporized
James M. Harnly • United States Department of Agriculture, Agricultural Research Ser-

vice, Beltsville Human Nutrition Research Center, Nutrient Composition Laboratory,
Beltsville, Maryland 20705. David L. Styris • Battelle, Pacific Northwest Laboratory,
Richland, Washington 99352. Philip G. Rigby • School of Biological and Chemical
Sciences, University of Greenwich, Woolwich, United Kingdom.
373
374 CHAPTER 9
into the discharge region, so the volatilization/atomization processes associ-

ated with the graphite furnace are decoupled from the excitation process in
the discharge.
The graphite furnace atomizer is well suited to this dual function con-
cept. It is a very efficient atomization source for trace metal determinations.
For most commercial furnaces, aqueous sampies (1 to 100 pI) are deposited,
dried, and then vaporized into a semirestricted furnace volume of less than
1 cm3 to produce a highly concentrated atom eloud. The transport efficiency
of analyte species into the analytical zone is elose to 100% for the graphite
furnace as opposed to 1-5% for other conventional atomizers (e.g.,
air-acetylene flame or inductively coupled plasma).(5) Residence times of
analyte atoms in the analytical zone in the furnace are approximately two
orders of magnitude greater than those found in other atomizers.(5) Con-
sequently, gas-phase analyte densities are orders ofmagnitude more concen-
trated in the furnace. The furnace also possesses an inherent capability of
providing matrix modifications in situ, since chemical and thermal pretreat-
ment of the sampie can be incorporated into the heating program.
Historically, the graphite furnace (GF) has been used primarily as an
atom reservoir for atomic absorption spectrometry (GF-AAS). The dimen-
sions of the furnace provide the longer optical path suitable for absorption
measurements. The furnace can reach 3300 K, although routine operation
is usually restricted to maxima of about 3000 K to enhance furnace life. At
these temperatures, atomization is achieved for most molecular species, and
thermal ionization is negligible.(33) GF-AAS exhibits excellent detection
limits (from tens to tenths of picograms)(34) but the technique lacks the
multielement capabilities inherent in other approaches such as inductively
coupled plasma-atomic emission and mass spectrometries (lCP-AES and
ICP-MS).
The graphite furnace has also been employed as an emission source
(GF_AES)(35,36) in an attempt to use the inherent multielement advantage of
the simplified optics of the emission mode and the available technology
of multichannel spectrometers. Effective excitation, however, can only be
achieved for transitions above 300 nm, i.e., those elements having excitation
energies less than the energies associated with the high-energy tail (-2 eV)
of the Maxwell-Boltzmann energy distribution of the thermionic electrons
or the Planck distribution for thermal radiation at temperatures near 3000 K.
Hence, for elements having ground state transitions at wavelengths less than
300 nm, the GF-AES detection limits are much poorer than those for GF-
AAS.
The F ANES concept extends the excitation energy range of the graphite
furnace to elements with transitions below 300 nm and thus gives the furnace
true multielement attributes. This concept was first proposed by Falk(I,2)
who used a low-pressure, glow discharge with the furnace as a hollow
D/SCHARGES W/TH/N GF ATOM/ZERS 375
3~-----------------------------------,
2 0
0
0 0
$ 0
0
~1 0 0
t
0
:::i(!) ~o
c-
o(/) 0 ()
i
-.:;W
uz 0 0
~<
Q)LJ..
0-
0 0
0 ... o
0> 0
0 -1
= ...
0
o
-2 0 o o
100 200 300 400 500 600 700 800

Wavelength (nm)
Figure 9-1. FANES detection limits taken from Table 9-1, normalized with respect to conven-
tional GF-AAS detection limits, as a function ofwavelength. (0) Falk et al., HC-FANES(15);
(A.) Naumann et al., HC-FANES(17); (0) Sturgeon et al., rf_FANES(27); (.) Hamly et al.,
HA-FANES.(25)
cathode (HC) and a point or a ring anode located external to the furnace;
the discharge, in Ar or He, filled the furnace volume. This source employs
operating parameters (1-25 Torr, 10-100 mA) and exhibits performance
characteristics similar to classical HC lamps. As shown in Fig.9-1, the
reported detection limits are, on the average, a factor of three lower than
those for GF-AAS.(5) We refer to this source as HC-FANES in future discus-
sions in this chapter in order to differentiate from later source developments.
Ballou et a/.,(I8 Harnly et a/.(25) and Riby et a/..(29) used a somewhat
different design concept and reversed the relative polarities associated with
HC-FANES. The furnace comprised the anode and the cathode was a central
graphite rod running the length of the furnace. The discharge in this design
is a bright corona constricted around the cathode. Operating pressures
(5-200 Torr in Ar and 10-600 Torr in He) are considerably higher than
those normally expected for dc glow discharges. The ability to maintain a
stable plasma at these pressures may result from the smaller cathode areas
associated with this design and, therefore, higher current densities. This type
ofhollow anode (HA) system is designated as HA-FANES in the discussions
that follow. The detection limits compare weIl with HC-FANES and conven-
tional GF-AAS (Fig. 9-1).
376 CHAPTER 9
The most recent developments in this field are the rf glow discharge
sources described by Blades and co-workers(20,24) and by Sturgeon and co-
workersYI,27,28,30,31) These sources have the same geometry as those used in
HA-FANES, but are operated in He at atmospheric pressure using rf
generators as power sources. The detection limits compare weIl with HA-
and HC-FANES and conventional GF-AAS. The spectrometric technique
that uses these sources has been designated furnace atomization plasma
excitation spectrometry (FAPES). For consistency in nomenclature in this
chapter, these systems will be referred to as rf-FANES.
The excitation sources that have just been described, HC-, HA-, and
rf-FANES, employ electrical discharges operated under conditions markedly
different from each other. In addition, all three sources are markedly
different from classical glow discharge conditions. First, all three F ANES
discharges have been optimized for the excitation process only, not for sput-
tering, etching, or plating. SampIe volatilization is accomplished thermally
and the rate of volatilization can be as much as three orders of magnitude
greater than sputtering rates. The sputtering process remains a significant
process, but as an interfering reaction leading to premature appearance and
loss of analyte. Second, all three F ANES discharges involve "hot" electrodes
having temperatures ranging from 800 to 2800 K, depending on the F ANES
source and the element being determined. Above 1800 K the generation of
thermionic electrons becomes important and can significandy reduce the
discharge potential. Discharges under these conditions have been referred to
as low-voltage arcs.(44-47) FinaIly, HA- and rf-FANES sources are operated
at considerably higher pressures than those used for most classical glow
discharge systems. These higher pressures result in longer residence times
for the analyte.
It is the purpose of this chapter to instill an awareness of and to provide
some physical perspective on this new and growing development in glow
discharge spectrometry and to share with the reader the present under-
standing associated with the various FANES sourees. To achieve this the
experimental results obtained for the three systems described above will
be reviewed. Comparisons will be made of their analytical attributes, their
geometries, and the nature of their discharges, and finally physical inter-
pretations will be presented and discussed.
9.2. System Designs
An overview of the physical designs (Table 9-1) and operating

parameters (Table 9-2) of HA-, HC-, and rf-FANES is presented in this
section. In Table 9-1 rf-FANES has been further partitioned according to
the designs of Blades et al. (20,24) and of Sturgeon et al. (21,27,28,30,31). In
I:)
Ci;
(")
~
~
Cl
~
Table 9-1. Furnace Designs Used for FANES ~
~
rf-FANES ~
HC-FANES HA-FANES Blades Sturgeon ~
:t-
Furnace el
Manufacturer Carl Zeiss Laboratory built Thenno Jarrell-Ash Perkin-Elmer ~
Type of furnace Conventional Integrated contact cuvette Conventional Conventional ~
(Ringsdorff-werke) ~
Dimensions· 28 x 6.5mm 19 x 6.5mm 28 x 6.5 mm 28 x 6.5mm
Electrical contacts Longitudinal Lateral Longitudinal Longitudinal
Support gas control Cont. pump Static Cont. flow Cont. flow
Furnace enc10sure Vacuum seal Vacuum seal Vacuum seal Positive pressure
Plasma
Power source dc dc rf rf
Cathode Furnace Axial rod Axial rod Axial rod
Anode External ring Furnace Furnace Furnace
Pressure 1-140 Torr (Ar, He) 70-200 Torr (Ar) 760 Torr (He) 760 Torr (He)
100-600 Torr (He)
aLength x diameter.
(,.)
::j
(,.)
Table 9-2. Operating Conditions

HC rf
HA
F" N D L H B S
Temperatures CC) 1850-2350 1845 2000 1100-2700
Heating rate CC S-I) 2000 2000 2000 2000 1100 1800 2000
Back. corr. no no no yes yes no no
Platform no no yes yes no no yes
Matrix modifiers no no yes no no no yes
Measurements ht. ht. ht. ht. and area ht. and area ht. and area ht. and area
Spectrometer MR&HRb HR MR MR&HR HR MR MR
Support gas Ar&He Ar Ar Ar&He Ar&He He He

Pressure (Torr) 1-4 40-140 10-20 1-30 190(Ar) 760 760
600(He)
Current (mA) 20-30 55-90 30-80 20-90 70-200 NA NA
Potential (V) 400-600 NA NA 300-600 200-300 N.A. NA
Max. power (W) 18 N.A. NA 54 60 20 75
aB, Blades et alyo.24); D, Dittrich et al.(7,11,13,16,22.23 ,26); F, Falk et aly..,;,9,10,12,I'); H, Harnly et al,('8,2,,29); L, Littlejohn et al,(8".,'9); N, Naumann et al,('7); S, Sturgeon et
al. (21,27,28,30,31).
bMR, medium-resolution monochromator; HR, high-resolution monochromator,
C")
~
"~
~
Co
Table 9-2, the operating parameters are presented for the seven major investi-
gators in the field.
Design similarities can be seen for all four F ANES devices (Table 9-1)
(Fig. 9-2 to 9-5). The furnaces are composed ofpyrolytically coated graphite,
are resistively heated, and, except for the integrated contact cuvette (lCC)
used with HA-FANES, are of conventional design, i.e., they are identical to
those used routinely for GF-AAS, cylinders 6 mm in internal diameter and
28 mm long. The lCC is also a cylinder 6 mm in diameter, but it is only
19 mm long and has wings attached externally to each side (Fig. 9-5). The
axial electrodes of HA- and rf-FANES are also made of pyrolytically coated
graphite.
Design differences can be seen in the furnace contacts (conventional
versus lCC) and the gas containment (four different approaches). The shapes
of the furnaces were described above. Conventional furnaces are clamped at
both ends of the cylinder and the electrical current flows longitudinally
h
h
-H--k
Figure 9-2. Schematic diagram of the HC-FANES source of Falk et al.(5) (a) Carrier gas
port; (b) pump port; (c) electricaI connector to heating transformer; (d) anode; (e) removable
Iid for sampIe injection; (f) graphite electrode; (g) graphite furnace and hollow cathode; (h)
window; (i) water-cooled vacuum vessel; (j) water-cooled part ofthe vacuum vessel and rota-
tion arm for changing the graphite tube; (k) pivot; (I) gasket.
380 CHAPTER 9
Graphite Furnace
Housing Pneumatic Sampie
Introd uction Port
~ /
Graphite Rod
====O:::::::j /Quartz Window
~~====/
Graphite Furnace
v
To RF Filter and
Plasma Gas Inlet
Graphite furnace
Power Supply
Figure 9-3. Schematic diagram of the rf-FANES source used by Liang and Blades. (20)
He plasma
l___ J __
R.F.
input Centre I
electrode
Quartz
window
SN sleeve "-.
Furnace
/
housing
Figure 9-4. Schematic diagram of rf-FANES source used by Sturgeon et al.(2!)
(Figs.9-2 to 9-4). The ICC is clamped on the wings (Fig.9-5) and the
electrical current flows transversely. This transverse flow produces a uniform
temperature over the length of the fumace.(37) For conventional fumaces
there is a significant temperature decrease between the center and each
end. (38,39) This temperature gradient can produce nonuniform atomization
I
a
* ,
Figure 9-5. Schematic diagram

ofHA-FANES. (a) Integrated con- d I
tact cuvette; (b) contacts for fur-

nace power supply; (c) sampIe i~I~~f
loading port; (d) Macor support II 9
block for cathode; (e) cathode; (f)
spring; (g) contact pin; (h) insul- ----I
--
ated wire from the dc discharge

high-voltage power supply.(18 l h
and excitation effects and can be a source of interferences, as will be discussed

in a later section.
The gas containment designs of the four F ANES devices are con-
siderably different. The HC-FANES devices employed by Falk,o-6,9,lO,12,15)
Littlejohn, (8,14,19) Dittrich, (7,11,13,16,22,23,26) and Naumann(17) are of the same
design. These devices resemble the conventional Perkin-Elmer (Ridgefield,
Conn.) design with vacuum seals added in the appropriate locations. HC-
FANES has been operated between 1 and 140 Torr, with the majority ofthe
work being done between 1 and 30 Torr. The vacuum seals are not of a high
quality, and continuous pumping is generally required to maintain a constant
reduced pressure. For some systems, a continuous gas flow is necessary to
stabilize the discharge. The leak rates and the gas flowrates necessary to
stabilize the plasmas appear to vary for each system.
The HA-FANES source has been constructed in a six-way vacuum
cross. (18,25,29) A constant pressure of a few Torr can be held without pumping.
The HA-FANES is run in the static mode, i.e., flow of the support gas is
stopped for the atomization step. The optimum conditions for most elements
have been found to be around 200 Torr of Ar and 400-600 Torr of He.(29)
Both rf-FANES devices are run at atmospheric pressure with He as the
support gas. Sturgeon et al.(21,27,28,30,31) use a conventional Perkin-Elmer
furnace (Model HGA-500). External air is prevented from reaching the
interior of the furnace by the positive pressure of the support gas through
an internal flow through the furnace, which can be halted during the atomiza-
tion cycle, and a continuous external flow around the furnace. The low mass
of He makes it necessary to maintain high support gas flows both internally
and externally throughout the atomization cycle. Infusion of atmospheric
air has been a problem as will be discussed later.
382 CHAPTER 9
The device employed by Blades et al. (20,24) is a conventional Thermo

Jarrell-Ash (Franklin, Mass.) furnace head (Model IL655) that has the
capability of operating at greater than atmospheric pressure. The furnace is
designed to prevent leakage outward, not inward. Considerably lower flows
are used than reported by Sturgeon et al. It is not clear whether the He flow
is needed to prevent infusion of the atmosphere or to enhance the discharge.
The operating parameters of each of the F ANES systems will be dis-
cussed in detail in appropriate sections of the chapter.
9.3. Fundamental Considerations
While the basic properties of glow discharges operated with "cold"

electrodes have been investigated extensively, there has been considerably
less effort directed toward understanding such discharges in the presence of
"hot" electrodes. We shall not endeavor to review such work, but will instead
address some of the basic characteristics of F ANES sources themselves. The
reader should be aware of Falk's excellent review of HC-FANES(15) and is
directed to recent reviews by SIevin and Harrison,(40) Caroli,(41) pillOW,(42)
and Broekaert(43) on "cold" HC discharges and the existing literature on
"hot" cathode discharges. (44-47)
9.3.1. Sputtering
Although sputtering is no longer the primary means of generating the

analyte gas phase, it is still one of the most important processes within the
discharge. Sputtering phenomena are divided into two classes: "chemical"
and "physical." Chemical sputtering involves incident ions that react chemi-
cally with the cathode to produce a volatile species. This can certainly be
important when sufficient reactive impurities are contained in the support
gas, but it is more controllable than the inherent physical sputtering associ-
ated with FANES. We will therefore discuss in this section only physical
sputtering processes as related to F ANES systems. More general details of
sputtering in glow discharges have been discussed most adequately in
Chapter 2.
Physical sputtering is a multiple sequence collision process involving a
cascade of translating subsurface target atoms ("knock-ons"). This cascade
can itself interact with the surface, in which case sufficient momenta can
be transferred from knock-ons to surface species for sputtering to occur.
Wehner(48) has shown that for C the sputter yields (atoms sputtered per
incident bombarding particle) are about 0.06 and 0.1 for 400 eV He and Kr,
respectively. If sputtering of analyte-containing species is the result of a
transfer of the collisions in the graphite to the analyte, we should generally
expect even smaller yields because of the mass mismatch between C atoms
and the analyte atom or molecule species. This mismatch prevents maximum
energy transfer between the cascading C atoms (mass ml) and the analyte
species (mass m2) on the surface. This can be seen for elastic collisions
where the maximum fraction of the energy transferred is given by 4mlm2/
(mI + m2)2. However, a cascade collision may be initiated directly in the
analyte crystallites on the surface of the furnace. Sputtering will then occur
directly from incident particle bombardment of the crystallite, Le., collision
cascades are induced directly into the crystallite to induce sputtering. The
smaller the crystallite, the greater the surface-to-volume ratio is, so the prob-
ability of the cascading knock-on atoms interacting with the surface to
produce sputtering should increase. A greater sputtering yield is, therefore,
expected for the smaller crystallites. However, some of the crystallite species
must couple with the graphite substrates; the smaller the crystallite, the
greater the fraction of crystallite atoms involved in the coupling. The general
result is that cascades created directly within the crystallite will produce
cascades in the substrate. The reverse is also true---cascades in the substrate
will produce cascades in the crystallite. The extent of this cascade interchange
depends on how weIl the mass impedance between crystallite substrate atoms
match one another.
Falk(12) estimated the sputtering rate in the HC-FANES system, assum-
ing a sputter yield of 0.2, to be approximately three orders of magnitude
smaller than the rate of thermal vaporization. But as discussed above, the
sputter yields in small crystallites may be enhanced because of crystallite
size. Other effects may offset any increased yields, however. Kaminsky(49)
makes the point that the mean free paths associated with the glow discharge
give rise to multiple collisions in the gas that can result in formation of
molecular and atomic ions of sputtered target species. If these ions diffuse
to the cathode fall region, they will return to the cathode and be redeposited
or possibly induce further (self) sputtering. Von Hippel(50) had indicated
that 90% of the sputtered species diffuse back to the cathode in a glow
discharge. The sputter yields in this case would indeed be smaller than the
thermal vaporization yields in the HC-FANES system.
Analyte sputtering does not occur in the dc HA-FANES system unless,
of course, sampies are deposited on the central probe. Sputter rates from
the probe, by the above arguments, will still be insignificant compared with
thermal vaporization rates unless self-sputtering becomes significant.
Certainly, the larger radial electric fields associated with the HA geometry
will provide the greater ion energy and would thus enhance the self-
sputtering yields. This will be of little consequence unless these yields are
greater than unity, in which case the sputtering rate could conceivably
increase monotonically.
Sputtering of analyte in the rf-FANES geometry should be somewhat
more significant. The secondary electrons can gain significant energy by
resonating in the rf field. Thus, ionization should increase and more ions
384 CHAPTER 9
should become available to induce sputtering. But just as for the dc case,
redeposition and self-sputtering must be considered. These processes could
be more severe because of the increased field strength and increased proba-
bility for ionization of the sputtered species that enter the glow region.
Consider, however, the mobilities of the ions. These are so small at 1 atm
that the ion can be considered immune to the presence of the rf field. Hence,
neither self-sputtering nor redeposition is involved, and analyte sputtering in
an rf discharge will be insignificant compared with the thermal vaporization.
9.3.2. Sheath Dimensions

The principal regions of interest associated with the F ANES-type
plasmas are the cathode fall region (sheath) adjacent to the cathode and the
negative glow that neighbors this region. The positive space charge that
accumulates in the sheath contributes to a steep voltage drop in this region.
Most of the discharge voltage drop occurs here with the drop being related
to the space charge density. As observed from the more conventional dc
discharges, the radial extent of both regions (glow and sheath) diminishes
with increasing pressure. For the negative glow this is a result of decreasing
electron mean free path; the range of excitation decreases with increasing
pressure. To understand the relationship between the sheath and pressure
requires a more detailed consideration of ion transfer into the sheath.
Bohm(51) has shown that for ions to enter the sheath region they must have
a velocity (vo) greater than (kT e/mi)I/2, Te being the electron temperature,
mi the mass of the ion, and k the gas constant. This so-called Bohm sheath
criterion simply means that the electron density gradient in the sheath is less
than the ion density gradient produced by the acceleration of the electrons
and ions in the sheath, (52) i.e., V2 V < 0, where V is the voltage at any point
within the discharge. Prior to entering the cathode fall region, the ions must
accelerate through a potential drop ~ V in order to acquire a velocity greater
than vo. Equating kinetic energy gained to the potential energy gives
1/2miv~ = e~V= eE~r (9-1)
where, for discussion purposes, the electric field Eis assumed uniform over
the radial distance ~r. Note that ion temperature and collisions in the sheath
are neglected. Using the Bohm sheath criterion, tbis distance becomes
~r = kTe/eE (9-2)
This is the minimum distance an ion must travel in a radial field, E, in order
to contribute to sheath formation. This distance, which is a quasi-neutral
transition region between the plasma and the positive space charge region,
accounts for the largest part of the sheath extension.(53) The much smaller
Debye shielding length associated with the sheath will not be considered
here.
It has been shown by von Engel(54) that Te is inversely proportional to
the pressure p, so
llr oc l/Ep (9-3)
Therefore, as a result of differences in the electron and ion density gradients

that establish the Bohm sheath criterion and because ofthe electron tempera-
ture pressure relationship to p, the cathode sheath thickness diminishes with
increasing press ure.
The successful operation of the HA-FANES as an emission source at
pressures greater than 104 Pa is probably due to the relatively small influence
of thermionic emission from the relatively small central cathode. This
thermally isolated cathode lags the tube temperature and will, from the
Richardson-Dushman equation, exhibit a thermionic electron emission cur-
rent density (Jha ) that is well below that of the tube wall. The temperature
lag is dependent on the ohmic heating produced by the discharge current.
The thermionic current density (Jhe ) for the HC-FANES is, in contrast, that
of the wall itself. The relative differences in thermionic current (i) for the
two cases will be even greater because of the smaller cathode surface area
associated with the cathode used in the HA configuration. The cathode in
the HC-FANES system is the tube itself and, therefore, attains a higher
temperature earlier in the temperature ramp than does the cathode in the
HA-FANES system. The resulting larger thermionic emission induces an
e1ectronic conductivity for the hot He geometry that exceeds that of the HA
case. Therefore, in the constant-current mode of operation, the discharge
voltage for the "hot" HC-FANES plasma will fall below that of the "hot"
HA-FANES plasma, assuming the same furnace temperature programs for
both cases. The plasma cannot be sustained if this discharge voltage falls
below that which can accelerate secondary electrons to ionizing energies
faster than elastic collisions associated with higher operating pressures can
soak up this energy. That is to say, the electron mean free path must remain
greater than the Debye length. The HA-FANES plasma can be quenched
more readily by the thermionic emission if the pressure is sufficiently high.
The constricted glow region of the higher-pressure discharges requires that
the monitoring of photons from this source be done very near the cathode
surface. The HA geometry is more suitable for achieving this because the
temperature of the central cathode lags behind that of the furnace wall. The
strong dependence of radiative emissive power on temperature suggests that,
for identical temperatures and ramp rates for the two cases, the spectral
interferences from cathode optical emission should be smaller for the HA
386 CHAPTER 9
case. High-pressure operation ofthe HC-FANES is therefore tube tempera-

ture limited because of radiative emission as weH as thermionic emission.
The HA-FANES is also limited by tube temperature, of course, but the
temperature limit is greater in this case because the cathode temperature lags
behind that of the tube.
9.3.3. Discharge Characteristics
The principal research efforts to elucidate discharge characteristics asso-

ciated with various diode configurations have been devoted to cold cathodes
where cathode emission is a secondary electron emission phenomenon. Dis-
charges involving thermionic cathodes have been investigated to a much
lesser degree. The reader is referred to Hemquist and Johnson's exceHent
summary of the early work on these "hot" cathodes.(44) Wittig(45) observed
an HC effect (rapidly increasing discharge current accompanied by a decreas-
ing voltage) in low-pressure (2 torr) "hot" HC discharges. The minimum
voltage necessary to sustain a given current was found to be dependent on
the cathode temperature. Salinger and Rowe(46) showed that thermionic
electrons that enter the glow in these discharges approach a near-MaxweHian
energy distribution through mutual interactions. Excitation and ionization
is induced by the electrons in the high-energy tail of this distribution. It has
been reported that cumulative ionization processes by collisions between
excited atoms are important.(47,48). Consequently, it is possible to sustain a
discharge even when the discharge potential drop is weH below the lowest
excitation energy of the gas atoms. Wittig(45) explained that this is because
cumulative ionization is proportional to the square of the plasma density. If
this density is increased, keeping current constant, it should be expected that
the ionization efficiency will increase rapidly and therefore aHow a decrease
in the electron temperature.
Characteristics of F ANES-type discharges are controHed by the type of
gas, gas pressure, voltage, and fumace temperature. The applied voltage
controls the current, which controls the charge distribution according to the
mobility of the charged species. The electric field is produced by this charge
distribution and satisfies Poisson's equation. The solution of this equation
for an HA geometry indicates the field is highly nonuniform.(54) Maximum
field strength occurs at the cathode and decreases rapidly to become a con-
stant value of (2i/ J.l )1/2, where i is the current and J.l is the mobility, at larger
distances from the cathode. Unfortunately, because of the nonsymmetry of
the HC-FANES source, a closed form solution to Poisson's equation is not
possible for this case.
Consider the principal differences between the HA- and the HC-F ANES
discharges. The HC system, because of its larger cathode area, will be more
strongly infiuenced by photoelectron emission from the cathode into the
plasma. But during the tube heating both systems can reach temperatures
where electrons from the cathode are supplied primarily from thermionic
emission. This supply satisfies Richardson-Dushman's equation for the
current density
je = AT2 exp[ -(eifJ - eJeE)lkT] (9-4)
where ifJ is the work function, e the electronic charge, and E the applied
field needed to account for the infiuence of external fields on this emission
(Schottky effect). The theoretical value of the constant A is about
120 A/cm2 _deg2YS) The temperature of the cathode in the HC-FANES
source increases at a greater rate than that ofthe HA-FANES because, for
the HA source (l) the cathode is thermally isolated and (2) thermal radiation
from the anode to the cathode will be negligible until the anode temperature
reaches approximately 1300 K. The cathode temperature can therefore lag
behind the tube temperature by more than 1000 K during a rapid heating
ramp of the tube. It is not unreasonable to assume that the thermionic
emission into the HC plasma will, because ofthe higher cathode temperature
and larger cathode surface area, be considerably greater than that achieved
for the HA plasma. This is true only if the field term in the exponential of
the Richardson-Dushman equation can be neglected, otherwise the applied
field can control the thermionic emission. But the applied fie1d is a com-
plicated function of the cathode dimensions and depends strongly on the
geometry. It is therefore not possible to predict apriori which of the two
geometries will exhibit the greater thermionic emission for all tube tempera-
tures. Note that Langmuir(S6) has shown that if an anode voltage V is applied
in the vicinity of an electron emitter in vacuum, the emission current (i)
follows the relation
i = SV 3 / 2 (9-5)
where s is a constant that depends on geometry. This derivation is for

vacuum and is, therefore, aspace charge-limited situation. The thermionic
emission becomes mobility limited at the higher pressures ; emission current
is then proportional to V 2 .(S3) The discharge current derived by von Engel
for the HA geometry shows this square dependence on the applied voltage,
so the current can be considered mobility limited in this case. Now, Schottky
effects become significant for fie1ds on the order of 106 VIm. Thermionic
emission current densities for the two geometries are, therefore, equal for
equal cathode temperatures if the sheath fie1ds are less than 106 VIm. The
assumption that the HC configuration exhibits the greater thermionic
emission current is then valid for these lower field strengths or for applied
potentials less than 500 V. Note that 500 V applied to a I-mm-diameter
388 CHAPTER 9
cathode will provide a 106 VIm fie1d when space charge is absent; the pres-
ence of positive space charge such as that responsible for the cathode fall
only increases the field.
The "hollow cathode effect," or enhanced electron density in an HC
glow, is responsible for the enhanced glow observed in HC systems.(57) Even
greater enhancement should be expected from the HC-FANES plasma
because of the stronger thermionic emission associated with this system.
These thermal electrons leave the cathode sheath, enter the glow, and become
trapped in the cylindrical region imposed by the retarding radial fields of
the sheath. The probability for elastic and for inelastic gas-phase collisions
is, therefore, enhanced, which further enhances the electron density. In con-
trast, the electrons emitted by the cathode in the HA geometry travel in
radially symmetric fie1ds. Most of the fast electrons lose their energy by
impacting the outer cylindrical anode, not in elastic collisions that will ionize
or excite the gas.
It is noted that the dc F ANES discharges differ from the more con-
ventional planar glow discharge. In fact, von Engel(54) c1assified the HA
geometry as a glow discharge without zones near the anode, i.e., without the
positive zones observed for the planar case. The HC plasma differs from the
planar glow discharge in that it exhibits an enhanced glow, the "hollow
cathode effect" that is discussed above.
The rf discharges described by Sturgeon et al.(21,27,28,30,31) and by Blades
et al. (20,24) are of the HA configuration. These rf corona discharges are not
yet understood. The rfmodulation ofthe (sheath) voltage and sheath thick-
ness must control the phase and intensity of the current, but a mathematical
description has not been developed.
The high power input associated with rf discharges at elevated pressure
implies enhanced ionization. Generally speaking, the impedance of the rf
plasma decreases as the frequency of the applied voltage increases. It has
been proposed by McDonald and Tatenbaum(58) that the rf field can drive
electrons to energies sufficient for ionization if e1astic collisions in the gas
are in phase with this field, Le., if collision-induced reversal of the electron
motion is in phase with the applied field. Energy lost by electrons through
ionizing collisions can then be reestablished by the primary and secondary
electrons through their interaction with the field, and the process continues.
This explains why the minimum operating pressure (maximum mean free
path) increases (decreases) with increasing frequency. Chapman reports the
suggestion of Holland et al. (59) that electrons can impact the electrodes to
produce secondaries that are rapidly acce1erated across the positive sheath
into the discharge. A resonance effect (multipacting) can then occur that
increases the electron energy, provides an efficient electron supply, and
induces efficient ionization. (58) It has also been suggested by Keller and
Pennebacker(60) that electrons mayaiso gain energy from the modulated
edge of the positive sheath, where sheath momentum is shared with the
electrons as they are refiected in resonance.
9.3.4. Voltage-Current Characteristics
Figures 9-6 and 9-7 show the current versus discharge voltage curves
for the HA- and the HC-FANES. In both cases the general slopes of these
curves decrease with pressure. The curves imply operation in the abnormal
discharge region since the magnitude ofthe slopes are finite. Such a discharge
extends over the entire inner surface of the cathode and any increase in
current results in an increase in current density (j) and discharge voltage
(V). The slope for the HA-FANES is about 18 times greater than that for
3.4,..----------------------,
3.2
2.8
_____ h
. -.......... i
~j
2.6 L..-_ _ _L--_ _ _L--_ _ _-'--_ _ _--'-_ _----'
-2 -1 o 2 3
Log [discharge curren! (mA))
Figure 9-6. Current-voltage eharaeteristics for HA-FANES operated with a I-mm-diameter

eathode and no fumace heating at pressures of: (a) 10, (0) 30, (e) 50, (d) 70, (e) 90, (f) 110,
(g) 130, (h) 150, (i) 170, and (j) 190 torr.(25)
390 CHAPTER 9
500
400
~
Q)
:~
s
Cl
Ci
>
300 c-
Figure 9-7. Current-voltage

characteristics for HC-FANES with
200-1--r----r--,-----,--...,----, no furnace heating at pressures of:
10 20 30 40 50 60 (a) 0.75, (b) 3.8, and (c) 7.5 Torr Ar
Curren! (mAl and (d) 6.8 Torr He.(IS)
the HC-FANES for currents between 20 and 30 rnA at nearly identical

pressures (7.5 and 10 Torr, respectively). If this ratio is multiplied by the
ratio of the respective cathode areas (0.1), it is seen that the (ßVjßj)p for
the HA-FANES is about 1.8 times that ofthe HC-FANES when operation
is in the above pressure range. This derivative represents the reciprocal of
the electrical "conductivity," (1', of the gas. Hence, for the above pressure
(9-6)
where p is the charged density and J.l is the mobility. This result confirms
that it is the charge density difference associated with maintaining a given
current in the two different geometries that is responsible for the differences
in current responses between the HA and HC cases. This might be expected
since the radial current per unit length is given by
i = 27rrp(r)v(r) = 27rrp(r)J.lE(r) (9-7)
where v(r) and E(r) are the radial velocity and electric field, respectively and
are highly dependent on system geometries. Equation (9-7) can be used with
the Poisson equation in cylindrical coordinates to obtain i in terms of the
voltage, v, applied to the HA system.(54) The current for the HA-FANES is
then described by
iha = (V - Vo)Vop/[rln(R/r)] (9-8)
where Rand rare the anode and cathode radii for this geometry, and Vo is
the breakdown (or running) potential for the HA discharge. Figure 9-6
shows departure from this predicted linearity with voltage. End effects are
probably responsible since the ca1culation is made for an infinitely long tube
and thus neglects such effects.
From Eq. (9-8):
V/i = [r In(R/r)]/ p Vo = 0/pEo) In(R/r) (9-9)
The denominator on the right-hand side of Eq. (9-9) is, by definition, the
drift velocity in the fie1d Eo, and the drift velocity is inversely proportional to
pressure.(54) Thus, taking the pressure derivative ofthe logarithm ofEq. (9-9)
gives:
%p In(V/O = k%p lnp = kl/p (9-10)
The observed decreases in (log V -log i) with increasing pressure (Fig. 9-6)
are therefore a result of the effects of pressure on mobility and on the
mobility-dependent breakdown field E o .
The He geometry is, unfortunately, not an amicable geometry for
c10sed form derivations of V(i). But the data of Fig. 9-7 imply a linear V(i)
relationship exists and that the slope decreases with increasing pressure,
similar to the HA case. It is expected then that the slope has a Vo dependence
that is similar to that described for the HA geometry in Eq. (9-9). The radial
(geometric) dependence cannot, of course, exhibit this similarity. It should
be noted that Vo is dependent on the uniformity of the electric field and
decreases with increasing deviation from this uniformity.(54) But the HA
geometry produces a hyperbolic1ike field distribution, so it should be
expected that
VO,(hOIlOW cathode) > VO,(hollowanode) (9-11)
This implies, from Eq. (9-9), that
(oV/oi)p, hollow anode> (oV/oOp, hollow cathode (9-12)
which is the observed geometric effect. It is conc1uded then that differences

observed for the V(i) curves in the two cases are caused by the dependence
392 CHAPTER 9
of the current on breakdown voltage, which is dependent onfield unifonnity

and hence on the geometries involved.
9.4. Operational Characteristics
9.4.1. Analytical Signals
The analytical emission signals associated with F ANES systems are

dependent on furnace-induced volatilization and atomization, on the dis-
charge excitation, and, to some extent, on discharge-induced atomization.
None of these processes is weIl understood. Furnace atomization is a well-
documented phenomenon, but low-pressure discharge-induced excitation
processes are, in general, not nearly as weIl characterized. Emissions from
these discharges are dependent on the support gas, the gas pressure, the gas
temperature, the discharge current, and the cathodejanode geometries. Each
of the first four factors will be discussed and the cathodejanode geometries
will be compared throughout. Literature on glow discharges is extensive,
and pertains mostly to use of these discharges in sputtering processes at low
pressures «10 Torr). Little infonnation is available for "hot" cathodes or
anodes and operation at higher pressures.
9.4.1.1. Support Gas
Ar and He are the only support gases that have been used to date with
FANES devices. For HA- and HC-FANES, the two gases can be readily
interchanged. A major analytical consideration is the higher excitation
energy ofHe. For HA-FANES it is also possible to operate at much higher
pressures with He (600 Torr) than with Ar (200 Torr). For rf-FANES, He
is used almost exclusively. Both Blades and Sturgeon have reported difficulty
in initiating an Ar rf plasma at atmospheric pressure.
9.4.1.2. Discharge Pressure
The operating pressure of any of the F ANES devices would appear to

represent a compromise between optimum conditions for furnace atomiza-
tion and glow discharge excitation. Higher pressures provide longer resi-
dence times in the furnace and, thus, larger integrated signals. Lower
pressures provide longer mean free paths and, consequently, contribute to
higher average kinetic energies for free electrons in the potential gradient of
the discharge. This kinetic energy increase can manifest itself as an increased
capability to ionize and to excite higher energy levels by electron impact.
First, consider the effect of pressure on furnace atomization. The graph-
ite furnace, as characterized by absorption measurements, induces transient
atomization of the analyte that is dependent on the convolution of the
supply and loss functions. The supply function is dependent primarily on the
chemical nature of the analyte, the interaction of the analyte with the chemi-
cal matrix and the furnace surface, and the heating rate and final temperature
of the furnace. The loss function, for operation in the static mode (gas flow
is terminated during the atomization step), is usually determined by the rate
of gaseous diffusion of the analyte from the furnace. U nder ideal, isothermal
conditions, the loss function remains constant during the presence of the
analyte and the appearance function has no effect on the integrated signal.
The conventional furnace differs from ideality, however, because of the lack
of spatial and temporal isothermality. (38,39) Consequently, for GF -AAS, con-
siderable emphasis is placed on maintaining a consistent supply function
(i.e., the same for sampies and standards) in order to ensure accurate
determinations.
As stated above, the analyte loss is usually controlled by diffusion,
provided the sampie matrix does not have a large mass. The diffusion rate
is inversely proportional to the pressure. The low operating pressure of
HC-FANES (routinely 1-30 Torr) increases the diffusion (loss) rate and
decreases the analyte residence time and the integrated analytical signal,
as compared with operation at atmospheric pressure. HC-FANES systems
generally use continuous pumping and a constant support gas flow to main-
tain the desired pressure and to reduce the effect of entrained air. It is not
c1ear whether the analyte residence is limited by diffusion or convection.
HA-FANES operates in the static mode (no pumping or internal gas
flow) with pressures of 150-200 Torr of Ar or 300-600 Torr of He. The
analyte residence times are much closer to those found for conventional
graphite fumace atomization than HC-FANES.
rf-FANES operates at atmospheric pressure in He and would be
expected to have residence times similar to conventional GF-AAS if Ar is
used as the support gas. Instead, rf-FANES uses exc1usively He, which has
a greatly reduced mass compared with Ar. Sturgeon et al.(21,27,28,30,33) use a
constant flow of He through the furnace in order to exc1ude air entrainment.
This flow has not altered the residence times.
Now consider the effect of pressure on glow discharge excitation and,
more specifically, HC excitation. The excitation process and, consequently,
the emitted intensities of low-pressure, HC discharges are well documented
as being highly pressure dependent. (40-42,49,61,62) This is apparent from an
examination of the emission intensities of the support gas (Fig. 9-8). At low
pressures the maximum intensity from the discharge is observed on the
cylindrical axis. The intensity increases (not shown in Fig. 9-8) and the
maximum region broadens radially as the pressure increases. Eventually, as
394 CHAPTER 9
I
(1010 cm- 3)
, I
" \ 8 ,-,
~ II \\ I I" \\
...E
.... " I \ 6 I \
'/--,
c:
I
, \\!'I' \\
\
,'/~ \\
~
8 4 0 4 8 8 4 8
mm mm
Figure 9-8. Radial distribution of intensity ratio, I/Imax , of the spectra1line of argon at
772.4 nm and of electron density, ne , in the negative glow of a cylindrical cathode 2 cm in
diameter and 10 cm long at a discharge current of 50 mA in Ar at: ( - ) 0.1 Torr, (- - -)
1.0 Torr, and (- - - -) 2.0 Torr.(61)
the pressure continues to increase, the maximum moves away from the center
and toward the cathode wall. A similar pattern is observed for the electron
density (Fig.9-8). As the pressure increases, the electron mean free path
decreases and the energy from the electrons, accelerated across the dark
space, is dissipated in inelastic collisions closer to the cathode wall. This
finding supports the assumption that electron collisions are primarily respon-
sible for excitation and that the electron density distribution will detennine
the maximum region of emission.
The exact position of the maximum emission region will depend on the
pressure and the mass of the support gas. The pressure dependence of the
intensity integrated over the entire HC cross-sectional area has not been
reported.(42) From Fig.9-8, it is clear that for a given pressure, there is a
maximum observation point within the cathode. For a fixed observation
point, at the cylindrical axis ofthe HC, the emitted intensity can be expected
to increase, pass through a maximum, and then decrease as a function of
pressure. The higher pressures, however, will give rise to lower diffusion
rates and longer residence times. Thus, it is not clear how fast the intensity
at the axial position will decline.
Falk(12) has reported results for the support gas emission that are
consistent with the data in Fig. 9-8. For observations made at the central
axis ofthe furnace, the intensities for Ne(I) and Ar(I) increase monotonically
from 0.5 to 6.0 Torr, and then decrease between 7 and 30 Torr. Falk(l2) also
showed that between 0.5 and 8.0 Torr the emitted intensities from He(II),
333.49 nm, and Ar(II), 454.51 nm, decreased with increasing pressure.
Transitions requiring high-energy electrons appear to be more sensitive to
increased pressure.(12,42,61) The number offast electrons (19-26 eV) shows a
sharp maximum between 1 and 3 Torr.
It should be noted that the HC-FANES is most convenientiy viewed

along the central axis. All of the data reported by Falk were obtained by
viewing the central position. This serves to minimize the background from
blackbody emission from the furnace wall at higher atomization tempera-
tures (above 2300 K). No attempt was made, for any of the HC-FANES
studies, to characterize the spatial dependence of the emission signal or to
correlate the total emitted intensity with the various parameter changes.
The plots of emission intensities for aseries of Fe lines from an Fe HC
lamp as a function of pressure in Fig. 9-9 show the expected behavior
described above. A broad emission maximum is observed between 0.1 and
10 Torr.(40,62) It must be remembered, however, that these data reflect the
maximization of the combined sputtering and excitation processes.
It is expected that a nonsputtered analyte species will show a similar
spatial dependence as the fill gas. Falk et al. (15) have reported experimental
detection limits for 22 elements using pressures ranging from 1 to 5 Torr for
HC-FANES. Detailed intensity versus pressure data have not been reported.
It must be assumed that there was a maximum for most elements within this
pressure range, otherwise a higher pressure range would have been used.
Conversely, Falk et al.(15) showed that for C(I), 247.8 nm, above 1800 K,
the emission intensities increased steadily over a range of 4-22 Torr in Ar
and 7-30 Torr in He. Above 1800 K, C is thermally volatilized from the
furnace surface and can therefore be considered a nonsputtered species.
Naumann et al.(17) have reported HC-FANES results that differ consid-
erably from those reported by Falk et al. The optimum pressures for Co,
Cr, Cu, and Ni, in the single-element mode, were 102, 121,40, and 116 Torr,
4.0
~
'"<:
~ 3.5
§.
.§'
3.0
0.2 0.5 1.0 2.0 5.0 10

Pressure (mm)
Figure 9-9. Dependence of spectral intensity, I, on the fill gas pressure of Ne in a hollow
cathode discharge at a constant current of90 mA for: (a) Fe, 268.7 nrn; (b) Fe, 368.2 nrn; and
(c) Fe, 379.4nrn.(62)
396 CHAPTER 9
respectively. Unfortunately, a description ofthe optics used and ofthe obser-

vation region of the discharge was not provided. Considering the tendency
of the maximum negative glow region to move toward the cathode wall with
increased pressure (Fig. 9-8), it would appear that these authors viewed a
large fraction ofthe off-axis region ofthe plasma. Overall, the signal response
as a function of pressure is still unclear for HC-FANES.
Optimum operating conditions for HA_FANES(IS,25) differ considerably
from those for HC-FANES. The HA geometry eliminates any true HC
effect. High-energy electrons accelerated across the potential drop of the
cathode dark space pass through the negative glow region, strike the furnace
wall or anode, and are lost. The spatial orientation of the glow discharge to
the cathode surface as a function of pressure, however, is similar to that of
the HC configuration or conventional planar glow discharges. As pressure
increases and the electron mean free path decreases, the region of maximum
emission intensity moves closer to the cathode surface (Fig. 9-8).
Ballou et aIY S) have shown that for Ar at low pressures «5 Torr) the
discharge uniformly fills the furnace. As pressure is increased to approxi-
mately 15 Torr, the discharge shrinks dramatically until it constitutes a
corona surrounding the cathode. As shown in Fig. 9-8 the region of maxi-
mum intensity is close to the cathode wall. Increasing the pressure up to
200 Torr serves to brighten and more clearly define the limits ofthe negative
region close to the cathode. The maximum signal-to-noise ratio (SNR) is
found within 1 or 2 rnrn of the cathode surface at all hut the lowest pressures.
Similar results are observed for He, except that a dramatic shift in pressures
is observed. A diffuse discharge is observed in He at pressures as high as
100 Torr and the discharge shrinks and brightens as pressures are increased
up to 600 Torr.
Table 9-3 shows the spatial dependence for the Cd and Cu analytical
signals in HA-FANES reported by Harnly et al.(25) at 30 Torr of Ar. The
Table 9-3. Peak Area versus Viewed Region of the Discharge·

Peak area (I -s )
Region viewed of corona b Cd Cu
I. Above 199. ±0.4 686. ±0.8
2. Top 191. ±0.5 1233. ±1.7
3. Center 152. ±0.4 78. ±0.6
4. Bottom 103. ±0.3 706. ±0.7
5. Left 155. ±0.3 1283. ±0.9
6. Right 151. ±0.4 611. ±1.8
"Results obtained at 70 Torr, 50 mA, atOlnization temperature of 1845'C, with a
1.0-mm-diameter cathode. Precisions determined as 3 sigma of the baseline noise.
b See Fig. 9-10 for diagram of viewed region.
Negative glow region
Figure 9-10. Schematic diagram of Furnace wall

the end-on view of the HA-FANES - (Anode)
source. The optics were adjusted so a
that the entrance slit (100 pm wide and b
500 pm high) viewed the six positions e
shown: (a) ahove, (h) top, (c) center,
(d) hottom, (e) left, and (f) right of
the eathode. (25)
regions examined are shown in Fig. 9-10. The analytical signals decrease
dramatically outside the corona. In general, the region of the maximum
analyte signal also gives the best SNR.
The effect of pressure on the integrated analytical signals for Cd and
Cu in Ar using HA-F ANES is shown in Fig. 9-11. (29) The optics remained
focused at position "e" (Fig. 9-10) for the entire experiment. In general, the
signals increase linearly with increasing pressure. This linear increase in the
integrated signal can be explained solely by the decrease in the diffusion
400 4000
Cd Cu
c
""'c
0
0
300 3000
Q)
b'"
'(ij
c 200
.2l 2000
,§.
ns
~
«
'"ns
Q) 100 1000
Q.
0 0
0 100 200 0 100 200
Discharge pressure (Torr) Discharge pressure (Torr)
Figure 9-11. Peak area for 2 ng ofCd and ofCu in HA-FANES with atomization at 2100 K
and viewing position "e" (see Fig. 9-10) as a funetion of the Ar pressure at diseharge eurrents
of: (a) 20, (h) 50, and (e) 70 mA.(25)
398 CHAPTER 9
coefficient. Aseries of currents was used since there is a finite range of

effective currents for each pressure (see Section 9.4.1.4.).
Tbe SNR for Cr in Ar as a function of pressure and discharge current is
shown in Fig. 9-12. Tbe shapes ofthe SNR plots in Fig. 9-12 are determined
primarily by the stability of the discharge, i.e., the background emission. A
comparison of the maximum of each plot reveals that the best SNR is found
at 160 Torr.
rf-FANES, operated at 1 atm ofHe, looks very similar to HA-FANES.
A bright region surrounds the central electrode and a less intense plasma
fitIs the remainder ofthe fumace volume. Although the rf potential is bipolar,
the self-biasing nature of the rf discharge, which arises from the difference
in the electron and ion velocities, appears to provide a negative bias to the
central electrode. Emission signals for the rf-FANES are obtained from the
region just outside the corona surrounding the central electrode. There have
been no measurements of the spatial dependence of the analyte signal.
Shorter residence times have been observed for rf-FANES than for
conventional GF-AAS. Tbis is most likely a result of the larger diffusion
coefficients for the analyte in He relative to Ar. Sturgeon et al.(2I,27,28,30,31)
reported a full width at half-height of 230 ms for Mn as compared with
approximately 400 ms for atomization from the wall of a conventional fur-
nace with 1 atm of Ar. Sturgeon also employs a constant internal flow of
400r-----------------------------------,
300
.2
iii
a:
.~
.*1
200
CI)
100
Current (mAl
Figure 9-12. SNR for 100 pg of Cr as a function of the discharge current at an atomization
teInperature of 2500 K and pressures of: (a) 20, (b) 40, (c) 80, (d) 120, (e) 160, and (f)
160 Torr.(29)
He to prevent entrainment of atmospheric air. The analyte signal shapes,

however, do not vary for He flows from 75 to 325 ml min- I . This suggests
that analyte loss is not influenced by the He flow and is still diffusion limited,
probably because of the low atomic mass of the He.
9.4.1.3. Atomization Temperature
In general, the atomization temperature for all the F ANES sources is

determined by the volatility of the element being determined. The lowest
temperature is used that provides complete volatilization of the analyte.
With respect to the excitation process, the critical temperature is 1800 K.
Below 1800 K, a "cold" cathode requires a potential ofseveral hundred volts
to sustain currents of 20-100 mA. Above 1800 K, the potential necessary to
sustain the same current with a "hot" cathode drops precipitously (Fig. 9-13)
because ofthe evolution ofthermionic electrons (as discussed in Section 9-3).
As established for GF-AAS, the atomization temperature refers specifi-
cally to the furnace wall temperature. It is generally synonymous, after a
few seconds into the atomization cyc1e, with the gas temperature, which
initially lags behind the rapidly rising wall temperature by several hundred
300
~
Q)
0> 200
S
"0
> d
/
100 c
b
a-
770 1270 1770 2270

Temperature (I<)
Figure 9-13. Diseharge voltage of HC-FANES at 3.8 Torr Ar as a funetion of the eathode
temperature for eonstant eurrents of: (a) 20, (h) 30, (e) 40, and (d) 60 mA.<'S)
400 CHAPTER 9
degrees.(39) The rate of diffusion of the analyte from the furnace is propor-
tional to T 3 / 2 • Consequently, the atomization temperature is generally kept
as low as possible to enhance residence time and the integrated signal.
The temperature necessary for complete volatilization varies with the
element and also varies significantly with the operating pressure of the
specific FANES system. In general, HC-FANES requires the lowest tem-
peratures, HA-FANES moderately higher temperatures, and rf-FANES, at
atmospheric pressure, uses temperatures comparable to those of conven-
tional furnace systems.
As mentioned above, cathode temperatures in excess of 1800 K can
have a significant influence on the discharge because of the considerable
quantity of thermionic electrons that are emitted. This has been weH docu-
mented for HC-FANES by Falk et aIY S) At 1800 K the electrical conductiv-
ity of the gas experiences a rapid increase and the potential across the
discharge rapidly decreases (Fig. 9_13).(IS) A similar effect is observed on the
emission intensity ofthe support gases (Fig. 9-14). Below 1800 K, the effect
of increasing the temperature is a slight enhancement of emission intensity.
This is likely caused by the reduction of gas density and an associated
increase in the electron mean free pathY S) The magnitude of the decrease
of the emission intensity of He at 1800 K in Fig. 9-14 appears to be less
10
~ ~ ~ 1~1~ 1~ 1~~~
Cathode temperature (K)
Figure 9-14. Intensity ofthe He 318.774-nm line in HC-FANES as a function ofthe cathode
temperature at a discharge current of 40 mA and pressures of: (a) 6.8, (b) 9.8, (c) 20.2, and
(d) 30.0 Torr.(IS)
significant at higher pressure. Similar potential-temperature profiles have

been observed for HA-FANES, however, at much higher pressures.
The analytical importance of the thermionic effect for F ANES is not
c1ear. Falk et aIY S) observed no effect of the thermionic electrons on the
detection of a wide range of elements by HC-FANES. It can be assumed,
however, that at the reduced pressures (1-30 Torr) of HC-FANES, all ele-
ments were volatilized at temperatures less than 1800°C. Naumann et al.(17)
reported atomization temperatures for Co, Cr, Cu, and Ni ranging from
2100 to 2600 K, at pressures from 40 to 140 Torr. For each element, the
detection limits were very competitive with Falk et al.(lS) and GF-AAS, i.e.,
the detection limits of elements atomized at temperatures greater than
1800 K did not deteriorate.
No discernible effect of thermionic electrons on analytical determina-
tions by HA-FANES has been observed. Like HC-FANES, however, the
cathodic atomization temperatures of HA-FANES are less than 1800 K.
Although the dramatically higher pressures of HA-FANES (up to 200 and
600 Torr of Ar and He, respectively) require furnace temperatures up to
2500 K, complete volatilization of the analytes (tested to date) has been
achieved before the temperature ofthe central cathode reached 1800 K. This
difference in temperatures arises from the temperature lag of the cathode.
WeIz et al.(39) have reported that a platform in a graphite furnace takes
about 2 s to reach 1800 K when the furnace power supply is set for a 2800 K
atomization step. The central electrode can be expected to exhibit a greater
lag than the platform since it is completely removed from the wall, is attached
to a large cool mass outside one end of the furnace, and is heated only
radiatively and convectively. Consequently, the analytical signal for Cr,
atomized at 2500 K, has returned to baseline before a decrease in the dis-
charge potential is observed.(29)
It is difficult to monitor the voltage potential across an rf plasma with
existing rf power supplies. Sturgeon et al., (21) however, have reported an
increase in reflected power for rf-FANES with increasing furnace tempera-
ture. Sturgeon suggested that this may be related to thermionic electron
emission. The increase in reflected power is delayed, occurring about 3 s
after initiating a 2800 K atomization step. This time lag may be associated
with the central electrode temperature lag. The increase in reflected power
is, therefore, a possible result of thermionic electrons emitted by the central
electrode at 1800 K.
Blades and co-workers have not mentioned a similar increase in reflected
power. This is most likely due to the automatic impedance matching circuit
associated with their power supply.
The central cathode of HA- and rf-FANES introduces some analytical
signal features that are not seen for transient GF-AAS signals. The tempera-
ture lag of the central electrode produces double peaks of the emission
402 CHAPTER 9
spectra in the atomization step. (24,27,29) Analyte atomized from the wall con-
denses on the cooler central electrode. Continued heating of the furnace
leads to eventual reatomization of the analyte from the cathode; hence, the
double peaks. Double peaks are observed for the less volatile elements (Cu
and Cr) at low discharge currents. At higher currents, only a single peak is
observed. The process of atomization, condensation, and reatomization may
still prevail, but the temperature difference between the cathode and the
furnace wall is sufficiently small that the peaks are not distinguishable.
For HA-FANES operation, the temperature ofthe cathode at the start
of the atoruization step is a function of the discharge current, resulting from
ohmic heating of the cathode. Because of the low mass of the rod-shaped
cathode, the temperature can exceed 1300 K (dull-red color temperature ) ·at
moderate currents (60 mA). Thus, at the start of the atomization step the
cathode temperature is higher than that of the furnace wall. Upon atomi-
zation, however, the wall heats rapidly and quickly exceeds the cathode
temperature. For a 3-s atomization step at 2300 K, the wall reached the set
temperature after only approximately 1.25 s. The cathode temperature at the
end of the 3-s cycle has just reached 1800 K, Le., the voltage drop across the
discharge started just prior to the end of the cycle.
The cathode ofthe HA-FANES source has been used as a platform for
sampie deposition instead of deposition on the furnace wall. Preliminary
data for Cu showed the peak maximum and width to be approximately the
same for deposition of the sampIe on the cathode or on the wall. Use of the
cathode as a platform produced double peaks for Cd. In this case one peak
occurred prior to the atoruization step after the discharge had been initiated.
Ohmic heating of the cathode by the discharge current resulted in thermal
atomization and nearly quantitative transfer of Cd to the cold furnace wall.
A second peak for Cd was then observed when the furnace was heated
during the atomization step.
9.4.1.4. Discharge Current
It is difficult to draw any generalities with respect to the discharge

current for the F ANES devices. The two best-characterized devices, HA-
and HC-FANES, display different integrated signal-discharge current pro-
files. HC-FANES signals increase monotonically with increasing current
while HA-FANES signals are independent ofthe current above a threshold.
There is insufficient information for rf-FANES to support any conclusions.
Falk et al.(IS) have observed consistent increases in the peak heights of
the analytical signals with increasing discharge currents for HC-FANES.
Littlejohn(19) has reported similar results. The analytical signal is propor-
tional to in, where i is the discharge current and n is an. integer representing
the number of collisions involved in the excitation process. Examining the
OISCHARGES WITHIN GF ATOMIZERS 403
support gas emission, it was observed that 3He (318.8 nm) had n values of
approximately 2 and 1 for hot and cold cathodes, respectivelyY 5) For Ar
(451.1 nm), n had values of approximately 0.5, 1, and 2, depending on
pressure and temperature. It was suggested that the value of 0.5 pointed to
a loss mechanism that was current dependent. Similar results were observed
for six atomized elements (Fig. 9-15). For Al, n was approximately 2.0; for
Cr and Ni, approximately 1; and for Co, Cu, and Fe, about 0.5.
The integrated analytical signals for HA-FANES, at significantly higher
pressures than HC-FANES, show a completely different dependence on the
discharge current (Fig. 9_16).(25) Integrated Cd and Cu signals, obtained at
70 Torr of Ar, initially increase with increasing current and then reach a
plateau. Initially the signals were proportional to in with n approximately
equal to 1 and 0.5 for Cd and Cu, respectively. The SNRs (Fig. 9-16) are
relatively comparable at all but the lowest and highest currents for both
elements. Cr showed similar results at pressures up to 200 Torr. At each
pressure, the integrated Cr signals did not vary as a function of current.
3.0
2.9
2.8
z:
'üi
c:
~ 2.7
~
Cl
0
-J
2.6 b
2.5
2.4
20 30 40 50 60
Curren! (mA)
Figure 9-15. Dependenee of the emitted analytieal intensities in HC-FANES on the diseharge
eurrent for: Ca) Co, Cu, and Fe; (h) Cr and Ni; and (e) A1.(15)
404 CHAPTER 9
1200r--------------------------------------,
"p--_Jo---<J C
_ - - -_ _- - - - - -.. a
o 20 40 60 80
Curren! (mA)
Figure 9-16. Dependence of peak areas (a and b) and signal-to-noise ratios (c and d) for
2 ng of Cd and Cu, respectively, on the discharge current in HA-FANES with atomization at
2100 K, 70 Torr Ar, and viewing position He" (see Fig. 9_10).(25)
The response shapes shown in Fig. 9-12 are primarily a reflection of the
background noise as a function of current.
For HA-FANES, the usable range of discharge currents is determined
by the pressure. Higher currents are achievable at higher pressures but the
usable range of currents is reduced. At any given pressure, the lower current
limit is the current necessary to cover the entire cathode surface with the
discharge. (15,25) At this minimum current, the discharge is just on the thresh-
old of the abnormal region. (40-42) From this minimum current, higher cur-
rents can only be achieved with an increase in the current density and
increased voltage. The current necessary to reach the abnormal discharge
mode increases with pressure. The upper current limit is the highest current
that can be used without excessive arcing within the system. This upper
current limit is determined by the electrical insulation of the system and the
contamination of the support gas. For HA-FANES, the upper current limit
was characterized by aseries of "minidischarges" between the anode (fur-
nace wall) and the point at which the cathode enters the Macor support
block (Fig. 9_5).(25) These "minidischarges" appear as random and rapidly
flickering arcs. The main discharge appears uninterrupted.
The limits of the current ranges described above further clarify the data
in Figs. 9-11 and 9-12. In Fig. 9-11, one current level was not possible over
50
•
40 •
~
c:
:::l
~ 30
~
:e-!!!.
i!' 20
"u;
c:
$
.!:
10
O~--------r-------.-------,--------.
x
--
x
--
3X
--
x
4 2 4
Matrix content
Figure 9-17. Dependence of the emission intensity of 100 pg of Cd (228.8 nm) on the com-
bined chloride matrix concentration, X, in HC-FANES atomized at 1800 K in 19 Torr Ar with
a discharge current of 30 mA. X = 1% rn/V NaCl, 1% rn/V KCl, 0.5% rn/V MgCl2 , and 0.5%
m/V CaCh.(8)
the range of pressures examined. Discharge currents of 50 and 70 mA caused

instability at the lower pressures while currents of 30 and 50 mA were
insufficient to reach the abnormal discharge mode at the higher pressures.
In Fig. 9-12, the integrated signals, at a given pressure, showed no variation
with current, but the background emission levels grew noisier toward the
upper and lower current limits. Consequently, the maximum SNRs are found
at intermediate current levels.
Two different cathode diameters have been employed for HA-
FANES. (25) The analytical signals appear to be dependent on the current
density. Comparable analytical signal levels for 1.0-mm- and 0.5-mm-
diameter cathodes were found for Cd and for Cu at currents of 50 and
25 mA, respectively.
The rf-FANES sources have been operated at 20-100 W.(20.21,24,27,28,30,31)
This compares with 60 W dissipated power for HA-FANES and 20 W for
HC-FANES. The lower power limit is that which is necessary to maintain
the plasma, and the upper limit, like that for the low-pressure discharges, is
determined by the appearance of the arcing between the electrodes. In gen-
eral, the analytical signal increases with increasing power. Sturgeon reported
the doubling of the Cd signal magnitude when power was increased from 50
to 75 W.(27)
406 CHAPTER 9
9.4.2. Noise Sources
Falk(l) predicted theoretically that the SNRs of nonlocal thermo-

dynamic equilibrium systems, such as HC-FANES, are greater than those
where LTE exists. In practice, Falk et al.(S) reported that the limiting noise
and, consequently, the detection limits of HC-FANES are determined not
by the recombinant continua but by extraneous molecular bands. These
bands provide intense and structured background and arise from molecules
from ambient gases that have entered the furnace through fiaws in the vac-
uum seal. Continuous pumping is required to maintain a constant pressure
during the atomization cycle. Although no specific molecules were identified
by Falk, the most likely species are those that Sturgeon et al.(2I) identified
for rf-FANES.
The rf-FANES source described by Sturgeon et al.(21,27) uses a commer-
cially available furnace that is not sealed against ambient atmosphere, but
depends on a positive internal fill-gas pressure. A positive internal
(125 ml min- I ) and external (1.5 liters min- I ) He fiow is used during the
atomization cycle. Sturgeon identified CO, OH, NH, and CN in the back-
ground spectra.
The CO and OH bands are not observed in HA-FANES emission spec-
tra, but CN and NH bands have been identified. The high vacuum enclosure
used for HA-FANES is superior to the vacuum enclosures of the other
sources discussed in this chapter. Slight changes in pressure have been
observed over extended periods, but these periods are orders of magnitude
greater than atomization cycle times. Consequently, the atomization step is
initiated under static conditions (constant pressure with no gas fiow). The
presence of CN bands appears to refiect an inherent problem arising from
frequent exposure of the source to ambient air when the sampIe is introduced
into the furnace. There is no apparent change when ultrapure Ar is substi-
tuted for the less pure and less expensive grade routinely used.
At higher analyte concentrations the reproducibility of the signals is
controlled by analyte flicker. The furnace atomization process is a major
source of this flicker. It appears that plasma excitation processes do not
introduce a measurable increase in signal variance when moderate current
levels are chosen. At higher current levels, where "minidischarges" are
observed, a significant decrease in reproducibility is noticed. The associated
increase in noise level is responsible for the deterioration of the SNR.
9.4.3. Interferences
Data on specific interferences for any of the F ANES sources are limited
at the present stage of development. It is conceivable that chemical, or
matrix, interferences will be more severe for F ANES than for conventional
GF-AAS since the technique combines the sophistication of the graphite
furnace atomizer with the complexity of sustaining a discharge process at
high temperatures and at reduced or at atmospheric pressure. For this reason
it is probable that chemical-matrix interferences will be more severe for
FANES than for conventional GF-AAS. The possibility exists, however,
that collisions with high-energy electrons from the glow discharge may pro-
vide the means of diminishing some of the common furnace atomization
interferences. Unquestionably, spectral interferences for F ANES will be
more severe than for GF-AAS, but less severe than for ICP-AES because of
the differences in the electron energies associated with the three sources.
The greatest potential for matrix interferences for any of the FANES
devices is the lack of stability of the discharge process throughout the atomiz-
ation cyc1e. Ideally, every atom must be subjected to the same excitation
potential or isoelectric conditions during its residence in the furnace. It has
been shown that the discharge potential is dependent on the discharge
pressure and furnace temperature (see Sections 9.4.1.2 and 9.4.1.3). Conse-
quently, if isothermal and isobaric conditions prevail during the atomization
process, the integrated signal will depend only on the analyte concentration.
In reality, furnace temperatures vary temporally and spatially (for the non-
ICC furnaces ofHC- and rf-FANES) and large localized pressure differenti-
als are produced within the furnace because of the rapid heating process. In
addition, the sampie matrix may introduce large numbers of ions into the
furnace atmosphere during the atomization step. Thus, the probability of
isoelectric F ANES operation seems unlikely.
At this time there is no evidence to demonstrate failure or dramatic
changes in the discharge during the atomization cyc1e for any of the devices.
Indirect evidence suggests that the various F ANES discharges may not be
as robust as hoped. This indirect evidence is the apparent need for the use
of the method of standard additions to achieve accurate determinations in
known materials. Falk et aIY S) have reported the successful determination
of Na in Al alloys, Ag in Au metal, and Cd in whole blood. In each case,
the method of standard additions was necessary to obtain accurate results.
Sturgeon et al. (27) found it necessary to use the method of standard additions
to determine Cd and Pb in marine sediment, dogfish musc1e, and lobster
hepatopancreas reference materials. Acceptable accuracy has yet to be
reported from analyses based on calibrations with aqueous or dilute acid
standards.
With respect to gas-phase interferences, the reduced pressure discharges
exhibit two distinct differences from GF-AAS. First, lower atomization tem-
peratures are associated with the reduced pressure. These result in lower
support gas temperatures and significantly fewer gas collisions. Reduced gas
collisions suggest the possibility of increased chemical interferences from
undissociated analyte molecules. Second, longer electron mean free paths
408 CHAPTER 9
are associated with reduced pressure. This allows for increased collisions
between nonfragmented analyte molecules and high-energy electrons.
Depending on the dissociation cross section relationship to this energy, ana-
lyte molecule dissociation can be enhanced.
Falk et al.(6) and Littlejohn et al.(8) reported that the flux of high-
temperature electrons in HC-FANES is useful in reducing halide inter-
ferences. They demonstrated that, for the determination of Cd, HC-FANES
tolerated concentrations ofNa, K, Ca, and Mg chlorides (1.0% ofNaCI and
KCl, and 0.5% of MgCh and CaCh) that are two orders of magnitude
greater than those acceptable for GF-AAS atomization from the wall (Fig.
9-17). In another study, Falk et al.(15) reported that a concentration of
0.025% NaCI was necessary to produce a 20% suppression of Cu and that
NaCI concentrations in excess of 0.065% were required to produce 20%
suppression of Co, Cr, Fe, and Ni signals. No suppression of the Al signal
was observed at NaCI concentrations extending to 0.25%. This study also
correlated the severity of the interference with the correspondence of the
atomization temperatures (overlap of the gas phases) of the analyte and the
interferent. The NaCl, having a low atomization temperature, most severe1y
affected the most volatile element, Cu. Effects of the NaCI on the other
elements diminished with decreased volatility. The same study determined
that recoveries of the same six elements were 100 ± 5% in up to 10% HN0 3 •
At 70% HN0 3, recoveries were within 100 ± 20%. These recoveries are
reasonable considering the acid concentration.
Blades and co-workers(24) reported that a concentration ofO.029% NaCI
produced a 20% suppression of the Ag signal in rf-FANES. The effect of
NaN0 3 was even greater with a concentration ofO.021% NaN0 3 yie1ding a
20% suppression of the Ag signal.
Another source of interference is the loss of analyte prior to initiation
of the atomization step. This is of greater concem for HC-FANES because
ofthe lower pressures. Falk et al.(15) reported early losses for standards and
sampies. Dittrich et al.(13.16) reported similar losses and improvements in
sensitivities and detection limits through the use of matrix modifiers. They
credited the improved signals to reduced preatomization analyte losses.
Analyte loss caused by sputtering prior to thermal atomization is unique
to the FANES process. This is of greatest concem for HC-FANES because
the sampie is placed directly on the cathode. Littlejohn et al. (8) reported
preatomization loss of Cd. Early loss of sulfur as carbon sulfide for standard
acid sulfide sampies has also been observed. A significant loss of P is
observed for rf-FANES if the discharge is ignited prior to initiating the
atomization temperature ramp. The severity of this type interference is
dependent on gas pressure, and the time interval between start of the dis-
charge and start of atomization.
Sputtering cannot contribute to analyte los ses in HA-FANES since the
analyte is deposited on the anode. An arc discharge from the cathode to the
sampie deposition site has been observed at the initiation of the discharge
in older furnaces subjected to several hundred firings. It is not clear whether
the dried sampie perturbed the graphite furnace sufficiently to act as an arc
admission point (electric field concentrator) or whether the pyrolytic coating
had been roughened in this area to expose a more conductive surface.
Reproducibility of the analytical signals appears to be worse when this
phenomenon occurs. New furnaces remedy the problem.
Preatomization sputtering losses for rf-FANES appear unlikely. Mean
free paths for the cations are small at atmospheric pressure and the low mass
ofHe makes it an inefficient primary ion for sputtering. Arcing to the sampie
deposition site at the initiation of the discharge has been reported. This
problem was alleviated by using a graphite platform within the furnace. The
edges of the platform probably served as field concentration sites for dis-
charge initiation.
Emission spectra of the F ANES devices are less complex than for ICP-
AES, but they are more complex than the simple absorption spectra observed
for GF-AAS. Even the broad support gas continuum can present a problem.
The magnitude of this continuum increases as a function of increasing
temperature. Thus, the background varies throughout the temperature atom-
ization ramp. There is, therefore, adefinite need for real-time background
correction.
Wavelength modulation has been successfully used with HA-
FANES. (18,15.29) The narrow duration of the analytical signals for F ANES
suggests that the frequency of modulation (56 Hz) used for GF-AES and
GF-AAS is not sufficiently rapid.(18,25) Falk et al.(4) employed three-step
square-wave modulation at 130 Hz, and sine-wave modulation at 200 Hz
has been used with HA-FANES. The square wave has a theoretical SNR
advantage of a factor of 1.8 relative to sine-wave modulation. (63)
Falk et al. (4) reported that wavelength modulation as compared with
intensity modulation yielded worse detection limits for aseries of spectral
resolutions. The intensity modulation, however, did not provide a back-
ground correction. If a background value is subtracted, the limiting noise
will increase by a factor of 1.4, assuming a quadratic addition of independent
noise sources. Correcting the intensity-modulated detection limits by this
factor yields comparable values for both methods.
This study just cited concluded that the best SNRs occurred at a resolu-
tion of 1-2 x 104 • This resolution corresponds to a slit width of 200 pm for
the echelle grating used. A consistent improvement of the SNRs with larger
slit widths was found for four elements. This is unexpected for either the
statisticalor fluctuation noise-limited cases. The improvement may be associ-
ated with the larger viewed region of the furnace.
Structured spectral interferences as opposed to broadband interferences
arise from the support gas and strong molecular spectra from entrained
air. The structured overlap can be reduced through use of high-resolution
410 CHAPTER 9
spectrometers. Wavelength modulation will also be more susceptible to

structured interferences because of the wider spectral range that is viewed.
Interferences for Cr at 357.9 nm (CN band) and at 425 nm [Ar (I) line]
were observed for a medium-resolution monochromator with wavelength
modulation. Sturgeon et al.(27) reported an interference from Fe at the Pb
line of 283.3 nm for rf-FANES. Sturgeon indicated that ICP wavelength
tables were useful for predicting the presence of potential spectral inter-
ferences but were not appropriate for rf-FANES with respect to the relative
intensities.
9.4.4. Figures of Merit
9.4.4. 1. Detection Limits
A tabulation of reported detection limits for the F ANES systems are

shown in Table 9-4. Detection limits for conventional GF-AAS have been
inc1uded for comparison. All results were obtained under conditions optim-
ized individually for each element. The operating conditions under which
the data in Table 9-4 were acquired are shown in Table 9-2. All values have
been corrected to be consistent with a definition of three standard deviations
(3a) of the background noise for the detection limit.
For the limited number of elements for which a comparison is possible,
the similarities among the detection limits for thc various F ANES methods
are remarkable considering the disparate operating parameters. It can be
seen that the detection limits for Ag, As, Bi, Cd, Co, Pb, Se, and Zn are
very similar. B, Cr, Cu, Fe, Ni, Tl, and V (tor HC- and rf-FANES) show
large discrepancies. It must be remembered that the results for HA- and rf-
F ANES are very preliminary.
All the F ANES detection limits can best be characterized as "compar-
able" to those for GF-AAS, with one notable exception. Falk et al.(5) have
shown, using a statistical approach, that the HC-FANES detection limits
are better than those for GF-AAS by a factor ofthree. A direct comparison
of the data in Table 9-4 shows that, relative to GF-AAS, the detection
limits of HC-FANES are better by a factor of three for eight elements
(predominantly the alkalis and alkaline earths), worse by a factor of three
for seven elements (predominantly the nonvolatiles), and comparable (less
than a factor ofthree different) for eight elements. The rf-FANES detection
limits reported by Sturgeon(21,27,28,30,31) and compared to GF-AAS arebetter
for two elements, worse for three elements, and comparable for eight
elements.
The notable exception mentioned above is B. HA- and HC-FANES
detection limits for Bare 750 times lower than that for conventional GF-
AAS. The reason for the large difference is not completely undetstood,
Table 9-4. Metal Detection Limits (pg)

PANES
HC HA rf
pa N D L H B S GP-AAS
Ag 0.6 4. 0.3 0.5* 0.8*
AI 22. 6.*
As 103.* 30.*
Au 4. 15.*
B 2. 80. 2. 1500.*
Be 5.* 2.*
Bi 45. 25. 15.
Ca 0.08 8.
Cd 0.6 1. 2. 0.9* 0.4*
Co 5. 4. 3.*
Cr 6. 0.5 0.8 2.*
Cu 1. I. b 0.5 5. 2.
Dy 200. 57.
Er 220. 100.
Eu 100. c 24.
Fe 7. 0.7 50. 3.*
Ga 0.6
Ho 440.
K 0.04 3.
Li 0.03 3.
Lu 10,800.
Mg 3.
Mn 0.9* 2.*
Mo 600. 32. 6.
Na 0.06 <75.
Ni 2. l.d 164. 8.*
Pb 4. 12. 7.2* 4.*
Rb 2. 8.
Sc 30.
Se 800. 1700. 45.
Sm 70. c 240.
Sn 10.* 30.*
Tb 3,100.
Tc 90.* 13.
Ti 1200. 75.
Tm 300.
TI 0.3 17.* 15.*
V 8. 184. 30.
Y 2,600. 13,000.
Yb 20. 2.
Zn 3. 4. 0.8*
aF, Falk et alY5); N, Naumann et al.(l7); D, Dittrich et al.(I3,16,26); L, Littlejohn et alY4); H, Harnlyet
al. (25,29); B, Blades et al. (24); S, Sturgeon et alY""JO)
b327.4nm,
'Ion line.
d352.4 nm.
*Matrix modifier used.
412 CHAPTER 9
although initial data suggest that collisions with the high-energy electrons
break down B20 3 , which is lost from the fumace using straight fumace
atomization with GF-AAS.
9.4.4.2. Reproducibi/ity
The reproducibility is defined as the relative standard deviation for
the analyte signals weH above (greater than ten times) the detection limit.
Reproducibility runs from 2% to 4% and is element dependent for GF-AAS
using autosampiers. At these levels (greater than ten times the detection
limit), the reproducibility is dependent on the concentration, i.e., signal
uncertainty is dependent on signal strength. The source of signal error is
generally attributed to the uncertainty of the sampie position after the drying
process. Platforms in the fumace restrict the sampie position but produce
smaller, broadened signals.
Standard deviations reported by Falk et al.(15) for HC-FANES are
2~3%. Hamly et al.(25) reported values of 2~6% for the HA-FANES.
Sturgeon et al.(21,27) reported rf-FANES relative precisions ranging from 3
to 10% at atmospheric pressure. Although the upper values in these last two
cases (6 and 10%, respectively) may seem high relative to GF-AAS, it must
be remembered that these values were obtained using manual sampIe delivery
without the use of platforms. The existing data show that the discharge only
has an adverse efIect on analytical precision at the extremes of the current
range (see Section 9.4.1.4).
9.4.4.3. Dynamic Range

The available data suggest that the dynamic ranges for the F ANES
devices are not comparable to those observed for ICP-AES (five to six orders
ofmagnitude). The major disadvantage for FANES is that the source is not
opticaHy thin. The fumace is 19~28 mm long and is occupied predominantly
by ground state atoms that are capable of absorbing the emitted photons.
This situation is made worse for HC- and rf-FANES by the nonuniformity
of the temperature along the fumace length. The temperature at the end of
the fumace can be several hundred degrees cooler than in the middle. (38,39)
This gives rise to condensation near the ends of the fumace and further
enhances self-absorption processes. This effect was so severe for GF-AES
that for some elements it is impossible to observe ground state transitions
that are not self-reversed, even at low concentrations. (64) This is not a prob-
lem for HA-FANES, which uses isothermal fumaces (lCC). This fumace is
shorter and has a uniform temperature along its length. The nonoptimum
geometry may be partiaHy remedied for HC-FANES by operation at reduced
pressures (l~5 Torr).
Working ranges for HC-FANES for Ag, Na, and Pb cover 5-6 orders
of magnitude of concentration with less than a 10% deviation from
linearity.(3.5) Data for Cd show curvature occurring, however, after 2 orders
of magnitude. (6) Dittrich et al. (16) reported calibration ranges of 3 orders of
magnitude for the rare earth elements. Dynamic ranges for rf-FANES also
show reduced linearity. Sturgeon has observed linear ranges of 2-4 orders
ofmagnitude, and Blades et al.(24) reported a usable concentration range for
Ag of 4 orders of magnitude with linearity over 2-3 orders of magnitude.
Initial data for HA-F ANES suggest dynamic ranges of 3-4 orders of magni-
tude at discharge pressures of 70-200 Torr in Ar.
9.5. Applications
Considering the body of literature to be found for the various F ANES

devices, there are relatively few applications of the technique. To a large
extent, this has been due to the general lack of availability of the instrumenta-
tion. In the last two years, however, the number of operational F ANES
devices has tripled. Hopefully, this trend will continue and the number of
applications-oriented research papers will increase dramatically.
9.5.1. Meta! Determinations
One ofthe earliest systematic applications ofHC-FANES was the deter-

mination of Cd in deproteinized whole blood. (6) The results, obtained using
the method of standard additions for two different sampies, were in excellent
agreement with those obtained by GF-AAS (97 and 99% recovery). These
determinations were made using 19 Torr Ar support gas pressure, a dis-
charge current of 30 mA, and an atomization temperature of 900 K, fairly
standard conditions for HC-FANES (Table 9-2).
A study of the effect of the metal chloride concentrations on the
analytical determination was discussed previously (Section 9.4.3). The data
suggested that the inorganic component (as chlorides) was unlikely to pro-
vide an interference. Spike recoveries in the deproteinized whole blood, how-
ever, gave only approximately 20% recovery. It was conc1uded that other
inorganic or organic components were responsible for the low recoveries
and necessitated the use of the method of additions.
Falk et al.(9) determined Cu, Fe, and Ni in grass and com (Soviet
Union reference materials) and orchard leaves (SRM 1571, USA). These
determinations were made simultaneously using solid sampIes and a specially
designed fumace. The results were very acceptable with elemental recoveries
of 92% to 103% in the orchard leaves and 92% to 140% in the grass and
414 CHAPTER 9
corn standards. Argon was used as the support gas at 25.4 Torr with a
discharge current of 30 mA and an atomization temperature of 2700 K.
In another simultaneous multielement study, Falk et al.(15) determined
Al, Co, Cr, Cu, Fe, and Ni doped into deionized water and HN0 3 • The
recoveries were generally low for the deionized water and were attributed to
the preatomization loss of the elements (see Section 9.4.3). These results led
to a detailed study on the optimum drying and ashing conditions. The
subsequent recoveries in HN0 3 were within ±10% of 100% for HN0 3 con-
centrations up to 15%. The analytical conditions used were 15.4 Torr of
Ar as the support gas, a discharge current of 30 mA, and an atomization
temperature of 2300 K.
Dittrich et al.(16) recently employed HC-FANES for the determination
of 11 rare earth elements (Dy, Er, Eu, Ho, Lu, Sc, Sm, Tb, Tm, Y, and Yb)
in the single-element mode using both atomic and ionic transitions. The
detection limits ranged from 20 to 530 pg for all the elements except Y, Tb,
and Lu, which were 2600, 3100, and 10,800 pg, respectively (Table 9-4). In
the presence of other rare earth elements the emitted intensities for each of
the elements were suppressed. This placed the detection limits in a rare earth
matrix at 10 to 1000 ng. Ar was used as the support gas at 12 Torr with
a discharge current of 30 mA and atomization temperatures of 2100 and
3000 K.
Dittrich et a1Y 3 ) also examined the determination of 99Tc using FANES.
They obtained a detection limit of90 pg using Ni(N0 3 h as a matrix modifier
and a tungsten platform. The matrix modifier helped prevent preatomization
losses and losses caused by molecular formation. The platform reduced the
formation of carbides. Optimum atomization conditions were obtained for
11 Torr Ar, a discharge current of 30 mA, and a temperature of 2900 K.
In another interesting study, Dittrich et al.(26) determined Sb using a
hydride HC-FANES method. The use of cold (external trap with liquid
nitrogen) versus hot (in the graphite furnace) trapping was investigated
along with atomization with and without cooling the furnace after the trap-
ping. A detection limit of 14 pg was obtained using a hot trap and immediate
atomization (no cooling) with an Ar pressure of 10 Torr, a discharge current
of 60 mA, and an atomization temperature of 1400 K. The detection limit for
GF-AAS is cited as 22 pg with straight atomization and matrix modification.
Using rf-FANES, Sturgeon et al.(27) determined Cd and Pb in three
reference materials from the National Research Council of Canada: marine
sediment (BCSS-l), dogfish mussei (DORN-I), and lobster hepatopancreas
(TORT-I). Aqueous calibration proved satisfactory for Cd, using peak area
measurements, but for Pb it was necessary to employ the method of standard
additions to obtain accurate results. Recoveries, with the method of standard
additions, ranged from 83% to 105% for peak area and 104% to 121% for
peak height. These results were obtained using a He support gas at 760 Torr,
an atomization temperature of 1400 K, and 50 and 75 W plasma powers for
Cd and Pb, respectively.
9.5.2. Nonmetal and Molecular Determinations
One of the attractive features of the F ANES devices is the ability to

determine the nonmetals under conditions similar to those used for metals.
Dittrich et al. (7,11) investigated the atomic emission of Br, Cl, and Fand
their tnolecular emission as In and Mg compounds. The molecular emission
detection limits (Table 9-5) are factors of 2 to 30 times better than those for
atomic emission. The detection limit for Cl was slightly better as a Mg
compound while the In complex was preferable for F. The atomic emission
detection limits agree with those reported by Littlejohn(14) but the detection
limit for Cl was considerably worse than that reported by Falk et al. (15) The
molecular emission detection limits for HC-FANES were also found to be
slightly better than those obtained using molecular absorption with furnace
volatilization.(7) The analytical sensitivity for each of the halides was found
to be significantly suppressed by the presence of the other halides. These
data were obtained using either Ar or He at 15 Torr, a 30 mA discharge
current, and atomization temperatures of 2100 K for In and 2400-2700 K
for Mg.
Dittrich and Fuchs(22.23) investigated the possibility of determining P in
the atomic and molecular emission modes. For atomic emission, the HC-
F ANES detection limit (600 pg) was improved by a factor of 7 with the use
of La as a matrix modifier (90 pg). The improved detection limit with the
matrix modifier was attributed to greater atomization efficiency, i.e., reduced
losses from preatomization evaporation and from moleeule formation. The
optimal atomization parameters for P were Ar at 15 Torr, 40 mA, and
Table 9-5. Nonmetal Detection Limits (pg)

HCa
D L
F
Element Atom Atom. Mole. Atom. Mole. GF-AAS
Br 12,000 6000· 23,000
Cl 120 8,000 240· 6,800 5100
F 250· 38,000
I 9,000 4,800
P 90· 700· 210 4500·
S 4000
Si 7,400
auc. UC-FANES; F, Falk et al.(lS); D, Dittrich and Fuchs(II.22.23); L. LittlejohnY4)
*Matrix modi/kation.
416 CHAPTER 9
2100 K. The optimal parameters for P in the presence of La were Ar at

15 Torr, 80 mA, and 2300 K. Dittrich and Fuchs(23) also examined the
molecular emission of P as PO and HPO. While the detection limits for both
PO and HPO were found to be better than those for GF-AAS, they were
not as good as those found for the atomic emission of P in the presence of
La. For these molecules the discharge current was found to represent a
compromise; increased current led to better excitation but also led to greater
dissociation of the molecules. For PO, the most intense emission was
observed at 15 Torr of Ar, 80 mA, and 1700 K.
Littlejohn has examined the feasibility of determining S as CS. No
signals were detectable for the atomic S. The CS band at 257.9 nm proved
the most sensitive. A major problem was the change in the background
continuum as a function of temperature. The emission bands were too broad
to permit use of wavelength modulation. A detection limit of 4 ng was
obtained.
9.6. Simultaneous Multielement Determinations
A predicted advantage for any of the F ANES devices, even if the detec-
tion limits are only comparable to GF-AAS, is the potential for multielement
determinations. The simplified optical requirements of the emission mode
and the available technology for multichannel spectrometers would seem to
place multielement fumace determinations within reach. At this time the
future of multielement F ANES is uncertain. This is the result of the lack of
data characterizing the dependency of the analytical signals and the SNR
on the atomization temperature and the discharge pressure and current.
Perhaps even more critical, data that characterize the compromise in the
accuracy of the determinations (the presence of interferences) with respect
to the operating parameters are not available.
Simultaneous multielement determinations have been investigated by
Falk et al.(IS) in two different experiments. The first study determined six
elements (Al, Cr, Cu, Fe, Mn, and Ni) doped in deionized water; this per-
mitted the effect of matrix interferences to be ignored. The second study
determined six elements (Al, Co, Cr, Cu, Fe, and Ni) doped in HN03 and
in NaCl.
In the first study the compromise atomization parameters were a 4-s
atomization at 2100 K, Ar support gas at 19 Torr, and a discharge current
of 30 mA. A conventional multielement spectrometer was used with the
adjunct computer-controlled data acquisition system. The recovery of 5 and
50 pg liter-I additions ranged from 60% to 124%, with all but two of the
values falling between 81% and 97%. Fe provided the highest recovery
(124%) and Al the lowest (60%). The generally low recoveries were attributed
to excessive pretreatment temperatures and system evacuation before
complete drying was accomplished.
The second study, in HN0 3 and NaCl, employed almost the same
atomization conditions (a support gas pressure of 15 Torr was used instead
of 19 Torr) and the same spectrometer and data acquisition system. The
drying and ashing temperature programs were carefully optimized for the
element to be analyzed. The multielement detection limits ranged from 5 to
30 times worse than the single-element values previously reported. These
poor detection limits were the result of the long integration time (1.5 s) of
the data acquisition system, which was an order of magnitude too large for
most elements. Optimum time gating for each element was not possible with
the adjunct computerized data acquisition system (the detection system was
designed for another source). The relative precisions of the elements ranged
from 2% to 3% for the short term (within experiment) and from 4% to 10%
for the long term (day-to-day). These values are consistent with results
observed for single-element operation. The linear range was 2.5 to 3.5 orders
of magnitude for all six elements using the poorer detection limits, and 3.5
to 4.5 orders of magnitude if extrapolated to the best detection limits. The
recoveries of the six elements in HN0 3 and NaCl have been discussed in
detail in Section 9.4.3. The severity of the interferences were related to the
elemental volatility and seemed little affected by the compromise multi-
element atomization conditions.
Naumann et a1Y 7 ) performed optimization studies for HC-FANES
using Co, Cr, Cu, and Ni. Optimum single-element conditions were found
to be temperatures of 2123,2161,2310, and 2623 K, pressures of 102, 141,
40, and 116 Torr, and currents of 59, 90, 90, and 90 mA for Co, Cr, Cu, and
Ni, respectively. The compromise multielement conditions, optimized from
the SNR, were an atomization temperature of 2315 K, an Ar pressure of
120 Torr, and a discharge current of 68 mA. The compromise detection
limits were degraded by less than a factor of 2. No analyses were undertaken
using these parameters.
Initial single-element results for HC- and rf-FANES suggest that the
optimum parameters found for the nonmetals are not significantly different
from those for the metals.
It is intriguing to consider the possibility of simultaneous, state-of-the-
art detection limits for both metals and nonmetals. An accurate assessment,
however, of the simultaneous multielement capabilities of the F ANES must
await further studies. These studies must include a wider selection of ele-
ments including the nonmetals, a spectrometer system optimized for the
rapid F ANES signals and associated wavelengths, and determinations in a
large variety of reference materials.
418 CHAPTER 9
Acknowledgment
Pacific Northwest Laboratory operated for the V.S. Department of

Energy at the Battelle Memorial Institute under Contract DE-AC06-76RLO
1830. .
References
1. H. Falk, Einige theoretiche uberlegungen zum vergleich der physikalischen grenzen therm-
ischer und nicht-thermischer spektroskopischer strahlungsquellen, Spectrochim. Acta 32B
(1977) 437.
2. H. Falk, E. Hoffmann, I. Jaeckel, and C. Ludke, Atomic emission trace analysis by non-
thermal excitation, Spectrochim. Acta 34B (1979) 333.
3. H. Falk, E. Hoffmann, and C. Ludke, F ANES (furnace atomic non-thermal excitation
spectrometry)-a new emission technique with high detection power, Spectrochim. Acta
36B (1981) 767.
4. H. Falk, E. Hoffmann, C. Ludke, J. M. Ottaway, and S. K. Giri, Furnace atomization
with non-thermal excitation----experimental evaluation of detection based on a high-resolu-
tion echelle monochromator incorporating automatie background correction, Analyst 108
(1983) 1459.
5. H. Falk, E. Hoffmann, and C. Ludke, A comparison offurnace atomic non-thermal excita-
tion spectrometry (FANES) with other atomic spectroscopic techniques, Spectrochim. Acta
39B (1984) 283.
6. H. Falk, E. Hoffmann, C. Ludke, J. M. Ottaway, and D. Littlejohn, Studies on the determi-
nation of cadmium in blood by furnace atomic non-thermal excitation spectrometry, Ana-
lyst 111 (1986) 285.
7. K. Dittrich, B. Hanisch, and H.-J. Stark, Moleeule formation in electrothermal atomizers:
Interferences and analytical possibilities by absorption, emission and fluorescence processes,
Fresenius Z. Anal. Chem. 324 (1986) 497.
8. D. Littlejohn, J. Carroll, A. M. Quinn, J. M. Ottaway, and H. Falk, Comments on the
characteristics of an atomizer for furnace atomic non-thermal excitation spectrometry
(FANES), Fresenius Z. Anal. Chem. 323 (1986) 762.
9. H. Falk, E. Hoffmann, C. Ludke, and K. P. Schmidt, Untersuchungen zur direktanalyse
fester pflanzlicher stoffe mittels FANES (furnace atomic non-thermal excitation spectro-
metry), Spectrochim. Acta 41B (1986) 853.
10. H. Falk and J. Tilch, Atomization efficiency and over-all performance of electrothermal
atomizers in atomic absorption, furnace atomization non-thermal excitation and laser-
excited atomic fluorescence spectrometry, J. Anal. At. Spectrom. 2 (1987) 527.
11. K. Dittrich and H. Fuchs, Molecular non-thermal excitation spectrometry (MONES) : A
procedure for the determination of non-metals using diatomic molecules in the non-thermal
(FANES) atomizer; Part 1. Determination of fluoride and chloride ions by magnesium
fluoride and magnesium chloride MONES, J. Anal. At. Spectrom. 2 (1987) 533.
12. H. Falk, Hollow-cathode discharge within a graphite furnace: Furnace atomic non-thermal
excitation spectrometry (FANES), in: Improved Hollow Cathode Lamps for Atomic Spectro-
scopy (S. Caroli, ed.), pp. 74-118, EIlis Horwood Ltd., Halstead Press, New York, 1987.
13. K. Dittrich, T. Glaubauf, H. Fuchs, and K. Mauersberger, Analytical applications of
furnace atomization non-thermal excitation spectrometry (FANES) and molecular non-
thermal excitation spectrometry (MONES) ; Part 2. Determination of technetium-99 by
FANES and electrothermal atomization atomic absorption spectrometry, J. Anal. At. Spec-
trom. 3 (1988) 89.
14. D. Littlejohn, Graphite fumace atomic emission spectrometry-the rediscovery of a tech-
nique, Anal. Proc. 25 (1988) 217.
15. H. Falk, E. Hoffmann, and C. Ludke, Experimental and theoretical investigations relating
to FANES, Prog. Anal. Spectrosc. 11 (1988) 417.
16. K. Dittrich, G. Eismann, and H. Fuchs, Analytical applications of fumace atomization
non-thermal excitation spectrometry (FANES) and molecular non-thermal excitation spec-
trometry (MONES) ; Part 3. Determination of rare earth elements by electrothermal atomi-
zation atomic emission spectrometry (ETA-AAS), FANES and fumace ionization non-
thermal excitation spectrometry (FINES), J. Anal. At. Spectrom. 3 (1988) 459.
17. B. Naumann, B. Knull, F. Kerstan, and J. Opfermann, Multivariate optimization ofsimul-
taneous multi-element analysis by fumace atomic non-thermal excitation spectrometry
(FANES), J. Anal. At. Spectrom. 3 (1988) 1121.
18. N. E. Ballou, D. L. Styris, and J. M. Hamly, Hollow-anode plasma excitation source for
atomic emission spectrometry, J. Anal. At. Spectrom. 3 (1988) 1141.
19. D. Littlejohn, Becoming absorbed and excited in atomic spectrometry, Anal. Proc. 26
(1989) 92.
20. D. C. Liang and M. W. Blades, An atmospheric pressure capacitively coupled plasma
formed inside a graphite fumace for atomic emission spectroscopy, Spectrochim. Acta 44B
(1989) 1059.
21. R. E. Sturgeon, S. N. Willie, V. Luong, S. Berman, and J. G. Dunn, Fumace atomization
plasma emission spectrometry (FAPES), J. Anal. At. Spectrom. 4 (1989) 669.
22. K. Dittrich and H. Fuchs, Analytical applications of fumace atomic non-thermal excitation
spectrometry (FANES) and molecular non-thermal excitation spectrometry (MONES) ;
Part 3. Determination oftrace amounts ofphosphorous by FANES, J. Anal. At. Spectrom.
4 (1989) 705.
23. K. Dittrich and H. Fuchs, Analytical applications of fumace atomic non-thermal excitation
spectrometry (FANES) and molecular non-thermal excitation spectrometry (MONES);
Part 5. Study of the MONES of PO and HPO for the determination of trace amounts of
phosphorous, J. Anal. At. Spectrom. 5 (1990) 39.
24. D. L. Smith, D. C. Liang, D. Steel, and M. W. Blades, Analytical characteristics of fumace
atomization plasma excitation spectrometry (FAPES), Spectrochim. Acta 45B (1990) 493.
25. J. M. Hamly, D. L. Styris, and N. E. Ballou, Fumace atomic non-thermal excitation
spectrometry with the fumace as a hollow anode, J. Anal. At. Spectrom. 5 (1990) 139.
26. K. Dittrich, B. Radziuk, and B. Welz, Investigations of the determination of chloride and
bromide by fumace atomic non-thermal excitation spectrometry and fumace ionic non-
thermal excitation spectrometry, Spectrometry 6 (1991) 465.
27. R. E. Sturgeon, S. N. Willie, V. T. Luong, and S. S. Berman, Determination of cadmium
and lead in sediment and biota by FAPES, J. Anal. At. Spectrom. 5 (1990) 635.
28. R. E. Sturgeon, S. N. Willie, V. Luong, and S. S. Berman, Figures of merit for fumace
atomization plasma emission spectrometry, Anal. Chem. 62 (1990) 2370.
29. P. G. Riby, J. M. Hamly, D. L. Styris, and N. E. Ballou, Emission characteristics of
chromium in hollow anode-fumace atomization non-thermal excitation spectrometry, Spec-
trochim. Acta 46B (1991) 203.
30. R. E. Sturgeon, S. N. Willie, V. T. Luong, and S. S. Berman, Application of platform and
palladium modification techniques with fumace atomization plasma emission spectrometry,
J. Anal. At. Spectrom. 6 (1991) 19.
31. R. E. Sturgeon, S. N. Wi11ie, V. T. Luong, and S. S. Berman, Characteristic temperatures
in a FAPES source, Spectrochim. Acta 46B (1991) 1021.
32. C. T. J. Alkemade, T. J. Hollander, W. Snelleman, and P. J. T. Zeegers, Metal Vapors in
Flarnes, Pergamon Press, Elmsford, N.Y., 1982.
33. R. E. Sturgeon and S. S. Berman, Analyte ionization in graphite fumace-atomic absorption
spectrometry, Anal. Chem. 53 (1981) 632.
420 CHAPTER 9
34. W. Slavin, Graphite Furnace AAS-A Source Book, Perkin-Elmer Corp., Norwalk, Conn.,
1984.
35. J. M. Ottaway and F. Shaw, Carbon fumace atomic-emission spectrometry: A pre1iminary
appraisal, Analyst 100 (1975) 438.
36. M. S. Epstein, T. C. Rains, and T. C. O'Haver, Wavelength modulation for background
correction in graphite furnace atomic emission spectrometry, Appl. Spectrosc. 30 (1976)
324.
37. W. Frech, D. C. Baxter, and B. Hutsch, Spatially isothermal graphite furnace for atomic
absorption spectrometry using side-heated cuvettes with integrated contacts, Anal. Chem.
58 (1986) 1973.
38. R. E. Sturgeon and S. S. Berman, Determination of the efficiency of the graphite furnace
for atomic absorption spectrometry, Anal. Chem. 55 (1983) 190.
39. B. Welz, M. Sperling, G. Schlemmer, N. Wenzel, and G. Marowsky, Spatially and tempor-
ally resolved gas phase temperature measurements in a Massman-type graphite tube furnace
using coherent anti-Stokes Raman scattering, Spectrochim. Acta 43B (1988) 1187.
40. P. J. SIevin and W. W. Harrison, The hollow cathode discharge as a spectrochemical
emission source, Appl. Spectrosc. Rev. 10 (1975) 201.
41. S. Caroli, Low-pressure discharges: Fundamental and applicative aspects, J. Anal. At.
Spectrom. 2 (1987) 661.
42. M. E. Pillow, A critical review of spectral and related physical properties of the hollow
43. J. A. C. Broekaert, State of the art of glow discharge lamp spectrometry, J. Anal. At.
Spectrom. 2 (1987) 537.
44. K. G. Hernquist and E. O. Johnson, Retrograde motion in gas discharge plasmas, Phys.
Rev. 98 (1955) 1576.
45. H. L. Wittig, Hollow cathode discharge with thermionic cathodes, J. Appl. Phys. 42 (1971)
5478.
46. S. N. Salinger and J. E. Rowe, Monte Carlo simulation of the low-voltage are mode in
plasma diodes, J. Appl. Phys. 39 (1968) 3933.
47. R. M. Martin and R. E. Rowe, Experimental investigations ofthe low-voltage arc in noble
gases, J. Appl. Phys. 39 (1968) 4289.
48. G. K. Wehner, J. Appl. Phys. 31 (1960) 1392.
49. M. Kaminsky, Atoms and Ionic Impact Phenomena on Metal Sur/aces, p. 143, Academic
50. A. von Hippei, Kathodenzerstaubungsprobleme III zur theorie der kathodenzerstaubung,
Ann. Phys. 81 (1976) 1043.
51. D. Bohm, Minimum ionic kinetic energy for a stable sheath, in: The Characteristics 0/
Electrical Discharges in Magnetic Fields (A. Guthrie and R. K. Walkerling, eds.), pp. 77-86,
McGraw-Hill, New York, 1949.
52. F. F. Chen, Introduction to Plasma Physics, Plenum Press, New York, 1949.
53. B. Chapman, Glow Discharge Processes, Wiley, New York, 1980.
54. A. von Engel, Ionized Gases, pp. 214-228, Oxford University Press, London, 1955.
55. C. P. Herring and M. H. Nichols, Thermionic emission, Rev. Mod. Phys. 21 (1949) 185.
56. I. Langmuir, The effect of space charge and residual gases on the thermionic current in
high vacuum, Phys. Rev. 2 (1913) 450.
57. P. F. Little and A. von Engel, The hollow cathode effect and the theory of glow discharges,
Proc. R. Soc. London Sero A 224 (1954) 209.
58. S. D. McDonald and S. J. Tatenbaum, High frequency and microwave discharges, in:
Gaseous Electronics (M. M. Hirsh and H. G. Oskam, eds.), Vol. I, pp. 173-217, Academic
59. L. Holland, W. Steckelmacher, and Y. Yarwood, Vacuum Manual, pp. 384-385, Spon,
London, 1974.
60. J. M. Keller and W. B. Pennebacker, Electrical properties of rf sputtering systems, IBM J.
Res. Dev. 23 (1979) 3.
61. V. P. Gofmeister and Y. M. Kagaan, Mechanism of excitation in a hollow cathode in
argon, Opt. Spectrosc. (USSR) (Engl. Transi.) 26 (1969) 379.
62. H. M. Crosswhite, G. H. Dieke, and C. S. Legagneur, Hollow iron cathode discharge as
a source for wavelength and intensity standards, J. Opt. Soc. Am. 45 (1955) 270.
63. T. C. O'Haver, Waveform effects in wavelength modulation spectrometry, Anal. ehern. 49
(1977) 458.
64. J. E. MarshalI, D. Littlejohn, J. M. Ottaway, J. M. Harnly, N. J. Miller-Ihli, and T. C.
O'Haver, Simultaneous multi-element analysis by carbon furnace atomic-emission spectro-
metry, Analyst 108 (1983) 178.
10
Laser-Based Methods
Kenneth R. Hess
10.1. Introduction
The development of cost-effective laser systems has generated a host of

laser hyphenated techniques that have been introduced into the analytical
laboratory. These hyphenated techniques take advantage ofthe laser's ability
to deliver a high photon Bux, high photon energies, and a narrow, tunable
photon wavelength range to optimize specific atomizationjexcitationjioniza-
tion processes in an analytical procedure. A laser system coupled to a glow
discharge is one such hybrid technique that offers unique opportunities for
both diagnostic and analytical investigations. This chapter will serve as an
introduction and overview of several reported methodologies that have
advantageously combined laser systems and glow discharges.
The analytical utility of the glow discharge results in part from a two-
step method of forming analytical species of interest. In the first step, the
solid cathode material is sputtered from the surface by the impacting ions
of the rare gas employed in the discharge .. The sputtering process is largely
nonselective, with relative elemental atomization rates for this step varying
by less than an order of magnitude. In addition, it has been postulated that
under steady-state conditions the elemental sputter rates become equivalent,
as discussed in Chapter 2. Overall, the sputtering process creates an atomic
population that is highly representative of the bulk cathode composition,
providing an excellent atom source for atomic spectroscopy. The second step
Kenneth R. Hess • Department of Chemistry, Franklin and Marshall College, Lancaster,

Pennsylvania 17604
423
424 CHAPTER 10
of the process involves the collisional excitation/ionization of the atomic

population, generating excited-state sputtered atoms for atomic emission and
ions for mass spectrometry. The ionization step is also relatively nonselective,
providing an ionic population that is representative of the atomic
population.
A laser system may be coupled to the discharge in order to enhance
either of these two steps. First, the high power and spatial resolution of the
laser may be advantageously employed to ablate material from the sampie
cathode, enhancing sampie atomization for subsequent excitation and ioniz-
ation in the discharge. Alternatively, the laser may ablate material from a
secondary sampie (not the cathode) into the discharge for excitation and
ionization. This method would not require the sampie to be conducting,
unlike the cathode of anormal dc glow discharge, which is required to be
electricaIly conducting. This expands the possible analytical applications of
glow discharge excitation and ionization to nonconducting materials without
employing rf discharges.(I,2) Second, the high photon energy and tunability
of the laser may be employed as a method for enhancing the excitation
processes in the discharge. Coupling of the laser to the discharge for sub-
sequent excitation of the discharge sputtered atoms has been effectively dem-
onstrated in a variety of application areas including atomic fluorescence,
optogalvanic effect spectroscopy, metastable atom depopulation studies,
resonance ionization, and degenerate four-wave mixing experiments. Either
method of interfacing laser systems to the glow discharge source, enhanced
atomization or enhanced excitation, provides opportunities to investigate
fundamental discharge processes, with several of the laser/discharge tech-
niques found also to be of analytical utility. Basic discussions of laser
enhancement of atomization and laser enhancement of excitation, with rep-
resentative applications, are provided in this chapter as an introduction to
the topic.
10.2. The Laser
Currently, there are many different types of laser systems available for
interaction with the glow discharge. The key characteristics that provide the
exceptional utility of the laser include: (l) the high spatial coherence and
directional properties of the beam, which allow easy manipulation and focus-
ing of the beam, providing high photon densities and spatial resolution, (2)
a high photon flux, which allows processes with low cross sections, such as
multiphoton events, to be investigated, (3) high photon energies combined
with the high photon flux, which will allow high energy densities for pro-
cesses requiring photon-induced damage, such as laser ablation, (4) with the
use of a dye laser and frequency doubling crystal, monochromatic, tunable
LASER-BASED METHODS 425
photon wavelengths from the UV to IR are available to optimize specific
excitation or atomization processes in an analytical procedure, and (5) the
laser pulse duration can be very fast, down to the femtosecond range, allow-
ing very fast chemical and physical processes to be investigated.
Detailed descriptions of the fundamentals of laser operation are beyond
the scope of this chapter and are covered in many excellent sources. (3-9)
Essentially, lasers result from the excitation of a gain medium to a condition
where a high energy level is populated to a greater extent than at a lower
level, termed a population inversion. Electrons in the higher energy level are
then stimulated to relax simultaneously, releasing a large number ofphotons
with the same wavelength at the same time, generating the coherent laser
pulse. The method for creating the population inversion and the medium
employed are varied, which gives rise to the large number of potential laser
systems, including gas lasers (e.g., excimer systems, N 2 , CO 2 , copper vapor),
solid-state lasers (Nd: YAG, ruby) , semiconductor lasers, and liquid dye
laser systems. Each of these systems has its own characteristic power levels,
pulse rates and durations, wavelength ranges, and expense. With this array
of choices oflaser instrumentation, many different lasers have been employed
in conjunction with the glow discharge. The specific choice of a laser system
to use is dependent on the specific application or investigation desired, and
each type of laser will have its own advantages and disadvantages.
The details of how to make such a choice are far too extensive to
elaborate on in this chapter, but a few general comments can be made. For
supplemental atomization in a glow discharge, the laser should be optimized
to generate a high power density. The repetition rate can be important ifthe
signal is to be accumulated or if a sequential scanning system, such as a
quadrupole mass spectrometer, is employed. Wavelength is of less import-
ance. For supplemental excitation, a broad range of wavelength tunability
is a great asset for selectively exciting particular atomic transitions. A high
power system is advantageous so as to saturate transitions and increase the
sensitivity of the techniques. A high pulse rate will allow for a higher duty
cycle and enhance the analytical utility of the system. Further information
on the various laser systems, their particular attributes, and their applica-
tions can be found in the previously referenced sources.
10.3. Supplemental Atomization
Although glow discharge sputtering effectively creates an atomic popu-

lation representative of the conductive cathode material, there are unique
advantages to using a high-powered laser as a source for supplemental or
secondary atomization. The ability to create an atomic population indepen-
dent of direct discharge processes is beneficial for the analysis of nonconduct-
ing materials and for investigations of fundamental glow discharge processes
426 CHAPTER 10
such as the mechanisms of ionization present in the discharge and the funda-
mental basis of relative sensitivity coefficients. These fundamental studies
of glow discharge processes have been the main application of the laser
ablation/glow discharge technique and will be the primary focus in this
section.
10.3.1. Introduction to Laser Ablation
When a high-power laser beam is focused onto a sampie surface, a laser-

generated plume of material can result. This plume consists of ground-state
atoms and multiatom clusters of the sampie material, and excited-state atoms
or ions of the sampie. The atoms formed in such an ablation process have
found application as a reservoir for atomic absorption measurements(IO,II)
and atomic fluorescence, (12-14) while the excited-state atoms have been moni-
tored by atomic emission(l5-17) and the ions extracted into mass spectromet-
ers for mass spectral analysis, (18-21) including the LAMMA instruments. (22-24)
In these latter methods (AE and MS), the laser is serving as the method
of both atomization and excitation. In addition, laser ablation has been
advantageously employed as asolid sampie introduction system for a variety
of subsequent ionization and excitation sources, including inductively cou-
pled plasma (ICP) emission,(25-29) ICP mass spectrometry,(30-34) microwave
discharges,(35) dc plasma devices,(36) arc/spark emission sources,<37) reson-
anee ionization methods, (38,39) and ion eyclotron resonanee ehambers for the
investigation of atom/ion reactions.(40) These methods may be extended to
the glow discharge source, with the laser serving as an ablation source for
injecting solid material into the glow diseharge for subsequent excitation
and ionization, or for direct laser vaporization and ionization ofthe cathode
material which can assist in fundamental studies of discharge processes.
10.3.2. Mechanisms of Laser Ablation
The mechanism of laser ablation is complex and not clearly defined,

with research into laser/solid interactions continuing.(41-49) The dominant
mechanism of the laser/solid interaction is believed to be athermal heating
of the sampie originating from the exchange of energy from the photons
absorbed by the solid and the solid's electrons, which then exchange energy
with the bulk lattice. (50) The basic idea behind laser ablation is to dump a
large amount of energy into the sampie surface in a very short period of
time, rapidly raising the temperature of the solid and vaporizing material
from the surface. Any parameters that impact the rate at which energy is
absorbed by the material will influence the effect of a laser pulse on asolid.
These parameters are most often divided into two categories: properties of
the solid and characteristics of the laser.
There are several properties of the material that can influence the
amount of energy dumped into the solid during the laser pulse. These inc1ude
the material's reflectivity and absorption coefficient for the wavelength of
incident photons, and the thermal properties of the material such as volume
specific heat, latent heat, and thermal conductivity. The reflectivity and
absorption coefficient of material can impact the number of photons
absorbed by the sampIe at lower power densities. As the power level is
increased to levels generally encountered in laser ablation studies, the sampIe
ultimately acts as a "blackbody" and absorbs all of the incident photons,
limiting the influence of the reflectivity and absorption coefficient on the
degree of ablation. The additional material parameters that can influence
the extent of laser interaction are those that control the rate of laser heating
of the material. The volume specific heat of the material will determine the
amount of energy required to raise the surface temperature of the sampIe to
the threshold of vaporization, with the latent heat determining the excess
energy required to undergo the phase change and vaporize the material. The
lower these values are, the lower the energy required for vaporization and
the greater impact of the laser pulse. Thermal conductivity is an important
parameter in determining how much heat can dissipate through the solid.
The greater the thermal conductivity, the harder a material is to ablate. At
sufficiently high power densities, enough energy will be available at the
sampie surface to rapidly vaporize the material regardless of the volume
specific heat, latent heat, or thermal conductivity and these material para-
meters will lose their importance. Ultimately, at sufficient laser power densi-
ties, material properties have little effect on the ablation process.
Laser characteristics that can impact the ablation process inc1ude wave-
length, pulse duration, and laser power density. At low laser power densities
«lOS W /cm2) , there is a laser wavelength dependence on the ablation pro-
cess. This dependence originates from a relationship between reflectance and
A, with the reflectance increasing with increasing A. After the laser power
density is sufficiently large to cause the sampIe to behave as a blackbody
(> 108 W/cm2 ) , there is almost no wavelength dependence on the ablation
process. The laser pulse length can influence the amount of energy dissipated
into the material, with a longer pulse allowing the thermal conductivity of
the sampie to transfer absorbed energy into the bulk sampIe and limit the
heat built up on the surface. Short pulses (nanosecond-microsecond) are
normally used in an effort to limit the energy losses caused by thermal
conductivity and to optimize the power density of the laser.
Laser power density appears to be the most important laser character-
istic to impact the ablation process, with three distinct regions of laser inter-
action occurring. At power levels <104 W /cm2, simple laser heating of the
surface occurs. When the laser power density is increased to the range of
104_106 W/cm2 the sampIe surface begins to melt with some differential
428 CHAPTER 10
boiling of elements with low vapor pressures. When the power density
reaches 106-1 OS W / cm2 the sampie undergoes laser vaporization with a
strong dependence on the thermal conductivity of the sampie observed. (51)
As the laser power density is increased above lOS W /cm2, the surface tem-
perature of the sampie rises to the vaporization temperature rapidly enough
to limit melting and the fractionalization of low-melting elements. Essen-
tially, a vapor front is set up that moves through the material, generating a
thermal shock wave that causes rapid heating of the sampie and further
vaporization.(42,52) Most of the material removed is in atomic form, with
some clusters, large aggregates, and a large number of electrons.(42,43) At
power densities above ~lOs W /cm2, the rate of sampie removal becomes
limited by the rate at which atoms leave the surface ofthe vapor. Continuing
to increase the laser power above this level does not appear to enhance
surface removal since the surface becomes shielded by the plasma, limiting
the interaction with laser photons. (53,54)
Laser ionization is also dependent on laser power densities with little
direct ionization observed below ~lOs W /cm2, and that which is observed
is thermal in origin and shows a strong dependence on the ionization poten-
tial ofthe element.(54) Ifthe power density is sufficiently high (> 109 W /cm2) ,
the atoms cannot leave the vapor before further interactions with the incident
laser photons occur. The laser light is then absorbed by these atoms, generat-
ing a vapor plasma that results in excitation and ionization. (55) At this point,
ionization is no longer thermal and may occur by several mechanisms includ-
ing multiphoton ionization and interaction with electric fields genera ted by
the laser.(42) Electrons present in the plasma can also undergo areverse
bremsstrahlung process (avalanche ionization) that can significantly increase
the energy of free electrons, causing a high degree of electron impact
ionization.(42,51,54) At power densities of 109 W /cm2 and above, nearly
uniform atomization with 100% ionization of the sampie occurs with little
dependence on either sampie or laser parameters. This results in relative
sensitivity coefficients for ionization that are close to unityYS,42,51,53,54) At
higher power densities, the plasma will absorb more of the laser photons,
resulting in the creation of multiply charged ions. Upper limits on the power
densities used for laser ablation/ionization are in the 10 10_10 11 W/cm2 range.
Some control of the ablation process is available by varying the laser power
density. Lower laser power densities (~10s W/cm2) will result in the produc-
tion of a largely atomic population of the sampie material, while higher
power densities (>5 x lOS W /cm2) will result in ionization of the ablated
material. (56) Depending on the type of investigation, either of these power
regimes may be employed. Further details on the mechanisms of laser
ablation may be found in the referenced sources, and especially arecent
book devoted to applications of laser ablation. (57)
10.3.3. Laser Ablation Coupled to Glow Discharge Devices
Coupling the laser ablation atomization process to the discharge for

subsequent ionization may provide several advantages over direct laser
ablation/ionization of the sampIe. Since the laser is used only to atomize
the sampIe, significantly lower laser power levels are required, possibly reduc-
ing laser complexity. In addition, the shot-to-shot irreproducibilities inherent
in laser operation will not have as great an effect on the low-power atomi-
zation process as they will on the high-power ionization processes, and
therefore the shot-to-shot reproducibility of the technique may be enhanced
relative to laser atomization/ionization alone. A combination oflaser ioniza-
tion and glow discharge ionization mayaIso result in an enhanced ionization
efficiency and enhanced sensitivity. Two fundamental configurations for
coupling laser ablation to the glow discharge can be envisioned. Either a
secondary sampIe can be positioned adjacent to a discharge plasma that is
sustained by another cathode and material ablated into the discharge, or the
laser may be used to ablate directly material from the discharge cathode.
Illustrations of these two methods are provided from doctoral research at
the University of Virginia. (58,59) To the best of the author's knowledge, these
are the only such studies of laser ablation into glow discharge devices cur-
rently available.
10.3.3.1. Laser Ablation of an Auxiliarv SampIe
In this configuration, the sampIe is positioned adjacent to the glow

discharge plasma, which is sustained by aseparate cathode, as illustrated in
Fig. 10-1. The laser beam is focused onto the sampIe surface and material
is ablated into the discharge for subsequent excitation and ionization. Several
potential problems must be taken into consideration with this configuration.
First, the ablated material must be transported from the laser impact region
to the negative glow of the glow discharge. Parameters that can impact the
transport processes indude discharge gas pressure and the distance between
the sampIe and the negative glow region of the discharge. A lower pressure
will allow greater diffusion of sampIe material into the negative glow; how-
ever, because of the operating pressure regime of the discharge, lower limits
on discharge pressure exist, below which the discharge voltage becomes too
large for stable discharge operation. An attempt to minimize the sampIe
distance from the cathode should also be made, but placing the ablated
sampIe too dose to the cathode can result in arcs and other discharge insta-
bilities as weIl as sputter deposition of the cathode material onto the ablated
sampIe. Judicious use of a discharge cathode with a low sputter yield, i.e.,
tantal um, can reduce the number of cathode sputtered atoms in the discharge
430 CHAPTER 10
Laser Generated
Plasma
Discharge
Negative Glow Cathode
Region
Laser
Figure 10-1. Diagram of the experimental configuration employed during laser ablation of
a sampIe positioned adjacent to the discharge plasma.
and lower redeposition and other potential interferences caused by the

cathode material. Another configuration envisioned to alleviate these prob-
lems would involve a separate ablation cell with the ablated material being
swept through an insertion tube into the discharge by the discharge gas,
similar to the configurations employed for the introduction of ablated mat-
erial into inductively coupled plasmas. (25-34) In this case, the ablation step
and discharge excitation/ionization steps are performed in separate cham-
bers and each can be optimized independently. With the exception of arecent
paper discussed in Seetion 10.3.5, no experimental results from this type
of configuration involving a glow discharge are available. The results of
supplemental ionization discussed in the following two sections were
obtained on a glow discharge mass spectrometer using a sampie placed in
the discharge source chamber and have been discussed in somewhat more
detail in other references. (58,59)
Initial investigations were performed to ascertain whether any dis-
charge-generated ion signal of the ablated material could be observed. In
these studies, the output from a XeCI excimer pumped dye laser operating
at the peak wave1ength for the rhodamine 590 dye was focused onto a 4-mm
copper disk positioned adjacent to the negative glow region of the discharge.
The laser beam was defocused to the point at which the laser-produced ion
signal (no discharge present) just disappeared. At this point, the power
density of the beam is sufficient for atomization without ionization and ions
formed from glow discharge processes should be readily discernible. Figure
10-2 depicts an analog signal profile of the 63CU ion signal with and without
~ ....
0.2 msec
LASER ONLY LASER WITH DISCHARGE
Figure 10-2. Temporal profile of the 63CU ion signal produced by laser ablation of a copper
disk placed adjacent to the discharge plasma, with and without a discharge. Argon discharge,
0.4 Torr, 4 mA, Ta cathode. [Reprinted with permission from: K. R. Hess and W. W. Harrison,
Laser ablation and ionization studies in a glow discharge, in: Lasers and Mass Spectrometry
(0. M. Lubman, ed.), p. 213, Oxford University Press, London, 1990.]
a supplemental glow discharge operated at 4 mA and 0.4 torr argon. As can

be seen, the discharge is ionizing copper material ablated from the cathode
by the laser.
In order to analyze the laser-generated ions, a data acquisition scheme
is employed to open a data gate during the time period of laser ablation,
with a second background gate opened during the time period when the
laser is off. Subtracting the background gate from the laser data gate will
provide an ion signal consisting only of the ions resulting from laser ablation
and discharge ionization. As previously mentioned, discharge pressure is one
parameter that can direct1y influence the amount of ablated material that
reaches the negative glow region for supplemental excitation and ionization.
The ablated ion signal was observed to maximize at low pressure, 0.3 Torr.
Unfortunately, discharge ionization processes in this type of source have
been observed to maximize at ~0.8 Torr(58) so that the pressure most suitable
for ablation of material into the discharge is not the optimum pressure for
further discharge ionization of that material. This limits the extent of
ionization observed for the ablated material, limiting the sensitivity of the
technique and any potential analytical applications. To improve this situ-
ation, a dual chamber source would be required in which each step in the
process could be optimized.
432 CHAPTER 10
As an illustration of the limited analytical utility of this method, a

National Institute ofStandards and Technology #410 steel was used as the
ablated sampIe and a mass spectrum of the resulting laser-ablated/discharge-
ionized material obtained. The mass spectrum over the iron mass region is
presented in Fig. 10-3. There are two major problems with this spectrum.
First, the relative insensitivity of the method can be judged by considering
the 52Cr peak which is present in the sampIe at a level of ~2%. Second, the
isotopic abundances and relative peak heights for the elements are not cor-
reet. The 52Cr peak is substantially higher than expected relative to the 56Fe
peak for a steel sampIe. This is most likely caused by a deterioration of the
sampIe as the laser ablates material from the sampie surface, generating a
laser-produced crater. By the time the mass scan reaches mass 56, a signifi-
cant alteration of the sampIe surface has occurred and the ablated material
is no longer reaching the discharge and the ion signal is significantly reduced.
As the mass of the monitored ions increases, the signal decreases, resulting
in skewed isotopic abundance measurements. Methods for alleviating this
problem inc1ude the use of a mass spectrometer that does not scan sequen-
tially but rather analyzes the entire mass range from a single event, such as
a time-of-flight instrument, ion trap, or Fourier transform mass spectro-
meter, and rotating the sampIe to provide a fresh surface for each laser/sam-
pIe interaction.
One of the major advantages of employing the laser as a supplemental
. method of atomization is the potential ability to analyze nonconducting
100
80
«
..J
z
(!)
ü5
z 60
Q
w
>
i= 40
«
..J
w
a:
20
0
50 52 54 56 58 60 62
M/Z
Figure 10-3. Mass spectrum produced by laser ablation of a NIST #410 steel sampie into
an adjacent glow discharge. Argon discharge, 0.4 Torr, 4 mA, Ta cathode. [Reprinted with
perrnission from: K. R. Hess and W. W. Harrison, Laser ablation and ionization studies in a
glow discharge, in: Lasers and Mass Spectrometry (D. M. Lubman, ed.), p. 214, Oxford Univer-
sity Press, London, 1990.]
materials. As an example, sampie disks 2.0 cm in diameter of CuC0 3 and
CUS04 were prepared from the powdered material using a press and die.
These disks were then mounted adjacent to the discharge which was operat-
ing with a tantalum cathode at 4 mA and 0.4 Torr argon. The laser was
defocused to apower density sufficient to cause only atomization, and the
laser-ablated/discharge-ionized ion signal was collected over the copper
mass range. The results are presented in Fig. 10-4. As with the NIST #410
100 A
BO
~0>
(/) 60
c
~
4)
> 40
i4)
a:
20
0
60 62 64 66 6B 70 72
M/Z
100 B
BO
(ij
C
0>
ii5 60
c
0
4)
> 40
§
4)
a:
20
0
60 62 64 66 68 70 72
M/Z
Figure 10-4. Mass spectra produced by laser ablation of (A) CUS04 and (B) CuC0 3 pressed
powdered sampies into an adjacent glow discharge. Argon discharge, 0.4 Torr, 4 mA, Ta
cathode.
434 CHAPTER 10
metal sampie, ions from higher masses, which are recorded later in time with
the quadrupole, are significantly reduced relative to their expected contribu-
tions. Because of cratering effects, there is a time dependence on the ablation
process, with the laser becoming less efficient at removing material as the
number of incident laser pulses increases. This generates a lower ion signal
from the 65CU isotope which is measured a number of laser shots after the
63CU. Instrumental methods for addressing this problem, time-of-tlight mass
spectrometer and rotating sampie, have been discussed previously. Ulti-
mately, although ablation of a sampie into a discharge for subsequent
ionization has been demonstrated, the method is far from analytically useful
and requires modification to reduce the effects of the time-dependent
behavior of the ion signals. One possible method for reducing the cratering
effects would be to ablate the discharge cathode directly, allowing discharge
sputtering and redeposition processes to "resurface" the sampie between
laser pulses. However, this method has the dis advantage of the inability to
analyze nonconducting material.
10.3.3.2. Laser Ablation of the Cathode Material
In this configuration the laser beam is focused directly onto a conduct-

ing sampie that also serves as the discharge cathode, as illustrated in Fig.
10-5. The material ablated from the sampie surface will directly diffuse into
Laser Generated
Plasma
lonL-
B&am~
Negative Glow
Region
Laser
Figure 10-5. Diagram of the experimental configuration employed during direct laser abla-
tion of a discharge cathode.
the negative glow region, more c10sely simulating the action of sputtered
material and reducing the impact of transport processes on the introduction
of material into the plasma. In addition, the discharge will continue to
sputter and erode the cathode surface, mitigating to some extent the effects
ofthe laser-produced crater, possibly enhancing the stability ofthe ion signal
from the laser-ablated material.
Experimentally, the laser was focused onto the surface of a copper
cathode (angled at 45°) with a laser power density sufficient to cause atomi-
zation without laser ionization. The glow discharge was then struck for
ionization of the ablated material. The laser impact results in the rapid
release of a substantial density of electrons and cathode material, causing
an instantaneous reduction in the resistance of the discharge. This rapid
reduction in resistance leads to arcs in the discharge, generating noise spikes
that interfere with the computerized data acquisition system, prohibiting the
acquisition of mass spectra due only to laser ablation/discharge ionization.
These noise problems limited the experimental investigations of this con-
figuration, and the only data available under these circumstances were
analog time profiles of the 63CU ion signals taken from an oscilloscope with
an XY recorder.
Figure 10-6 provides the analog time profiles for a 63CU ion signal at
various discharge pressures. The sharp initial peak is the noise signal gener-
ated by the ablation of material into the plasma, corresponding to the firing
of the laser. As can be observed, there is a laser-dependent ion signal on top
0.2 torr 0.3 torr
0.5 torr 0.7 torr

Figure 10-6. Analog time profiles of a 63CU ion signal generated from the ablation of a
copper cathode during discharge operation versus discharge pressure. Argon discharge, 2 mA.
436 CHAPTER 10
of a background ion level due to the glow discharge. The available results
do show that the ablation of an atomic population from a discharge cathode
can be subsequently ionized in the discharge. With further instrumental
developments to limit the effect of noise spikes, mass spectra of the ablated
material could be obtained. Future investigations in this direction may lead
to some analytical utility of the method, but currently the main application
of laser ablationjglow discharge excitation is as a method of investigating
fundamental discharge processes.
10.3.4. Fundamental Discharge Investigations
The ability of the laser to ablate material independent of the discharge

provides an opportunity to directly inject an atomic population into the
discharge to study various discharge processes. Illustrative examples of
fundamental investigations of the glow discharge employing laser ablation
are presented in the following sub-section.
10.3.4.1. Mechanism of Formation of ArM+ Species
The formation of an atomic population independent of the discharge

can also provide insight into fundamental discharge mechanisms. As an
example, the laser has been used to investigate the mechanism of formation
of species of the type ArM+, where Ar is the discharge gas and M is the
sputtered metal atom. These argides are common in low-pressure glow dis-
charge sources and can generate a signal as high as 10% of the base ion
signal from the bulk metal. (60) These ions are common isobaric interferents
in glow discharge mass spectrometry and several methods of overcoming
these interferences have been proposed, including monitoring of a different
isotope of the element of interest (if possible), variation of the discharge gas
to shift the mass of the rare gas-metal ion,(61) use of a collision cell to
dissociate these molecules, (62,63) and selective laser resonance ionization. (64)
Information on the mechanism offormation ofthese ions would be ofbenefit
in adjusting discharge conditions or developing methods for reducing these
interferences.
Two mechanisms of formation of the ArM+ species have been pro-
posed: associative ionization
(10-1)
where Ar* is a metastable state of the argon, and a three-body collision

mechanism(60)
(10-2)
These mechanisms are argon pressure dependent, with the associative ioniza-
tion process dominating at low pressures and the three-body collision process
dominating at pressures of approximately 1.0 Torr and above. (60) The
pressure regime in which a standard diode geometry discharge operates,
0.3-1.5 Torr, is such that a contribution from both mechanisms may be
expected for the formation of ArM+. Laser ablation of a sampIe in the
discharge chamber under an argon pressure identical to that used for a
discharge will allow the formation of M+ with Aro. This should permit
the formation of ArM+ through the three-body collision mechanism in the
absence of any associative ionization. This assumes little Ar* formation with
laser ablation in a low-pressure argon environment, an assumption that is
supported by the lack of argon emission from laser-generated plasmas under
a 1 Torr argon environment. (65,66) If a glow discharge is then opera ted under
the same conditions, both associative and three-body collision mechanisms
can play a role in the formation of the ArM+ and this signal would be
expected to increase relative to the base signal from the metal ion. Overall,
through laser ablation, mechanism (10-2) can be created in the relative
absence of mechanism (10-1), allowing the extent of each mechanism to be
determined, provided the signals are normalized in some manner. Mecha-
nism (10-1) is indirectly dependent on the M+ signal. The amount of ArM+
formed by this mechanism is dependent on the Ar* and MO populations,
which, in turn, directly impact the M+ signal observed for this discharge
configuration.(58) Mechanism (10-2) is directly dependent on the M+ popula-
tion. Since both mechanisms are related to the metal ion signal, this signal
serves as a good standard against which the laser ablation and glow discharge
ArM+ signals can be normalized. A ratio of M+ / ArM+ should remove the
metal atom and ion concentration dependence on the amount of ArM+
formed with both the ablation process and the discharge. Since the discharge
allows a contribution from both mechanisms and laser ablation only a contri-
bution from the three-body collision process, a comparison ofthe M+ / ArM+
ratios for each process will provide a relative value for the extent of ArM+
formed by each mechanism.
These processes were studied by ablating a copper pin that also served
as a sampIe cathode under different pressures of argon, while monitoring
the Cu +/ ArCu + ratio. (58.59) The results are presented in Table 10-1. The ratio
Table 10-1. Cu+ /ArCu+ Ratios during Laser Ablation and Discharge Opera-
tion at Various Pressures of the Argon Fill Gas.
Pressure Ablation Discharge Ablation/discharge ratio
0.25 torr 529.1 371.7 1.42
0.45 torr 483.4 366.8 1.32
0.60 torr 857.6 689.3 1.24
0.80 torr 646.9 582.6 1.11
438 CHAPTER 10
of the results for the discharge (both mechanisms present) to the results for
the ablation [mechanism (10-2) only] provides an estimation of the relative
importance of each ArM+ formation process. For example, at 0.25 torr the
ablation-only measurements show a ratio of Cu+ to ArCu + of 529 while for
the discharge this ratio is 372. This shows there is a larger ArCu+ signal with
the discharge, as would be expected since both mechanisms of formation are
contributing to the observed ArCu+ signal. The ArCu+ signal for the dis-
charge is approximately 30% larger than for ablation, indicating that the
associative ionization mechanism contributes approximately 30% to the total
observed ArCu + signal in the discharge, with the three-body collision process
contributing the other 70%. The relative importance of the three-body pro-
cess is plotted as a function of pressure in Fig. 10-7 and shows that as the
pressure increases, so does the extent of the three-body collision process.
The formation of ArCu + becomes completely dependent on the three-body
mechanism at approximate1y I Torr, as was previously postulated. (60)
10.3.4.2. Sputter Redeposition Studies

A second type of experiment that illustrates the potential application
of laser ablation for fundamental studies of discharge processes involves the
investigation of relative sensitivity factors (RSFs). RSFs are a measure of
the relative difference in e1emental sensitivity caused by differences in the net
100,--------------------------------------------------------------,
N-
o
c +
80
~
..
C
.I: +
+
il 70
~
i-
80
50
0.0 0.2 0.4 0.' 0.8 1.0
Discharge Pressure (Torr)
Figure 10-7. Relative extent of the three-body collision mechanism for the formation of
ArM+ species in the discharge versus discharge pressure. (See text for more detail.)
ionization rate of the various species arising from various physical, chemical,
or instrumental factors. Variations in the sputter atomization rates for
different elements, different ionization rates due to the different elemental
ionization potentials, and differential mass throughput through a quadru-
pole can cause the observed differences in ionization rate. For rapid
semiquantitative analysis of the entire periodic chart, RSFs of approximately
1 would work best and allow direct interelement comparisons of relative ion
signals. RSFs in the glow discharge have been observed to be on the order
of 3_5,(67,68) indicating little selective enhancement of elemental ion signals.
Elemental sputter yields with argon as the discharge gas are known to vary
within tbis range,(69,70) indicating the observed RSF values arise from differ-
ences in sputtering, with uniform elemental ionization efficiencies. Experi-
ments employing a pulsed discharge with laser ablation have provided
evidence that observed differences in ionization of sputtered elements are in
fact due mostly to differences in elemental sputter yields and not to a selective
ionization process. (58,59)
Direct ablation of the discharge cathode during operation of the dis-
charge is hampered by the sudden pulse of material into the discharge which
can generate electrical arcs and interfere with the electronic components of
the system. If a pulsed discharge is employed, and the laser pulses timed to
impact the cathode surface between discharge pulses, laser ablation of the
cathode can be performed with no deleterious effects. Figure 10-8 illustrates
the temporal response of ion signals observed for the pulsed discharge/laser
ablation combination. Each discharge pulse sputters the cathode surface and
Laser
ablation
Discharge
pulse
~ 1---1
2 msec 2 msec
Figure 10-8. Temporal response of discharge/ablation 63CU ion signals during laser ablation
between the pulses of a pulsed glow discharge. [Reprinted with permission from: K. R. Hess
and W. W. Harrison, Laser ablation and ionization studies in a glow discharge, in: Lasers and
Mass Spectrometry (D.M. Lubman, ed.), p. 215, Oxford University Press, London, 1990.]
440 CHAPTER 10
creates an atomic population whose elemental composition will be dependent

on the relative sputter yields of the elements present in the cathode. A
significant portion of the sputtered material, estimated to be 99% at
1.0 Torr,oI,72) is then redeposited back on the cathode surface through
collisions with gas atoms and ions present above the sampie surface. At the
termination of the discharge pulse, the cathode will be covered by a thin
layer of the redeposited material. This redeposited material is then ablated
and ionized by the subsequent laser pulse.
If the material ablated results from the redeposition of the sputtered
species, then the ablated ion signal can be used to determine the composition
of the redeposited material. Elements with a higher sputter yield in the
discharge should show an enhanced concentration in the redeposited mat-
erial relative to the bulk concentrations, while those species that do not
sputter readily will show no enhancement in the redeposited material. The
elemental ratios (RSFs) from the ablated ion signal should be similar to the
discharge sputter yields and different from the bulk concentration ratios. If
the RSFs in glow discharge mass spectrometry arise from differences in
sputter rates and not differential ionization, then the glow discharge RSFs
should match those from the ablated redeposited material.
A NIST reference steel sampie, # 41 OA, was employed to investigate
the respective RSFs as discussed above. Table 10-2 presents the certified
ratio of 62Ni to 63CU along with the ratio observed for laser ablation only,
laser ablation with a pulsed discharge, and the glow discharge only. As can be
seen, the ratio for laser ablation closely matches the certified ratio, indicating
nonselective ionization of the material by laser ablation. At these power
densities, complete ionization of ablated material is expected. (18,42,51,53,54,56)
When the laser is timed to impact the cathode between the pulses of the
discharge, the 62Nij63Cu ratio changes. Since the laser power is sufficient for
complete ionization of the ablated material, this indicates the cathode surface
composition has been altered by redeposition of the sputtered discharge
species. The ratio observed in the glow discharge is approximately the same
as that observed with laser ablation of the redeposited material, evidence to
support the conclusion that the discharge ionizes sputtered material non-
selectively and that the observed glow discharge RSF values arise from
Table 10-2. 62Ni/3Cu Ion Ratios Observed in Various Ion

Sources for NIST #410 Steel
Ion Source 62Ni/3Cu ratio
Ablation only 0.25
Ablation between discharge pulses 0.43
Discharge only 0.39
Certified NIST value 0.24
differences in elemental sputter yields. Further experimental evidence that

the redeposited material is being ablated can be found in other sources. (58,59)
10.3.5. Analyticallnvestigations
The combination of laser ablation/ discharge excitation of atomic emis-

sion analysis has been reported recently.(73) This paper, reportedly the first
to investigate the analytical utility of the combination of laser vaporization
followed by excitation in a glow discharge, was directed to overcome
disadvantages of each method when employed independently. Emission
spectroscopy based solelyon laser vaporization and excitation allows micro-
sampling and the ability to analyze both conducting and nonconducting
materials, but suffers from a lack of precision, with poor shot-to-shot repro-
ducibility largely due to fluctuations in the power density of the beam and
sampIe cratering effects, and poor linearity of analytical curves. Glow dis-
charges, on the other hand, are stable with good precision and employ a
low-pressure inert gas to reduce linewidths and background interferences.
However, discharges suffer from the inability to perform microanalysis and
the need for a conducting sampIe. The combination of laser vaporization as
a method for the creation of an atomic population followed by excitation
in the discharge for atomic emission would then take advantage of each
technique's strengths while reducing their shortcomings.
In this work, a Q-switched Nd: YAG laser (250 mJ/pulse, lO-ns pulses)
was focused to 1.5 mm on an aluminum alloy. The vaporized material was
then transported by a stream of argon 5 mm into a supplemental hollow-
cathode glow discharge with an aluminum cathode. The influences of a
variety of experimental factors, such as current, argon flow rate, cathode size,
and discharge pressure, were studied to maximize the analytical potential of
the system. The reported results showed relatively good precision and lin-
earity, with relative standard deviations of 2-7% in the determination of a
variety of transition metals at concentrations of 0.01-0.20%. Future studies
on specific applications and possible source modifications may further
improve the technique, making it an analytically useful method for the
analysis of nonconducting materials or for applications requiring high spatial
resolution of the analysis.
10.4. Supplemental Excitation
The plasma region of a low-pressure glow discharge device contains a

diverse population of species which may interact with laser photons, provid-
ing a variety of methods to monitor supplemental laser excitation in the
discharge plasma. There are sputtered atoms of the cathode material,
442 CHAPTER 10
molecules such as dimers and molecular clusters, metastable atoms of the

rare gas, and ions from the sputtered material and discharge gas, all of which
can be excited by laser photons. This excitation can lead to emission of a
photon, direct ionization of the species, or changes in the fundamental
properties of the discharge due to enhanced or reduced ionization. These
effects may be monitored optically, electronically, or mass spectrometrically,
generating a variety of analyticallaserjdischarge excitation methodologies.
Abrief introduction to several of these methodologies is provided in this
section.
Glow discharges have several advantages for their use as atom reservoirs
for atomic spectroscopy. First, the cathodic sputtering process creates an
atomic population direct1y from the solid cathode material, eliminating the
need for dissolution and possible solvent effects, contamination, and dilution
of the sampIe. The sputtering process is stable, reproducible, and the atomic
population can be easily controlled through the manipulation of discharge
current. Second, as opposed to flames or ICP devices, most discharges
operate in a reduced pressure « 10 Torr) environment of an inert gas. The
low-pressure inert environment reduces the number and variety of back-
ground molecular species formed by collisional processes, although some still
exist. The relatively low-pressure discharge also reduces collisional pressure
broadening of atomic transitions, enhancing the resolution of spectroscopic
techniques. Areduction in source pressure, and the associated reduction in
collision rates, will also significantly reduce quenching reactions, allowing
spectroscopy of transient excited-state species. Finally, discharge devices
operate effectively in a pulsed mode which allows synchronization with a
pulsed laser, increasing the effective duty cycle of the system. Proper timing
sequences will allow time-resolved laser interactions with the discharge,
permitting laser interaction at any point in a discharge pulse. A particular
advantage of the time-resolved interactions is the ability to make spectro-
scopic measurements after the discharge excitation processes cease and the
corresponding discharge emission has decayed, but before the atomic
population has diffused from the laser interaction volume. This provides an
effective method for reducing background emission interferences from the
glow discharge, enhancing the measurements generated by laser interaction.
10.4.1. Atomic Fluorf;1scence
Atomic fluorescence measurements employing glow discharges were the

first reported combination of lasers and discharge devices and laser-excited
atomic fluorescence spectroscopy (LEAFS) techniques continue to be the
most widely employed analyticallaserjdischarge interaction. The fundamen-
tals of atomic fluorescence methods are covered in several excellent
sources, (74--84) including Chapter 3 of tbis volume, and will only be treated
briefly here. The method is based on the absorption of radiation by an
atomic vapor, producing an excited-state atomic population which subse-
quently relaxes by emitting characteristic radiation which is monitored spec-
troscopically. Depending on the wavelengths of excitation and relaxation,
several types of atomic fluorescence are possible. A few of the most common
pathways are illustrated graphically in Fig. 10-9. In Fig. 1O-09A, the Aexcitation
and Aemitted are the same and this process is referred to as aresonant pathway.
In Fig. 10-9B, A.excitation is shorter than A.emitted, giving rise to a Stokes direct
line fluorescence process. In Fig. 10-ge, A.excitation is longer than Aemitted with
the supplemental excitation to level 1 provided by the discharge plasma, a
process called anti-Stokes direct line fluorescence. Finally, in Fig. 1O-9D, the
emitted photon is observed after another relaxation process (level 2 to level
1), termed Stokes stepwise line fluorescence. These are only a few of a
number of potential fluorescence pathways available with a combination of
discharge and photon excitation, and other sources provide a more complete
~out
o
A B
~In
- I---r-r-
l.. out ~out
o
c D
Figure 10-9. Diagram of several possible atomic fluorescence pathways. (See text for more
detail.)
444 CHAPTER 10
discussion of the potential mechanisms. (74-84) Since fiuorescence measure-

ments involve two wavelength-dependent resonant steps, excitation and
emission, the method is very selective and not as prone to background
interferences as is atomic emission spectroscopy. In addition, the two-step
process significantly decreases the spectral complexity of the resulting
emission spectra since fiuorescence photons will only be observed for certain
excitation transitions. The technique is also inherently very sensitive,
reaching toward single atom detection, as discussed in arecent paper.(85)
The basic instrumentation for fiuorescence measurements is dia-
grammed in Fig. 10-10. The system essentially consists of a source for excita-
tion photons, an atom reservoir, a monochromator for selection and
isolation of the emitted photons, and agated detection system such as a
boxcar averager. The detection system is normally placed at 90° with respect
to the excitation source to reduce the effects of scattered radiation from the
incident beam. In addition, scattered radiation can be reduced by employing
a nonresonant method of fiuorescence in which the excitation wavelength
differs from the emitted wave1ength.
Advantageous properties of the laser excitation sources inc1ude high
power for efficient optical saturation of the excitation transition, high
wavelength resolution to enhance method selectivity, tunability to increase
excitation versatility, and the ability to operate at a high pulse repetition
rate or continuously to enhance the technique's duty cyc1e. Various types of
laser systems deady meet most, if not all, of these characteristics. Specific
discussions of the choice of laser systems for LEAFS are provided in other
Fluorescence
Photons
Exclatlon
Photons
Detectlon
J. SO\.I'Ce System
(Laser)
Readout
Devlce
Figure 10-10. Basic instrumentation employed for atomic fiuorescence measurements iIIus-
trating the fundamental components of the system.
sources. (82,83) Advantageous properties of the atom source indude the ability
to produce a large and stable atomic population with little background
emission, the ability to be pulsed synchronous to the laser to enhance the
duty cyde of the method, an inert environment to prevent chemical reactions
and lower molecular background, and preferably a low press ure to limit
collisional quenching of the excited state, pressure broadening of the atomic
transitions, and laser light scatter. Again, details on source selection are
covered in other references(82,83) and will not be presented here but it should
be noted that pulsed glow discharges appear to be very attractive sources
for LEAFS. (85)
10.4. 1. 1. Ana/vtica/ App/ications
Analytical applications of LEAFS have developed significantly since

their origins in the 1970s with most applications employing fiames or electro-
thermal atomizers as atomic sources. Many of these applications are covered
in arecent review(83) and Omenetto(86) has provided an excellent summary
of the developments of LEAFS with various atomic sources. The use of
glow discharges for atomic fiuorescence measurements evolved naturally
from their use as sources in atomic absorption and atomic emission
spectroscopy.(87) Early applications of glow discharge atomic fiuorescence
used a dual discharge design in which separate discharges served as the
excitation radiation source and as the reservoir of the atomic vapor. Limits
of detection for these initial investigations were reported to be on the order
of 20-400 ppm for a variety of elements in an iron matrix, with a reproduc-
ibility of ± 1%. (88) Detection limits from this early work were improved, to
levels of 1-100 ppm for various elements in alloys of iron, aluminum, and
copper and an initial discussion of the possibility of using fiuorescence for
depth profiling of metal sampIes was also provided in a follow-up report. (89)
One reason for the enhancement in detection limits originated by pulsing the
discharge and carefully timing the excitation source and detector response in
such a manner as to discriminate against background emission from the glow
discharge atomizer, improving the signal-to-noise ratios in the experiments.
Utilization ofthe discharge's ability to be pulsed by synchronizing excitation
in the period after the decay of atomizer emission has become the standard
method of glow discharge operation for LEAFS. The availability of tunable
dye laser systems in the 1970s resulted in a further improvement in sensitivity
and versatility of LEAFS, with more recent applications of the technique
employing lasers as excitation sources in fiuorescence techniques.
Two articles from J. D. Winefordner's group at the University of
Florida appeared in the mid-1980s detailing the analysis of lead(90) and
indium.(91) The first of these reports employed a pulsed discharge with laser
446 CHAPTER 10
excitation applied 300 ms after the termination of the discharge. The dis-
charge cathodes were either NIST reference copper rods containing certified
lead concentrations ofO.4-128 pg/g, or aqueous solutions oflead dried onto
graphite electrodes. Output from a nitrogen-pumped dye laser was frequency
doubled to provide 283.31-nm excitation photons, with the lead's fluor-
escence radiation at 405.78 nm isolated by a 0.5 m monochromator. Results
from the analysis of the NIST material suffered from inexplicably poor
precision, with relative standard deviations of 8-20% observed, especially
for two of the sampies. Detection limits based on 3a of the background
were calculated to be approximately 0.1 pg/g. When lead solution residues
were analyzed on a graphite electrode, the behavior was similar to that
observed for the copper cathodes, only with substantially improved preci-
sion. The detection limits for these analyses were calculated to be 20 pg of
lead, illustrating the analytical potential of glow discharge sources for
LEAFS. In the second paper, indium sampies were prepared by dissolving
pure indium metal in nitric acid and depositing 10 pI ofthe resulting solution
on a cathode prepared from either graphite or copper. (91) Excitation of the
indium employed a nitrogen pumped dye laser, frequency doubled to the
303.94-nm In(I) transition. Typical pulse energies of the 15-mm unfocused
beam traversing the discharge were on the order of 5 pJ. Nonresonant
indium fluorescence was observed at 325.61 nm with limits of detection of
indium determined to be 8 ng at 3a. When the copper electrodes were
employed as the substrate for the deposited solution, the results were not
quite as good, with a greater deviation in the data and a reported detection
limit of 11 ng. One reason postulated for the decreased analytical perform-
ance with the copper electrodes is the potentially higher background fluo-
rescence from molecular species formed from sputtered cathode material.
Copper has a higher sputter atomization rate than graphite and is known
to form a variety of molecular species in discharges. Fluctuations in the
background emission from these molecular species could increase the devia-
tions in the data and reduce the analyte number densities due to molecule
formation, possibly reducing the sensitivity of the technique. The composi-
tion of the discharge cathodes seems to have an effect on the LEAFS signals
detected and should be considered when using solutions deposited on
cathodes in the glow discharge.
More recent investigations have shown the LEAFS technique to provide
remarkable sensitivity, possibly approaching single atom detection. In a 1989
paper, Winefordner et al. used a commercial hollow cathode lamp as an
atomization source for LEAFS analysis of lead, calculating reasonable
theoreticallimits of detection to be 1.8 ag within the laser beam volume. (92)
This study employed a copper vapor laser to pump a dye laser, which was
subsequently frequency doubled to 283.3 nm, as an excitation source for the
lead atoms. Fluorescence emission from the lead was monitored at 405.7 nm.
Using the fundamental collection and detection efficiencies of the instru-
mentation, a system responsivity to lead was calculated to be 1.64 x
10-8 V/pulse-atom ± 50%. Based on this system responsivity and the laser
probe volume, 1 V of signal was calculated to correspond to an atomic
density of7.77 x 109 atoms/cm3. This value was checked by substituting the
hollow cathode lamp with a quartz cell containing 0.2 mg lead heated to
646 K, providing accurate vapor pressure data from which atomic density
could be calculated. A lead density of 7.7 x 108/atoms/cm3 provided a
fluorescence signal ofO.0194 V which will correspond to 4 x 1010 atoms/ern3
for a I-V signal. A final estimate of system responsivity was given as
5.3 x 1O-9V/atom. Based on background noise considerations and this sys-
tem responsivity, a detection limit of 1100-6000 lead atoms per probe volume
was calculated. These studies clearly illustrate the potential for the detection
of extremely low numbers of lead atoms with glow discharge LEAFS and
provide the foundation for a subsequent paper in which experimental detec-
tion limits for lead and iridium were compared with a calculated intrinsic
detection limit. (93)
In this subsequent investigation, 50-pI sampies of lead and iridium
solutions were deposited in a disposable graphite cup which served as the
discharge cathode for the sputter atomization of the dried solution. The
discharge was pulsed and the excitation pulse from a frequency-doubled,
excimer-pumped dye laser was timed to interact with the atomic population
100 ps after the discharge was extinguished. At this point in the discharge
pulse sequence, no emission from the discharge was observed, lowering the
background and decreasing the detection limit by about a factor of three.
The complete instrumental system is diagrammed in Fig. 10-11. The boxcar
integrator gate was set to collect data during the time period of the lead
fluorescence signal. The theoretical detection limit for lead was calculated
to be 25 atoms per probe volume. When extrinsic factors such as noise and
probing efficiencies are taken into account, the detection limit for lead is
calculated to be 300 atoms in the probe volume. Although impressive, these
calculated detection limits in the probe volume are of limited value and more
conventional sample detection limits were measured using dried solution
residues. These values are provided in Table 10-3. These experimental values
are higher than those calculated from the theoretical considerations, indicat-
ing incomplete atomization of the sampie or inefficient probing of the
sampie. When atomization and probing efficiencies were considered in the
theoretical calculations, an intrinsic detection limit in the sampie of 1.6 x 106
atoms, or 0.5 fg, was determined. This is only about three orders of magni-
tude lower than the experimental value of 500 fg, with the difference most
likely originating from difficulties in complete atomization and efficient
experimental probing. Further modifications of the system to reduce back-
ground fluorescence and increase collection efficiency (both spatially and
448 CHAPTER 10
~~J:;-----c----l PREAMP
LONOPASS ~
1000 Cl
FILTER~ ~
r1J
VACUWI
CHAMBER ~ PHOTODIODE
TRIGGER
CATHODE ,!
~
V- ---
.~ .... "
FREO.OOUBLED
L-.._D_Y_E_LA_S_E_R_...J
Figure 10-1'. Instrumentation employed for the determination ofindium employing a pulsed
discharge. [Reprinted with permission from: M. Glick, B. W. Smith, and J. D. Winefordner,
Anal. ehern. 62 (1990) 158. Copyright 1990 American Chemical Society.]
Table 10-3. Limits of Detection for Pb and Ir"

Element A"x (nm) ;.,."" (nm) LODb (ppb) LOD' (pg)
Pb 283.3 405.8 0.1 0.5
Ir 285.0 357.4 6 20
"Taken from M. Glick, B. W. Smith, and J. D. Winefordner, Anal. ehern. 62 (1990) 160. Copyright 1990
American Chemical Society.
bCalculated by using 50-111 sampies.
'Calculated by using 2-111 sampies.
temporally) are expected to generate experimental detection limits that

approach the intrinsic detection limits even closer, providing a method for
elemental analysis that truly does approach single atom detection.
Other recent analytical applications of LEAFS with glow discharges
include areport on the possibility of using the technique for depth profiling
of microelectronic materials, reporting sodium analysis in molybdenum at a
limit of detection of 24 ngj g, (94) and the analysis of iron in brass at a level of
150 ngjg.(95) In the second ofthese articles, the authors indicate a substantial
source of noise is laser scatter from clusters of the sputtered material, limiting
the sensitivity of the technique. Evaluation of cluster formation processes
and the plasma conditions that minimize cluster formation will also have
to be considered in order to approach the intrinsic detection limits of the
technique.
An alternative application of glow discharges to laser-induced fluor-
escence spectroscopy does not rely on the discharge's ability to sputter
atomize the cathode material, but rather advantageously employs the
discharge to convert gaseous species to forms that can be monitored by
fluorescence methods. These methods are particularly attractive for the
analysis of species whose major transitions from the ground state lie in the
vacuum ultraviolet range, inaccessible to laser wavelengths. The discharge
plasma can overcome this problem by one oftwo methods. The first employs
the discharge to preexcite the sampIe species to a metastable level whose
transitions are accessible to laser fluorescence excitation. Examples provided
for this method included the analysis of neon in helium and argon, with
detection limits of 0.3 ppb and 100 ppm reported, and ionization of N 2 to
excited state Ni which is then analyzed to about 10 ppb in helium. (96) The
second method uses chemical reactions in the discharge plasma to convert the
analytical species to a form that is amendable to laser-induced fluorescence
methods. Examples ofthis method included the analysis ofN2 and NO after
oxidation to N02 in the plasma, with detection limits of approximately 1 ppb
for analysis in helium, argon, and air. (96) The potential of the discharge for
conversion of species to more analytically useful forms can be expected to
be more fully exploited in the future.
10.4.1.2. Diagnostic Studies bV Laser-/nduced F/uorescence
In addition to analytical applications, laser-induced fluorescence can

also be used as a diagnostic tool for fundamental investigations of glow
discharge plasmas. The technique is particularly attractive as a diagnostic
tool because of the method's ability to monitor both ground and excited-
state populations of atoms, ions, or molecules and its high degree of spatial
and temporal resolution. Information available from fluorescence measure-
ments can also be used to estimate the absolute number densities of atomic
species. (97) More comprehensive discussions of the diagnostic applications
of laser fluorescence are contained in other references, (82-84,86) and only a few
representative applications are presented here. Although several of these
illustrative applications do not employ glow discharge devices as the atom
reservoir, they are included in this section since the diagnostic methodologies
discussed are applicable to glow discharge systems as weIl.
One of the primary applications of laser-induced fluorescence for
diagnostic studies has been to study combustion processes. (80) Fluorescence
450 CHAPTER 10
methods can detect a variety of flame or plasma species, including radicals

such as OH, 0, and H, and various molecular compounds such as CO2 and
N02 , with high sensitivity and spatial resolution. The fast photon pulses
provided by the laser will also allow detection of rapidly reacting intermedi-
ates in flames or plasmas. Fluorescence measurements can also be employed
to determine flame or plasma excitation temperatures by measuring the
relative signal intensities of fluorescence transitions. (82,98) There are a number
of possible experimental variations of the methodology, each with a set
of specific equations and conditions to be used for the determination of
temperature. An excellent tabulation of the details of many of these methods
is available, and should be consulted for further information on the specific
requirements for this type of diagnostic application of laser fluorescence. (99)
A particular advantage of laser fluorescence measurements is the spatial
resolution of the method. Since only the volume of the plasma intersected
by the laser beam and the solid angle of detection will provide fluorescence
signals, laser interaction provides an excellent method for spatially profiling
plasma species, both atomic and molecular. An illustrative example of the
technique is provided by Gillson and Horlick(IOO,IOI) who used atomic
fluorescence measurements to spatially map the population of ground-state
neutral atom and ion species in an ICP. Information available from such
extensive mapping includes locating the regions of the plasma that will
contain the most useful analytical information, investigations of the effects
of instrumental parameters such as argon flow rate and torch power on
observed signals, and insight into how and where plasma processes such as
recombination occur. Applications of spatial profiling of species by laser-
induced fluorescence in a glow discharge have been reported by Winefordner
et al. (102) These studies provided information on the population densities of
various species within the discharge plasma, determining the best region of
the discharge to view for analytical information. In addition, this paper
reported on the use of temporal fluorescence monitoring of sodium atoms
as a method of determining the diffusion rate of sodium atoms through the
argon discharge.
Closely related to the determination of diffusion rates are experiments
designed to determine the gas velocities in plasmas. 82 An illustrative method
for the determination of the velocity of individual atoms using laser saturated
fluorescence is contained in a publication by She et al. (103) In this example,
a laser beam is split into two parallel beams which interact with the sampie
cello If the concentration of atoms in the cell is such that less than one
resonant atom is contained within the viewing region of the system, then
this atom will interact with each laser beam as it travels across the sampie
cello This should give rise to two bursts of fluorescence, separated by the
time it takes the atom to traverse the distance between the beams. From
these measurements, the atom's velocity can be determined.
The spatial resolution can be further enhanced using a two-photon

excitation fluorescence method where the intersection of two lasers in a
volume is required for the generation of a fluorescence signal, termed fluor-
escence dip spectroscopy.(104,I05) In this technique, two laser beams with a
spatial and temporal overlap are directed into the sampIe region. If one laser
wavelength is tuned to an atomic transition, the second laser wavelength can
be scanned to further excite the atom from the energy level reached in the
first excitation step. When the second photon is resonant with a transition
originating from the first excited level, the fluorescence from the first excita-
tion step will decrease due to the depopulation of the level by the absorption
of the second photon. In addition to enhanced spatial resolution afforded
by the overlapping laser beams, theoretical studies have shown that funda-
mental information on the second transition, such as the absorption oscilla-
tor strength, can also be derived from the experimental measurements.
The capability of the fluorescence method to analyze molecular species
in glow discharges has been illustrated by Winefordner et al. who reported
on the use of fluorescence techniques to investigate the spectroscopy of lead
dimers sputtered in a low-pressure glow discharge device.(106,107) In addition
to the fundamental information on lead dimer spectroscopy obtained by
these studies, information on the sputtering and dimer formation processes
in glow discharges was also obtained. These studies indicate that the mecha-
nism of molecule formation in the plasma is cluster formation with some
molecules sputtered from the cathode surface. (107) The intensity of the
molecule fluorescence was observed to be dependent on the operating para-
meters ofthe glow discharge, indicating the formation processes are sensitive
to discharge conditions. Glow discharge mass spectrometry is known to
suffer from isobaric interferences generated by clusters and molecular species
formed in the discharge. Fundamental information on how, and under what
discharge conditions, these molecules are formed would be useful in attempts
to reduce isobaric interferences. Molecular fluorescence studies such as these
could be a useful method for obtaining this fundamental information. In
addition to moleeule formation in the discharge, fluorescence measurements
have also been employed for fundamental investigations of sputtering pro-
cesses and the effect of target oxidation on these processes employing a
pulsed argon ion beam and stainless steel sampIes. (108) Other illustrative
applications oflaser fluorescence for fundamental studies ofvarious plasmas
include the use of atomic fluorescence to investigate possible chemical effects
of a reactive atmosphere of oxygen as compared with argon in laser ablation
of copper and lithium targets(109) and the use of LEAFS to study the effect
of easily ionized elements on the atomic densities in dc plasma devises.(llO)
Laser-induced fluorescence has also been valuable for fundamental
mvestigations and characterizations of chemical vapor deposition (CVD)
glow discharge devices. These devices have found significant application for
452 CHAPTER 10
the production of thin-film coatings of silicon for semiconductor devices.

Ho and Breiland used laser-excited fiuorescence to observe an intermediate
species of Sh in a CVD device as a thin film of silicon was prepared from a
silane discharge system.(III) Observation of this intermediate provided
insight into the mechanisms of deposition of the silicon. Further laser-
induced fiuorescence investigations of the silane discharge system have been
reported by Yamamoto et al. who used the technique to spatially profile
various species in a dc glow discharge device.(II2) The densities of both
ground-state and excited-state species were monitored as a function of dis-
charge CUTrent and spatiallocation in the discharge, and absolute densities
of the species were calculated from the experimental measurements. These
experiments also genera ted data allowing for the calculation of the radiative
lifetime of silicon atoms. These lifetimes were monitored versus pressure to
show that the lifetime of the excited atomic state was less than the collisional
rate in the discharge, indicating the absence of collisional quenching by the
buffer gas atoms.
Laser excitation can provide a method to simultaneously create an
excited-state population of atoms or molecules at a specific point in time in
glow discharge plasmas. If instrumental effects such as the time constant of
the detection system are known, they can be deconvoluted from the fiuor-
escence decay curve ofthe excited species, providing information on collision
rates and state lifetimes.Complete details of these methods can be found in
other sources.(82,86,1I3,1I4)
If a laser is sufficiently intense that saturation of the excitation step in
the fluorescence process occurs, information on transition probabilities and
quantum efficiencies can be developed. An illustrative example of such a
method has recently been described for the determination of the atomic
transition probability of the argon 430.0-nm transition.(1I5) In this experi-
ment, the population of a Iss level of argon is determined by a self-absorption
method. A laser tuned to 419.1 nm then saturates the transition to the 3P8
level. Based on the population ofthe Iss level, the population ofthe 3p8level
can be determined, provided the transition is saturated. The fiuorescence at
430.0 nm is then observed, and based on the number density of the upper
3p8 level and the intensity of the fiuorescence emission, a transition proba-
bility for the 430.0-nm transition can be calculated.
Although an attractive method for the characterization of plasmas,
relatively few direct references of diagnostic applications of fiuorescence
spectroscopy with glow discharge devices have been reported. Most of the
examples provided above employed fiames, ICPs, or CVD devices. With
the expanding application of glow discharge devices in various analytical
methodologies, more diagnostic studies of the discharge will be performed.
These studies can be expected to advantageously employ laser fiuorescence
methods, and an increase in the reported diagnostic applications of laser
fiuorescence in glow discharges can be anticipated.
10.4.2. Optogalvanic Effect Spectroscopy
As an alternative to optical monitoring of laserjdischarge interactions,

changes in the electrical properties of the discharge mayaIso be used to
detect optical transitions, a technique commonly referred to as optogalvanic
spectroscopy. General descriptions of the method are provided
elsewhere(J16--120) and a thorough review of optogalvanic effect (OGE) spec-
troscopy has recently appeared.(121) Briefly, the current flow through the
discharge is essentially an electrical circuit that obeys Ohm's law. The current
through the system is dependent on both the applied voltage and the dis-
charge resistance, with the discharge resistance determined by parameters
such as discharge pressure and the net ionization rate in the plasma. Altera-
tions in the ionization rate of the discharge will lead to variations in discharge
resistance and changes in the electrical properties of the discharge circuit.
Enhancements in the ionization rate of the plasma will result in a decrease
in discharge resistance and an increase in discharge current if the system is
operating in a constant-voltage mode. If the system is operating under
constant-current conditions, then the discharge voltage will decrease with
the enhancement of the plasma ionization rate. Alternatively, the plasma
ionization rate can be decreased, resulting in an increase in discharge
resistance and a corresponding decrease in discharge current (with constant
voltage) or increase in discharge voltage (with constant current). The cur-
rentjvoltage characteristics of the discharge may then be monitored by a
simple resistorjcapacitor circuit as illustrated in Fig. 10-12. Laser photons
provide one method of altering the ionization rate of the discharge, and in
combination with the electronic detection system, provide the foundation for
OGE spectroscopy. Advantages of electrical detection of optical transitions
include possible reductions in noise from stray light, the ability to monitor
nonluminescent species or transitions that are normally difficult to observe
opticallY' and the ability to generate an electrical signal directly without the
need of an optoelectrical transducer.
Optogalvanic signals resulting from increases in discharge resistance
were historically the first observed. (122) These signals correspond to net
decreases in the ionization rate of the discharge and arise principally through
areduction in the population of the rare gas metastable atoms. The meta-
stable gas atoms are relatively long-lived excited-state atoms whose radiative
decay to the ground state is forbidden. They playa significant role in the
net ionization rate of the discharge through two processes. First, they serve
as an excited-state precursor to ionization, reducing the energy required for
ionization. Second, metastables are an integral part of the Penning ionization
process in which there is a potential energy transfer from the metastable
atom to a discharge species, resulting in ionization provided the metastable
energy is greater than the ionization potential of the colliding species (see
Chapter 2). The metastable argon levels are the 3P 2 at 11.55 eV and the 3Po
454 CHAPTER 10
Trigger Out
Tunable Laser System

I"
Discharge High Voltage

Power Supply
Laser Beam
<
< Current Limiting
( Reslstor
Discharge Plasma
dc Blocking Capacitor
Oscilloscope
OGE Signal
.--------,
Detection Eleclronics ~
Figure 10-12. Generic instrumental arrangement employed for optogalvanic effect spectros-
copy. R = 2.2 kil, C= I JJF.
at 11.72 eV, while the neon metastable levels are the 3P2 at 16.62 eV and the
3 Po at 16.71 eV. Upon laser interaction, absorption of a photon by a meta-
stable rare gas atom can occur resulting in the excitation of the rare gas
atom into a nonmetastable state from which it can radiatively decay, decreas-
ing the net population of metastable atoms. The net depopulation of meta-
stable atoms will then reduce the overall rate of ionization, leading to a
decrease in discharge current under constant-voltage conditions.
Enhancements in the ionization rate of the plasma are generally
obtained by use of the laser to excite a species to a higher energy level. One
proposed mechanism for enhanced ionization is a simple model where the
excited species undergoes ionization more readily than the unexcited species.
For example, if the dominant mechanism of ionization of iron species in a
glow discharge was electron impact, only those iron atoms that undergo
collisions with electrons possessing >7.9 eV (first ionization potential for
ground-state iron) would be ionized. The average electron energies in glow
discharge devices have been reported to be on the order of 1-4 eV, with a
fraction possessing higher energies as approximately defined by a
Maxwell-Boltzmann distribution.(123.124) If an average electron energy of
2 eV is assumed, 4.8% ofthe electrons would have energies above the ioniza-
tion threshold of iron.(124) If the iron atom now absorbs radiation corre-
sponding to its transition at 344.06 nm, it will be excited to an energy level
::::i3.6 eV above the ground state. The iron atom is now only 4.3 eV below
its ionization potential, and approximate1y 27% of the electrons have the
requisite energy for ionization, (124) an increase of sixfold as compared with
the number of electrons able to ionize iron from the ground state. Since
there is a substantially greater population of electrons with sufficient energy
to cause ionization, the probability of ionization from excited state has
increased relative to the ground state, and an increase in the net ionization
rate ofthe discharge would be observed. Under this mechanism, the enhance-
ment should be dependent on both the energy of the absorbed photon and
the ionization potential of the species.
Keller and Zalewski, working with a hollow cathode discharge, did not
find an ionization potential dependence on the monitored OGE signal, lead-
ing them to propose that a second mechanism for the generation of the
enhanced OGE signal was operative.(125,126) This second mechanism involves
use of the atomic system merely as a method of transferring laser energy to
the electron population in the discharge by superelastic collisions, thereby
increasing the electron energy distribution of the plasma and lowering dis-
charge impedance. In electrical plasmas such as hollow cathode glow dis-
charges, the electron temperature and the atomic excitation temperatures
are intimately tied together due to the equilibrium established by the many
electron/atom collisions. When an atom is excited by the laser, the electron
collision frequency is sufficient to prevent a significant change in energy level
populations. Instead, through electron/atom collisions the laser energy is
redistributed among all of the electrons, leading to a slightly higher average
electron energy and a decrease in discharge resistance. Dreze et al. simulta-
neously monitored the OGE effect and variations in atomic emission from
a variety of uranium and xenon transitions when a U(I) transition was
illuminated by laser photons.(127) They found a small enhancement in the
measured excitation temperature of the discharge along with an increase in
emission from U(I), U(II), and Xe(I) transitions, indicating a more universal
mechanism of enhanced ionization than the simple excitation of uranium.
456 CHAPTER 10
These results seem to confirm the postulated mechanism of energy transfer

to the electron population. Arecent evaluation of rate constants for various
possible collisional processes potentially responsible for OGE signals in a
high-pressure Hg discharge have also indicated that lowered impedance
effects observed when atomic transitions are matched by the laser are due
to superelastic collisions involving electrons, resulting in increases in the net
electron energy of plasma and enhanced ionization. (128) This process is highly
dependent on the electron density and collision rate in the plasma, only
being dominant when the superelastic collision rate constant is greater than
the rate constant for ionization from the excited state, and may not be
operative under all discharge conditions, so that the ionization enhancement
mechanism may be operative in some discharge sources. Another report
has indicated that apressure pulse caused by the laser interaction with the
discharge may be responsible for the OGE effect, (129) but this effect was
originally postulated for observed signals in an 12 discharge in which the 12
can dissociate into 21 with laser interaction, increasing the pressure and
lowering discharge resistance, (130) and does not appear to be applicable to
atomic systems. The exact mechanism of the OGE effect remains unclear,
and further detailed theoretical discussions and applications oflaser-induced
optogalvanic spectroscopy can be found elsewhere.(121) As an introduction
to the application of the method, a few specific examples of OGE spectros-
copy are presented here.
10.4.2.1. Optogalvanic Studies of Discharge lonization Processes
Monitoring the optogalvanic signal as a function of discharge para-

meters and discharge gases can provide fundamental information on the
ionization processes occurring in a glow discharge device. Several such
examples of mechanistic studies using OGE spectroscopy have been
reported, a few of which are discussed here as a representative introduction
to the method. Zalewski and co-workers(l3l) carried out a mechanistic study
of the laser-induced impedance changes in a neon hollow cathode. They
found a variety of transitions originating from the 2p neon states, all of
which resulted in resistance decreases (enhanced ionization) upon laser irrad-
iation. The magnitude of the resistance decrease was always proportional to
the product of the wavelength, degeneracy, and oscillator strength of the
transition, indicating the impedance change due to each of these atomic
transitions is proportional to the probability for absorption of a laser pho-
ton. In addition to transitions originating in the 2p levels, transitions origin-
ating in the ls neon metastable level were also observed. These transitions
gave rise to the strongest signals observed and displayed both increases and
decreases in resistance, depending on the transition. The magnitude of these
changes did not correlate to the Ag! values as did the 2p transitions. These
impedance changes originate from the depletion of the metastable popula-
tion by excitation from the ls level to the 2p levels, and would be expected
to show a dependence on the ability of the excited state to decay back to
the ground state. This dependence would be related to the branching ratio
of the radiative decay of the excited 2px state to the various lsx states and
the quantum yield for the decay of the lsx state back to the ground state.
Quantitative values of these effects are difficult to determine, but qualitative
agreement was found by Zalewski et al.(131) and these investigations provide
evidence for the role of metastable atoms in the ionization process of the
glow discharge. Similar studies of the OGE effect in a neon glow discharge
were carried out by Doughty and Lawler who also found a variety of transi-
tions that exhibited both positive and negative resistance changes upon laser
irradiation. (132)
Smyth and Schenck(l33) also used the OGE to investigate the ionization
mechanisms present in a neon hollow cathode lamp. They observed 75 transi-
tions in the 572-654 nm region, a majority (57) of which showed decreased
discharge resistances corresponding to transitions originating in the 2p states.
The other transitions (18) are Is-2p adsorptions and these showed a combi-
nation of increased and decreased resistances, dependent on the transition
and discharge current. All 18 transitions exhibited decreased resistances (ion-
ization enhancements) at high discharge currents. They explained these
results in terms of competing mechanisms of ionization present in the hollow
cathode discharge. At low discharge currents, the metastable neon atoms
playa disproportionate role in the ionization of the discharge. Depopulation
of these states by Is-2p transitions will then decrease the net ionization in
the plasma, increasing the discharge resistance. As the current is increased,
electron impact ionization becomes dominant relative to the metastable
dependent processes. Under these conditions, absorption of a laser photon
by the neon atom will enhance ionization by electron impact with a resultant
increase in the ionization rate of the discharge and a decrease in discharge
resistance. In a follow-up paper, Smyth et al. (134) used a mass spectrometer
to directly monitor the neon ion signal upon laser irradiation in an effort to
correlate direct1y the observed OGE signals with changes in Ne ionization
in the plasma. As expected, they found the Ne + signals to increase for the
2p transitions (which resulted in a decrease in discharge resistance) and
found the Ne + to decrease for the metastable depopulation transitions
originating from the ls levels. These investigations were extended to the
mass spectrometric monitoring of discharge species other than the discharge
gas, such as sputtered material and background gases. Smyth et al.(I35) and
Bentz(\36) monitored OGE effects and the ion signals mass spectrometrically
from a neon hOllOW cathode discharge operating with a carbon cathode.
Upon irradiation of the neon lS5-2p4 transition (594.48 nm), ion signals
from 2~e+, 2~e+,2~e;, 12C+, and 14N+ all decreased by approximately
458 CHAPTER 10
20%. Further studies were performed with a copper cathode, monitoring

both the OGE signals and decreases in ion signal as the discharge current
was varied. The 63CU + exhibited a decrease of 6.6% at 4 mA, which fell to no
change at 40 mA. The OGE signal corresponded to an increase in discharge
resistance, and also decreased as discharge current was increased. These
studies provided evidence that at low discharge currents in hollow cathode
lamps, the primary mechanism of ion formation was through metastable
impact (Penning ionization), shifting to ionization by electron impact at the
higher discharge currents.
Similar investigations employing a coaxial cathode glow discharge (pin-
type geometry) were performed to allow a comparison of depopulation
effects for argon versus neon in order to provide further insight into the role
of Penning ionization in the discharge.(I37) If the Penning process is impor-
tant, species with ionization potentials less than the metastable energy ofthe
discharge gas should show decreases in ion signal with laser depopulation,
while those species with ionization potentials greater than the metastable
energy will be unaffected by metastable depopulation. In these investigations,
discharge ion signals were monitored with a quadrupie mass spectrometer
as the laser was tuned to rare gas transitions which result in the depopulation
of the metastable states (594.48 nm for neon, 696.54 nm for argon). Upon
laser irradiation in a neon discharge, those species with ionization potentials
less than the metastable energy of 16.62 eV (OH+, H 2 0+, N~, Cu+, Zn+)
showed ion signal decreases. In addition, the Ne + signal also decreased
with laser irradiation due to rem oval of metastable atoms which serve as a
precursor to neon ionization. Other species which are believed to be ionized
through an associative ionization mechanism, such as CuNe+, N 2H+, Cu~,
and Zn~, also show decreases, possibly due to the reduced ion populations
resulting from metastable depopulation. A difference mass spectrum for a
brass cathode in a neon discharge, resulting from subtraction of signals with
the laser off from those with the laser on, is presented in Fig. 10-13. This
"negative mass spectrum" clearly illustrates the reduced sputtered ion
intensity with metastable depopulation. When argon was employed as the
discharge gas, the results were substantially different for those species whose
ionization potentials are greater than the argon metastable energy
(11.55 eV). Species such as OH+ (LP. = 13.2 eV), H 20+ (LP. = 12.8 eV),
N~ (LP. = 15.6 eV), N+ (LP. = 14.55 eV), and 0+ (LP. = 13.62 eV) did not
exhibit decreases in ion signal with metastable depopulation. These experi-
ments provided evidence for the assertion that Penning ionization is an
important mechanism of ionization in the coaxial glow discharge.
10.4.2.2. Other App/ications of OGE Spectroscopy
In addition to the study of discharge ionization processes, the OGE in

glow discharges has found other applications in a variety of areas. One of
..J
<C
Z
(!)
iii
z
Q
w
>
~
w
a:
60 62 64 66 68 70 72
M/e
Figure 10-13. Example of a mass spectrum obtained by subtracting a glow discharge ion
signal from an ion signal obtained during the time period of laser depopulation of neon
metastable atoms at 594.48 nm. Neon discharge, 0.4 Torr, 3 mA, brass cathode. [Reprinted
with permission from: K. R. Hess and W. W. Harrison, Anal. ehern. 60 (1988) 694. Copyright
1988 American Chemical Society.]
the first applications of OGE spectroscopy involved use of the technique as

a method for wavelength calibration of laser systems.(138,139) Laser photons
are directed into a commercial hollow cathode lamp which provides a variety
of possible OGE transitions. As the laser wavelength is scanned across an
atomic transition, of either the sputtered cathode material or the discharge
gas, a resulting OGE signal can be detected. Accurate knowledge of the
exact transition wavelengths provides the information necessary to accu-
rately calibrate the laser scan control. In addition, the electrical detection of
the OGE effect provides a direct electrical signal which can be used in
conjunction with some type of a feedback mechanism to direct1y lock the
laser on a specific transition for an extended period of time. Another c1ear
advantage of this method is its relative simplicity, with no need for complex
optical isolation and detection instrumentation.
Other fundamental studies of discharges in which optogalvanic spectro-
scopy has also been successfully employed inc1ude the generation of detailed,
high-resolution, atomic spectraY 40) The OGE has also found application in
a variety ofhyperfine-structure studies, (141-143) as a method for the determina-
tion of term values of neon,(I44) for measuring He and Ne line profiles,<145)
460 CHAPTER 10
for observing collisional mixing of atoms in the discharge,(l46) as a method

of detection in state selective laser photodissociation of TI 2,(147) and as a
method for determining the threshold energy of sputtering.(148)
Another potentially useful application of the OGE for discharge
diagnostic investigations employs Stark shifts and splittings for electric field
strength measurements in hollow cathode discharges. Initial observations of
a correlation between Stark shifts of a krypton OGE signal and proximity
to the cathode, and hence electric field strength, were reported in 1983.(149)
This technique was exploited in a detailed study of a krypton hollow cathode
discharge in which a linear decrease in electric field versus distance from the
cathode was observed. In addition, this study employed the Stark-shifted
OGE signals to calculate the number density of charge carriers, and to
investigate the interrelationship of various discharge parameters such as
current, voltage, cathode dark space distance, and pressure.(150) The theory
behind these measurements was subsequently outlined in another report(151)
and the measurements extended to a neon discharge.(152)
Reports of direct analytical applications of optogalvanic spectroscopy
are limited, although the method has shown some promise as a possible
analytical method by observation of trace amounts of sodium in a lithium
hollow cathodeY53) Keller et al. have also illustrated the potential analytical
utility of the method by observing an OGE signal from 235U in a hollow
cathode lamp, with the 235U level calculated to be ::::::3 x 109 atoms/cm3Y54)
In another report, Keller and Zalewski studied noise and other considera-
tions in OGE spectroscopy using a hollow cathode lamp and calculated a
theoretical detection limit of ::::::3 x 106 atoms/cm3 for minority sputtered
species and ::::::102 atoms/cm3 for metastable species of the discharge gas. (125)
In a somewhat different analytical application, monitoring of the OGE
in a secondary discharge cell served as adetector for laser atomic absorption
measurements.(155,156) The instrumental setup is diagrammed in Fig. 10-14.
In these applications, the laser is directed through a sampie cell (flame) into
the discharge containing the same species as the potential analyte which
serves as the detector. The laser power must be in a range in which changes
in power produce a linear response in the detected OGE signal. If this
condition is met, then if no laser photons 'ire absorbed by the sampie, a
baseline OGE signal is measured with the detector discharge. As analyte
atoms or moleeules are added to the sampie cell, they will absorb a fraction
of the incident laser light and the power density of light arriving at the
detector discharge will be reduced. This reduction in power should result in
a reduction in the monitored OGE signal that is proportional to the number
of photons absorbed, which in turn is proportional to the concentration of
sampie species in the cell, providing the analytical utility of the technique.
The method has been demonstrated for both atomic(155) and molecular
species. (156)
Power Supply
. - - - - .----V--f -
Burner
Pulsed
-- C
L--_L~_~_:r_-,
I
HCL
Photodiode
(Trigger)
I Scope I
I I
Boxcar Averager
I I
SCR
Figure 10-14. Diagrammatic representation of the experimental system employed for opto-
galvanic effect detection of laser atomic absorption: R = 25 kn, C = 1 JJ F; SCR, strip chart
recorder. [Reprinted with permission from: M. A. Nippoldt and R. B. Green, Appl. Opt. 20
(1981) 3207. Copyright 1981 Optical Society of America.]
Most of the previous discussions of OGE applications have involved

atomic transitions, although the OGE can be observed for molecular transi-
tions as weIl. Examples include a study of OGE spectra generated from the
illumination of iodine(I3O) and the monitoring of the transient intermediate
species, SiH, by the OGE in a (CVD) device.(157) Further applications of
OGE spectroscopy are sure to arise as research in this area continues.
10.4.3. Laser-Enhanced lonization
A method similar to OGE spectroscopy is laser-enhanced ionization

(LEI). Rather than monitoring the change in discharge voltage or current
with the interaction oflaser photons, LEI methods directly measure electrons
released when the ionization of the absorbing species is enhanced due to
462 CHAPTER 10
photon excitation. Since the electrons are measured directly, a discharge

circuit is not required, permitting LEI to be performed in flames. A diagram
of the general instrumentation employed for LEI measurements is presented
in Fig. 10-15. There have been many reports of the LEI process, with a few
illustrative examples provided in a variety of sources.(158-168)
The LEI process has been largely limited to flames although the tech-
nique could be performed in a glow discharge. The main difficulties with
LEI in discharges deal with the difficult detection of the relatively few elec-
trons re1eased as a result of the ionization enhancement, relative to the
large background electron population generated by current flow through the
system. In addition, LEI assumes an ionization enhancement is observed
due to excitation of a discharge species. This. may not occur if ionization
proceeds by the Penning ionization process, or if the observed reduction in
discharge resistance with laser interaction is due to a change in the electron
Trigger
Photodiode
Photodiode
Ose ilioseope
Samp~ Samp~
&. &.
Ho~ Ho~
AID Converter
Figure 10-15. Representative instrumental arrangement employed for laser-enhanced ioniza-

tion spectroscopy. [Reprinted with permission from: G. C. Turk, J. C. Travis, J. R. DeVoe,
and T. C. O'Haver, Anal. ehern. 50 (1978) 818. Copyright 1978 American Chemical Society.]
energy distribution and not necessarily due to enhanced ionization, as dis-
cussed in Section 9.4.2. In fact, under certain circumstances, the OGE signal
in discharge plasmas has not been found to depend on either the ionization
potential of the atom or how elose the terminal state lies to the ionization
continuum, (125) as would be expected under a mechanism of enhanced ioniza-
tion. This is in contrast to LEI signals in flames which have shown an
ionization potential dependence.(169) Apparently, the atmospheric pressure
flame provides a higher collisional environment for electron excitation and
ionization than the low-pressure discharges, generating a distinct difference
in OGE and LEI mechanisms and possibly serving as an explanation for the
lack of LEI signals observed in plasma devices.
10.4.4. Resonance lonization Mass Spectrometry

A laser mayaiso be used for selective ionization enhancement of sput-
tered species in the discharge, possibly increasing both the sensitivity and
the selectivity of glow discharge mass spectrometry. In these applications, a
high-power tunable dye laser is focused into the negative glow region of the
discharge, where the photons may interact with the substantial population
of sputtered atoms. An atom may resonantly absorb one or more photons,
generating either an ion directly or an excited-state atom which requires less
energy for ionization, enhancing the ionization rate. Direct ionization due
to the resonant absorption of laser photons is generally termed resonance
ionization spectroscopy (RIS) ifthe ejected electron is detected, or resonance
ionization mass spectrometry (RIMS) if the positive ion is monitored.
The resonance ionization process involves the sequential absorption of
photons having energies which match bound-bound transitions, followed
by a bound-continuum transition. The bound-bound transition renders the
process highly wavelength selective, with only elements whose bound-bound
transitions correspond to the wavelength of the laser preferentially ionized
during the laser pulse. When combined with mass spectrometry, the method
has two degrees of selectivity, wavelength and mass. In addition to possible
enhanced sensitivity with laser ionization, the selective photoionization step
in the RIMS process will allow the reduction of potential isobaric interfer-
ences present in normal mass spectra and provides unique analytical utility.
Many detailed descriptions and reviews of various resonance ionization
techniques, ineluding theoretical and practical discussions of the parameters
necessary for application of the methodology, are available.(170-181) Briefly,
in a collisionally active environment such as the glow discharge, many
pathways exist for selective, photon-induced ionization of a species. Any
combination of laser and discharge excitation that results in a net energy
sufficient to cause ionization can be a viable pathway for resonance ioniza-
tion. Those pathways involving only the absorption of photons are generally
464 CHAPTER 70
referred to as multiphoton ionization (MPI), while pathways that advan-

tageously employ discharge energy in the ionization process are referred to
as LEI methods. A few representative MPI pathways are diagrammed in
Fig. 10-16, where ro represents a laser photon. In scheme 1 of Fig. 10-16, an
atom, A, absorbs a photon, rot, causing excitation to a true energy level
greater than halfway to the ionization potential of the atom. While in the
excited state, the atom absorbs a second (coincident) photon of the same
wavelength as the first, resulting in ionization. In scheme 2, the laser is
frequency doubled to provide a photon of higher energy, 2rot, for the initial
photon absorption to a true excited state, followed by the absorption of a
nondoubled photon to cause ionization. If two lasers are available and their
pulses interact the discharge at the same time, multiple color experiments
can be designed where one laser is tuned to one atomic transition, and the
second laser is tuned to a different transition originating from the excited
state of the first transition, with multiple photon adsorptions resulting in
ionization, as represented by schemes 3 and 4 in Fig. 10-16. Scheme 5
involves a two-photon absorption through a virtuallevel, followed by the
absorption of a third photon for ionization. More detail on these pathways
can be found elsewhere.(171) The key feature common to all ofthese pathways
®
A[co ll co1, coe-]A+
@
A[2C01, co 2 , coe-JA+
® A[co1, C0 2 ,coe-JA+
C0 2 co1(or C0 2)
CD A[2co ll co 1e-]A+ _
A[2co 1,co 1e-]A+ co =co 1,?r C02

C0 1
C0 2
C0 1
2C0 1 C01
2C0 1
C0 1 C0 1
Figure 10-16. Example ofvarious pathways available for resonance ionization. See text for
details on each pathway. [Reprinted with permission from: G. S. Hurst, M. G. Payne, S. D.
Kramer, and J. P. Young, Rev. Mod. Phys. 51 (1979) 774. Copyright 1979 American Physical
Society.]
is the presence of at least one resonant step to provide selectivity. The exact
pathway chosen is highly dependent on a multitude of instrumental factors
such as laser power densities, collisional excitation rates in the atom source,
absorption cross sections, the species under investigation, and so forth. A
variety of illustrative examples of the resonance ionization method have
appeared, employing a wide range of atomic sources including furnaces,(182)
laser ablation,<'83-'85) filaments,(186-188) and ion guns.(189-191) Each of these
applications is in a sense unique, with specific instrumental details that
influence the methodology and pathway chosen. For the purposes here, the
detailed discussion of resonance ionization is confined to the application of
the glow discharge as the source of atoms for RIMS.
Resonance lonization in the Glow Discharge
Resonance ionization in the glow discharge appears to have the poten-

tial to enhance the effectiveness of glow discharge mass spectrometry.
Although the glow discharge provides an ample sputtered ion signal for
mass spectrometric analysis, it has been estimated that < I % of the atomic
population is ionized. (192) There remains a large nonionized atomic popula-
tion so that efficient supplemental ionization by the laser may increase the
sensitivity of glow discharge mass spectrometry. In addition, glow discharge
mass spectrometry suffers from the potential for isobaric interferences from
other elements and a host of molecular species. Many of these interferences
can be discriminated against by using a high-resolution magnetic sector mass
spectrometer, as in a commercial glow discharge mass spectrometer, (193) but
many potential interferences require resolutions above those currently
available.(194) The resonant nature of the laser ionization process provides a
means for selectively ionizing an element and subtracting all nonresonant
(steady state) ions, including contributions from molecular species and other
sputtered atoms, discriminating against isobaric interferents. The enhanced
selectivity of the RIMS process is an attractive feature for analytical
investigations.
The glow discharge offers several possible advantages as an atom reser-
voir for resonance ionization. These include the simplicity of the discharge,
its ability to create an atomic population representative of the solid cathode
material, the low differential elemental sputter rates observed in the dis-
charge, its stability, and the ability to pulse the discharge to couple more
effectively with a pulsed laser system. Possible disadvantages include the
relatively high pressure in region where ions are formed, resulting in a low-
ered extraction efficiency of laser-formed ions, the electric fields associated
with the glow discharge, and the relatively high-background ion signal from
the discharge which must be subtracted from the laser-produced signal.
466 CHAPTER 10
Previous investigations of resonance ionization in a gIow discharge have

been reported in more detail elsewhere.(58,64,195,196) These investigations
employed an excimer pumped dye laser system with a frequency doubler to
generate photons that subsequently interacted with a coaxial discharge which
was operated with argon fill at pressures ofO.3-2.0 torr, currents of 1-5 mA,
and voltages of 300-3500 V.
When the dye laser is tuned to an appropriate atomic transition which
results in enhanced ionization, aresonant, laser-generated ion signal is
observed on top of the "background" ion signal from glow discharge ioniza-
tion. These signals are illustrated in Fig. 10-17. The laser-generated signal
may be isolated by setting a data sampling window over the time period of
laser interaction, collecting the "laser + discharge ions," then setting a
second "background" data gate during a time period when the laser is off,
and finally subtracting the "background" signal from the "laser + discharge
ions" signal. The ions observed after "background" subtraction should be
only those formed during the laser pulse, direct1y attributable to laser-
induced ionization, either through a multiphoton process or through an LEI
process.
The ions formed by laser interaction will exhibit a wavelength depen-
dence corresponding to atomic transitions of the element of interest. Fig.
10-18 is a wavelength ionization scan for iron in the glow discharge. The
signals observed were obtained by setting the mass spectrometer on the
56Fe isotope and scanning the dye laser wavelength. Each observed peak
corresponds to an atomic transition of iron, as tabulated in Table 10-4. A
partial energy diagram for iron is provided in Fig. 10-19 to illustrate the
collisional and photon-induced ionization processes occurring for several of
the observed signals. These transitions may originate from the ground state
or from any one of severallow-Iying energy levels which have been populated
Resonant
....
'"'7
-='"
...
-l
:z:
~
rn
~
......
Discharge
TIME
Figure 70-77. Temporal profile ofresonant and nonresonant (discharge) ionization in a glow
discharge.
-
.-
306.0
Figure 10-18. Wavelength scan of 56Fe ions produced by resonant ionization in a glow
discharge. Argon discharge, 0.3 torr, 2 mA, pure iron cathode.
Tab/e 10.4. lron Transitions Which Generated a Resonance

Ionization Signal in the Glow Discharge for the Wavelength
Region 299-306 nm
Wavelength (nm) Transition energies (eV)
299.44 0.05-4.19
300.10 0.09-4.22
300.81 0.11-4.23
301.76 0.11-4.22
302.05 0.09-4.19
302.06 0.00-4.11
302.11 0.05-4.16
302.40 0.11-4.21
302.58 0.12-4.22
303.74 0.11-4.19
304.76 0.09-4.16
305.74 0.86-4.91
305.91 0.05-4.11
by collisional excitation in the glow discharge. The availability of supplemen-

tal excitation from the discharge expands the number of transitions amend-
able to resonance ionization and provides a multitude of pathways for
ionization.
For the transitions observed in Fig. 10-18, simultaneous absorption of
two photons will result in direct photoionization of the iron atom and a
multiphoton ionization process. Altematively, the absorption of a single
photon may result in an excited-state iron atom which may subsequently be
468 CHAPTER 10
I.P
7.9 eV
4.16 eV
4.11 eV 4.11 eV
304.76 n.
302.06 ßI
305.91,.
0.09 eV
6.5. J.-c. ,. 0.011 eV
~
Figure 10-19. Partial iron energy level diagram for several of the transitions detected in Fig.
10-18 showing transitions which rely on collisional excitation (C.,.) in addition to laser photon
absorption for ionization.
ionized by a collisional process in the glow discharge. If electron impact is

a dominant mechanism of ionization in the discharge, the ionization of an
excited-state species will be enhanced relative to a ground-state atom as
discussed previously for OGE spectroscopy (Section 10.4.2). The net result
in a single photon enhancement of ionization is indicative of the LEI process.
The photon-induced ion signals observed in the discharge may arise from
either the MPI or the LEI process and it is not certain wbich is responsible
for the enhanced ionization observed. Energy dependence studies on the
process have been performed in an effort to distinguish the two processes,(64)
but these studies were inconclusive. In any event the signal is resonant pro-
viding an enhanced degree of selectivity and sensitivity.
A potential problem with resonance ionization in the glow discharge is
the relatively bigh-pressure region in wbich the laser-generated ions are
formed. The rare gas pressure necessary for discharge operation provides a
collisionally rich environment wbich contributes to a decreased efficiency of
resonant ion collection because of collisional scattering, recombination, and
decreased extraction of the ions. In order to most efficiently collect ions
formed by laser interaction, the laser beam must be positioned directly adja-
cent to the exit orifice of the discharge cello As the beam is moved from the
ion exit orifice, the laser-generated ion signal falls dramatically. A possible
solution would be to construct a discharge system in wbich the laser beam
interacts external to the discharge source, as illustrated in Fig.lO-20. Since
no discharge is operating on tbis side of the ion exit orifice, the pressure in
tbis region may be reduced to a few millitorr or less, reducing the number
of collisions and increasing the collision-free paths. A substantial population
of sputtered atoms may diffuse through the ion exit orifice and subsequently
interact with the laser to form ions, wbich may then be more efficiently
High VacuLm ( < 1 0 - 4 Torr)
Discharge
Cathode
Ion
Beam
~
to ~ AtomE-
Mass ~ Beam
Spectrometer IL.'
Negative
GlowRegion
Window
Laser
Discharge Chamber (0.3 - 1 .3 Torr)
Figure 10-20. Schematic diagram of an experimental arrangement for resonance ionization

of an atomic population external to the discharge source.
collected and analyzed. Preliminary work with such a system has been
attempted, but no significant RIMS signal was observed. (196) The instrumen-
tal configuration for these studies was not optimized, with the long distance
between the laser interaction zone and the entrance to the ion optics signifi-
cantly reducing the collected ions. Also, atomic absorption signals showed
a negligible atom population external to the source, necessitating system
modifications to provide a higher atom throughput from the source. No
further investigations of resonance ionization external to a discharge source
have been reported, but this remains a possible area of future investigation.
Although the glow dischargejresonance ionization technique may not
be optimized for analytical applications, previous results have shown the
inherent selectivity of resonance ionization as a possible method for discrimi-
nating against isobaric interferences in glow discharge mass spectrometry.
Figure lO-21A shows a portion of a glow discharge mass spectrum for a
470 CHAPTER 10
60 62 64 66 68 70
MlZ
60 62 64 66 68 70
MlZ
Figure 10-21. Representative example of the potential of resonance ionization to reduce
isobaric interferences in glow discharge mass spectrometry. (A) Glow discharge mass spectrum
showing ions from copper isotopes (Cu at 0.29% in an aluminum matrix) and a 27AI40Ar
species. (B) Resonance ionization mass spectrum generated using a Cu(I) transition at
324.75 nm illustrating the reduction of the 27AI40Ar species. Argon discharge, 0.3 Torr, 2 mA.
[Reprinted with permission from: K. R. Hess and W. W. Harrison, Anal. ehem. 58 (1986)
1700. Copyright 1986 American Chemical Society.]
NIST # 603 aluminum sampIe. The total copper concentration in this sampIe
is 0.29%. As can be seen, a large peak at m/z 67 arises from an AIAr+
"argide" species. Species formed from a combination of argon and both the
sputtered material and background gases such as nitrogen are commonly
found in the glow discharge. They can often create problems in elemental
analysis if the only isotope of the element of interest occurs at an m/z
value corresponding to an argide species. Resonance ionization methods
can selectively ionize the atomic species in the presence of the molecular
compounds, generating a method to eliminate the isobaric interference.
Figure 10-21B shows the same portion of the aluminum mass spectrum
employing resonance ionization of the copper atoms at 324.75 nm and
subtracting the background glow discharge ions. In this spectrum, the copper
isotopes are enhanced while the AlAr+ has been essentially removed.
Another problem potentially eliminated by resonance ionization is the
etfect of a large spectral peak directly adjacent to a trace constituent, swamp-
ing any ion signal from the trace material. Although this is not a problem
for high-resolution mass spectrometers, it can be a significant problem when
lower-resolution quadrupole instruments are employed. For example, Fig.
10-22A shows a portion of a NIST # 626 zinc glow discharge mass spectrum.
60 62 64 66 68 70 60 61 62 63 64 65 66
MlZ MIZ
I I I i I I I
60 62 64 66 68 70
MIZ
Figure 10-22. Illustrative example of the use of resonance ionization to measure small con-
centrations of copper (0.058%) in a zinc matrix (NIST #626 reference sampie). (A) Low-gain
glow discharge mass spectrum showing the zinc isotopes. (B) Expansion of the glow discharge
mass spectrum in A showing the 63CU ion signal relative to the much larger zinc ion signals.
(C) Resonance ionization mass spectrum of the copper at 324.75 nm. Argon discharge, 0.3 Torr,
2 mA. [Reprinted with permission from: K. R. Hess and W. W. Harrison, Anal. ehern. 58
(1986) 1701. Copyright 1986 American Chemical Society.]
472 CHAPTER 10
Copper is present at 0.058% total in this sampie, but the 65CU isotope is
masked by a 64ZnH+ interferent while the 63CU peak is barely resolved from
the much larger 64Zn signal, as illustrated in Fig. 10-22B. Resonance ioniza-
tion of the copper atoms at 324.75 nm allows subtraction of the zinc peaks
while continuing to ionize and coHect the copper. In Fig. 10-22C, resonance
ionization of copper in the zinc sampie aHows both copper isotopes to be
visualized in their proper isotopic ratios. These are currently the only
reported applications of resonance ionization in a glow discharge, though
the combination of discharge atomization and laser ionization oifers oppor-
tunities for furtherstudy.
10.4.5. Additional Laser-Based Techniques
In addition to the applications of glow discharges discussed previously,

the plasmas have also found use as atom reservoirs for several specialized
laser spectroscopic techniques. These methods have been employed largely
for physical measurements of atomic properties such as transition linewidths
or atomic lifetimes, rather than for analytical applications. Brief discussions
of the techniques with representative literature citations are provided in the
foHowing sections.
10.4.5.1. Dopp/er-Free Spectroscopy
The combination of the discharge's ability to create a substantial and

stable atomic population of most elements in a low-pressure environment,
which limits collisional broadening and quenching of the atomic population,
has resulted in their application as atomic sources for a variety of Doppler-
free spectroscopic experiments. The elimination of the velocity broadening
component of the atomic linewidth available with these investigations
provides high-resolution atomic spectra which generate a wealth of spectro-
scopic information, including isotopic shifts and hyperfine coupling con-
stants. An illustrative example of a Doppler-free spectrum as compared with
the Doppler-broadened spectrum is provided in Fig. 10-23. This spectrum
was obtained by the technique of saturation absorption spectroscopy.(84,197)
A variety of other specific Doppler-free spectroscopic techniques are avail-
able and these are too numerous and experimentaHy complex to discuss in
detail. Included in this list are intermodulated fluoroscence, (198)
optical-optical double resonance, (199) two-photon absorption methods, (199)
and laser polarization spectroscopy.(200) The previously discussed OGE
method of recording atomic spectra has also been employed as a method
for generating Doppler-free atomic spectra.(84,121)
...... 6.SMHz
(8 x 10- 6nm)
94 Zr 92 Zr
I I
200 MHz
91c
---+ 'V
Figure 10-23. Comparison of a Zr 614.3-nm Doppler-free spectrum (lower trace) with a

Doppler-broadened spectrum (top trace) employing a discharge as an atom source. The three
strongest 91Zr hyperfine components are labelIed a, b, and c in the lower trace. [Reprinted with
permission from: P. Hannaford and A. Walsh, Spectrochim. Acta 43B (1988) 1062. Copyright
1988 Pergamon Press.)
10.4.5.2. Laser Atomic Absorption
As the cost of tunable dye laser systems decreases and they become more
widely available, lasers may find increasing application as a cost-effective
alternative to hollow cathode lamps as line sources for atomic absorption
experiments. The main advantage of a tunable laser system over hollow
cathode lamps is the availability of narrow linewidths over a wide range of
wavelengths, which would be suitable for multielement analysis as compared
with the limited elemental application of specific hollow cathode lamps. The
lasers also provide a substantial number of incident photons, are coherent,
and easily directed, enhancing the spatial resolution of atomic absorption
experiments. Other than the cost and complexity, the main disadvantages of
the laser atomic absorption source are an incident power density that is too
high to measure the small differences in power due to photon absorption and
the generally poor shot-to-shot reproducibility of the laser. A representative
example of laser atomic absorption with a discharge atom reservoir is the
determination of thorium atom densities in a hollow cathode discharge as
474 CHAPTER 10
reported by Gagne et al.(20I) In this work, an Ar+ laser pumped a dye laser,
whose output was reduced and adjusted by the use of optical filters. More
recent developments in the field of semiconductor diode lasers could provide
further applications of lasers to atomic absorption. The diode lasers are
significantly less expensive, of lower power, and are generally more stable
than other laser types, all attractive features for atomic absorption measure-
ments. The main drawback is the very limited wavelength range currently
available with diode lasers and the narrow tuning range of most systems.
Recent publications have indicated that semiconductor diode lasers are
better spectralline sources than hollow cathode lamps, (202,203) of course pro-
vided an atomic transition is accessible to the laser wavelength range. The
current range of available wavelengths is approximately 700-900 nm, and
atomic number densities can be ca1culated for atomic transitions which lie
in this range. (204) Continued advances in the laser technology can be expected
to extend the accessible range of diode lasers making them more attractive
as line sources for atomic absorption investigations, both in flames and in
glow discharge plasmas.
10.4.5.3. Degenerate Four-Wave Mixing
Low-pressure discharges have also been employed as an atom source

for nonlinear spectroscopic techniques. Several degenerate four-wave mixing
(DFWM) experiments have been reported in flames,(205-207) and recently the
methodology has been employed in conjunction with a glow discharge source
of atoms.(208,209) Detailed theoretical discussions of the DFWM technique
are provided in these references and in arecent book, (210) and only a short
introduction to the technique will be discussed here. The analytical signal in
DFWM arises from the interaction of three coherent and degenerate (identi-
cal frequencies) laser beams, generating a fourth beam which is detected. A
schematic of the overall instrumentation is provided in Fig.IO-24, while Fig.
10-25 provides a simplified illustration of the laser interaction zone. The
forward pump beam, Ir, and the probe beam, lp, will overIap in a region of
the source and since the two beams have identical wavelengths, an inter-
ference pattern can be generated, creating a spatial variation in the photon
intensity of the overIap region. The interference pattern can act as a grating,
so that when the third beam, the backward pump beam lb, interacts with the
region, a portion of the radiation can be Bragg-scattered to form the output
phase conjugate beam, lpc, which is subsequently detected. Conservation of
momentum will require that the output beam is counterpropagating to the
input beam, providing a beam with a field strength that is the complex
conjugate of the pump beam. The analytical utility of the method originates
from the dependence of Ipc on the intensities of the other three laser beams
and the physical properties ofthe interacting medium. The medium provides
Ring Dy& Las&!' Argon Ion LaS&r

C699-21
Figure 10-24. Instrumentation employed in degenerate four-wave mixing experiments.

[Reprinted with permission from: G. A. Luena and W. G. Tong, Appl. Spectrosc. 44 (1990)
1669. Copyright 1990 The Society for Applied Spectroscopy.]
Ir
~ NONlINEAR
3( ~ MEDIUM
~L...----.--J
Figure 10-25. Laser beam alignment in degenerate four-wave mixing experiments. [Reprinted
with permission from: J. M. Ramsey and W. B. Whitten, Anal. ehern. 59 (1987) 168. Copyright
1987 American Chemical Society.]
a significant contribution to the nonlinear refractive index, and hence the

ability to scatter lb, only when the light frequency is close to an allowed
transition, providing the ability to generate spectra of atomic species in the
medium. In addition, the signal is also proportional to the square of the
concentration of absorbing species, providing the potential for quantitative
analysis. (205)
There are several attractive features of the DFWM technique reported
to be advantageous relative to other spectroscopic methods such as atomic
fluorescence. (205) First, in theory no signal should exist without an absorbing
species, providing a very low background against which to measure the
signal. In addition, the signal photons are all coherent, allowing for very
476 CHAPTER 10
efficient detection, even over long distances or through fiber-optic cables,

and for discrimination against background photons arising from incoherent
processes, enhancing the signal-to-background characteristics ofthe method.
Further discrimination from background photons can also be obtained
through polarization of the laser beams, generating a polarized output from
a "real" signal. The interaction of the forward and backward pump beams
provides Doppler-free spectroscopic signals, providing a method for high-
resolution spectroscopy. The high spectral resolution available with DFWM
can be effectively employed to enhance the selectivity of the method and,
with a low-pressure source which limits transition linewidths, to observe
hyperfine splitting patterns of atomic transitions.(20S) The DFWM inter-
action is essentially instantaneous, reducing problems associated with
quenching in a collisional environment. The technique is relatively simple
compared with other nonlinear methods, requiring only one laser and
wavelength-independent phase-matching conditions. Due to the fact that
analytical signals arise only from the region of overlapping beams, the
method also has a very high degree of spatial resolution which is beneficial
for spatial mapping of species in various sources.
The application of DFWM with a glow discharge atom reservoir has
been reported by Tong et al. in two recent publications.(208,209) In the first of
these reports, I mg of sodium was deposited onto the surface of a de-
mountable hollow cathode by drying a sodium carbonate solution and the
discharge was then struck. The DFWM signal from the sodium was moni-
tored with sufficient resolution to observe the hyperfine splitting of the Na
D2 transition. Although no direct analytical studies of the method were
performed in this work, the authors illustrate the potential analytical sensitiv-
ity of the method by commenting that the signal beam was visible to the
naked eye when only I mg of sodium was placed in the cathode. The second
of these papers provides further details and an expansion of the studies
reported in the first paper. Possible future applications of the DFWM tech-
nique may inc1ude further analytical development of the method, an
increased application in the area ofhigh-resolution spectroscopy, and advan-
tageously employing the high degree of spatial resolution afforded by
DFWM in spatial mapping studies of discharge species.
10.5. Conclusions
The combination of a glow discharge's ability to generate a copious

atomic population from asolid sampie, a low-pressure operating regime,
and the ability to be pulsed synchronous with a laser provides unique
opportunities to exploit laserjdischarge combinations for both fundamental
and analytical investigations. Previous applications of glow discharges and
laser interactions have inc1uded laser ablation, laser fluorescence, resonance
ionization, OGE spectroscopy, and nonlinear absorption techniques. As the
instrumentation for both lasers and discharges improves, and as novel
spectroscopic techniques are developed, glow discharge sources will find
increasing use as atom sources for laser spectroscopic measurements.
References
1. D. Duckworth and R. K. Marcus, Anal. Chem. 61 (1989) 1879.

3. O. Svelto and D. C. Hanna, Principles o[ Lasers, Plenum Press, New York, 1976.
4. D. C. O'Shea, W. R. Callen, and W. T. Rhodes, Introduction to Lasers and Their Applica-
tions, Addison-Wesley, Reading, Mass., 1977.
5. J. C. Wright and M. J. Wirth, Anal. Chem. 52 (1980) 1087A.
6. W. Demtroder, Laser Spectroscopy: Basic Concepts and Instrumentation, Springer, Berlin,
1981.
7. P. W. Milonni and J. H. Eberly, Lasers, Wiley, New York, 1988.
8. J. Wilson and J. F. B. Hawkes (eds.), Lasers: Principles and Applications, Prentice-Hall,
Englewood Cliffs, N.J. 1987.
9. L. J. Radziemski, R. W. Solarz, and J. A. Paisner (eds.), Laser Spectroscopy and Its
Applications, Dekker, New York, 1987.
10. T. Ishizuka, Y. Uwamino, and H. Sunahara, Anal. Chem.49 (1977) 1339.
11. R. Wennrich and K. Dittrich, Spectrochim. Acta 39B (1984) 657.
12. A. Quentmeier, W. Sdorra, and K. Niemax, Spectrochim. Acta 45B (1990) 537.
13. W. Sdorra and K. Niemax, Spectrochim. Acta 45B (1990) 917.
14. S. Grazhulene, V. Khvostikov, and M. Sorokin, Spectrochim. Acta 46B (1991) 459.
15. G. Dimitrov and T. Zheleva, Spectrochim. Acta 39B (1984) 1209.
16. K. J. Mason and J. M. Goldberg, Anal. Chem. 59 (1987) 1250.
17. Y. I1da, Appl. Spec. 43 (1989) 229.
18. R. J. Conzemius and J. M. CapelIen, Int. J. Mass Spectrom. Ion Phys. 34 (1980) 197.
19. H. Van Doveren, Spectrochim. Acta 39B (1984) 1513.
20. R. J. Cotter, Anal. Chem. 39B (1984) 1513.
21. F. P. Novak, K. Balasanmugam, K. Viswanadham, C. D. Parker, Z. A. Wilk, D. Matern,
and D. M. Hercules, Int. J. Mass Spectrom. Ion Phys. 53 (1983) 135.
22. E. Denoyer, R. Van Grieken, F. Adams, and D. F. S. Natusch, Anal. Chem. 54 (1982)
26A.
23. R. W. Odom and B. Schueler, in: Lasers and Mass Spectrometry (D. M. Lubman, ed.),
p. 103, Oxford University Press, London, 1990.
24. A. H. Verbueken, F. J. Bruynseels, R. Van Grieken, and F. Adams, in: Inorganic Mass
Spectrometry (F. Adams, R. Gijbels, and R. Van Grieken, eds.), p. 173, Wiley, New York,
1988.
25. T. Ishizuka and Y. Uwamino, Spectrochim. Acta 38B (1983) 519.
26. A. Aziz, J. A. C. Broekaert, K. Laqua, and L. Leis, Spectrochim. Acta 39B (1984) 1091.
27. M. Thompson, S. Chenery, and L. Brett, J. Anal. At. Spectrom. 4 (1989) 11.
28. S. Darke, S. E. Long, C. J. Pickford, and J. F. Tyson, J. Anal. At. Spectrom. 4 (1989)
715.
29. S. Lin and C. Peng, J. Anal. At. Spectrom. 5 (1990) 509.
30. P. Arrowsmith, Anal. Chern. 59 (1987) 1437.
478 CHAPTER 10
31. J. W. Hagar, Anal. ehem. 61 (1989) 1243.

32. J. MarshalI, J. Franks, I. Abell, and C. Tye, J. Anal. At. Spectrom. 6 (1991) 145.
33. H. Pang, D. R. Wiederin, R. S. Houk, and E. S. Yeung, Anal. ehem. 63 (1991) 390.
34. E. R. Denoyer, K. J. Fredeen, and J. W. Hager, Anal. ehem. 63 (1991) 445A.
35. F. Leis and K. Laqua, Spectrochim. Acta 34B (1979) 307.
36. P.G. MitchelI, J. Sneddon, and L. J. Radziemski, Appl. Spectrosc. 41 (1987) 141.
37. W. van Deijcke, J. Balhe, and F. J. M. Maesser, Spectrochim. Acta 34B (1979) 359.
38. D. W. Beekman, T. A. Callcott, S. D. Kramer, E. T. Arakawa, G. S. Hurst, and E.
Nussbaum, Int. J. Mass Spectrom. Ion Phys. 34 (1980) 89.
39. N. S. Nogar and R. C. Estler, in: Lasers and Mass Spectrometry (D. M. Lubman, ed.),
40. B. S. Freiser, Talanta 32 (1985) 697.
41. J. W. Carr and G. Horlick, Spectrochim. Acta 37B (1982) 1.
42. B. E. Knox, in: Dynamic Mass Spectrometry, Vol. 2 (D. Price, ed.) p. 61, Heyden & Son,
Philadelphia, 1970.
43. Y. P. Raizer, Sov. Phys. Usp. 8 (1966) 650.
44. D. Cremers and L. J. Radziemski, in: Laser Spectroscopy and Its Applications (L. J.
Radziemski, R. W. Solarz, and J. A. Paisner, eds.) p. 351, Dekker, New York, 1987.
45. M. Thopson, S. Chenery, and L. Brett, J. Anal. At. Spectrom. 5 (1990) 49.
46. P. Arrowsmith, in: Lasers and Mass Spectrometry (D. M. Lubman, ed.), p. 179, Oxford
University Press, London, 1990.
47. Z. Hwang, Y. Teng, K. Li, and J. Sneddon, Appl. Spectrosc. 45 (1991) 435.
48. A. Vertes, M. De Wolf, P. Juhasz, and R. Gijbels, Anal. ehem. 61 (1989) 1029.
49. L. Balazs, R. Gijbels, and A. Vertes, Anal. ehem. 63 (1991) 314.
50. N. Bloembergen, in: Laser-Solid Interactions and Laser Processing-1978 (S. D. Ferris,
H. J. Leamy, and J. M. Poate, eds.), p. 1, American Institute ofPhysics, New York, 1979.
51. R. A. Bingham and P. L. Salter, Int. J. Mass Spectrom. Ion Phys. 21 (1976) 133.
52. J. F. Ready, J. Appl. Phys. 36 (1965) 462.
53. K. D. Kovalev, G. A. Maksimov, A. I. Suchkov, and N. V. Larin, Int. J. Mass Spectrom.
Ion Phys. 27 (1978) 101.
54. I. Opauszky, Pure Appl. ehem. 54 (1982) 879.
55. J. T. Luxan and D. E. Parker, Industrial Lasers and Their Applications, Prentice-Hall,
Englewood Cliffs, N.J., 1985.
56. K. P. Seltzer and J. J. Kunze, Phys. Scripta 25 (1982) 929.
57. L. J. Radziemski and D. A. Cremers (eds.), Laser-Induced Plasmas and Applications,
Dekker, New York, 1989.
58. K. R. Hess, Ph.D. dissertation, University of Virginia, Charlottesville, 1986.
59. K. R. Hess and W. W. Harrison, in: Lasers and Mass Spectrometry (D. M. Lubman, ed.),
60. J. W. Coburn and W. W. Harrison, Appl. Spectrosc. Rev. 17 (1981) 95.
61. N. Jakubowski and D. Stuewer, 1990 Winter Conference on Plasma Spectrochemistry,
St. Petersburg, Fla., 1990.
62. F. L. King and W. W. Harrison, Int. J. Mass Spectrom. Ion Proc. 89 (1989) 171.
63. D. C. Duckworth and R. K. Marcus, Appl. Spectrosc. 44 (1990) 649.
64. K. R. Hess and W. W. Harrison, Anal. ehem. 58 (1986) 1696.
65. K. Kagawa and S. Yokoi, Spectrochim. Acta 37B (1982) 789.
66. K. Kagawa, M. Ohtani, S. Yokoi, and S. Nakajima, Spectrochim. Acta 39B (1984) 525.
67. F. L. King and W. W. Harrison, Mass Spectrom. Rev. 9 (1990) 285.
69. N. Laegrreid and G. K. Wehner, J. Appl. Phys. 32 (1961) 365.
70. G. Carter and J. S. Colligan, Ion Bombardment 0/ Solids, American Elsevier, New Y ork,
1968.
71. E. Nasser, Fundamentals 01 Gaseous Ionization and Plasma Electronics, Wiley-Interscience,
New York, 1971.
72. H. Mase, S. Nakaya, and Y. Hatta, J. Appl. Phys. 38 (1967) 2960.
73. Y. Iida, Spectrochim. Acta 45B (1990) 427.
74. L. M. Fraser and J. D. Winefordner, Anal. Chern. 44 (1972) 1444.
75. N. Omenetto, Anal. Chem. 48 (1976) 75A.
76. N. Omenetto and H. G. C. Human, Spectrochim. Acta 39B (1984) 1333.
77. H. G. C. Human, N. Omenetto, P. CavalJi, and G. Rossi, Spectrochim. Acta 39B (1984)
1345.
78. S. J. Weeks and J. D. Winefordner, in: Lasers in Chemical Analysis (G. M. Hieftje, J. C.
Travis, and F. E. LytIe, eds.), p. 159, Humana Press, Clifton, N.J., 1981.
79. J. C. Wright, in: Lasers in Chemical Analysis (G. M. Hieftje, J. C. Travis, and F. E. LytIe,
eds.), p. 185, Humana Press, Clifton, N.J., 1981.
80. R. P. Lucht, in: Laser Spectroscopy and Its Applications (L. J. Radziemski, R. W. Solarz,
and J. A. Paisner, eds.) p. 623, Dekker, New York 1987.
81. J. D. Ingle, Jr., and S. R. Crouch, Spectrochemical Analysis, Chap. ll, Prentice-Hall,
Englewood Cliffs, N.J., 1988.
82. N. Omenetto and J. D. Winefordner, Prog. Anal. At. Spectrosc. 2 (1979) I.
83. D. J. Butcher, J. P. Dougherty, F. R. Preli, A. P. Walton, G. Wei, R. L. Irwin, and R.
G. Michel, J. Anal. At. Spectrom. 3 (1988) 1059.
84. P. Hannaford and A. Walsh, Spectrochim. Acta 43B (1988) 1053.
85. J. D. Winefordner, B. W. Smith, and N. Omenetto, Spectrochim. Acta 44B (1989) 1397.
86. N. Omenetto, Spectrochim. Acta 44B (1989) 131.
87. G. F. Kirkbright, Analyst 96 (1971) 609.
88. D. S. Gough, P. Hannaford, and A. Walsh. Spectrochim. Acta 28B (1973) 197.
89. D. S. Gough and J. R. Meldrum, Anal. Chem. 52 (1980) 642.
90. B. W. Smith, N. Omenetto, and J. D. Winefordner, Spectrochim. Acta 39B (1984) 1389.
91. B. M. Patel and J. D. Winefordner, Spectrochim. Acta 41B (1986) 469.
92. B. W. Smith, J. B. Womack, N. Omenetto, and J. D. Winefordner, Appl. Spectrosc. 43
(1989) 873.
93. M. Glick, B. W. Smith, and J. D. Winefordner, Anal. Chem. 62 (1990) 157.
94. S. Grazhulene, V. Khvostikov, and M. Sorokin, Spectrochim. Acta 46B (1991) 459.
95. J. C. Travis, G. C. Turk, R. L. Watters, Jr., L. Yu, and J. L. Blue, J. Anal. At. Spectrom.
6 (1991) 261.
96. A. Bol'shakov, N. V. Golovenkov, S. V. Oshemkov, and A. A. Petrov J. Anal. At.
Spectrom. 5 (1990) 549.
97. M. J. Rutledge, B. W. Smith, and J. D. Winefordner, Anal. Chem. 59 (1987) 1794.
98. D. Sethi and H. Sethi, Spectroscopy 2 (1987) 26.
99. G. Zizak, N. Omenetto, and J. D. Winefordner, Opt. Eng. 23 (1984) 749.
100. G. Gillson and G. Horlick, Spectrochim. Acta 41B (1986) 1299.
101. G. Gillson and G. Horlick, Spectrochim. Acta 41B (1986) 1323.
102. C. van Dijk, B. W. Smith, and J. D. Winefordner, Spectrochim. Acta 37B (1982) 759.
103. C. Y. She, W. M. Fairbank, Jr., and K. W. Billman, Opt. Lett. 2 (1978) 30.
104. N. Omenetto, G. C. Turk, M. Rutledge, and J. D. Winefordner, Spectrochim. Acta 42B
(1987) 807.
105. J. B. Simeonsson, B. W. Smith, J. D. Winefordner, and N. Omenetto, Appl. Spectrosc. 45
(1991) 521.
106. B. M. Patel, B. Smith, and J. D. Winefordner, Spectrochim. Acta 40B (1985) 1195.
108. A. Elbern and P. Mioduszewski, J. Vac. Sei. Technol. 16 (1979) 2090.
109. A. L. Lewis, II, and E. H. Piepmeier, Appl. Spectrosc. 37 (1983) 523.
480 CHAPTER 10
110. M. S. Hendrick, M. D. Seltzer, and R. G. Michel, Spectrochim. Acta 41B (1986) 335.
11 1. P. Ho and W. G. Breiland, Appl. Phys. Leu 44 (1984) 51.
112. Y. Takubo, Y. Takasugi, and M. Yamamoto, J. Appl. Phys. 64 (1988) 1050.
113. P. Hannaford and R. M. Lowe, Opt. Eng. 22 (1983) 532.
114. P. Hannaford and R. M. Lowe, Aust. J. Phys. 39 (1986) 829.
115. A. Hirabayashi, S. Okuda, Y. Nambu, and T. Fujimoto, Phys. Rev. A 35 (1987) 639.
116. G. J. Beenen, B. P. Lessard, and E. H. Piepmeier, Anal. Chem. 51 (1979) 172.
11 7. D. S. King and P. K. Schenck, Laser Focus 18(3) (1982) 50.
118. J. E. M. Goldsmith and J. E. Lawler, Contemp. Phys. 22 (1981) 235.
119. J. C. Travis and J. R. DeVoe, in: Lasers in Chemical Analysis (G. M. Hieftje, J. C. Travis,
and F. E. Lytle, eds.), p. 185, Humana Press, Clifton, N. J., 1981.
120. J. C. Travis, G. C. Turk, J. R. DeVoe, P. K. Schenck, and C. A. van Dijik, Prog. Anal.
At. Spectrosc. 7 (1984) 199.
121. B. Barbieri, N. Beverini, and A. Sasso, Rev. Mod. Phys. 62 (1990) 603.
122. F. M. Penning, Physica (The Hague) 8 (1928) 137.
124. B. Chapman, Glow Discharge Processes, Wiley, New York, 1980.
125. R. A. Keller and E. F. Zalewski, Appl. Opt. 19 (1980) 3301.
126. R. A. Keller and E. F. Zalewski, Appl. Opt. 21 (1982) 3392.
127. C. Dreze, Y. Demers, and J. M. Gagne, J. Opt. Soc. Am. 72 (1982) 912.
128. J. Kramer, J. Appl. Phys. 64 (1988) 1758.
129. P. A. Fleitz and C. J. Se\iskar, Appl. Spectrosc. 43 (1989) 293.
130. C. T. Rettner, C. R. Webster, and R. N. Zare, J. Phys. Chem. 85 (1981) 1105.
131. E. F. Zalewski, R. A. Keller, and R. Engleman, Jr., J. Chem. Phys. 70 (1979) 1015.
132. D. K. Doughty and J. E. Lawler, Appl. Phys. Leu. 42 (1983) 234.
133. K. C. Smyth and P. K. Schenck, Chem. Phys. Leu. 55 (1978) 466.
134. K. C. Smyth, R. A. Keller, and F. F. Crim, Chem. Phys. Lett. 55 (1978) 473.
135. K. C. Smyth, B. L. Bentz, C. G. Bruhn, and W. W. Harrison, J. Am. ehern. Soc. 101
(1979) 797.
136. B. L. Bentz, Ph.D. dissertation, University of Virginia, Charlottesville, 1982.
137. K. R. Hess and W. W. Harrison, Anal. Chem. 60 (1988) 691.
138. R. B. Green, J. C. Travis, and R. A. Keller, Anal. Chern. 48 (1976) 1954.
139. R. B. Green, R. A. Keller, G. G. Luther, P. K. Schenck, and J. C. Travis, IEEE J. Quant.
Elec. QE-13 (1976) 63.
140. W. B. Bridges, J. Opt. Soc. Am. 68 (1978) 352.
141. G. Chevalier, J. M. Gagne, and P. Pianarosa, J. Opt. Soc. Am. B Opt. Phys. 5 (1988)
1492.
142. M. N. Reddy and G. N. Rao, J. Opt. Soc. Am. B Opt. Phys. 6 (1989) 1481.
143. R. Singh, G. N. Rao, and R. K. Thareje, J. Opt. Soc. Am. B Opt. Phys. 8 (1991) 12.
144. N. Ami, A. Wada, Y. Adachi, and C. Hirose, Chem. Phys. Leu. 153 (1989) 118.
145. T. F. Johnston, Jr., Laser Focus 14(3) (1978) 58.
146. N. Uchitomi, T. Nakajima, S. Maeda, and C. Hirose, Opt. Commun. 44 (1983) 154.
147. A. S. Naqvi, K. N .. Ullah, and M. 1. Rehmatullah, Opt. Commun. 80 (1991) 331.
148. S. Fujimaki, Y. Adachi, and C. Hirose, Appl. Spectrosc. 41 (1987) 1243.
149. K. Kawakita, T. Nakajima, Y. Adachi, S. Maeda, and C. Hirose, Opt. Commun. 48 (1983)
128.
150. S. Fujimaki, Y. Adachi and C. Hirose, Appl. Spectrosc. 41 (1987) 567.
151. C. Hirose and T. Masaki, Appl. Spectrosc. 42 (1988) 811.
152. N. Ami, A. Wada, Y. Adachi, and C. Hirose, Appl. Spectrosc. 43 (1989) 245.
153. R. B. Green, R. A. Keller, G. G. Luther, P. K. Schenck, and J. C. Travis, Appl. Phys.
LeU. 29 (1976) 727.
154. R. A. Keller, R. Engleman, and E. F. Zalewski, J. Opt. Soc. Am. 69 (1979) 738.
155. M. A. Nippoldt and R. B. Green, Appl. Opt. 20 (1981) 3206.
156. M. A. Nippoldt and R. B. Green, Anal. Chem. 55 (1983) 1171.
157. W. G. Tong and R. W. Shaw, Appl. Spectrosc. 40 (1986) 494.
158. J. C. Travis, G. C. Turk, and R. B. Green, Anal. Chem. 54 (1982) 1006A.
159. G. C. Turk, J. R. DeVoe, and J. C. Travis, Anal. Chem. 54 (1982) 643.
160. N. Omenetto, T. Berthoud, P. Cavalli, and G. Rossi, Anal. Chem. 57 (1985) 1256.
161. M. J. Rutledge, M. E. Tremblay, and 1. D. Winefordner, Appl. Spectrosc. 41 (1987) 5.
162. I. Magnusson, Spectrochim. Acta 43B (1988) 727.
163. K. C. Lin, S. H. Lin, P. M. Hunt, G. E. Leroi, and S. R. Crouch, Appl. Spectrosc. 43
(1989) 66.
164. K. C. Lin and Y. S. Duh, Appl. Spectrosc. 43 (1989) 20.
165. O. Axner and H. Rubinsztein-Dunlop, Spectrochim. Acta 44B (1989) 835.
166. O. Axner and S. Sjostrom, Appl. Spectrosc. 44 (1990) 144.
167. O. Axner, M. Norberg, M. Persson, and H. Rubinsztein-Dunlop, Appl. Spectrosc. 44
(1990) 1117.
168. K. C. Ng, M. J. Angebranndt, and J. D. Winefordner, Anal. Chem. 62 (1990) 2506.
169. G. C. Turk, J. C. Travis, and J. R. de Voe, Anal. Chem. 50 (1978) 817.
170. J. P. Young, G. S. Hurst, S. D. Kramer, and M. G. Payne, Anal. Chem. 51 (1979) 1050A.
171. G. S. Hurst, M. G. Payne, S. D. Kramer, and J. P. Young, Rev. Mod. Phys. 51 (1979)
767.
172. G. S. Hurst, Anal. Chem. 53 (1981) 1448A.
173. G. S. Hurst, J. Chem. Educ. 59 (1982) 895.
174. G. I. Bekov, V. N. Radayev, and V. S. Letokhov, Spectrochim. Acta 43B (1988) 491.
175. E. Sekreta, K. Owens, A. Engel, and J. P. Reilly, Spectrochim. Acta 43B (1988) 679.
176. J. D. Fasset and J. C. Travis, Spectrochim. Acta 43B (1988) 1409.
177. G. S. Hurst and M. G. Payne, Spectrochim. Acta 43B (1988) 715.
178. J. P. Young, R. W. Shaw, and D. H. Smith, Anal. Chem. 61 (1989) 1271A.
179. G. S. Hurst and M. G. Payne, Principles and Applications 0/ Resonance Ionization Spectros-
copy, Hilger, Philadelphia, 1988.
180. D. H. Smith, J. P. Young, and R. W. Shaw, Mass Spectrom. Rev. 8 (1989) 345.
181. C. H. Chen, M. G. Payne, G. S. Hurst, S. D. Kramer, S. L. Allman, and R. C. Phillips,
in: Lasers and Mass Spectrometry (D. M. Lubman, ed.), p. 3, Oxford University Press,
London, 1990.
182. A. S. Gonchakov, N. B. Zorov, Y. Y. Kuy zyakov, and I. O. Matveev, J. Anal. Chem.
USSR (Engl. TransI.) 34 (1980) 192.
183. S. Mayo, T. B. Lucatorto, and G. G. Luther, Anal. Chem. 54 (1982) 553.
184. M. W. Williams, D. W. Beekman, J. B. Swan, and E. T. Arakawa, Anal. Chem. 56 (1984)
1348.
185. N. S. Nogar and R. C. Estler, in: Lasers and Mass Spectrometry (D.M. Lubman, ed.), p.
65, Oxford University Press, London, 1990.
186. D. L. Donahue, J. P. Young, and D. H. Smith, Int. J. Mass Spectrom. Ion Phys. (1982)
293.
187. J. P. Young and D. L. Donohue, Anal. Chem. 55 (1983) 88.
188. D. L. Donohue, J. P. Young, and D. H. Smith, Appl. Spectrosc. 39 (1985) 93.
189. D. L. Donohue, W. H. Christie, D. E. Goeringer, and H. S. McKown, Anal. Chem. 57
(1985) 1193.
190. J. D. Blum, M. J. Pellin, W. F. Calaway, C. E. Young, D. M. Green, I. D. Hutcheon,
and G. J. Wasserburg, Anal. Chem. 62 (1990) 209.
191. J. E. Parks, in: Lasers and Mass Spectrometry (D. M. Lubman, ed.), p. 37, Oxford
University Press, London, 1990.
482 CHAPTER 10
192. W. Gerhard and H. Oechsner, Z. Phys. B. 22 (1975) 4l.

193. D. J. Hall and P. K. Robinson, Am. Lab. 19 (1987) 74.
194. F. L. King, A. L. McConnack, and W. W. Harrison, J. Anal. At. Spectrom. 3 (1988) 883.
195. P. J. Savickas, K. R. Hess, R. K. Marcus, and W. W. Harrison, Anal. Chem. 56 (1984)
817.
196. P. J. Savickas, Ph.D. dissertation, University of Virginia, Charlottesville, 1984.
197. T. W. Hansch, I. S. Shahin, and A. L. Schawlow, Phys. Rev. Let!. 27 (1971) 707.
198. D. S. Gough and P. Hannaford, Opt. Commun. 55 (1985) 9l.
199. R. J. McLean, P. Hannaford, H. A. Bachor, P. T. H. Fisk, and R. J. Sandernan, Z. Phys.
D. 1 (1986) 253.
200. G. A. Luena and W. G. Tong, Appl. Spectrosc. 44 (1990) 1668.
201. P. Pianarosa, Y. Demers, and J. M. Gagne, Spectrochim. Acta 39B (1984) 761.
202. J. Lawrenz and K. Niemax, Spectrochim. Acta 44B (1989) 155.
203. R. Hergenroder and K. Niemax, Spectrochim. Acta 43B (1988) 1443.
204. T. E. Barber, P. E. Walters, and J. D. Winefordner, Appl. Spectrosc. 45 (1991) 524.
205. J. M. Ramsey and W.B. Whitten, Anal. Chem. 59 (1987) 167.
206. W. G. Tong, J. M. Andrews, and Z. Wu, Anal. Chem. 59 (1987) 896.
207. J. M. Andrews and W. G. Tong, Spectrochim. Acta 44B (1989) 1Ol.
208. W. G. Tong and D. A. Chen, Appl. Spectrosc. 41 (1987) 586.
209. D. A. Chen and W. G. Tong, J. Anal. At. Spectrom. 3 (1988) 531.
210. R. A. Fisher (ed.), Optical Phase Conjugation, Academic Press, New York, 1983.
11
Laser-Based Diagnostics of
Reactive Plasmas
Bryan L. Preppernau and Terry A. Miller
11.1. Introduction
There should be little doubt to readers of this volume that reactive plasma
processing applied to the development of electronic devices and novel
materials has considerable significance for current and future technology.
Furthermore, it is fairly apparent that the design and control of plasma
processes has been empirical in nature over the years, as is the case in
the analytical glow discharge plasmas discussed in the previous chapters.
However, we have now perhaps reached a point where it has become neces-
sary to have a more comprehensive diagnostic and theoretical interpretation
of discharge reactor systems in order to advance the promise of this
technology. (I)
With the advent and promise of extraordinary computing ability and
sophisticated plasma models, in principle one could describe in detail the
complex interactions present in low-temperature, nonequilibrium processing
plasmas. Thus, it would seem that in the very near future the empirical
approach could so on be discarded, save for one important element to this
problem. There is presently a lack of in situ nonintrusive high-resolution
experimental diagnostic data for the major interactive components of
processing plasmas as weIl as for the predominate driving mechanism, the
Bryan L. Preppernau and Terry A. Miliar. Laser Spectroscopy Facility, Department

of Chemistry, The Ohio State University, Columbus, Ohio 43210.
483
484 CHAPTER 11
electric field. Only through confirmation of the calculated models by correla-

tion with diagnostic data can a comprehensive description and hence the
possibility of logical design of plasma processes become a reality.
In tbis chapter we survey the current in situ plasma diagnostics with
emphasis on those techniques that are laser-based. Given that the relative
populations of component species in processing plasmas are for the most
part dictated by the continuous velocity distributions of e1ectrons, one
cannot hope to direct1y measure all possible fractional energy distributions
internal to each component species. Couple this difficulty with the added
complexity presented by the addition of surface boundary conditions, and
competing processes such as recombination, charge transfer, etc., one can
quickly see that a full diagnosis of even the simplest of discharge systems is
basically an intractable problem. Therefore, one usually confines the scope
of diagnostic measurements to major reactive species that would be expected
to be of special significance in processing plasmas. Indeed, these species will
have greater interest for processing technology since they are the components
of the plasma that would most likely find their way into the boundary layers
near surfaces and mainly determine the characteristics of the plasma-surface
interactions. We expect that the trend of future experimental work will be
to make measurements that more fully describe the plasma-surface bound-
ary layer. Hence, it is our intention to discuss the detection of those major
gas-phase reactive species that would be ofinterest precisely because oftheir
influence on surfaces exposed to the discharge.
We first discuss the various reactive species expected in processing
plasmas and briefly review the associated production mechanisms. Next the
requirements needed to make accurate profiling measurements in both space
and time are reviewed along with the inherent difficulties in making these
sorts of measurements especially near surfaces. In addition, we discuss the
necessity and requirements for electric field profile measurements. The aim
here is to critically examine profiling measurements for simple geometric
situations.
In Section 3 (experimental techniques) we describe the basic experi-
mental arrangements, advantages, and disadvantages of a variety of laser-
based probe diagnostics. Measurements of reactive species and plasma
parameters by these techniques are compared with those obtained by other
familiar diagnostics. It is our bias that the refinement of laser-based probes
for species detection provides the best means of making reliable and precise
in situ, nonintrusive measurements on reactive species and plasma para-
meters. However, it is important to recognize the complementary nature of
other diagnostic techniques.
Having discussed the basis of the experimental techniques, Section 4
reviews the experimental findings by researchers utilizing these techniques
and the species and parameters thus far studied by laser-based probes.
LASER-BASED DIAGNOSTICS OF REACTIVE PLASMAS 485
Finally, we conclude with a discussion of future trends for reactive plasma
research and point to some of the more difficult problems still to be solved.
11.2. An Overview of Reactive Plasmas
11.2.1. Species Identity

Reactive plasmas can be characterized by four qualities: fairly low total
pressure (0.1-10 Torr), moderate electric fields (1-5000 VIcm), moderate
electron energies (1-100 eV), and near-thermal translational velocity
distributions for neutral and ionie heavy-particle species. Taken together
these qualities ensure an environment that is complicated in the number of
competing production mechanisms. (2)
The production of heavy-particle species can be grouped via two main
pathways: those species resulting from chemical reactions between neutral
species, either in the gas phase or through surface-assisted reactions, and
those produced by electric field-driven charged species (usually electron)
collisions. Certainly, these two pathways are never c1early separated in space
and time in a discharge, but each may dominate under certain conditions or
regions in the plasma discharge. From the point of view of plasma diagnos-
ties, one would like to be able to account for the role of the each production
path by analysis of careful population measurements for each species present
as a function of spatial position and temporal behavior. Thus, one might
ask: what are the important species to be expected in a reactive plasma
and whieh ones could hopefully be detected and their population profiles
measured?
11.2.1.1. Atoms
Atomic species production via electron impact dissociation of parent

molecules ranks as one of the most prevalent processes in the bulk of a
reactive plasma along with internal molecular electronic excitation and ioniz-
ation caused by electron impact. As an example, for hydrogen, the electron
collision cross sections are all comparable (10- 17 cm2 at 30 eV) and thus one
would expect to measure a high degree of molecular dissociation in the H 2
discharge. (3) Atoms produced in this manner radiate rapidly to their ground
e1ectronic state or to excited long-lived metastable states. Most atoms of
processing interest have in their ground state considerable chemieal reactivity
while metastable states can have, in addition, considerable energy content;
both properties can have great importance in surface and gas-phase
interactions.
486 CHAPTER 11
The atoms produced by electron impact dissociation usually result from

the breaking of a single molecular bond and as such are readily produced in
the plasma. At typical operating pressures, the neutral atom recombination
lifetimes are long (milliseconds) and thus the atoms can diffuse over the
entire volume of the reactor system with large concentration gradients in
sheath regions and near surfaces. The dissociation fraction of atoms to
molecules in simple diatomic gas discharges can be very high, on the order of
50%. Given the high dissociation fraction and long recombination lifetimes,
atomic production is not expected to be temporally modulated except in
very low audio frequency or pulsed discharges.
11.2.1.2. Radicals
Highly reactive radicals can be readily produced in plasmas and have

significant concentrations. Radical production is sustained by electron
impact dissociation, charge transfer collisions between neutrals and ions,
and photodissociation of neutral molecules. The spatial and temporal
behavior are similar to those of atoms. Radicals readily diffuse throughout
the discharge and can strongly influence surface deposition and etching
processes.
For discharges in molecular gases such as fluoro- or hydrocarbons,
many intermediate radical forms can be produced. The interpretation of
radical production in discharges is complicated by uncertainties in the break-
down pathways for the parent molecule. One thing that is certain is that
application discharges with high concentrations of radical species (compar-
able to atomic species) are finding more and more uses, so that measurements
on radical production and distributions will be vitally important.
11.2.1.3. Ions
Molecular positive ions and negative ions obviously playamajor role in

sustaining reactive species production and determining plasma parameters.
Positive ion production is maintained by electron impact ionization, charge
transfer collisions, photoionization, and fast heavy particle or metastable
collisions. Negative ions are mainly produced by slow electron impact attach-
ment or by electron impact dissociative attachment. The average ionization
fraction (over the entire discharge ) of ions to neutral particle number
densities is typicaIly 10-4 or 10-3 for these low-temperature nonequilibrium
reactive plasmas; however, the ionization fraction may be strongly modu-
lated in sheath regions.
Positive ions are important to ion etching and implantation plasmas and
can be detrimental to certain plasma deposition processes. Unfortunately,
measurements of positive ion concentration profiles are sometimes difficult
for laser-based techniques. Closed-shell positive ions usually can be thought
of as arising from proton addition to a closed-shell neutral. Thus, for small-
molecular-weight ions typically found in plasma environments, their elec-
tronic transitions typically lie in the VUV and the upper states are often
dissociative making laser-induced ftuorescence (LIF) detection difficult.
Indeed, only IR laser absorption techniques using ground electronic state
vibrational transitions seem widely applicable. Yet as we shall see later,
such line-of-sight techniques suffer in terms of both spatial resolution and
sensitivity. It is, however, important to point out that open-shell ions, e.g.,
Ni, often have bound-bound electronic transitions in regions easily access-
ible to dye lasers and can conveniently be detected by LIF.
Negative ion densities are known to be very high (even exceeding
electron densities) in discharges in electronegative gases. Negative ions can
be easily detected by laser photodetachment of the extra electron and sub-
sequent measurement of the laser-induced transient current in the discharge.
The identity of the negative ion is more difficult to ascertain, but can some-
times be inferred by the wavelength dependence of the photodetachment or
by LIF of the photodetached neutral. Measurements on negative ions in
reactive plasmas are important in order to analyze their inftuence on the
local electric field as weIl as their interactions with surfaces exposed to
electronegative plasmas.
11.2.1.4 Excited Mo/ecu/ar States
The relative populations of excited molecular electronic, rotational, and

vibrational states are primarily determined by electron impact excitation and
recombinative cross-sections. Most excited electronic states rapidly radiate
because of short radiative lifetimes. Metastable electronic states which are
not effectively coupled to lower states can have lifetimes as long as several
milliseconds even in the plasma. Hence, as with atomic metastables, sus-
tained molecular metastable concentrations develop, accounting for an
appreciable percentage of the discharge energy content.
The high electron temperature distribution, along with photoemission
from the discharge, can lead to nonthermal rotational and vibrationallevel
populations in ground and metastable electronic states. These result from a
variety of processes, including electron impact excitation, v' -v" pumping,
and even translational to vibrational excitation through neutral collisions.
Laser-based techniques such as LIF and resonantly enhanced multipho-
ton ionization (REMPI) can be used to probe excited molecular states with
high spectral resolution. This allows the assigning of approximate rotational
or vibrational temperatures to a molecule as weIl as elucidating excitation
kinetics.
488 CHAPTER 11
11.2.2. Sampling Considerations For Concentration Measurements
The ability to make laser-based concentration measurements in reactive

plasmas depends on the investigator's interpretation of instrumental factors
that critically affect photoluminescence or optogalvanic detection. While
relative populations or concentrations may be based on the relative response
ofthe detection apparatus, absolute measurements are much more difficult to
achieve. Reliable absolute concentration determinations without a standard
require the following: (1) accurate information regarding spectroscopic
parameters such as excitation cross sections, (2) instrumental factors affect-
ing the excitation and collection of ftuorescence or optogalvanic signal, (3)
analysis of systematic perturbations caused by nonlinear effects that may be
convoluted into the collected data, and (4) careful error analysis and error
propagation.
It should be clear that the total observed voltage change, A. V at the
input of a data collection system for a LIF experiment can be written
A.V= Kn2)l (11-1)
where n2 = number of atoms or molecules excited to the upper state (2)

during the laser pulse, y = fraction of atoms or molecules that emit photons,
Le., quantum yield, and K = instrument and apparatus constant (V jphoton).
If we assume that only a small fraction of the laser's photons are absorbed
and that only a small fraction of the atoms or molecules in the lower state
(l) are excited, n2 can be written
(11-2)
where (J' = cross section for excitation ofnl to n2 (cm2 jphoton), t = effective
absorption length (cm), CI = concentration of atoms or molecules in state 1
(speciesjcm3), and Nt = number of photons in the laser pulse into the
sample's volume. Generally, the experimental apparatus will record the num-
ber of ftuorescence photons as a charge pulse Q from a photomultiplier tube
(PMT). Typically, this charge is translated to a voltage using a boxcar or
integrator. In this case Kin Eq. (11-1) becomes
(11-3)
where k = boxcar or integrator conversion factor (V jcharge pulse per pho-

ton), T/c = fraction of emitted photons striking the active surface ofthe PMT,
T/g = fraction of photons striking the active surface to yield a photoelectron,
e = electron charge, G = gain oftube (electron multiplication factor). k, T/g,
and G can be calibrated in the lab using a known photon source or less
reliably gleaned from manufacturer's specifications. The collection efficiency
1]c is apparatus dependent and can be written
where 1]T = wavelength-dependent transmission efficiency of the collection

optics and Qp/41t = the fractional solid angle of collection for fluorescence.
1]T can also be calibrated in the lab for a given set of lenses, filters, etc.
Perhaps the most difficult quantity to determine is the collection solid
angle, Qp/41t, which is in general a function of the relative positions of the
laser probe interaction volume in the discharge, the focal point of the collec-
tion optics, and the reactor surfaces; mainly the electrodes. For most experi-
mental apparatus, the geometry is arranged such that the laser probe beam
path and the focal axis of the collection optics intersect either at right angles
or are parallel. Typically, both axes are parallel to a planar electrode surface
and either the collection optics and laser probe are translated simultaneously
across the discharge or the reactor vessel is moved relative to the excitation
point. The geometry for a plane parallel electrode cylindrical discharge reac-
tor is shown in Fig. 11-1.
LASER
PROBE
ELECTRODE
LENS
Figure 11-1. Geometry in a planar eleetrode cylindrieally symmetrie diseharge reaetor (sym-
bols defined in text).
490 CHAPTER 11
For most positions between the electrodes the fractional collection solid
angle for an assumed point source is
(11-4)
where dis the lens or aperture diameter at the entrance of the optical collec-
tion system andfis the focallength ofthe first lens and the last approximate-
equality assumes f» d. However, near an electrode surface the collection
solid angle is occulted. The critical distance at which this begins to happen
is
rd
x=- (11-5)
c 2f
where r is the difference in the radial position from the edge of the electrode
to the radical position of the excitation point, P (see Fig. 11-1). Thus, for
lateral positions, x, doser than Xc to an electrode the collection solid angle
becomes
(11-6)
For values of X more negative than the origin, the collected signal should
drop discontinuously to zero as the probe beam then strikes the electrode
surface. Clearly, this simple geometric analysis does not take into account
reflections and fluorescent scattering from the electrode which could be
sampled by the collection optics and assumes ideal optical properties for the
collection system.
An alternative way of determining the collection solid angle variation
is to fill the reactor system with a test gas that can be excited and will
fluoresce at wavelengths similar to those used in the actual diagnostic experi-
ments. By using the same excitation/collection optics and geometry, the
relative position of the reactor and the excitation point can be translated in
order to collect a signal that represents the true variation of any actual
diagnostic data as a function of the collection solid angle across the discharge
system. The advantage here is in being able to automatically account for
many systematic optical effects that can influence the LIF signal.
In spatial and temporal profiling of species concentration, the resolution
of the measurements is typically determined by the probe laser interaction
LASER-BASEO OlAGNOSrlCS OF REACrlVE PLASMAS 491
volume and the experimental timing. For LIF experiments, the laser inter-
action volume or focal spot size can be made as small as 100 pm in diameter.
F or most practical applications, this is ideal as spatial concentration profiles
would not be expected to vary discontinuously over such a small dimension.
Near electrode surfaces, where concentration gradients are expected to be
the most significant, the variation in collection solid angle as shown in Eq.
(11-6) contributes to larger uncertainties in spatial resolution and concentra-
tion assessment. Careful consideration of probe beam geometry and instru-
mental factors can help to minimize these larger systematic uncertainties
near surfaces, and allow refined measurements of concentration gradients.
Most LIF experiments in plasma discharges utilize pulsed lasers because
of their high peak output power, which helps in the discrimination of LIF
signal against ambient background contributions. Timing electronics today
can accurately synchronize the firing trigger of a pulsed laser system with a
particular phase of the current cyc1e of a discharge (rf or pulsed) with
very low relative jitter. Thus, the limiting factor that determines temporal
resolution is the temporal width of the laser probe light pulse. For most
systems, this pulse width can be IOnsec or less. This implies that if one
wanted to make temporally resolved concentration measurements of say 10
points per rf current cyc1e, the discharge frequency could be as high as
10 MHz. Other factors that influence temporal resolution are the radiative
lifetime of the LIF excited state and the species production kinetics. For
many situations, the steady-state populations of important reactive species
(as mentioned above) can be maintained throughout the rf current cyc1e and
temporal variations are observed only for low-frequency or pulsed
discharges.
Electric field profile measurements can also be influenced by the same
factors as for species concentration measurements. Large electric field varia-
tions are observed near electrode surfaces in the sheath regions. Thus, excita-
tion geometry will mostly determine the spatial resolution in these regions.
For electric field measurements made by optogalvanic spectroscopy (OGS),
the temporal resolution can be limited by the laser-induced photocurrent
transient time which can couple strongly with the electron energy distribution
function variation across the discharge. However, in most instances, the
temporal resolution is again limited (for pulsed laser excitation) by the laser
pulse temporal width and appropriate choice of impedance matching to
measuring electronics can ensure a temporal resolution approaching this
characteristic time.
11.3. Experimental Techniques
In this section we discuss the general experimental practices and tech-

niques used in laser-based diagnostics for plasma species and characteristics.
492 CHAPTER 11
A comparison to other known diagnostics is meant to support the basic

utility of usjng state-selective laser probe techniques for making in situ non-
perturbative measurements. However, in dealing with laser spectroscopic
diagnostics, care must be taken to have a comprehensive understanding of
the spectroscopy of any particular set of transitions used as a diagnostic or
else one's results may be misleading. Furthermore, every effort should be
made to delineate and minimize spurious systematic effects caused by the
experimental apparatus in order to make a qualified interpretation of the
data. For example, because ofthe high photon flux levels available from dye
lasers, one should always verify assumptions concerning power-broadening
and saturation effects. It is also important when using fluorescence tech-
niques to avoid saturation of detector elements such as photomultipliers,
which could lead to a false analysis of species profile data.
11.3.1. One-Photon LlF
One-photon LIF techniques make use of three quantum states of the

species in question (see Fig. 1I-2a), with the possibility of the initial and
IONIZATION
///J/J///// / / ///J//I//
CONTINUUM
GROUND STATE
One· Photon Two- Photon REMPI,OGS

LlF LlF
(2a) (2b) (2c)
Figure 11-2. Laser-based diagnostic excitation and detection schemes.
final states being the same. The initial excitation transition from the lower
(typically ground) state is pumped by the probe laser to an excited upper
state with a fiuorescence photon detected as this state radiates to an inter-
mediate level. One-photon LIF is a well-developed technique and can typ-
ically take advantage of repetitive pulsed visible dye laser systems yielding
intense probe beams with good spatial and temporal resolution.(4,5) Standard
excitation wavelengths range from the visible to the near UV. The detected
fiuorescence typically appears to the longer wavelength side of the excitation
wavelength, usually in the visible region to the near infrared, though some
fiuorescence transitions are in the near UV.
One-photon LIF as typically used for plasma measurements involves
transitions in free radicals, molecular ions, and metastable states of atoms,
and molecules. By analyzing the line shapes of transitions, information can
be obtained concerning the velocity of the species from the Doppler broaden-
ing and the electric field strength by analyzing the Stark splitting compo-
nents, in addition to species concentrations. For radicals and molecular ions,
measurements over successive rovibronic transitions can yield information
about relative populations between levels as weIl as determining rotational
and vibrational temperatures. In discharges this information is useful since
discharge kinetics do not necessarily preclude local thermal equilibrium
population distributions. Diagnostic LIF transitions can be chosen to have
a large transition moment between the lower (ground) and excited upper
states in order to enhance the probe discrimination against the background
plasma-induced emission. The excited upper state is usually weakly popu-
lated by the discharge. This minimizes spontaneous emission from nonlaser
excited molecules and stimulated emission by the laser thereby providing a
true measure of the ground state population. However, one-photon LIF
laser probes are absorbed along the length of the probe path. If the probe
laser is either a collimated or softly focused beam, the collected fiuorescence
corresponds to an integration of the lower state population along the beam
path. By strongly focusing the laser probe and limiting the photon fiux in
the probe, it is possible to restrict the contribution of the probe interaction
with the discharge to a specific small volume which can be translated around
the discharge region. However, saturation and power-broadening effects are
enhanced by this technique.
In arecent extension to typical concentration profiling experiments,
Hargis and Greenberg(6) have demonstrated a one-photon LIF diagnostic
scheme that uses a thin uniform sheet of laser probe light and an orthogonal
CCD array detector to image the LIF in thin strips across the entire discharge
gap as shown in Fig. 11-3. The advantage here is in being able to quickly
measure the relative LIF spatial profile from electrode to electrode provided
one is using strongly allowed transitions and has sufficient laser power
available.
494 CHAPTER 77
2-Dimensional Laser-Induced-Fluorescence Imaging
ImagHntenllifled
GltedCCD
CaIMf'l
Figure 11-3. Strip LIF laser diagnostic detection. The experimental arrangement for detecting
LIF across a reactor gap in thin strips is shown. The laser probe is formed into a uniform
collimated thin sheet that bridges the gap. The collection optics can image a thin strip ortho-
gonal to the laser probe onto the entrance slit of a spectrometer which provides wavelength
se1ection. The exit slit output is measured along a diode array, recording the variation in LIF
signal across the discharge.
11.3.2. Two-Photon and Multiphoton LlF
Measurements using transitions from ground states to higher electronic

states in atoms or molecules usually require more than one photon, since
the energy differences between such states would require the use of a single
VUV photon which is very difficult to generate and propagate in appreciable
quantities. An extension of the basic one-photon LIF technique (see Fig.
11-2b) makes use of nonlinear multiphoton and/or multicolor excitations
to the excited upper state. Indeed, the two excitation arrangements are so
closely related that often it is important to examine the role of multiphoton
processes even for one-photon LIF; particularly since high photon densities
caused by focusing of the probe beam can result in nonlinear processes such
as multiphoton ionization thereby depleting the upper state population.
Multiphoton processes usually involve much smaller absorption cross
sections for excitation than those in one-photon LIF. This fact implies it is
much more difficult to ensure a sufficient transition rate to the upper level
and hence to achieve a significant fluorescence photon yield that can be
detected against the plasma-induced emission. However, from another per-
spective multiphoton techniques can have a decided advantage over their
one-photon counterpart since the typically small absorption cross section
means that the laser probe must be focused to a very small spot size in order
to reach the high photon densities required for a transition. This means that
multiphoton LIF can in practice achieve high spatial resolution (100 pm or
less) without the adverse effects of saturation mentioned above. Further-
more, the requirement that the excitation transitions utilize two or more
photons implies the possibility, by using particular excitation geometries, of
making Döppler-free measurements or of measuring a particular directed
velocity component for a species from the Doppler profile.
Currently, the most practical form of diagnostic multiphoton LIF is
an excitation involving two photons of a single frequency as shown in
Fig. 11-2b. Because the excitation absorption ofthe photons must be simul-
taneous, one of the photons is absorbed either to a specific allowed state or
to an intermediate virtual state. The atomic selection rule for excitation
conserves parity, (7) which allows the technique to probe transitions that are
normally electric dipole forbidden. The detected fluorescence is emitted from
a highly localized volume which gives the technique its excellent spatial
resolution.
Two-photon LIF can be applied to measurements initiating from
ground states of the lighter atoms such as hydrogen or oxygen or between
the ground electronic state and higher electronic states in molecules. As
mentioned above, this technique is usually required since a single photon
excitation would require a coherent VUV source and increased experimental
complications. For multiphoton LIF, photons with tunable wavelengths
down to 200 nm can now be easily genera ted by using nonlinear mixing
crystals such as beta-barium borate (BBO)(8) or Raman shifting in a hydro-
gen gas cello
An alternative technique is two-photon stimulated emission
spectroscopy, (9) which has been applied to the detection of carbon, (10)
chlorine,(II) and hydrogen(12) in flames and cold gases. Here the advantage
is in the directionality of the stimulated emission along the probe beam
direction. This allows for a high collection efficiency of the stimulated fluor-
escence either in the forward or backward direction of the line of sight of
the probe beam path. However, the relationship between emission intensity
and species population is very nonlinear, making it quite difficult to infer
species concentrations from these measurements.
11.3.3. Multiphoton lonization

Absorption of additional photons to very high lying states in an atom
or moleeule and into the ionization continuum leads to the process of
496 CHAPTER 11
REMPI (see Fig. 11-2c).(\3) In this arrangement the ejected photoelectron is

detected upon avalanche multiplication in the discharge as a transient change
in the measured discharge impedance correlated in time with the firing of
the pulsed excitation laser probe. The data obtained with REMPI can be
complicated by a convolution with the position and energy dependence of
the avalanche multiplication process making the interpretation of profile
measurements essentially intractable.
11.3.4. Optogalvanic Detection
A technique related to REMPI is OGS in which a laser-induced im-

pedance change is observed across the discharge.(14) Typically, upper elec-
tronic states of an atom or molecule are excited from either the ground or
metastable states with ionization induced by absorption of an additional
photon from the probe beam, photons from the discharge, and energetic
collisions of the excited state molecule with other plasma species. (15)
The most reliable application of OGS has been for the measure of
electric fields by observing Stark effect perturbations and mixing of energy
levels in Rydberg atoms and molecules. The influence of Stark effect pertur-
bations can generally be ca1culated in a straightforward manner,(16,17) so that
a comparison of the degree of Stark splitting in a resonance manifold can
yield a value of the magnitude of the local electric field strength as weIl as
the e1ectric field vector directionYS) Since the wavelength location of the
components of the Stark split resonance manifold are influenced only by the
local electric field at the measurement point, this technique does not suffer
the same drawback present with REMPI concentration determinations,
because of the convolution of the abundance with the complicated avalanche
process. OGS for measurement of electric fields has been demonstrated to
have excellent spatial and temporal resolution. Indeed, the temporal
response of OGS has been shown to be excellent, approaching the minimum
time resolution set by the laser temporal pulse width.(19)
11.3.5. Absorption Techniques
Reliable measurements of species concentrations in reactive plasma sys-

tems using laser absorption techniques have been very difficult to achieve in
the past. A most notable success of this technique has been the recent detec-
tion and measurement of rovibronic population of X I:E; H 2 in a plasma
by VUV laser absorption. (20) The success of this experiment depended on the
generation via harmonie techniques of coherent radiation at approximately
112nm.
Nonetheless, for the most part, the absorption technique has thus far
depended on the use of tunable IR lasers for absorption within a ground
state energy level manifold of the atom or molecule. The difficulties that
exist with this technique are varied. Since reactive plasmas at low pressures
and species concentrations are optically thin, the observed fractional absorp-
tion tends to be exceedingly small and requires long absorption path lengths,
which limits both the dynamic range and minimum detectable concentration
for this approach. As the change in absorption is detected in the forward
direction along the line of sight of the probe beam, this causes an integration
of the species concentration behavior across the discharge necessitating the
use of Abel inversions and assumptions that limit the spatial resolution. An
additional complication results from the necessity of observing IR transitions
which can coincide with a large IR background because of the thermal
signature of the discharge and the surrounding environment. Similarly, the
previous assumption of a weakly populated upper state may no longer
always be valid.
Despite these difficulties, reactive halogen atoms have been detected in
plasmas by this technique. Concentrations and translational temperatures
for atomic chlorine have been measured in a chlorine glow discharge using
diode laser absorption at 882.4 cm- 1 by Wormhoudt and StantonYl) A
successful detection measurement for one ofthe more elusive reactive plasma
species, atomic fluorine, was demonstrated by Stanton and Kolb(22) using a
tunable lead-salt diode laser to probe a ground state fine structure transition
at 404 cm- 1 (25 J-lm).
11.3.6. Comparison with Other Diagnostic Methods
We can compare the basic laser-based techniques mentioned above with

other standard diagnostic methods such as plasma emission detection and
actinometry for species concentration measurements, ex situ titration detec-
tion of transient species, nonlaser absorption methods, electric probe
measurements of charged species, and mass spectrometric analysis of the
reactive plasma constituents. Some or all of the following general problems
can arise in utilizing these latter techniques. First, there can be a reduction
in the precision or resolution of the data; the collected data may not truly
reflect the presence or profile distribution of a species or parameter because
of interference by other plasma processes. Second, the diagnostic technique
itself could be significantly perturbative of the discharge environment.
Finally, some methods make measurements ex situ to the plasma environ-
ment adding further assumptions in order to extrapolate data back into the
plasma.
11.3.6.1. Emission Speetroseopy
Direct emission measurements from discharges are extremely useful

particularly in conjunction with LIF techniques. (23) Indeed, emission
498 CHAPTER 11
measurements are still the only manner in which to detect some reactive
species. Recent research efforts have been directed toward analyzing emission
from chlorocarbon, (24) fluorocarbon, (25) and hydrocarbon(26) deposition and
etching discharges. Carbon-bearing plasmas have extremely complicated
emission spectra, as a large variety of hot intermediate radical and ionic
species are formed, many of whose spectra have not been fully interpreted
even in less complicated environments.
11.3.6.2. Actinometry
Actinometric concentration measurements are basically an emission-

based technique where the concentration of a reactive species [r] is given by
[r] = Ir)(aair) [I].

(L (11-7)
where [i] is the known controlled concentration of an inert trace gas species,
Ir
such as argon, added to the discharge. and I i denote emission intensities
for specific transitions in the reactive and inert species (normally in the
same wavelength region), and aare constants of proportionality between
concentration and emission intensity for each species. (27)
Actinometry is primarily used to measure concentrations of light reac-
tive atoms and some radicals. (See Table 11-1.) An important criterion for
the successful use of actinometry is that the ratio arj
ai must be constant
regardless of position within the discharge or discharge conditions so long
as the concentration ratio [r]j[i] does not change. Typically, it is assumed
that the addition of an inert gas in moderate trace amounts will not perturb
the discharge environment. However, this assumption may not always be
justified. Heavy inert gas ions will impact on electrode surfaces changing
secondary electron and ion emission rates as well as possibly participating
Table 17-1. Various Reactive species

rand Insert Gases i That Have Been
Used for Actinometric Diagnostics
r
F Ar
0 N2
Br
CO
CO+
CF
CF 2
Cl
CCI
in three-body reactions and thus changing recombination rates or act as
excited state quenchers in the bulk of the plasma. Moreover, in practice this
technique often has reduced spatial and temporal resolution compared with
LIF results.
11.3.6.3. Ex Situ Titration
Downstream chemiluminescent titration is used to detect transient

atoms and radicals generated in flowing gas reactive plasmas. The method
works by introducing a controHed amount of a chemical reactant down-
stream of a discharge or in the flowing afterglow. Through a known chemi-
luminescent reaction, the transient species combines to form an excited state
product which may decay with a characteristic emission proportional to
concentration or an alternative chemiluminescent reaction initiates at a
known concentration after the transient species is nearly or completely
scavenged from the flow discharge products. (28)
Titration has been used to detect atomic aluminum,<29) chlorine, (30)
fluorine,(31) nitrogen,(28) and oxygen,<32) as weH as radicals such as CN.(29)
The obvious drawbacks of chemiluminescent titration are the complete loss
of spatial concentration profile-dependent information and the possibility of
further reactions and interactions with surfaces downstream which can alter
the relative populations of species from those in the reactive plasma.
11.3.6.4. Absorption Spectroscopy
Absorption measurements using nonlaser sources, such as discharge

lamps, can probe directly for the rotational and vibrational temperature of
reactive species. High-resolution Fourier transform infrared (FTIR) tech-
niques are able to measure simultaneously the complicated cold and hot band
structures of radicals and molecules present in discharges using broadband
sources. FTIR absorption and emission has been applied to silane (SiH4)
discharges to measure the rotational and vibrational temperature of the SiH
radical as weH as attempting to detect SiH2 and SiH3 radicals. (33)
VUV absorption spectroscopy has been used to measure concentrations
of ozone, ground state, and metastable molecular oxygen in a dc glow
discharge.(34) Absorption techniques generaHy require sophisticated instru-
mentation to handle the large amount of convoluted spectral information
present in a reactive plasma and, as mentioned with laser-based absorption
probes, lack good spatial resolution because of the lengthy absorption paths.
11.3.6.5. Charged Particle Methods
Two related measurements are in situ electric or Langmuir probes and

quasi-ex situ mass spectrometry. Electric probes are used to measure positive
500 CHAPTER 11
ion density, electron temperature, and dc localized space potential in rf

discharges. (35.36) In principle, electric probes could provide a wealth of well-
resolved spatial and temporal information. However, there are systematic
perturbations introduced by the very presence of the e1ectric probe in the
plasma environment. These effects make the accurate interpretation of
Langmuir probe data difficult.
Mass spectrometry can be performed in very elose proximity to a
reactive plasma as a very general means of measuring ionic species and with
the addition of electron- or photo-ionization neutral moleeules can also be
detected, often with excellent sensitivity.(37) The primary disadvantage of
mass spectrometry lies in its ex situ nature, meaning that one is never quite
sure that the species in the mass spectrometer are the same as in the plasma
itself, with a corresponding loss of spatial and temporally resolved informa-
tion. Additionally, fragmentation patterns that are not well known, upon
ionization, particularly for radical species, lead to ambiguities in species
identification.
11.4. Species Detected and Studied
Having reviewed the basic experimental techniques and practices for

laser-based plasma diagnostics we now turn to a survey of reactive species
and plasma characteristics that have been measured by these techniques.
The material presented, while fairly inc1usive, is probably not exhaustive
but should serve as an excellent introductory survey for familiarizing the
researcher with demonstrated measurement capabilities.
11.4.1. One-Photon LlF Detection
A number of reactive plasma species that have been measured with one-
photon LIF techniques are listed in Table 11-2, which presents the particular
species, the plasma source, the excitation wavelength and transition, the
observed fluorescence wavelength and transition, and arecent reference.
Note from Table 11-2 that the majority of the species are radicals which
have particular relevance to reactive plasma deposition or etching. As indica-
ted in Seetion 11-3, absolute concentration measurements for radicals in
discharges are both demanding and important for a complete description
of these processes. However, current methods of determining instrumental
factors can have large uncertainties (and vary from experiment to experi-
ment), making the interpretation of data for absolute measurements difficult.
Along with measurements for radicals, several measurements are presented
for species that find use in processing discharges for ion impact etching and
actinometry .
Table 11-2. Reactive Plasma Species Detected by One-Photon LIF
Excitation Fluorescence
Source wavelength Excitation wavelength Fluorescence
Species discharge (nm) transitiona (nm) transition Ref.
CF2 Pulsed, CF, 248,266 Ä'A,-X'A, 257-271 A-X 38
CF Pulsed, CF, 193 B2i\_X2rr 194-220 B-X 38
BCI rf, BCh 272 A'rr-X':E+ 272 A-X 39
OH Afterglow 281-284 A2:E+ _x2rr 312 A-X 40
Nt Pulsed, N2 337 C3rru -B 3rrg 380-467 C-B 41
N; Pulsed, N2 428 B2:E~_X2:E; 391 B-X 41
F* CF,/02/Ar 690 'P3 / 2- 'D S/ 2 677 'D S/ 2- 'PS / 2 42
Ar* CF,/02/Ar 696 Is s-2p, 727,772 2p,-ls3 ,ls, 42
CI; rf,Ch 386 A2rr_x2rr 396 A-X 43
CCl rf, CCI, 278 A2i\_X 2rr 278 A-X 44,45
aFollowing standard convention, for molecules the upper state is Iisted first; for atoms the convention is
reversed, the lower state is listed first.
It should also be noted that many other important species have been
detected either downstream from reactive plasmas or in chemical vapor
deposition (CVD) systems. CVD can produce reactive species with concen-
trations comparable to those generated directly in reactive plasmas. The
important point is that given a similar concentration in a reactive plasma
and the ability to discriminate against the ambient background plasma-
induced emission, the LIF detection schemes for these reactive species should
be equally as straightforward for direct detection in the plasmas as from
other sources.
In addition to the species listed in Table 11-2, LIF detection of rare
gas metastable states has been demonstrated for He, Ne, Ar, Kr, and Xe
downstream from a dc discharge. (46) Metastable helium has also been
detected using optogalvanic techniques in dc discharges.(47) Rare gas atomic
metastable detection and profiling in discharges is important because of the
considerable energy content such metastables can supply during collisional
processes in the bulk plasma and with surfaces. Detection of SiF2 and CF 2
radicals has been performed during downstream silicon etching from a
CF4/0 2 microwave discharge.(48) The important observation ofOH and SiO
by LIF from CVD deposition of Si02 provides rotational and vibrational
temperatures present during the reaction processes.(49)
11.4.2. Two-Photon LlF Detection
Species detected by two-photon allowed LIF (TALIF) results are shown

in Table 11-3. The data shown here are presented in the same manner as for
Table 11-2, with the addition of a column for the total excitation energy for
the two-photon excitation transition. As can be seen, TALIF is weH suited
502 CHAPTER 17
Table 11-3. Reactive Plasma Species Detected by TAUF

Excitation Etotal Fluorescence
Source wavelength Excitation 2-photon wavelength Fluorescence
Species discharge (nm) transition (ern-I) (nm) transition Ref.
Cl rf, CClF3 233 3p2jiJ-4p4SJ 85726 725-775 4p4SJ-4s4p 50
H rf,H2 205 I?S-3d 2D 97492 656 3d 2D-2ip 51
N Afterglow 211 2p 34SJ_ 94787 869 2p23p 4Do_ 52
2i3i3p4Do 2p23s4p
0 rf, CF40 2 226 2p43p_3p 3p 88495 845 3p 3p_3s3S 19
to concentration measurements for light reactive atoms. Indeed, light atoms

are perhaps the dominant reactive species to interact with surfaces on the
boundary of the discharge. Thus, their detection and profiling in plasma
processing systems is critical to a detailed interpretation of the plasma-
surface interactions. The TAUF measurements for these atoms require the
use of two UV photons typically near the short-wavelength generation limit
of most tunable dye laser systems and as such are more demanding experi-
mental techniques than those used for one-photon UF.
As with the one-photon UF measurements discussed above, several
atomic species important to reactive plasmas have been detected by TAUF,
though not as yet directly in plasmas. Multiphoton dissociation of molecules
and the subsequent detection of an atomic fragment by TAUF has been
exhibited for carbon,(53) silicon,(54) and sulfur.(55) Given the caveats
mentioned for the ex situ species detected by one-photon UF, the TAUF
detection and profiling of these atoms directly in reactive plasmas should be
possible.
11.4.3. OGS and REMP! Detection
In a plasma, it is sometimes difficult to distinguish between OGS and

REMPI processes. Let us refer to these processes in general as laser-ion
detection, since they both involve excitation of a bound-bound transition in
an atom or molecule followed by detection of charged particles, as a change
in the discharge current.
Perhaps the most direct use of laser-ion techniques was the pioneering
work on Ni and CO + in an rf plasma by Walkup et al. (56) These experiments
were made possible by the fact that the charge-exchange cross section was
a strong function of the electronic state of the ion. Thus, by laser pumping
the ions from one state to another, a significant change in the discharge
current or impedance could be created and detected. The technique was
particularly sensitive in the cathode dark space.
In the case of molecular hydrogen, recent work has shown the possibility
of detecting highly vibrationally excited, ground electronic state H2 by 2 + 1
REMPI.(57,58) Hydrogen-bearing plasmas have very nonthermal vibrational
level population distributions for H 2 • Analysis of these expected high-Iying
vibrational population distributions has shown that such distributions can
be an appreciable conversion source of vibrational to translational energy
during collisions,(59) as weIl as perhaps the predominant source of the nega-
tive ion, H-, from dissociative attachment. (60) Detection of excited electronic
states by REMPI has been demonstrated for H 2 .(61)
11.4.4. Plasma Characteristic Measurements by Laser-Based

Techniques
Laser-based diagnostic probes are weIl suited for measuring plasma

characteristics other than atomic and molecular species concentrations.
Some of these characteristics are localized electric fields, negative ion densi-
ties, atomic or molecular velocity vectors via Doppler profiles, and micro-
scopic particulate contamination.
11.4.4. 1. E/ectric Fie/d Strength
Laser-based diagnostics of electric fields have been performed using

both LIF and optogalvanic detection. Both techniques rely on measuring
Stark effect perturbations to molecular energy levels due to the local electric
field. (62) The Stark effect can perturb the coulombic interactions in either an
atom or molecule thereby shifting energy levels and causing mixing of elec-
tronic wave functions. The alterations to the wave functions lead to changes
in transition probabilities between states. Thus, the Stark effect both induces
energy shifts in allowed transitions and permits normally forbidden transi-
tions to occur. The utility of detecting these Stark effect perturbations is that
one can generally calculate these effects as a function of electric field. The
predictions of the calculated model can then be fitted to the real data in an
iterative fashion as a function of electric field. Both LIF and aas techniques
generate data showing wavelength shifts and intensity modulations as a
function of the local electric field at the probe point. Analysis of influences
of the excitation source polarization on the perturbed transitions is impor-
tant and can yield further confirmation of the electric field magnitude and
even vector direction.
LIF-based detection of electric fields has been very successfully demon-
strated in electronegative BCh rf etching discharges. (63) Here, Stark effect
mixing of rotational wave functions with opposite parity in diatomic BCI
permits forbidden transitions between two electronic states which are
detected by LIF. The method can measure electric fields as small as 40 VIcm
to 1-2 kV lern with a spatial resolution of 100 pm along an axis orthogonal
to the planar electrode surfaces. Since the excitation laser is repetitively
504 CHAPTER 17
pulsed, the diagnostic can be phase synchronized with the rf driving voltage
across the discharge with a temporal resolution of 20 ns, thus allowing time-
resolved measurements of the electric field temporal profile. (64)
A comparable technique to LIF measurements of electric fields is the
method ofRydberg state Stark OGS.(65) High-Iying Rydberg states ofatoms
or molecules exhibit a pseudohydrogenic behavior under the influence of the
Stark effect; in other words, Rydberg molecules can be modeled as a singly
charged positive ion core with a single valence electron in an extended orbit
that is perturbed by the electric field in the same manner as hydrogen. (62)
Usually excited intermediate or metastable states of atoms or molecules
which can be populated by discharge kinetics are used as a starting point in
the excitation scheme. A pulsed probe laser beam excites the metastable
species to a perturbed Rydberg level. Again the excited Rydberg states are
mixed as a function of the electric field so that line positions and intensities
relative to the unperturbed lower state can be compared with calculated
values. Further excitation of this level to the ionization continuum is pro-
vided by additional photons, collisions with other species, or field ionization.
At this point the transient change in the discharge impedance is measured
as an optogalvanic signal incorporated in the discharge current waveform.
Rydberg state OGS has been shown to be capable of measuring very
small electric fields from 5 V Icm to fields over 5 kV Icm. As with LIF
measurements, the technique has excellent spatial resolution (100 pm). In
addition, analysis of forbidden transition intensity patterns as a function of
excitation probe laser polarization can determine the electric field vector
direction.(65) Furthermore, these measurements have been performed in a
time-resolved fashion as a function of rf driving voltage phase. (66)
11.4.4.2. Photodetachment and Negative Ions

Negative ion densities have been measured by laser photodetachment
techniques. Determination of negative ion production and kinetics in reactive
plasmas has considerable current interest. (67) Negative ions in electronegative
discharges used for etching of materials can reach high densities comparable
to those for electrons. The role of negative ions in many reactive plasmas is
not weIl understood, particularly since calculations of production rates often
fall short of measured values. (68,69)
Photodetachment of electrons from negative ions usually require only
a single relatively low"energy photon absorption. Once liberated in the dis-
charge, the laser-induced electron current is detected either by optogalvanic
detection across the discharge(70) or by measurements made from a metal
electrode placed very near the laser probe interaction region.(7I) Negative
ion densities down to 108 cm -3 have been measured as weIl as electron affin-
ities for various molecules and radicals(72) in reactive plasmas and negative
ion kinetic energy. (71)
11.4.4.3. Energy Content by Doppler Profiles
The translational energies of reactive species in discharges can be

determined by careful analysis of line-broadening effects cause by relative
Doppler motion along a probe direction. Photodissociation of molecules by
lasers and subsequent detection of molecular fragments by LIF or mass
spectrometry has been used to determine bond dissociation energies.(73) Like-
wise, similar analysis of LIF measurements on molecular or atomic frag-
ments in discharges yie1ds translational energies for the species along a
particular direction. (74) A full analysis of line shapes must take into account
possible power-broadening effects and the contribution of the laser line pro-
file to the data.
11.4.4.4. Particulates
Finally, recent advances have been made in laser scattering measure-

ments for detection of microscopic scale particulate contamination in dis-
charges. Significant amounts of particulate contaminates are formed from
etching or sputtered products from the bounding surfaces or from gas-phase
c1ustering. Their presence can adversely affect the production and per-
formance of electronic devices and nove1 materials. Laser scattering can
be used to determine the density profiles and size of the particulates. (75)
Measurements using two-photon LIF on neutral atomic species, which have
been formed from photodetached electrons from the particulates, indicate
the particulates are negative1y charged and can be suspended near sheath
boundaries. (76)
11.5. Conclusions and Future Trends
The overall conc1usion concerning laser-based diagnostics is very posi-

tive, portending increased use of this technique in the future. This prediction
is based on several facts. The number of plasma species for which laser-
based detection and characterization methods now exist is substantial as
evidence by Tables 11-2 and 11-3. In every like1ihood, it will continue to
grow. Furthermore, the perception that laser-based diagnostics are too com-
plicated and expensive is eroding. The large number of successfullaser-based
diagnostics that have been reported, some extending to the production line,
belie most ofthe negative arguments concerning their complexity. Similarly,
while the cost of the lasers is not low, this cost must be weighted against the
much greater cost, both capital and operating, of plasma processing in gen-
eral. Viewed in this manner, perhaps the better question is not who can
506 CHAPTER 11
afford laser-based diagnostics, but who can afford not to have laser-based
diagnostics.
Predicting future technical advances for laser-based diagnostics is like
all prognostications-uncertain at best. Clearly additional specific spectral
applications will be found to monitor important intermediates. Probably the
strongest driving force in this respect will not be spectroscopic advances, but
new processes themselves, which require the identification and monitoring
of different species.
Another growth area for laser-based diagnostics is their extension
beyond the traditional analytical applications, qualitative and quantitative
analysis of plasmas in terms of species and concentrations. It has recently
been recognized that laser-based diagnostics offer nonperturbative probes
for other characteristics of plasma. Examples of quantifiable entities inc1ude
electric field measurements from Stark effects, kinetic energy or local tem-
perature measurements from Doppler effects, and correlation of rnicroscopic
particu1ate formation with negative ion detection. One would expect to see
further developments involving measurements of other plasma-related
properties via their effect on the spectral transitions involved in laser-based
diagnostics.
Acknowledgment
The authors gratefully acknowledge the support of this work by USAF

Contract No. F33615-89-C-2921.
References
1. S. Veprek, Plasma Chem. Plasma Process. 9 Suppl. (1989) 31S.

2. B. Chapman, Glow Discharge Processes. Wiley, New York, 1980.
3. R. K. Janev, W. D. Langer, K. Evans, and D. E. Post, Elementary Processes in
Hydrogen-Helium Plasmas. Springer-Verlag, Berlin, 1987.
4. A. Corney, Atomic and Laser Spectroscopy, Clarendon Press, Oxford, 1988.
5. W. Demtroder, Laser Spectroscopy, Springer-Verlag, Berlin, 1982.
6. P. J. Hargis and K. E. Greenberg, 42nd Annual Gaseous Electronics Conference, Pal0
AIto, 1989.
7. K. D. Bonin and T. J. McIlrath, J. Opt. Soc. Am. B 1 (1984) 52.
8. C. Chen, Laser Focus World 129 (Nov. 1989).
9. M. AIden, U. WestbIom, and J. E. M. Goldsmith, Opt. Lett. 14 (1989) 305.
10. M. AIden, P.-E. Bengtsson, and U. WestbIom, Opt. Commun. 71 (1989) 263.
11. A. D. Sappey and J. B. JefIries, Appl. Phys. Lett. 55 (1989) 1182.
12. J. E. M. Goldsmith, J. Opt. Soc. Am. B 6 (1989) 1979.
13. E. W. Rothe, G. S. Ondrey, and P. Andersen, Opt. Commun. 58 (1986) 113.
14. J. E. M. Goldsmith and J. E. Lawler, Contemp. Phys. 22 (1981) 235.
15. M. Maeda, Y. Nomiyama, and Y. Miyazoe, Opt. Commun. 39 (1981) 64.
16. P. H. Heckmann and E. Trabert, Introduction to the Spectroscopy of Atoms, Chapter 11,
North-Holland, Amsterdam, 1989.
17. C. DeIsart, L. Cabaret, C. BlondeI, and R.-J. Champeau, J. Phys. B 20 (1987) 4699.
18. B. N. Ganguly and A Garscadden, Phys. Rev. A 32 (1985) 2544.
19. C. E. Gaebe, T. R. Hayes, and R. A Gottscho, Phys. Rev. A 35 (1987) 2993.
20. G. C. Stutzin, A. T. Yang, A S. Schlacter, K. N. Leung, and W. B. Kunkel, Chem. Phys.
Lett. 155 (1989) 475.
21. J. Wormhoudt and A C. Stanton, J. Appl. Phys. 61 (1987) 142.
22. AC. Stanton and C. E. Kolb, J. Chem. Phys. 72 (1980) 6637.
23. R. E. Walkup, K. L. Saenger, and G. S. Selwyn, J. Chem. Phys. 84 (1986) 2668.
24. P. E. Clarke, D. Field, and D. F. Klemperer, J. Appl. Phys. 67 (1990) 1525.
25. H. Kawata, Y. Takao, K. Murata, and K. Nagami, Plasma Chem. Plasma Process 8 (1988)
189.
26. T. Kokubo, F. Tochikubo, and T. Makabe, J. Phys. D 22 (1989) 1281.
27. J. W. Cobum and M. Chen, J. Appl. Phys. 51 (1980) 3134.
28. S. Meikle and Y. Hatanaka, Appl. Phys. Lett. 54 (1989) 1648.
29. S. Meikle, H. Nomura, Y. Nakanishi, and Y. Hatanaka, J. Appl. Phys. 67 (1989) 483.
30. D. A. Daner and D. W. Hess, J. Appl. Phys. 59940 (1985)
31. K. Ninomiya, K. Suzuki, S. Nishimatsu, and O. Okada, J. Appl. Phys. 58 (1985) 1177.
32. C. Vinckier, P. Coeckelberghs, G. Stevens, M. Heyns, and S. De Jaegere, J. Appl. Phys. 62
(1987) 1450.
33. J. C. Knights, J. P. M. Schmitt, J. Perrin, and G. Guelachvili, J. Chem. Phys. 76 (1982)
3414.
34. G. Gousset, P. Panafieu, M. Touzeau, and M. Vialle, Plasma Chem. Plasma Process 7
(1987) 409.
35. A von Engel, Electric Plasmas: Their Nature and Uses, Taylor & Francis, London, 1983.
36. H. Sabaldi, S. Klagge, and M. Kammeyer, Plasma Chem. Plasma Process 8 (1988) 425.
37. J. E. Cantle, E. F. Hall, C. J. Shaw, and P. J. Turner, Int. J. Mass Spectrom. Ion Phys. 46
(1983) 11; J. W. Cobum, Thin Solid Films 171 (1989) 65.
38. S. G. Hanson, G. Luckman, and S. D. Coison, Appl. Phys. Lett. 53 (1988) 1588.
39. c. A. Moore, G. P. Davis, and R. A. Gottscho, Phys. Rev. Lett. 52 (1984) 538.
40. N. G. Adams, C. R. Herd, and D. Smith, J. Chem. Phys. 91 (1989) 963.
41. R. F. Wuerker, L. Schmitz, T. Fukuchi, and P. Straus, Chem. Phys. Lett. 150 (1988) 443.
42. S. G. Hansen, G. Luckman, G. C. Nieman, and S. D. Colson, Appl. Phys. Lett. 56 (1990)
719.
43. V. M. Donnelly, D. L. Flamm, and G. Collins, J. Vac. Sei. Technol. 21 (1982) 817.
44. R. A Gottscho, R. H. Burton, and G. P. Davis, J. Chem. Phys. 77 (1982) 5298.
45. R. A Gottscho, G. P. Davis, and R. H. Burton, J. Vac. Sei. Technol. A 1 (1983) 622.
46. V. E. Bondybey and T. A Miller, J. Chem. Phys. 66 (1977) 3337.
47. B. N. Ganguly, J. Appl. Phys. 60 (1986) 571.
48. Y. Matasumi, S. Toyoda, T. Hayashi, and M. Miyamura, J. Appl. Phys. 60 (1986) 4102.
49. P. Van der Weijer and B. H. Zwerver, Chem. Phys. Lett. 163 (1989) 48.
50. G. S. Selwyn, L. D. Baston, and H. H. Sawin, Appl. Phys. Lett. 51 (1987) 898.
51. B. L. Preppemau, D. A Dolson, R. A Gottscho, and T. A Miller, Plasma Chem. Plasma
Process. 9 (1989) 157. See also B. L. Preppemau and T. A Miller, J. Vac. Sei. Technol. A.
8 (1990) 1673.
52. W. K. Bischel, B. E. Perry, and D. R. Crosley, Chem. Phys. Lett. 82 (1981) 85.
53. P. Das, G. Ondrey, N. van Veen, and R. Bersohn, J. Chem. Phys. 79 (1983) 724.
54. P. D. Brewer, Chem. Phys. Lett. 136 (1987) 557.
55. P. Brewer, N. van Veen, and R. Bersohn, Chem. Phys. Lett. 91 (1982) 126.
508 CHAPTER 11
56. R. Walkup, R. W. Dreyfus, and P. Avouris, Phys. Rev. Lett. 50 (1983) 1846.
57. D. C. Robie, L. E. Jusinski, and W. K. Bischel, Appl. Phys. Lett. 56 (1990) 722.
58. W. Meier, H. Zacharias, and K. H. Welge, ehem. Phys. Lett. 163 (1989) 88.
59. M. Cacciatore, M. Capitelli, and G. D. Billing, ehem. Phys. Lett. 157 (1989) 305.
60. K. N. Leung and W. B. Kunkel, Phys. Rev. LeU. 59 (1987) 787.
61. E. E. Marinero, R. Vasudev, and R. N. Zare, J. ehem. Phys. 78 (1983) 692.
62. H. A. Bethe and E. E. Salpeter, Quantum Mechanics 0/ One and Two Electron Atoms,
Plenum Press, New York, 1977.
63. C. A. Moore, G. P. Davis, and R. A. Gottscho, Phys. Rev. Leu. 52 (1984) 538.
64. M. L. Mandich, C. E. Gaebe, and R. A. Gottscho, J. ehem. Phys. 83 (1985) 3349.
65. B. N. Ganguly, J. R. Shoemaker, B. L. Preppernau, and A. Garscadden, J. Appl. Phys. 61
(1987) 2778.
66. B. L. Preppernau and B. N. Ganguly, Paper KB-5, 39th Annual Gaseous Electronics
Conference, 1986.
67. R. A. Gottscho and C. E. Gaebe, IEEE Trans. Plasma Sei. PS-14 (1986) 92.
68. M. Bacal and G. W. Hamilton, Phys. Rev. LeU. 42 (1979) 1538.
69. A. Garscadden and W. F. Bailey, Paper No. 188, 12th Int. Symp. Rarefied Gas Dynamies,
1980.
70. C. E. Gaebe, T. R. Hayes, and R. A. Gottscho, Phys. Rev. A 35 (1987) 2993.
71. D. Devynck, A. M. Bacal, P. Berelmont, J. Bruneteau, R. Leroy, and R. A. Stern, Rev.
Sei. Instrum. 60 (1989) 2873.
72. K. E. Greenberg, G. A. Hebner, and J. T. Verdeyen, Appl. Phys. Leu. 44 (1984) 299.
73. B. Koplitz, Z. Xu, and C. Wittig, ehem. Phys. Leu. 137 (1987) 505.
74. J. Dunlop, A. Tserepi, B. L. Preppernau, T. Cerny, and T. A. Miller, Plasma ehem. Plasma
Process. 12 (1992) 1.
75. G. S. Selwyn, J. Singh, and R. S. Bennett, J. Vac. Sci. Technol. A. 7 (1989) 2758.
76. G. S. Selwyn, J. S. McKillop, and K. L. Haller, Proc. SPIE Integ. Process. (1. Bondur and
A. Reinburg, eds.), Santa C1ara, Calif., 1989.
Index
Actinometry, 498 Atomic diffusion (Cont.)

Alloys, analysis by transport theory (Cont.)
atomic absorption, 104 radial profiles, 83
atomic emission, 154 transient sampIes, 85
atomic fluorescence, 446 Atomic emission spectroscopy
hollow cathode emission, 242 analytical performance, 146-153
mass spectrometry, 197 applications, 153-160
Analyte Atomsource T" , 69-71, 79 bulk metals and alloys, 154
Atomic absorption spectrophotometry depth profiling, 155
analyte transport, 78-86 nonconducting solids, 158-160
applications solution residues, 157
alloys, 104 background radiance, 130
nonconductive sampIes, 105 comparisons with other methods, 169
solution residues, 105 detection limits, 147-149
background correction, 94 detectors, 135-139
commercial instrumentation, 69 Grimm-type discharge, 128
effect of line source characteristics, 76 linewidths, 120
effect of sputtering conditions, 89 matrix effects, 150
flow-assisted atomizers, 69-71, 79, 285 microwave boosting, 157, 161-164
historical, 67-71 molecular components, 119
instrumentation, 68-71, 99 noise power spectra, 152
isotopic analysis, 107 noise sources, 146
laser line source, 460, 472-474 precision, 152
multielement determinations, 99 radiation sources, 122-129
precision, 93 radio-frequency sources, 310, 322
pulsed atomizers, 96-98 rotational temperatures, 120
spectral profiles, 71-77 spectral characteristics, 113-122
standardization, 100-104 background, 130
Atomic diffusion, 86 Bohr model, 113
transport theory, 80 Boltzmann factors, 116
axial profiles, 84 line intensities, 115, 148
kinetic modeling, 81 molecular, 119
optimum gas velocity, 82 selection mIes, 115
509
510 INDEX
Atomic emission spectroscopy (Cont.) Degenerate four-wave mixing, 474-476

spectral characteristics (Cont.) Depth profiling
self-reversal, 121 atomic emission, 155
spectrometers, 128-135 mass spectrometry, 203, 321; see also Thin
Fourier transfonn, 164-166 films analysis
photodiode array, 167 Die press, 268
Atomic fluorescence spectroseopy Differential sputtering, 34-37
"analytieal" plasma diagnostics, 449-452 Doppler-free spectroscopy
applications hollow eathode, 249
alloys, 446 laser saturation absorption, 472
solution residues, 446-448
historical, 67-71 Electron
influence of spectra1 profiles, 77 densities, 45-52
instrumentation, 68-71 energy distribution function (EEDF), 47-51,
118, 223
Background eorrection Maxwellian nature, 50
atomie absorption, 94 sarnpling teehniques, 45
atomic emission, 132 spatial distribution, 48, 52
FANES, 409 temperature, 46
Boltzmann factors, 116 Elemental analysis
Boosted glow discharges, 160-164 aqueous sarnples, 3
ideal solid sampie techniques, 5
Catalyst materials, analysis of 100, 105, 283, scope, 2
285-288 solid sarnple, 5-10
Cathode dark space, 21 Excitation mechanisms, 38, 53, 57-59, 117
FANES systems, 384 Grimm diseharges, 59, 115-120, 144
hollow cathodes, 44 hollow cathodes, 227
Stark broadening in OGE spectra, 460 Excitation temperatures, 250
thickness, 22, 42
thickness-ion energy relationship, 352 Flow-assisted devices, 69, 79, 168, 192, 285
Cathode fall potential, 37 FANES (furnace atomization nonthennal
Cathodic sputtering , 26, 37 excitation spectrometry), 126
crater shapes, 143 applications
redeposition, 39 metals, 413-415
Ceramics, analysis of, 263 nonmetals, 415
Charge exchange, 58; see also Ionization background correction, 409
mechanisms comparison with graphite furnace-AAS, 375
Collision-induced dissociation, 315 comparison of geometries, 376-379,410
Compacted sarnples current-voltage relationships, 389-392
analysis by detection limits, 410
atomic absorption, 100, 105 effect of atomization temperature, 399-402
atomic emission, 158-160, 242 effect of discharge current, 398, 402-405
mass spectrometry, 205-207 effect of diseharge pressure, 392
analyte migration, 275-278 electrical characteristics, 386
effect of particle size, 273 furnace designs, 376-382
effeet of trapped gases/water vapor, 273 HA-FANES (hollow anode), 375
host matrix identity/composition, 270-273 HC-FANES (hollow cathode), 375
presputtering, 278 moleeular species determinations, 415
sarnple preparation, 158-160, 265-270 noise sources, 406
standardization, 281 operating eonditions, 378
radio-frequency-powered, 376; see also rf-
de ares, 123 FANES; FAPES
INDEX 511
FANES (furnace atomization nonthermal Hollow cathode discharges (Cont.)
excitation spectrometry) (Cont.) applications (Cont.)
sheath (dark space) structure, 384 compacted sampies, 241
spatially-resolved optical sampling, 396 gases, 244
sputtering considerations, 382 solution residues, 242
thermionic electron effects, 385, 399-401 atom/ion beam sources, 247
FAPES (fumace atomization plasma emission boosted output, by
spectrometry), 376; see also FANES auxiliary electrodes, 237
Faraday dark space, 23 filaments, 239
Floating anode (restrictor), 157, 331 magnetic fields, 239
Flowing afterglow, 499 microwave radiation, 240
Fourier transforrn electron energy distribution, 224
mass spectrometers, 185 excitation mechanisms, 227
optical spectrometers, 164-166 fundamentals, 383
historical, 215-218
Geological materials, analysis by ionization processes, 223
atomic emission, 158, 270 laser media, 245
mass spectrometry, 206 nonconductive sampie preparation, 268
Getter agents, 271 optogalvanic spectroscopy, 244
Glow discharge plasma diagnostics, 248-251
abnormal, 19 plasma formation, 221
boosted, plasma structure, 222
dC,16O radial emission distribution, 394
microwave, 157 source geometries, 228-240
current-voltage relationships, 18-19, 124 Hollow cathode effect, 24, 127
effect of cathode identity, 140
FANES systems, 389 Integrated contact cuvette, 379
Grimm-type devices, 140 Internal standardization,
hollow cathodes, 43 atomic absorption, 100
planar devices, 42 atomic emission, 150, 153
FANES, 373 mass spectrometry, 188-191
fundamental processes, 17 nonconductive sampies, 281, 285-288
operating characteristics, 18-19 Ion
planar geometries: see Grimm discharges density, 48
sampling modes, 11-14 spatial profiles, 41, 314
spatial characteristics, 21-22, 124 Ion beam ratios, 191
Graphite fumace atomizers, 126 Ion energies
atOinic absorption spectrophotometry, 374 analyte ions, 302
atomic emission spectroscopy, 374 sputtering ions, 39
See also FANES Ion sources, 191-193
Grimm-type discharge lamps, 128 Ionization mechanisms, 38, 53, 190
mass spectrometry, 192 charge exchange, 58
microwave boosted, 157 diagnostics, 59-62
thin films analysis, 331 hollow cathode discharges, 60, 223
Lovett model, 60
High-purity metals, analysis by Penning, 53, 59
atomic emission, 155 radio-frequency discharges, 313
mass spectrometry, 195
Hollow cathode discharges Langmuir probes, 45,499
advantageous characteristics, 220 Laser ablation, 127
applications, 220-248 laser ablationlatomization
bulk metals, 242 auxiliary e1ectrode ablation, 429-434
512 INDEX
Laser ablation (Cont.) Mass spectrometry (glow discharge) (Cont.)

laser ablationlatomization (Cont.) mass analyzers, 178-187
catbode ablation, 434-441 comparison of, 178, -200
preferential volatilization, 432 Fourier transformlion cyclotron resonance,
temporal response, 431, 435 185
mechanisms, 426-428 magnetic sector, 178
Laser-enhanced ionization (LEI), 461-463 quadrupole ion trap, 184
Laser-induced ftuorescence (LIF) quadrupole mass filter, 181
"analytical" plasma diagnostics, 449-452 time-of-ftight, 179
applications quantitation
alloys,446 ion beam ratios, 191
solution residues, 446-448 relative ion yields, 189
CVD plasma diagnostics, 451, 500 relative sensitivity factors, 188, 199
energy level diagrams, 443, 492 radio-frequency sourees, 302-322
instrumentation, 444 reactive plasma sampling, 500
metastable atom sampling, 501 sampling considerations, 187
multiphoton processes, 494, 501 sputtered neutral MS, 304
optical sampling considerations, 488-491 Metal argides/formation, 436-438
radio-frequency discharges, 324 Metastable atoms, 38, 53
study of dimer species, 324, 451 energy levels, 58
See also Atomic ftuorescence production rates, 54
Laser media, hollow catbodes, 245 quenching mechanisms, 53
Line broadening spatial profiles, 55
collisional (pressure/Lorentzian), 74, 121 Microwave boosting,
Doppler, 74, 121 Grimm-type discharges, 157
hollow catbode discharge, 249 hollow cathode discharges, 240, 243
lifetime (natural), 73,121
linewidtbs (emission), 144 Negative glow, 21
Stark, 496, 503 Nonconducting sampie types
Line selection analyte migration, 275-278
atomic absorption, 87 applications
atomic ftuorescence, 88 catalytic materials, 285-288
geological materials, 288-290
Mass spectrometry (glow discharge) nuclear materials, 291
applications atomic absorption analysis, 100, 105
atmospheric sampling, 210 atomic emission analysis, 158-160
alloys, 197 effect of particle size, 273
compacted sampies, 205 effect of trapped gases/water vapor, 273
deptb profiling, 203 hollow catbode preparation, 268
high-purity metals, 195 host matrix identity/composition, 270-273
liquid chromatography, 211 pulsed discharges, 283
nuclear materials, 291 radio-frequency discharges, 291
semiconductors, 199 sampie preparation, 265-270
solution residues, 207
characteristics, 194 Optogalvanic (OGE) spectroscopy, 244, 453-
comparison witb spark sourees, 198 461,502
comparison witb secondary ion MS, 201 enhanced analyte ionization, 455
effect of trapped gases/water vapor, 274 high resolution spectroscopy, 459
fundamentals, 177 role of Penning ionization, 457
historieal, 176 Stark broadening in tbe cathode dark space, 460
ion kinetic energy, 302 study of metastable populations, 453
ion sources, 191-193 use in wavelengtb calibration, 459
INDEX 513
Penning ionization, 53 Reactive plasmas (Cont.)

Photon detectors laser sampling considerations, 488-491
charge-coupled devices, 138 molecular ion sampling, 486
image dissector tubes, 137 radical species, 486
photodiode arrays, 136 Redeposition
photomultipliers, 135 laser diagnostics, 438-441
Preferential sputtering, 34 relative extent, 39
Presputtering (preburn) procedures, 91, 143, Relative ion yields (RIYs), 188
150, 266, 278 Relative sensitivity factors (RSFs), 188, 198,
Pulsed discharges 288
atomic absorption, 96-98 Resonance-enhanced multiphoton ionization
mass spectrometry, 193 (REMPI), 495, 502
nonconductive sampies, 283 Resonance ionization mass spectrometry, 463-
redeposition diagnostics, 438 472
alleviation of isobaric interferences, 469-
Radio frequency (rf) glow discharges, 25, 291 472
applications energy level diagrams, 464
glasses, 319 glow discharge atomization, 465
MACOR,323 sampling of post-source atom beam, 468
metal oxides, 318 studies of Fe species, 467
thin films, 321 Resonance monochromators, 77
dc bias/self-bias, 295, 297-300
diagnostic techniques, 297 Secondary electron yield (-y), 21, 28, 38, 351
direct insertion probe, 308 Self-reversal, 121
effect of operating frequency, 301 Semiconductor materials, analysis by
electrode geometry, 297-300 mass spectrometry, 199
external sampie mount (AES/MS), 309-312 Sigmund theory, 27
historical, 294 Solution residues, analysis by
hollow cathode discharges, 305 atomic absorption, 85, 98, 105, 157
ionization mechanisms, 313 atomic emission, 157, 242-244
laser-induced fluorescence (LIF), 324 FANES, 413-417
mass spectrometry, 302-322 mass spectrometry, 207
metal oxides, 318 Spark sources, 124
optical spectroscopies, 323-325 Spectral profiles, 72-78
secondary excitation sources, 304 broadening, 73-75
sheath capacitance, 299-301 Doppler, 74, 121
spatial distribution of ions, 314 hyperfine structure, 72
theory of operation, 295 isotope shifts, 72
Rare earth elements, analysis by linewidths (emission), 144
FANES, 414 Stark, 496, 503
mass spectrometry, 207 Spectrometers
Reactive plasmas depth-profiling requirements, 332-336
absorption techniques (laser), 496 dispersive, 129-133
actinometry, 498 mass, 177-187
atom densities, 485 multiplex, 133-135
charged particle sampling, 499 optical, 129-135
chemiluminescence, 499 Sputter deposition, 247
electric field strength, 503 Sputter rate
emission spectroscopy, 497 compensation for differences, 100
excited-state molecules, 487 current dependence, 32, 350
flowing afterglow, 499 depth-profiling applications, 346-354
laser-induced fluorescence, 492-495 effect of reduced power, 34
514 INDEX
Sputter rate (Cont.) Thin films analysis

effect of trapped gases/water vapor, 273 advantages, 369
effect of water vapor, 54, 88 applications, 364
Grimm-type devices, 144, 346 carbide coatings, 356
hollow cathodes, 32 corrosion products, 336
mass spectrometry, 203 hot nitriding tool steels, 364
modeling, 351-355 ion implantation, 359, 367
planar cathodes, 33 passive layers, 363
power dependence, 89 zinc-plated steel, 360
voltage dependence, 350 concentration calibration
Sputter yield, 28, 189, 350 empirical approaches, 356
alloys, 36, 354 theoretical approaches, 358
effect of crater shapes, 340-342
angle of incidence, 29 historical, 329
incident ion energy, 29 instrumentation
sputtering ion mass, 28, 189 optical spectrometerrequirements, 332-336
surface binding energy, 28, 31, 189 source design, 331
target material, 30 !imitations, 370
target temperature, 31, 354 line selection, 333
hollow cathode discharges, 226 mass spectrometry, 203
mode!ing, 351-355 plasma parameters, 338
Sigmund theory, 27 sampie geometries, 342
Yamamura model, 353 sputtering rate
Sputtered atoms crater-depth measurement, 346
spatial profile, 39, 81-85 effect of sampie composition, 346
Sputtered neutral mass spectrometry (SNMS), effect of sampie temperature, 348
304 experimental variables, 349-351
Standardization modeling, 351-355
atomic absorption, 99-104
mass spectrometry, 188-191 VG 9000, 179, 196
nonconductive sampies, 281-283
Stark broadening, 496, 503 Yamamura model, 353

Untitled

Uploaded by

Copyright:

Available Formats

Untitled

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Untitled

Uploaded by

Copyright:

Available Formats

Glow Discharge

ADVANCES IN COAL SPECTROSCOPY

ANALYTICAL ATOMIC SPECTROSCOPY

CHEMICAL DERIVATIZATION IN ANALYTICAL CHEMISTRY

COMPUTER-ENHANCED ANALYTICAL SPECTROSCOPY

GLOW DISCHARGE SPECTROSCOPIES

ION-SELECTIVE ELECTRODES IN ANALYTICAL CHEMISTRY

LIQUID CHROMATOGRAPHY/MASS SPECTROMETRY

MASS SPECTROMETRY: CLiNICAL AND BIOMEDICAL APPLICATIONS

PRINCIPLES OF CHEMICAL SENSORS

TRANSFORM TECHNIQUES IN CHEMISTRY

Springer Science+Business Media, LLC

Glow discharge spectroscoples I edlted by R. Kenneth Marcus.

1. Gas dlscharges. 2. Spectrum analysis. 3. Solid state physlcs.

ISBN 978-1-4899-2396-7 ISBN 978-1-4899-2394-3 (eBook)

© 1993 Springer Science+Business Media New York

All rights reserved

No part of this book may be reproduced, stored in a retrieval system, or transmitted

J. A. C. Broekaert • Department ofChemistry, University ofDortmund,

Philip G. Rigby • School ofBiological and Chemical Sciences, University

One of the greatest challenges remaining in the area of analytical atomic

area of atomic spectroscopy. Given the wide variety of source geometries,

1.1. Rationale ........................................ 1

2. Fundamental Plasma Processes

Duencheng Fang and R. Kenneth Marcus

2.1. Gaseous Discharges ................................ 17

3. Atomic Absorption and Fluorescence

3.1. Introduction ...................................... 67

4. Atomic Emission Spectrometry

4.1. Optical Emission Spectrometry ...................... 113

4.1.3.Spectrometers .............................. 129

5. Glow Discharge Mass Spectrometry

5.1. Introduction ...................................... 175

5.2.2.4. Selective m/z Ejection (Quadrupole Ion

6. Hollow Cathode Discharges

Sergio Caroli and Oreste Senofonte

6.1. Introduction ...................................... 215

6.4.3. Analysis of Liquids . . . . . . . . . . . . . . . . . . . . . . . . .. 242

7. Analysis of Nonconducting Sampie Types

Michael R. Winchester, Douglas C. Duckworth,

7.1. Introduction ...................................... 263

8. Thin Film Analysis

8.1. Historical Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 329

8.8.2.2. Theoretical Approach. . . . . . . . . . . . . . .. 358

9. Discharges within Graphite Furnace Atomizers

James M. Harnly, David L. Styris, and Philip G. Rigby

9.1. Introduction ...................................... 373

10.1. Introduction ...................................... 423

11. Laser-Based Diagnostics of Reactive Plasmas

Bryan L. Preppemau and Terry A. Miller

11.1. Introduction ...................................... 483

11.2. An Overview of Reactive Plasmas ... . . . . . . . . . . . . . . . .. 485

Index .................................................... 509

Driven by advances in materials research, government regulations, and inter-

R. Kenneth Marcus • Department of Chemistry, Howard L. Hunter Chemical Labora-

1.1.1. Scope of Elemental Analysis

Table 1-1. Sampie Types Requiring Elemental Analysis

1.1.2. Aqueous SampIe Elemental Analysis

of absolute analyte mass. Inductively coupled plasmas (lCP) are generated

1.1.3. Direct Solids Elemental Analysis