Scanning Electron Microscopy and X-Ray Microanalysis
Scanning Electron Microscopy and X-Ray Microanalysis
Scanning Electron Microscopy and X-Ray Microanalysis
Goldstein
Dale E. Newbury
Joseph R. Michael
Nicholas W.M. Ritchie
John Henry J. Scott
David C. Joy
Scanning Electron
Microscopy and
X-Ray Microanalysis
Fourth Edition
Scanning Electron Microscopy and X-Ray Microanalysis
Joseph I. Goldstein
Dale E. Newbury
Joseph R. Michael
Nicholas W.M. Ritchie
John Henry J. Scott
David C. Joy
Scanning Electron
Microscopy and
X-Ray Microanalysis
Fourth Edition
Joseph I. Goldstein Dale E. Newbury
University of Massachusetts National Institute of Standards and Technology
Amherst, MA, USA Gaithersburg, MD, USA
Joseph R. Michael Nicholas W.M. Ritchie
Sandia National Laboratories National Institute of Standards and Technology
Albuquerque, NM, USA Gaithersburg, MD, USA
John Henry J. Scott David C. Joy
National Institute of Standards and Technology University of Tennessee
Gaithersburg, MD, USA Knoxville, TN, USA
Preface
This is not your father’s, your mother’s, or your is that a reader seeking specific information can
grandparent’s Scanning Electron Microscopy and select a topic from the list and obtain a good
X-Ray Microanalysis (SEMXM). But that is not to understanding of the topic from that module
say that there is no continuity or to deny a family alone. While each topic is supported by informa-
resemblance. SEMXM4 is the fourth in the series of tion in other modules, we acknowledge the like-
textbooks with this title, and continues a tradition lihood that not all users of SEMXM4 will “read
that extends back to the “zero-th edition” in 1975 it all.” This approach inevitably leads to a degree
published under the title, “Practical Scanning of overlap and repetition since similar informa-
Electron Microscopy” (Plenum Press, New York). tion may appear in two or more places, and this is
However, the latest edition differs sharply from its entirely intentional.
predecessors, which attempted an encyclopedic
approach to the subject by providing extensive In recognition of these fundamental changes, the
details on how the SEM and its associated devices authors have chosen to modify SEMXM4 exten-
actually work, for example, electron sources, lenses, sively to provide a guide on the actual use of the
electron detectors, X-ray spectrometers, and so on. instrument without overwhelming the reader with
the burden of details on the physics of the opera-
In constructing this new edition, the authors have tion of the instrument and its various attachments.
chosen a different approach. Modern SEMs and the Our guiding principle is that the microscopist-
associated X-ray spectrometry and crystallography microanalyst must understand which parameters
measurement functions operate under such exten- can be adjusted and what is an effective strategy to
sive computer control and automation that it is select those parameters to solve a particular prob-
actually difficult for the microscopist-microanalyst lem. The modern SEM is an extraordinarily flex-
to interact with the instrument except within care- ible tool, capable of operating over a wide range
fully prescribed boundaries. Much of the flexibility of electron optical parameters and producing
of parameter selection that early instruments pro- images from electron detectors with different sig-
vided has now been lost, as instrumental operation nal characteristics. Those users who restrict them-
functions have been folded into software control. selves to a single set of operating parameters may
Thus, electron sources are merely turned “on,” with be able to solve certain problems, but they may
the computer control optimizing the operation, or never know what they are missing by not explor-
for the thermally assisted field emission gun, the ing the range of parameter space available to them.
electron source may be permanently “on.” The user SEMXM4 seeks to provide sufficient understand-
can certainly adjust the lenses to focus the image, ing of the technique for a user to become a com-
but this focusing action often involves complex petent and efficient problem solver. That is not to
interactions of two or more lenses, which formerly say that there are only a few things to learn. To
would have required individual adjustment. More- help the reader to approach the considerable body
over, the nature of the SEM field has fundamentally of knowledge needed to operate at a high degree
changed. What was once a very specialized instru- of competency, a new feature of SEMXM-4 is the
ment system that required a high level of training summary checklist provided for each of the major
and knowledge on the part of the user has become areas of operation: SEM imaging, elemental X-ray
much more of a routine tool. The SEM is now sim- microanalysis, and backscatter-diffraction crystal-
ply one of a considerable suite of instruments that lography.
can be employed to solve problems in the physical
and biological sciences, in engineering, in technol- Readers familiar with earlier editions of SEMXM
ogy, in manufacturing and quality control, in fail- will notice the absence of the extensive material
ure analysis, in forensic science, and other fields. previously provided on specimen preparation.
Proper specimen preparation is a critical step in
The authors also recognize the profound changes solving most problems, but with the vast range of
that have occurred in the manner in which peo- applications to materials of diverse character, the
ple obtain information. The units of SEMXM4, topic of specimen preparation itself has become
whether referred to as chapters or modules, are the subject of entire books, often devoted to just
meant to be relatively self-contained. Our hope one specialized area.
VI Preface
Throughout their history, the authors of the Finally, the author team sadly notes the passing in
SEMXM textbooks have been closely associated 2015 of Professor Joseph I. Goldstein (University
as lecturers with the Lehigh University Summer of Massachusetts, Amherst) who was the “found-
Microscopy School. The opportunity to teach and ing father” of the Lehigh University Summer
interact with each year’s students has provided a Microscopy School in 1970, and who organized
very useful experience in understanding the com- and contributed so extensively to the microscopy
munity of users of the technique and its evolution courses and to the SEMXM textbooks throughout
over time. We hope that these interactions have the ensuing 45 years. Joe provided the stimulus to
improved our written presentation of the subject the production of SEMXM4 with his indefatigable
as a benefit to newcomers as well as established spirit, and his technical contributions are embed-
practitioners. ded in the X-ray microanalysis sections.
Dale E. Newbury
Nicholas W.M. Ritchie
John Henry J. Scott
Gaithersburg, MD, USA
Joseph R. Michael
Albuquerque, NM, USA
David C. Joy
Knoxville, TN, USA
VII
a b
100 nm
EHT - 15.00 kV Signal A = InlLens Mag - 94.28 K X Width - 1.213 mm Date: 19 Oct 2015
YK I Probe - 135 pA ESB Grid = 800 V Image Pixel Size - 1.184 nm
WD - 1.7 mm Signal B = InlLens
c
HV mag HFW WD 500 nm
800.00 V 100 000 x 1.49 µm 990.7 µm Helios
10nm JEOL
x500,000 0.30kV UED GBSH WD 2.0mm
.. Fig. 1 a High resolution SEM image taken at high (image courtesy of Trevan Landin, FEI); Bar = 500 nm. c
beam energy (E0 = 15 keV) of a finFET transistor (16-nm Mesoporous silica nanosphere showing 5-nm-diameter
technology) using an in-lens secondary electron detector. pores imaged with a landing energy of 0.5 keV (specimen
This cross section was prepared by inverted Ga FIB milling courtesy of T. Yokoi, Tokyo Institute of Technology; images
from backside (Zeiss Auriga Cross beam; image courtesy courtesy of A. Muto, Hitachi High Technologies); Upper image
of John Notte, Carl Zeiss); Bar = 100 nm. b High resolu- Bar = 200 nm, Lower image Bar = 100 nm. d Si nanoparticle
tion SEM image taken at low beam energy (E0 = 0.8 keV) of imaged with a landing energy of 0.3 keV; Bar = 10 nm
zeolite (uncoated) using a through-the-lens SE detector (image courtesy V. Robertson, JEOL)
threshold. This threshold can only be lowered 3. The beam divergence angle can be minimized
by increasing beam current, pixel dwell time, to increase the depth-of-field (e.g., . Fig. 3).
and/or detector efficiency. Selecting higher With optimized selection of aperture size and
beam current means a larger beam size, caus- specimen-to-objective lens distance (work-
ing resolution to deteriorate. Thus, there is a ing distance), it is generally possible to achieve
dynamic contest between resolution and visibil- small beam convergence angles and therefore
ity leading to inevitable limitations on feature effective focus along the beam axis that is at
size and feature visibility that can be achieved. least equal to the horizontal width of the image.
IX
Scanning Electron Microscopy and Associated Techniques: Overview
0.5 nA 20 nA
20 mm 20 mm
BSE MAG: 1000 x HV: 20.0 kV WD: 11.0 mm BSE MAG: 1000 x HV: 20.0 kV WD: 11.0 mm
.. Fig. 2 Effect of increasing beam current (at constant pixel dwell time) to improve visibility of low contrast features.
Al-Si eutectic alloy; E0 = 20 keV; semiconductor BSE detector (sum mode): (left) 0.5 nA; (right) 20 nA; Bar = 20 µm
1
2
10 mm
BSE MAG: 750 x HV: 20.0 kV WD: 11.0mm
20 mm 20 mm
SE MAG: 500 X HV: 20.0 kV WD: 11.0 mm BSE MAG: 500 X HV: 20.0 kV WD: 11.0 mm
.. Fig. 5 Topographic contrast as viewed with different detectors: Everhart–Thornley (positive bias) and semiconductor
BSE (sum mode); silver crystals; E0 = 20 keV; Bar = 20 µm
combinations of elements, for example, Ti and standard spectra can be effectively used if a quality
Ba; S, Mo, and Pb; and many others, especially measurement program is implemented to ensure
when the peaks of major constituents interfere the constancy of measurement conditions, includ-
with the peaks of minor or trace constituents. ing spectrometer performance parameters. With
Operator knowledge of the physical rules govern- such a standards-based measurement protocol and
ing X-ray generation and detection is needed to ZAFc matrix corrections, the accuracy of the anal-
perform a careful review of software-generated ysis can be expressed as a relative deviation from
peak identifications, and this careful review must expected value (RDEV):
always be performed to achieve a robust mea-
surement result. RDEV ( % ) = [ ( Measured − True ) / True ] × 100% (2)
After a successful qualitative analysis has been
performed, quantitative analysis can proceed.
Based on extensive testing of homogeneous
The characteristic intensity for each peak is auto-
binary and multiple component compositions,
matically determined by peak fitting procedures,
the distribution of RDEV values for major con-
such as the multiple linear least squares method.
stituents is such that a range of ±5 % relative cap-
The intensity measured for each element is pro-
tures 95 % of all analyses. The use of stable, high
portional to the concentration of that element,
integrated count spectra (>1 million total counts
but that intensity is also modified by all other ele-
from threshold to E0) now possible with the sili-
ments present in the interaction volume through
con drift detector EDS (SDD-EDS), enables this
their influence on the electron scattering and
level of accuracy to be achieved for major and
retardation (“atomic number” matrix effect, Z),
minor constituents even when severe peak inter-
X-ray absorption within the specimen (“absorp-
ference occurs and there is also a large concen-
tion” matrix effect, A), and X-ray generation
tration ratio, for example, a major constituent
induced by absorption of X-rays (“secondary flu-
interfering with a minor constituent. Trace con-
orescence” matrix effects, F, induced by charac-
stituents that do not suffer severe interference
teristic X-rays and c, induced by continuum
can be measured to limits of detection as low as
X-rays). The complex physics of these “ZAFc”
C = 0.0002 (200 parts per million) with spectra
matrix corrections has been rendered into algo-
containing >10 million counts. For interference
rithms by a combined theoretical and empirical
situations, much higher count spectra (>100
approach. The basis of quantitative electron-
million counts) are required.
excited X-ray microanalysis is the “k-ratio proto-
An alternative “standardless analysis” protocol
col”: measurement under identical conditions
uses libraries of standard spectra (“remote stan-
(beam energy, known electron dose, and spec-
dards”) measured on a different SEM platform
trometer performance) of the characteristic
with a similar EDS spectrometer, ideally over a
intensities for all elements recognized in the
wide range of beam energy and detector parame-
unknown spectrum against a suite of standards
ters (resolution). These library spectra are then
containing those same elements, producing a set
adjusted to the local measurement conditions
of k-ratios, where
through comparison of one or more key spectra
(e.g., locally measured spectra of particular ele-
k = I Unknown / I Standard (1)
ments such as Si and Ni). Interpolation/extrapola-
tion is used to supply estimated spectral intensities
for each element in the unknown. Standards are for elements not present in or at a beam energy not
materials of known composition that are tested to represented in the library elemental suite. Testing
be homogeneous at the microscopic scale, and of the standardless analysis method has shown
preferably homogeneous at the nanoscale. that an RDEV range of ±25 % relative is needed to
Standards can be as simple as pure elements—e.g., capture 95 % of all analyses.
C, Al, Si, Ti, Cr, Fe, Ni, Cu, Ag, Au, etc.—but for High throughput (>100 kHz) EDS enables col-
those elements that are not stable in a vacuum lection of X-ray intensity maps with gray scale rep-
(e.g., gaseous elements such as O) or which resentation of different concentration levels (e.g.,
degrade during electron bombardment (e.g., S, P, . Fig. 7a). Compositional mapping by spectrum
and Ga), stable stoichiometric compounds can be imaging (SI) collects a full EDS spectrum at each
used instead, e.g., MgO for O; FeS2 for S; and GaP pixel of an x-y array, and after applying the quanti-
for Ga and P. The most accurate analysis is per- tative analysis procedure at each pixel, images are
formed with standards measured on the same created for each element where the gray (or color)
instrument as the unknown(s), ideally in the same level is assigned based on the measured concentra-
measurement campaign, although archived tion (e.g., . Fig. 7b).
XII Scanning Electron Microscopy and Associated Techniques: Overview
Al Fe
20µm
Ni Al Fe Ni
BSE Al
20 mm
BSE MAG: 1000 x HV: 20.0 kV WD: 11.0 mm
Ni Fe
20µm
0.001 0.01 0.1 1.0
b
0.1 1.0 10 100 wt%
.. Fig. 7 a EDS X-ray intensity maps for Al, Fe, and Ni and color overlay; Raney nickel alloy; E0 = 20 keV. b SEM/BSE (sum)
image and compositional maps corresponding to a
XIII
Scanning Electron Microscopy and Associated Techniques: Overview
a b
40 mm
BSE MAG: 400 x HV: 20.0 kV WD: 11.0 mm
.. Fig. 8 a Electron channeling contrast revealing grain boundaries in Ti-alloy (nominal composition: Ti-15Mo-3Nb-3Al-
0.2Si); E0 = 20 keV. b Electron backscatter diffraction (EBSD) pattern from hematite at E0 = 40 keV
XIV Scanning Electron Microscopy and Associated Techniques: Overview
.. Fig. 9 Directionally-solidified Al-Cu eutectic alloy after ion beam milling in a dual-beam instrument, as imaged by the
SEM column (left image); same region imaged in the HIM (right image)
Contents
2 Backscattered Electrons�������������������������������������������������������������������������������������������������������������������������������� 15
2.1 Origin������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������ 16
2.1.1 The Numerical Measure of Backscattered Electrons����������������������������������������������������������������������������������������������������� 16
2.2 Critical Properties of Backscattered Electrons������������������������������������������������������������������������������������������������������������ 16
2.2.1 BSE Response to Specimen Composition (η vs. Atomic Number, Z)����������������������������������������������������������������������� 16
2.2.2 BSE Response to Specimen Inclination (η vs. Surface Tilt, θ)������������������������������������������������������������������������������������� 20
2.2.3 Angular Distribution of Backscattering���������������������������������������������������������������������������������������������������������������������������� 22
2.2.4 Spatial Distribution of Backscattering������������������������������������������������������������������������������������������������������������������������������� 23
2.2.5 Energy Distribution of Backscattered Electrons������������������������������������������������������������������������������������������������������������ 27
2.3 Summary������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������ 27
References���������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 28
3 Secondary Electrons��������������������������������������������������������������������������������������������������������������������������������������� 29
3.1 Origin������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������ 30
3.2 Energy Distribution��������������������������������������������������������������������������������������������������������������������������������������������������������������� 30
3.3 Escape Depth of Secondary Electrons���������������������������������������������������������������������������������������������������������������������������� 30
3.4 Secondary Electron Yield Versus Atomic Number����������������������������������������������������������������������������������������������������� 30
3.5 Secondary Electron Yield Versus Specimen Tilt���������������������������������������������������������������������������������������������������������� 34
3.6 Angular Distribution of Secondary Electrons�������������������������������������������������������������������������������������������������������������� 34
3.7 Secondary Electron Yield Versus Beam Energy����������������������������������������������������������������������������������������������������������� 35
3.8 Spatial Characteristics of Secondary Electrons����������������������������������������������������������������������������������������������������������� 35
References���������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 37
4 X-Rays�������������������������������������������������������������������������������������������������������������������������������������������������������������������� 39
4.1 Overview������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������ 40
4.2 Characteristic X-Rays������������������������������������������������������������������������������������������������������������������������������������������������������������� 40
4.2.1 Origin������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 40
4.2.2 Fluorescence Yield������������������������������������������������������������������������������������������������������������������������������������������������������������������� 41
4.2.3 X-Ray Families��������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 42
4.2.4 X-Ray Nomenclature��������������������������������������������������������������������������������������������������������������������������������������������������������������� 43
4.2.5 X-Ray Weights of Lines����������������������������������������������������������������������������������������������������������������������������������������������������������� 44
4.2.6 Characteristic X-Ray Intensity���������������������������������������������������������������������������������������������������������������������������������������������� 44
4.3 X-Ray Continuum (bremsstrahlung)��������������������������������������������������������������������������������������������������������������������������������� 47
4.3.1 X-Ray Continuum Intensity��������������������������������������������������������������������������������������������������������������������������������������������������� 49
4.3.2 The Electron-Excited X-Ray Spectrum, As-Generated�������������������������������������������������������������������������������������������������� 49
4.3.3 Range of X-ray Production���������������������������������������������������������������������������������������������������������������������������������������������������� 50
4.3.4 Monte Carlo Simulation of X-Ray Generation����������������������������������������������������������������������������������������������������������������� 51
4.3.5 X-ray Depth Distribution Function, ϕ(ρz)������������������������������������������������������������������������������������������������������������������������� 53
XVI Contents
6 Image Formation����������������������������������������������������������������������������������������������������������������������������������������������� 93
6.1 Image Construction by Scanning Action����������������������������������������������������������������������������������������������������������������������� 94
6.2 Magnification��������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 95
6.2.1 Magnification, Image Dimensions, and Scale Bars�������������������������������������������������������������������������������������������������������� 95
6.3 Making Dimensional Measurements With the SEM:
How Big Is That Feature?������������������������������������������������������������������������������������������������������������������������������������������������������ 95
6.3.1 Calibrating the Image������������������������������������������������������������������������������������������������������������������������������������������������������������� 95
6.4 Image Defects�������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 98
6.4.1 Projection Distortion (Foreshortening)���������������������������������������������������������������������������������������������������������������������������� 98
6.4.2 Image Defocusing (Blurring)���������������������������������������������������������������������������������������������������������������������������������������������� 100
6.5 Making Measurements on Surfaces With Arbitrary Topography:
Stereomicroscopy����������������������������������������������������������������������������������������������������������������������������������������������������������������� 102
6.5.1 Qualitative Stereomicroscopy������������������������������������������������������������������������������������������������������������������������������������������� 103
6.5.2 Quantitative Stereomicroscopy���������������������������������������������������������������������������������������������������������������������������������������� 107
References�������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 110
28 Cathodoluminescence��������������������������������������������������������������������������������������������������������������������������������� 481
28.1 Origin���������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 482
28.2 Measuring Cathodoluminescence��������������������������������������������������������������������������������������������������������������������������������� 483
28.2.1 Collection of CL���������������������������������������������������������������������������������������������������������������������������������������������������������������������� 483
28.2.2 Detection of CL����������������������������������������������������������������������������������������������������������������������������������������������������������������������� 483
28.3 Applications of CL���������������������������������������������������������������������������������������������������������������������������������������������������������������� 485
28.3.1 Geology������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 485
28.3.2 Materials Science������������������������������������������������������������������������������������������������������������������������������������������������������������������� 485
28.3.3 Organic Compounds������������������������������������������������������������������������������������������������������������������������������������������������������������� 489
References�������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 489
Index���������������������������������������������������������������������������������������������������������������������������������������������������������������������������������������� 547
1 1
Electron Beam—Specimen
Interactions: Interaction
Volume
1.1 What Happens When the Beam Electrons Encounter
Specimen Atoms? – 2
References – 14
Au
20
Cu
10
Al
5
C
0
5 10 15 20 25 30
10000
C
Bethe range (nm)
8000
Al
6000
4000 Cu, Ag
2000
Au
0
5 10 15 20 25 30
Incident beam energy (keV)
kNote the change of scale cross section and the Bethe range as its altitude, the volume
The Bethe range for Au with an incident beam energy of of a cylinder 1 nm in diameter and 1200 nm deep would be
20 keV is approximately 1200 nm, a linear change in scale of approximately 940 nm3, and the number of gold atoms it
a factor of 1200 over an incident beam diameter of 1 nm. If contained would be approximately 7.5 × 104, which can be
the beam–specimen interactions were restricted to a cylin- compared to the incident beam footprint surface atom count
drical column with the circular beam entrance footprint as its of approximately 12.
4 Chapter 1 · Electron Beam—Specimen Interactions: Interaction Volume
( )
Qelastic ( >φ 0 ) =1.62 ×10−20 Z 2 / E 2 cot 2 (φ0 / 2 ) P1
(
[events > φ0 / electron atom / cm 2
) (1.2)
P2
1.3.1 ow Frequently Does Elastic
H
Scattering Occur? P3
C
10
Al
0.1
0.01
5 10 15 20 25 30
Beam energy (keV)
Given a specific set of these parameters, the Monte Carlo elec- dimension y projected onto the x-z plane. (An example of the
1 tron trajectory simulation utilizes geometrical expressions to true three-dimensional trajectories, simulated with the Joy
calculate the successive series of locations P1, P2, P3, etc., suc- Monte Carlo, is shown in . Fig. 1.6, in which a small number
cessively determining the coordinate locations (x, y, z) that the of trajectories (to minimize overlap) have been rendered as
energetic electron follows within the solid. At each location P, an anaglyph stereo representation with the convention left
the newly depreciated energy of the electron is known, and after eye = red filter. Inspection of this simulation shows the y
the next elastic scattering angle is calculated, the new velocity motion of the electrons in and out of the x-z plane.) The sto-
vector components vx, vy, vz are determined to transport the chastic nature of the interaction imposed by the nature of
electron to the next location. A trajectory ends when either the elastic scattering is readily apparent in the great variation
electron energy falls below a threshold of interest (e.g., 1 keV), among the individual trajectories seen in . Fig. 1.5a, b. It
or else the path takes it outside the geometric bounds of the quickly becomes clear that individual beam electrons follow a
specimen, which is determined by comparing the current loca- huge range of paths and simulating a small number of trajec-
tion (x, y, z) with the specimen boundaries. The capability of tories does not provide an adequate view of the electron beam
simulating electron beam interactions in specimens with com- specimen interaction.
plex geometrical shapes is one of the major strengths of the
Monte Carlo electron trajectory simulation method.
Monte Carlo electron trajectory simulation can pro- 1.4.2 onte Carlo Simulation To Visualize
M
vide visual depictions as well as numerical results of the the Electron Interaction Volume
beam–specimen interaction, creating a powerful instructional
tool for studying this complex phenomenon. Several power- To capture a reasonable picture representation of the electron
ful Monte Carlo simulations appropriate for SEM and X-ray interaction volume, which is the region of the specimen in
microanalysis applications are available as free resources: which the beam electrons travel and deposit energy, it is nec-
CASINO [7 http://www.gel.usherbrooke.ca/casino/What.html]
essary to calculate many more trajectories. . Figure 1.5c
Joy Monte Carlo [7 http://web.utk.edu/~srcutk/htm/
shows the simulation for copper, E0 = 20 keV at 0° tilt
simulati.htm] extended to 500 trajectories, which reveals the full extent of
NIST DTSA-II [7 http://www.cstl.nist.gov/div837/837.02/
the electron interaction volume. Beyond a few hundred tra-
epq/dtsa2/index.html] jectories, superimposing the three-dimensional trajectories
to create a two-dimensional representation reaches dimin-
While the static images of Monte Carlo simulations pre- ishing returns due to overlap of the plotted lines. While sim-
sented below are useful instructional aids, readers are ulating 500 trajectories provides a reasonable qualitative
encouraged to perform their own simulations to become view of the electron interaction volume, Monte Carlo calcu-
familiar with this powerful tool, which in more elaborate lations of numerical properties of the interaction volume and
implementations is an important aid in understanding criti- related processes, such as electron backscattering (discussed
cal aspects of SEM imaging. in the backscattered electron module), are subject to statisti-
cally predictable variations because of the use of random
numbers to select the elastic scattering parameters. Variance
1.4.1 hat Do Individual Monte Carlo
W in repeated simulations of the same starting conditions is
Trajectories Look Like? related to the number of trajectories and can be described
with the properties of the Gaussian (normal) distribution.
Perform a Monte Carlo simulation (CASINO simulation) for Thus the precision, p, of the calculation of a parameter of the
copper with a beam energy of 20 keV and a tilt of 0° (beam interaction is related to the total number of simulated trajec-
perpendicular to the surface) for a small number of trajecto- tories, n, and the fraction, f, of those trajectories that produce
ries, for example, 25. . Figure 1.5a, b show two simulations of the effect of interest (e.g., backscattering):
25 trajectories each. The trajectories are actually determined
−1/ 2
p = ( f n) / ( f n) = ( f n)
1/ 2
in three dimensions (x-y-z, where x-y defines the surface (1.4)
plane and z is perpendicular to the surface) but for plotting
are rendered in two dimensions (x-z), with the third
1.4 · Simulating the Effects of Elastic Scattering: Monte Carlo Calculations
7 1
.. Fig. 1.5 a Copper, E0 = 20 keV; 0 tilt; 25 trajec- a
tories (CASINO Monte Carlo simultion). b Copper, Cu E0 = 20 keV 0.0 nm
E0 = 20 keV; 0 tilt; another 25 trajectories. c Copper,
E0 = 20 keV; 0 tilt; 200 trajectories 200 nm
200.0 nm
400.0 nm
600.0 nm
800.0 nm
200 nm
180.0 nm
360.0 nm
540.0 nm
720.0 nm
200 nm
233.5 nm
466.9 nm
700.4 nm
933.8 nm
25%
412.6 nm
10%
5% 618.9 nm
5.0%
10.0%
825.2 nm
25.0%
50.0%
75.0%
90.0%
b
10%
90%
25%
75%
50%
200 nm
10 Chapter 1 · Electron Beam—Specimen Interactions: Interaction Volume
0.0 nm
1 C
Cu
892.6 nm
1785.3 nm
Ag
2677.9 nm
3570.6 nm
755.3 nm
E0= 20 keV
0° tilt
1510.6 nm
1 µm
2265.9 nm
3021.3 nm
.. Fig. 1.8 Monte Carlo simulations for an incident beam energy of 20 keV and 0° tilt for C, Si, Cu, Ag, and Au, all shown at the same scale
(CASINO Monte Carlo simulation)
1.4 · Simulating the Effects of Elastic Scattering: Monte Carlo Calculations
11 1
.. Fig. 1.9 Monte Carlo simula- 0.0 nm
tions for Cu, 0° tilt, incident beam 30 keV
energies 5, 10, 20, and 30 keV
(CASINO Monte Carlo simulation)
432.6 nm
865.2 nm
1297.8 nm
1730.4 nm
10 keV
5 keV
0.0 nm
20 keV
233.5 nm
466.9 nm
700.4 nm Cu
0° tilt
933.8 nm
500 nm
-680.0 nm -340.0 nm 0.0 nm 340.0 nm 680.0 nm
12 Chapter 1 · Electron Beam—Specimen Interactions: Interaction Volume
1 0.0 nm 0.0 nm
233.5 nm
254.1 nm
466.9 nm
508.1 nm
700.4 nm
762.2 nm
933.8 nm
0° tilt 1016.2 nm
60° tilt 500 nm
-680.0 nm -340.0 nm 0.0 nm 340.0 nm 680.0 nm
Cu
0.0 nm 0.0 nm
219.7 nm
240.0 nm
439.5 nm
480.0 nm
659.2 nm
720.0 nm
878.9 nm
45° tilt 960.0 nm
.. Fig. 1.10 Monte Carlo simulations for Cu, 20 keV, with various tilt angles (CASINO Monte Carlo simulation)
Carbon Aluminum
1000 nm 1000 nm
Copper Gold
1000 nm
1000 nm
.. Fig. 1.11 Kanaya–Okayama range (gold arrow) superimposed on the interaction volume for C, Al, Cu, and Au at E0 = 20 keV and 0° tilt (Joy
Monte Carlo simulation)
is shown superimposed on the Monte Carlo simulation of should only be considered as a “gray” number useful for esti-
the interaction volume in . Fig. 1.11 and is plotted graphi- mation purposes. Nevertheless, the Kanaya–Okayama range
cally in . Fig. 1.12. It is, of course, simplistic to use a single is useful as a means to provide scaling to describe the spatial
numerical value of the range to describe such a complex phe- distributions of the signals produced within the interaction
nomenon as the electron interaction volume with its varying volume: secondary electrons, backscattered electrons, and
contours of energy deposition, and thus the range equation X-rays.
14 Chapter 1 · Electron Beam—Specimen Interactions: Interaction Volume
10000
8000 C
Range (nm)
Al
6000
4000
Cu
2000 Ag
Au
0
5 10 15 20 25 30
References
Bethe H (1930) Theory of the transmission of corpuscular radiation
through matter. Ann Phys Leipzig 5:325
CASINO 7 http://www.gel.usherbrooke.ca/casino/What.html
Joy Monte Carlo 7 http://web.utk.edu/~srcutk/htm/simulati.htm
Kanaya K, Okayama S (1972) Penetration and energy-loss theory of elec-
trons in solid targets. J Phys D Appl Phys 5:43
NIST DTSA-II 7 http://www.cstl.nist.gov/div837/837.02/epq/dtsa2/index.
html
15 2
Backscattered Electrons
2.1 Origin – 16
2.1.1 The Numerical Measure of Backscattered Electrons – 16
2.3 Summary – 27
References – 28
0.0 nm
Cu
E0 = 20 keV
0° Tilt
200.0 nm
Absorbed Electrons
(lost all energy and are 400.0 nm
absorbed within specimen)
600.0 nm
500 nm
800.0 nm
a b
0.0 nm 0.0 nm
624.9 nm 755.3 nm
1249.7 nm 1510.6 nm
1874.6 nm 2265.9 nm
C Si
E0 = 20 keV E0 = 20 keV
2499.4 nm 3021.3 nm
1 µm
1 µm
-1820.0 nm -910.0 nm -0.0 nm 910.0 nm 1820.0 nm -2200.0 nm -1100.0 nm -0.0 nm 1100.0 nm 2200.0 nm
c d
0.0 nm 0.0 nm
233.5 nm 137.3 nm
466.9 nm 274.7 nm
700.4 nm 412.0 nm
Au
Cu E0 = 20 keV
E0 = 20 keV 933.8 nm 549.3 nm
250 nm
500 nm
-680.0 nm -340.0 nm -0.0 nm 340.0 nm 680.0 nm -400.0 nm -200.0 nm -0.0 nm 200.0 nm 400.0 nm
.. Fig. 2.2 a Monte Carlo simulation of 500 trajectories in carbon c Monte Carlo simulation of 500 trajectories in copper with an incident
with an incident energy of E0 = 20 keV and a surface tilt of 0° (CASINO energy of E0 = 20 keV and a surface tilt of 0°. d Monte Carlo simulation
Monte Carlo simulation). b Monte Carlo simulation of 500 trajectories of 500 trajectories in gold with an incident energy of E0 = 20 keV and a
in silicon with an incident energy of E0 = 20 keV and a surface tilt of 0°. surface tilt of 0°. Red trajectories = backscattering
the transition elements, e.g., Z = 26 (Fe), the slope progressively η mixture = Σηi Ci (2.3)
decreases until at very high Z, e.g., the region around Z = 79
(Au), the slope becomes so shallow that there is very little where C is the mass (weight) fraction and i is an index that
change in η between adjacent elements. Plotted in addition to denotes all of the elements involved.
the experimental measurements in . Fig. 2.3a is a mathemati- When measurements of η vs. Z are made at different
cal fit to the 20 keV data developed by Reuter (1972): beam energies, combining the experimental measurements
of Heinrich and of Bishop in . Fig. 2.3b, little dependence on
η = − 0.0254 + 0.016 Z − 1.86 ×10−4 Z 2 + 8.3×10−7 Z 3 (2.2) the beam energy is found from 5 to 49 keV, with all of the
measurements clustering relatively closely to the curve for
This fit provides a convenient estimate of η for those elements the 20 keV data shown in . Fig. 2.3a. This result is perhaps
for which direct measurements do not exist. surprising in view of the strong dependence of the dimen-
Experimental measurements (Heinrich 1966) have shown sions of the interaction volume on the incident beam energy.
that the backscattered electron coefficient of a mixture of The weak dependence of η upon E0 despite the strong depen-
atoms that is homogeneous on the atomic scale, such as a dence of the beam penetration upon E0 can be understood as
stoichiometric compound, a glass, or certain metallic alloys, a near balance between the increased energy available at
can be accurately predicted from the mass concentrations of higher E0, the lower rate of loss, dE/ds, with higher E0, and
the elemental constituents and the values of η for those pure the increased penetration. Thus, although a beam electron
elements: may penetrate more deeply at high E0, it started with more
18 Chapter 2 · Backscattered Electrons
Backscatter coefficient
0.4
0.3
Reuter Fit
Heinrich 20 keV data
0.2
0.1
0.0
0 20 40 60 80 100
Atomic number
b
Electron backscattering vs. atomic number
0.6
0.5
Backscatter coefficient
0.4
0.0
0 20 40 60 80 100
Atomic number
energy and lost that energy at a lower initial rate than an elec- EM Image Contrast with BSE: “Atomic
S
tron at a lower incidence energy. Thus, a higher incidence Number Contrast”
energy electron, despite penetrating deeper in the specimen, Whenever a signal that can be measured in the SEM, such as
retains more energy and can continue to scatter and progress backscattered electrons, follows a predictable response to a
through the target to escape. specimen property of interest, such as composition, the phys-
ical basis for a “contrast mechanism” is established. Contrast,
Ctr , is defined as
2.2 · Critical Properties of Backscattered Electrons
19 2
Ctr = ( S 2 − S1 ) / S 2 with S 2 > S1 (2.4) The monotonic behavior of η vs. Z establishes the physical
basis for “atomic number contrast” (also known as
where S is the signal measured at any two locations of interest “Z-contrast” and “compositional contrast”). When an SEM
in the image field. As shown in . Fig. 2.4, examples include BSE image is acquired from a flat specimen (i.e., no topogra-
the contrast between an object P1 and the general back- phy is present, at least on a scale no greater than about 5 % of
ground P2 or between two objects that share an interface, P3 the Kanaya–Okayama range for the particular material com-
and P4. By this definition, contrast can range numerically position and incident beam energy), then local differences in
from 0 to 1. composition can be observed as differences in the BSE inten-
sity, which can be used to construct a meaningful gray-scale
SEM image. The compositionally-different objects must have
dimensions that are at least as large as the Kanaya-Okayama
range for each distinct material so that a BSE signal charac-
teristic of the particular composition can be measured over at
least the center portion of the object. The BSE signal at beam
•P 2 locations on the edge of the object may be affected by pene-
•P 1 tration into the neighboring material(s).
From the definition of contrast, Ctr, atomic number con-
trast can be predicted between two materials with backscat-
ter coefficients η1 and η2 when the measured signal S is
proportional to η:
2 1
3
10 µm
20 Chapter 2 · Backscattered Electrons
a b
AI 0.0 nm AI 0.0 nm
487.5 nm 330.0 nm
975.0 nm 660.0 nm
1462.6 nm 990.0 nm
Al Al
E0 = 15 keV E0 = 15 keV
0∞ tilt 45∞ tilt
1950.1 nm 1320.0 nm
500 nm
500 nm
-1420.2 nm -710.0 nm -0.0 nm 710.0 nm 1420.0 nm -961.2 nm -480.6 nm -0.0 nm 480.6 nm 961.2 nm
c d
AI 0.0 nm AI 0.0 nm
420.0 nm 240.0 nm
840.0 nm 480.0 nm
1260.0 nm 720.0 nm
Al Al
E0 = 15 keV E0 = 15 keV
60∞ tilt 75∞ tilt
1680.0 nm 960.0 nm
500 nm 500 nm
-1223.3 nm -611.7 nm -0.0 nm 611.7 nm 1223.3 nm -699.0 nm -349.5 nm -0.0 nm 349.5 nm 699.0 nm
.. Fig. 2.6 a Monte Carlo simulation for aluminum at E0 = 15 keV for E0 = 15 keV for a tilt angle of 60°. d Monte Carlo simulation for alumi-
a tilt angle of 0°. b Monte Carlo simulation for aluminum at E0 = 15 keV num at E0 = 15 keV for a tilt angle of 75°
for a tilt angle of 45°. c Monte Carlo simulation for aluminum at
2.2 · Critical Properties of Backscattered Electrons
21 2
. Figure 2.7 shows the results of similar Monte Carlo topography. . Figure 2.8a shows an example of a pure
simulations for various elements as a function of surface material (polycrystalline silver) with grain faces inclined
inclination. As the surface tilt increases, η increases for all at various angles. The higher the inclination of the local
elements, converging toward unity at high tilt and grazing surface to the incident beam, the higher will be the BSE
incidence for the incident beam. signal, so that highly inclined surfaces appear bright, while
dark surfaces are those nearly perpendicular to the beam.
EM Image Contrast: “BSE Topographic
S This image was prepared with a backscattered electron
Contrast—Number Effects” detector (discussed in the Electron Optics—Detectors
This regular behavior of η vs. θ provides the basis for a module), which has a very large solid angle, so that back-
contrast mechanism by which differences in the relative scattered electrons are collected with high efficiency
numbers of backscattered electrons depend on differences regardless of the direction that they travel after leaving the
in the local surface inclination, which reveals the surface specimen.
Au
0.6
0.4
0.2
0.0
0 20 40 60 80
Tilt angle (degrees)
a b
5 µm 5 µm
.. Fig. 2.8 a SEM backscattered electron image of a topographically prepared with a small collection angle BSE detector placed at the top
irregular surface of pure silver prepared with a large collection angle of the image looking down
BSE detector. b SEM backscattered electron image of the same area,
22 Chapter 2 · Backscattered Electrons
2.2.3 Angular Distribution of Backscattering Thus, the largest number of BSEs follow a path parallel to the
surface normal (φ =0°, cosine = 1), while virtually no BSEs
eam Incident Perpendicular to the Specimen
B exit along a trajectory nearly parallel to the surface (φ =90°,
2 Surface (0° Tilt) cosine = 0). The angular distribution seen in . Fig. 2.8a is also
For a flat, bulk target, backscattered electrons emerge rotationally symmetric around the beam: the same cosine
through the surface along a wide range of possible angular shape would be found in any section through the distribution
paths measured relative to the surface normal. When the in any plane perpendicular to surface containing the beam
incident beam is perpendicular to the specimen surface (0° vector and surface normal.
tilt), experimental measurements and Monte Carlo simula-
tions show that the angular distribution of the trajectories is eam Incident at an Acute Angle
B
such that the fraction along any given angle of emission is to the Specimen Surface (Specimen
proportional to the cosine of that angle of emission, φ, Tilt > 0°)
between the electron trajectory and the surface normal, as When a flat, bulk target is tilted so that the beam is incident at
shown in . Fig. 2.9a: an acute angle to the surface, the angular distribution of back-
scattered electrons changes from the rotationally symmetric
η (ϕ ) ~ cos (ϕ ) (2.6) cosine function of . Fig. 2.9a to the asymmetric distribution
0.2
270 0.0 90
1.0 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0
Flat specimen
90
0.25
120 dh/dW 60
0.20
BSE
0.15 Surface
150 normal 30
0.10
Beam
0.05
180 0.00 0
0.25 0.20 0.15 0.10 0.05 0.00 0.05 0.10 0.15 0.20 0.25
2.2 · Critical Properties of Backscattered Electrons
23 2
seen in . Fig. 2.9b, with this distribution measured for a tilt of 2.2.4 Spatial Distribution of Backscattering
60° (angle of incidence = 30°). The angular distribution is
peaked in the forward direction away from the incident beam Model a small number of trajectories (~25) so that the indi-
direction, with the maximum BSE emission occurring at an vidual trajectories can be distinguished; e.g., for a copper target
angle above the surface close to the value of the angle of inci- with an incident beam energy of 20 keV and 0° tilt, as seen in
dence above the surface of the beam. This angular asymmetry . Fig. 2.10 (Note: because of the random number sampling,
develops slowly for tilt angles up to approximately 30°, but the repeated simulations will differ from each other and will be dif-
asymmetry becomes increasingly pronounced with further ferent from the printed example.) By following a number of
increases in the specimen tilt. Moreover, the rotational sym- trajectories from the point of incidence to the point of escape
metry of the 0° tilt case is also progressively lost with increas- through the surface as backscattered electrons, it can be seen
ing tilt, with the asymmetric distribution seen in . Fig. 2.9b that the trajectories of beam electrons that eventually emerge
becoming much narrower in the direction out of the plotting as BSEs typically traverse the specimen both laterally and in
plane. See 7 Chapter 29 for effects of crystal structure on depth.
backscattering angular distribution.
Depth Distribution of Backscattering
EM Image Contrast: “BSE Topographic
S By performing detailed Monte Carlo simulations for many
Contrast—Trajectory Effects” thousands of trajectories and recording for each trajectory
The overall effects of specimen tilt are to increase the number the maximum depth of penetration into the specimen before
of backscattered electrons and to create directionality in the the electron eventually escaped as a BSE, we can determine
backscattered electron emission, and both effects become the contribution to the overall backscatter coefficient as a
increasingly stronger as the tilt increases. The “trajectory function of the depth of penetration, as shown for a series of
effects” create a very strong component of topographic con- elements in . Fig. 2.11a. To compare the different elements,
trast when viewed with a backscattered electron detector that the horizontal axis of the plot is the depth normalized by the
has limited size and is placed preferentially on one side of the Kanaya–Okayama range for each element. From the depth
specimen. . Figure 2.8b shows the same area as . Fig. 2.8a distribution data in . Fig. 2.11a, the cumulative backscatter-
imaged with a small solid angle detector, located at the top ing coefficient as a function of depth can be calculated, and as
center of the image. Very strong contrast is created between shown in . Fig. 2.11b, this distribution follows an S-shaped
faces tilted toward the detector, i.e., facing upward, and those curve. To capture 90 % of the total backscattering, which cor-
tilted away, i.e., facing downward. These effects will be dis- responds to the region where the slope of the plot is rapidly
cussed in detail in the Image Interpretation module. decreasing, the backscattered electrons are found to travel a
439.5 nm
659.2 nm
Cu
E0 = 20 keV
0º tilt
878.9 nm
500 nm
Backscatter fraction
0.04
0.02
0.00
0.0 0.1 0.2 0.3 0.4 0.5
Depth/range (Kanaya-Okayama)
0.8
Cumulative backscattering (normalized)
C
AI
Cu
0.6 Ag
Au
0.4
0.285
0.250
0.205
0.2
0.185
0.155
0.0
0.0 0.1 0.2 0.3 0.4
Depth/range (Kanaya-Okayama)
significant fraction of the Kanaya–Okayama range into the Monte Carlo study performed for a target at 45° tilt. For a
target. Strong elastic scattering materials with high atomic tilted target, all materials show a slightly smaller fraction of
number such as gold sample a smaller fraction of the range the Kanaya–Okayama range to reach 90 % backscattering
than the weak elastic scattering materials such as carbon. compared to the normal incidence case.
. Table 2.2 lists the fractional range to capture 90 % of back- When the beam energy is increased for a specific mate-
scattering at normal beam incidence (0° tilt) and for a similar rial, the strong dependence of the total range on the incident
2.2 · Critical Properties of Backscattered Electrons
25 2
adial Distribution of Backscattered
R
.. Table 2.2 BSE penetration depth (D/RK-O) to capture 90 % of
total backscattering
Electrons
The Monte Carlo simulation can record the x-y location at
0º tilt 45º tilt which a backscattered electron exits through the surface
plane, and this information can be used to calculate the
C 0.285 0.23
radial distribution of backscattering relative to the beam
Al 0.250 0.21 impact location. The cumulative radial distribution is shown
Cu 0.205 0.19 in . Fig. 2.14 for a series of elements, as normalized by the
Kanaya–Okayama range for each element, and an S-shaped
Ag 0.185 0.17
curve is observed. . Table 2.3 gives the fraction of the range
Au 0.155 0.15 necessary to capture 90 % of the total backscattering. The
radial distribution is steepest for high atomic number ele-
ments, which scatter strongly compared to weakly scattering
low atomic number elements. However, even for strongly
beam energy leads to a strong dependence of the sampling scattering elements, the backscattered electrons emerge over
depth of backscattered electrons, as shown in the depth dis- a significant fraction of the range. This characteristic impacts
tributions of backscattered electrons for copper over a wide the spatial resolution that can be obtained with backscat-
energy range in . Fig. 2.12. The substantial sampling depth tered electron images. An example is shown in . Fig. 2.15 for
of backscattered electrons combined with the strong depen- an interface between an aluminum-rich phase and a copper-
dence of the electron range on beam energy provides a useful rich phase (CuAl2) in directionally solidified aluminum-
tool for the microscopist. By comparing a series of images of copper eutectic alloy. The interfaces are perpendicular to the
a given area as a function of beam energy, subsurface details surface and are atomically sharp. The backscattered electron
can be recognized. An example is shown in . Fig. 2.13 for an signal response as the beam is scanned across the interface is
engineered semiconductor electronic device with three- more than an order-of-magnitude broader (~300 nm) due to
dimensional layered features, where a systematic increase in the lateral spreading of backscattering than would be pre-
the beam energy reveals progressively deeper structures. dicted from the incident beam diameter alone (10 nm).
5 keV 10 keV
20 keV 30 keV
.. Fig. 2.13 BSE images at various incident beam energies of a semiconductor device consisting of silicon and various metallization layers at dif-
ferent depths
0
0 500 1000 1500 2000 2500
Position (nm)
0.02
0.01
0.00
0.0 0.2 0.4 0.6 0.8 1.0
E/E0
0.8
C
AI
Cumulative backscattering (E/E0)
Cu
0.6 Ag
Au
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
E/E0
Secondary Electrons
3.1 Origin – 30
References – 37
200
100
0
0 5 10 15 20 25 30
Energy (eV)
0.6
0.4
0.2
4 eV 8.4 eV
0.0
0 5 10 15 20 25 30
Secondary electron energy (eV)
representative measurement of a property that results from the “ideal” pure element or pure compound value. If the
very low energy electrons generated within and escaping specimen is pre-cleaned by ion bombardment in an ultra-
from a very shallow layer below the surface. Thus, a surface high vacuum electron beam instrument (chamber pres-
modified by accumulations of oxide and contamination sure maintained below 10−8 Pa) which preserves the clean
(e.g., adsorbed water, chemisorbed water, hydrocarbons, surface, and if the surface composition is confirmed to be
etc.) is likely to produce a value of δ that is different from that of the pure element or compound by a surface-specific
32 Chapter 3 · Secondary Electrons
0.6
0.4
0.2
0.0
0 2 4 6 8
Depth (nm)
0.8
Cumulative SE intensity
0.6
0.4
0.2
measurement method such as Auger electron spectroscopy pressure ~ 10−4 Pa) or a variable pressure SEM (chamber
or X-ray photoelectron spectroscopy, then the measured pressure from 10−4 Pa to values as high as 2500 Pa) are not
secondary electron coefficient is likely to be representative likely to be that of pure substances, but are almost inevitably
of the pure substance. However, the surfaces of most speci- covered with a complex mixture of oxides, hydrocarbons,
mens examined in the conventional-vacuum SEM (chamber and chemisorbed water molecules that quickly redeposit
3.4 · Secondary Electron Yield Versus Atomic Number
33 3
.. Fig. 3.3 Mean secondary electron
escape depth for various materials as Secondary electron escape depth
modeled by Kanaya and Ono (1984) 10
Li
Kanaya-Ono (1984) calculations
8 K
Na
4
AI
Be Bi
2 Hf Th
Ni Rh Re
Ce
0
0 20 40 60 80
Atomic number
0.8
0.6
0.4
0.2
0.0
0 20 40 60 80 100
Atomic number
at such elevated pressures even when ion beam cleaning is Thus, while compositionally dependent secondary electron
utilized to expose the “true” surface. The effective second- signals may be occasionally observed, they are generally
ary electron coefficient of a “real” material under typical not predictable and reproducible, which is the critical basis
SEM or VP-SEM vacuum conditions is unlikely to pro- for establishing a useful contrast mechanism such as that
duce a consistent, predictable response as a function of the found for backscattered electrons as a function of atomic
composition of the nominal substance under examination. number.
34 Chapter 3 · Secondary Electrons
E0 = 1 keV
1.5
1.0
E0 = 10 keV
0.5
0.0
0 10 20 30 40 50 60 70
Tilt angle (degrees)
3.8 · Spatial Characteristics of Secondary Electrons
35 3
0.6
300 60
0.4
S φ PL
cos φ = s/PL 0.2
PL = s/cos φ
270 0.0 90
1.0 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0
c SE
S
φ
PL
SE
.. Fig. 3.7 a Dependence of the secondary electron escape path normal as simulated by Monte Carlo calculations (Koshikawa and
length on the angle relative to the surface normal. The probability of Shimizu 1974) compared to a cosine function. c The escape path length
escape decreases as this path length increases. b Angular distribution situation of . Fig. 3.7a for the case of a tilted specimen. A cosine
of secondary electrons as a function of the angle relative to the surface dependence relative to the surface normal is again predicted
secondary electrons is expected to follow a cosine relation electrons is reduced so more of that energy is deposited and
with the emergence angle relative to the local surface nor- more secondary electrons are generated in the near surface
mal. Behavior close to a cosine relation is seen in the Monte region from which secondary electrons can escape. This is a
Carlo simulation of Koshikawa and Shimizu (1974) in general behavior found across the Periodic Table, as seen in
. Fig. 3.7b. the plots for C, Al, Cu, Ag, and Au in . Fig. 3.8c.
Even when the surface is highly tilted relative to the beam,
the escape path length situation for a secondary electron gen-
erated below the surface is identical to the case for normal 3.8 patial Characteristics of Secondary
S
beam incidence, as shown in . Fig. 3.7c. Thus, the secondary Electrons
electron trajectories follow a cosine distribution relative to
the local surface normal regardless of the specimen tilt. As the beam electrons enter the sample surface, they begin to
generate secondary electrons in a cylindrical volume whose
cross section is defined by the footprint of the beam on the
3.7 econdary Electron Yield Versus Beam
S entrance surface and whose height is the escape depth of the
Energy SE, as shown schematically in . Fig. 3.9 These entrance surface
SE, designated the SE1 class, preserve the lateral spatial resolu-
The secondary electron coefficient increases as the incident tion information defined by the dimensions of the focused
beam energy decreases, as shown for copper in . Fig. 3.8a for beam and are similarly sensitive to the properties of the near
the conventional beam energy range (5 keV ≤ E0 ≤ 30 keV) surface region due to the shallow scale of their origin. As the
and in . Fig. 3.8b for the low beam energy range (E0 < 5 keV). beam electrons move deeper into the solid, they continue to
This behavior arises from two principal factors: (1) as the generate SE, but these SE rapidly lose their small initial kinetic
beam electron energy decreases, the rate of energy loss, dE/ds, energy and are completely reabsorbed within an extremely
increases so that more energy is deposited per unit of beam short range. However, for those beam electrons that subse-
electron path length leading to more secondary electron gen- quently undergo enough scattering to return to the entrance
eration per unit of path length; and (2) the range of the beam surface to emerge as backscattered electrons (or reach any
36 Chapter 3 · Secondary Electrons
0.4
0.3
0.2
0.1
0 5 10 15 20 25 30
Beam energy (keV)
Secondary electron yield vs. beam energy for copper
b 2.2
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0 1 2 3 4 5
Beam energy (keV)
1.4
1.2
Au
1.0 Ag
Cu
0.8 AI
C
0.6
0.4
0.2
0.0
0 5 10 15 20 25 30
Beam energy (keV)
References
37 3
.. Fig. 3.9 Schematic diagram
showing the origins of the SE1, SE2, and
SE3 classes of secondary electrons. The
SE1 class carries the lateral and
near-surface spatial information defined
by the incident beam, while the SE2 and SE3
SE3 classes actually carry backscattered
electron information. The blue rectangle
represents the escape depth for SE, and
SE1 SE2
the cylinder represents the volume from
which the SE1 escape
other surface for specimens with more complex topography that for gold the sum of the SE2 and SE3 classes which actually
than a simple flat bulk target), the SE that they continue to gen- carry BSE is nearly ten times larger than the high resolution,
erate as they approach the surface region will escape and add to high surface sensitivity SE1 component. These three classes of
the total secondary electron production, as shown in . Fig. 3.9. secondary electrons influence SEM images of compositional
This class of SE is designated SE2 and they are indistinguishable structures and topographic structures in complex ways. The
from the SE1 class based on their energy and angular distribu- appearance of the SE image of a structure depends on the
tions. However, because of their origin from the backscattered details of the secondary electron emission and the properties
electrons, the SE2 class actually carries the degraded lateral spa- of the secondary electron detector used to capture the signal,
tial distribution of the BSE: because the relative number of SE2 as discussed in detail in the image formation module.
rises and falls with backscattering, the SE2 signal actually car-
ries the same information as BSE. That is, the relative number
of the SE2 scales with whatever specimen property affects elec- References
tron backscattering. Finally, the BSE that leave the specimen
Bongeler R, Golla U, Kussens M, Reimer L, Schendler B, Senkel R, Spranck
are energetic, and after traveling millimeters to centimeters in
M (1993) Electron-specimen interactions in low voltage scanning
the specimen chamber, these BSE are likely to hit other metal electron microscopy. Scanning 15:1
surfaces (objective lens polepiece, chamber walls, stage compo- Joy D (2012) Can be found in chapter 3 on SpringerLink: http://link.
nents, etc.), generating a third set of secondary electrons desig- springer.com/chapter/10.1007/978-1-4939-6676-9_3
nated SE3. The SE3 class again represents BSE information, Kanaya K, Ono S (1984) Interaction of electron beam with the target in
including the degraded spatial resolution, not true SE1 infor- scanning electron microscope. In: Kyser DF, Niedrig H, Newbury DE,
Shimizu R (eds) Electron interactions with solids. SEM, Inc, Chicago,
mation and resolution. The SE1 and SE2 classes represent an pp 69–98
inherent property of a material, while the SE3 class depends on Koshikawa T, Shimizu R (1973) Secondary electron and backscattering
the details of the SEM specimen chamber. Peters (1984) mea- measurements for polycrystalline copper with a retarding-field
sured the three secondary electron classes for thin and thick analyser. J Phys D Appl Phys 6:1369
gold targets to estimate the relative populations of each class: Koshikawa T, Shimizu R (1974) A Monte Carlo calculation of low-energy
secondary electron emission from metals. J Phys D Appl Phys 7:1303
Incident beam footprint, high resolution, SE1 (9 %) Peters K-R (1984) Generation, collection and properties of an SE-I
BSE generated at specimen, low resolution, SE2 (28 %) enriched signal suitable for high resolution SEM on bulk specimens.
BSE generated remotely on lens, chamber walls, SE3 (61 %) In: Kyser DF, Niedrig H, Newbury DE, Shimizu R (eds) Electron beam
interactions with solids. SEM, Inc, AMF O’Hare, p 363
A small SE contribution designated the SE4 class arises from Moncrieff DA, Barker PR (1976) Secondary electron emission in the
scanning electron microscope. Scanning 1:195
pre-specimen instrumental sources such as the final aperture Reimer L, Tolkamp C (1980) Measuring the backscattering coefficient
(2 %) that depends in detail on the instrument construction and secondary electron yield inside a scanning electron micro-
(apertures, magnetic fields, etc.). These measurements show scope. Scanning 3:35
39 4
X-Rays
4.1 Overview – 40
References – 63
K-shell
vacancy Eν = EK – EL = 277eV
Ekin = EK – 2EL
4.2 · Characteristic X-Rays
41 4
absorption energy,” and the “K-edge energy”) of 284 eV and ment of its characteristic kinetic energy can identify the
four electrons in the L-shell, two each in the L1 and the L2 sub- atom species of its origin, forming the physical basis for
shells bound to the atom, with an ionization energy of 7 eV. An “Auger electron spectroscopy.”
incident energetic beam electron having initial kinetic energy 2. The right branch in . Fig. 4.1 involves the creation of an
Ein > Ec can scatter inelastically with a K-shell atomic electron X-ray photon to carry off the inter-shell transition energy:
and cause its ejection from the atom, providing the beam elec-
tron transfers to the atomic electron kinetic energy at least Eν = EK − EL = 277 eV (4.2b)
equal to the ionization energy, which is the minimum energy
necessary to promote the atomic electron out of the K-shell Because the energies of the atomic shells of an element are
beyond the effective influence of the positive nuclear charge. sharply defined, the shell difference is also a sharply defined
The total kinetic energy transferred to the K-shell atomic elec- quantity, so that the resulting X-ray photon has an energy that is
tron can range up to half the energy of the incident electron. characteristic of the particular atom species and the shells
The outgoing beam electron thus suffers energy loss corre- involved and is thus designated as a “characteristic X-ray.” Char-
sponding to the carbon K-shell ionization energy EK = 284 eV acteristic X-rays are emitted uniformly in all directions over the
plus whatever additional kinetic energy is imparted: full unit sphere with 4 π steradians solid angle. Extensive tables
of characteristic X-ray energies for elements with Z ≥ 4 (beryl-
Eout = Ein − EK − Ekin (4.1) lium) are provided in the database embedded within the DTSA-
II software. The characteristic X-ray photon energy has a very
The ionized carbon atom is left with a vacancy in the K-shell narrow range of just a few electronvolts depending on atomic
which places it in a raised energy state that can be lowered number, as shown in . Fig. 4.2 for the K–L3 transition.
through the transition of an electron from the L-shell to fill
the K-vacancy. The difference in energy between these shells
must be expressed through one of two possible routes: 4.2.2 Fluorescence Yield
1. The left branch in . Fig. 4.1 involves the transfer of this
K–L inter-shell transition energy difference to another The Auger and X-ray branches in . Fig. 4.1 are not equally
L-shell electron, which is then ejected from the atom probable. For a carbon atom, characteristic X-ray emission
with a specific kinetic energy: only occurs for approximately 0.26 % of the K-shell ioniza-
tions. The fraction of the ionizations that produce photons is
Ekin = EK − EL − EL = 270 eV (4.2a) known as the “fluorescence yield,” ω. Most carbon K-shell
ionizations thus result in Auger electron emission. The fluo-
This process leaves the atom with two L-shell vacancies rescence yield is strongly dependent on the atomic number
for subsequent vacancy-filling transitions. This ejected of the atom, increasing rapidly with Z, as shown in . Fig. 4.3a
electron is known as an “Auger electron,” and measure- for K-shell ionizations. L-shell and M-shell fluorescence
10
K-L3 Peak Width (eV, FWHM)
0
0 5 10 15 20 25 30
0.4
0.3
Fluorescence yield
Fluorescence yield
0.3
0.2
4 0.2
0.1
0.1
0.0 0.0
0 5 10 15 20 25 0 20 40 60 80 100
Atomic number Atomic number
c M5-shell fluorescence yield d Fluorescence yield
0.0020 1
0.1
0.0014
0.0012 K-shell
L3-shell
0.0010 0.01 M5-shell
0.0008
0.0006
0.001
0.0004
0.0002
0.0000 0.0001
30 40 50 60 70 80 90 100 0 20 40 60 80 100
Atomic number Atomic number, Z
.. Fig. 4.3 a Fluorescence yield (X-rays/ionization) from the K-shell. b Fluorescence yield (X-rays/ionization) from the L3-shell. c Fluorescence yield
(X-rays/ionization) from the M5-shell. d Comparison of fluorescence yields from the K-, L3- and M5- shells (Crawford et al. 2011)
yields are shown in . Fig. 4.3b, c; and K-, L-, and M-shell ″K − L2,3 ″ ( ″K α″ ) EX = EK − EL =1041eV (4.3a)
yields are compared in . Fig. 4.3d (Crawford et al. 2011).
From . Fig. 4.3d, it can be observed that, when an element ″K − M ″ ( ″K β″ ) EX = EK − EM =1071eV (4.3b)
can be measured with two different shells, ωK > ωL > ωM.
The shell transitions for carbon are illustrated in the shell For atoms with higher atomic number than sodium, addi-
energy diagram shown in . Fig. 4.4a. Because of the small tional possibilities exist for inter-shell transitions, as shown
number of carbon atomic electrons, the shell energy values are in . Fig. 4.4b, leading to splitting of the K − L2,3 into K − L3
limited, and only one characteristic X-ray energy is possible for and K − L2 (Kα into Kα1 and Kα2), and similarly for Kβ into
carbon with a value of 277 eV. (The apparent possible transi- Kβ1 and Kβ2, which can be observed with energy dispersive
tion from the L1-shell to the K-shell is forbidden by the quan- spectrometry for X-rays with energies above 20 keV.
tum mechanical rules that govern these inter-shell transitions.) As these additional inter-shell transitions become possi-
ble, increasingly complex “families” of characteristic X-rays
are created, as shown in the energy diagrams of . Fig. 4.4c for
4.2.3 X-Ray Families L-shell X-rays, and 4.4d for M-shell X-rays. Only transitions
that lead to X-rays that are measurable on a practical basis
As the atomic number increases, the number of atomic elec- with energy dispersive X-ray spectrometry are shown. (There
trons increases and the shell structure becomes more com- are, for example, at least 25 L-shell transitions that are possi-
plex. For sodium, the outermost electron occupies the ble for a heavy element such as gold, but most are of such low
M-shell, so that a K-shell vacancy can be filled by a transition abundance or are so close in energy to a more abundant tran-
from the L-shell or the M-shell, producing two different sition as to be undetectable by energy dispersive X-ray
characteristic X-rays, designated spectrometry.)
4.2 · Characteristic X-Rays
43 4
a c N7
L2 N6
N5
L1 N4
N3
N2
N1
M5
M4
K M3
M2
L2-M4
M1
b N7 L3-M5 L3-N5
L2-N4
L2-M1
Lb1 L g1 L1-N2 L3-M1
L1-M3 L1-M2 Lh
N6 La1 L3-M4 Lb2
L g2
L1-N3
Ll
Lb3 Lb4
N5 La2 L g3
N4 L3
N3 L2
N2
L1
N1
M5
M4 K
M3
M2
M1 d N7
K-M3 N6
Kb1 N5
K-N3 N4
Kb2 N3
N2
L3 N1
M5-N7 M5-N6 M4-N6 M3-N5 M4,5-N2,3
L2 Mα1 Mα2 Mβ Mγ Mζ
M5
L1 M4
K-L3 M3
M2
Ka1 K-L2 M1
Ka2
K
L3
L2
L1
.. Fig. 4.4 a Atomic shell energy level diagram for carbon illustrating energy level diagram illustrating possible L-shell vacancy-filling transi-
the permitted shell transition K–L2 (shown in green) and the forbidden tions. d Atomic shell energy level diagram illustrating some possible
transition K–L1 (shown in red). b Atomic shell energy level diagram M-shell vacancy-filling transitions
illustrating possible K-shell vacancy-filling transitions. c Atomic shell
4.2.4 X-Ray Nomenclature replaced by the International Union of Pure and Applied
Chemistry (IUPAC) labeling protocol in which the first
Two systems are in use for designating X-rays. The term denotes the shell or subshell where the original ion-
traditional but now archaic Siegbahn system lists the ization occurs while the second term indicates the sub-
shell where the original ionization occurs followed by a shell from which the electron transition occurs to fill the
Greek letter or other symbol that suggests the order of vacancy; for example, Kα 1 is replaced by K-L3 for a K-shell
the family members by their relative intensity, α > β > γ ionization filled from the L3 subshell. . Table 4.1 gives
> η > ζ. For closely related members, numbers are also the correspondence between the Siegbahn and IUPAC
attached, for example, Lβ1 through Lβ15. Additionally, labeling schemes for the characteristic X-rays likely to
Latin letters are used for the complex minor L-shell fam- be detected by energy dispersive X-ray spectrometry.
ily members: l , s, t, u, and v. While still the predominant Note that for the M-shell, there are minor family mem-
labeling system used in commercial X-ray microanalysis bers detectable by EDS for which there are no Siegbahn
software systems, the Siegbahn system has been officially designations.
44 Chapter 4 · X-Rays
.. Table 4.1 Correspondence between the Siegbahn and IUPAC nomenclature protocols (restricted to characteristic X-rays observed
with energy dispersive X-ray spectrometry and photon energies from 100 eV to 25 keV)
Lη L2-M1 M2-N4
Ll L3-M1
4.2.5 X-Ray Weights of Lines The relationship of the energy of the exciting electron to
the ionization energy of the atomic electron is an important
Within these families, the relative abundances of the charac- parameter and is designated the “overvoltage,” U:
teristic X-rays are not equal. For example, for sodium the
ratio of the K-L2,3 to K-M is approximately 150:1, and this U = E / Ec (4.5a)
ratio is a strong function of the atomic number, as shown in
. Fig. 4.5a for the K-shell (Heinrich et al. 1979). For the
The overvoltage that corresponds to the incident beam
L-shell and M-shell, the X-ray families have more members, energy, E0, which is the maximum value because the beam
and the relative abundances are complex functions of atomic electrons subsequently lose energy due to inelastic scattering
number, as shown in . Fig. 4.5b, c. as they progress through the specimen, is designated as U0:
U 0 = E0 / Ec (4.5b)
4.2.6 Characteristic X-Ray Intensity
For ionization to occur followed by X-ray emission, U > 1.
Isolated Atoms With this definition for U, Eq. (4.4) can be rewritten as
When isolated atoms are considered, the probability of an
energetic electron with energy E (keV) ionizing an atom by (
QI ionizations / e − ( atom / cm 2 ) )
ejecting an atomic electron bound with ionization energy Ec = 6.51× 10−20 ( ns bs ) / U Ec 2 log e ( csU ) (4.6)
(keV) can be expressed as a cross section, QI:
0.10
Kβ/(Kα+Kβ)
0.08
0.06
0.04
0.02
0.00
10 15 20 25 30 35
Atomic number(Z)
L-shell weights of lines [relative to L3-M4,5 (Lα1,2) = 1]
b 1
Relative weights of lines
0.1
0.01
L3-M1 (L )
L2-M1 (Lη)
0.001 L2-M4 (Lβ1)
L3-N5 (Lβ2)
L2-N4 (Lγ1)
L1-N3 (Lγ3)
0.0001
20 30 40 50 60 70 80 90 100
Atomic number (Z)
M5-N3 (Mζ)
Relative weights of lines
M4-N6 (Mβ)
M3-N5 (Mγ)
0.1
M2-N4
0.01
0.001
65 70 75 80 85 90 95
Atomic number (Z)
46 Chapter 4 · X-Rays
1.2e-20
4 8.0e-21
6.0e-21
4.0e-21
2.0e-21
0.0
0 5 10 15 20 25 30
Beam energy (keV)
25 K-shell
L3 shell
M5 shell
Critical ionization energy (keV)
20
15
10
0
0 10 20 30 40 50 60 70 80 90 100
Atomic number (Z)
energy loss. The X-ray production in a thin foil of thick- where A is the atomic weight and N0 is Avogadro’s number.
ness t can be estimated from the cross section by calculat- When several elements are mixed at the atomic level in a
ing the effective density of atom targets within the foil: thin specimen, the relative production of X-rays from differ-
ent elements depends on the cross section and fluorescence
nX photons / e − = QI ionizations / e − ( atom / cm 2 ) yield, as given in Eq. 4.7, and also on the partitioning of the
× ω [ X-rays / ionization ] × Ν 0 [ atoms / mole ] X-ray production among the various possible members of
the X-ray families, as plotted in . Fig. 4.5a–c. The relative
× (1/ A ) [ moles / g ] × ρ g / cm3 production for the most intense transition in each X-ray fam-
× t [ cm ] = QI × ω × N 0 × ρ × t / A (4.7) ily is plotted in . Fig. 4.8 for E0 = 30 keV. . Figure 4.8 reveals
4.3 · X-Ray Continuum (bremsstrahlung)
47 4
.. Fig. 4.8 Product of the ionization
cross section, the fluorescence yield, Product of ionization cross section, relative transition
and the relative weights of lines for the probability and fluorescence yield
most intense member of the K-, L-, and 1
M-shells for E0 = 30 keV
K-shell
L-shell
M-shell
0.1
0.001
0.0001
0 20 40 60 80
Atomic number (Z)
strong differences in the relative abundance of the X-rays rises rapidly from a zero value at U = 1. For a reasonably effi-
produced by different elements. This plot also reveals that cient degree of X-ray excitation, it is desirable to select E0 so
over certain atomic number ranges, two different atomic that U0 > 2 for the highest value of Ec among the elements of
shells can be excited for each element, for example, K and L interest.
for Z = 16 to Z = 50, and L and M for Z = 36 to Z = 92. For
lower values of E0, these atomic number ranges will be
diminished. 4.3 X-Ray Continuum (bremsstrahlung)
40
20
10
0
0 2 4 6 8 10
Overvoltage, U = E0/Ec
Eν
Eν =E0
direction of the electron travel. In thin specimens where the However, in thick specimens, the near-randomization of
beam electron trajectories are nearly aligned, this anisotropy the beam electron trajectory segments by elastic scattering
can result in a different continuum intensity in the forward effectively smooths out this anisotropy, so that the X-ray con-
direction along the beam relative to the backward direction. tinuum is effectively rendered isotropic.
4.3 · X-Ray Continuum (bremsstrahlung)
49 4
4.3.1 X-Ray Continuum Intensity 4.3.2 he Electron-Excited X-Ray Spectrum,
T
As-Generated
The intensity of the X-ray continuum, Icm , at an energy Eν is
described by Kramers (1923) as The electron-excited X-ray spectrum generated within the
target thus consists of characteristic and continuum X-rays
I cm ≈ ip Z ( E0 − Eν ) / Eν (4.9) and is shown for pure carbon with E0 = 20 keV in . Fig. 4.12,
as calculated with the spectrum simulator in NIST Desktop
where ip is the incident beam current and Z is the atomic Spectrum Analyzer (Fiori et al. 1992), using the Pouchou and
number. For a particular value of the incident energy, E0, the Pichoir expression for the K-shell ionization cross section and
intensity of the continuum decreases rapidly relative to lower the Kramers expression for the continuum intensity (Pouchou
photon energies as Eν approaches E0, the Duane–Hunt limit. and Pichoir 1991; Kramers 1923). Because of the energy
An important parameter in electron-excited X-ray micro- dependence of the continuum given by Eq. 4.10, the generated
analysis is the ratio of the characteristic X-ray intensity to the X-ray continuum has its highest intensity at the lowest photon
X-ray continuum intensity at the same energy, Ech = Eν, often energy and decreases at higher photon energies, reaching zero
referred to as the “peak-to-background, P/B.” The P/B can be intensity at E0. By comparison, the energy span of the charac-
estimated from Eqs. (4.8) and (4.9) with the approximation teristic C–K peak is its natural width of only 1.6 eV, which is
that Eν ≈ Ec so that Eq. (4.9) can be rewritten as— related to the lifetime of the excited K-shell vacancy. The
energy width for K-shell emission up to 25 keV photon energy
I cm ≈ ip Z ( E0 − Ec ) / Ec ≈ ip Z (U − 1) (4.10) is shown in . Fig. 4.2 (Krause and Oliver 1979). For photon
energies below 25 keV, the characteristic X-ray peaks from the
Taking the ratio of Eqs. (4.8) and (4.10) gives K-, L-, and M- shells have natural widths less than 10 eV. In
the calculated spectrum of . Fig. 4.12, the C–K peak is there-
P / B ≈ (1 / Z )(U − 1)
n −1
(4.11) fore plotted as a narrow line. (X-ray peaks are often referred to
as “lines” in the literature, a result of their appearance in high-
The P/B is plotted in . Fig. 4.11 with the assumption that energy resolution measurements of X-ray spectra by diffrac-
n = 1.7, where it is seen that at low overvoltages, which are tion-based X-ray spectrometers.) The X-ray spectra
often used in electron-excited X-ray microanalysis, the char- as-generated in the target for carbon, copper, and gold are
acteristic intensity is low relative to higher values of U, and compared in . Fig. 4.13, where it can be seen that at all photon
the intensity rises rapidly with U, while the P/B increases energies the intensity of the X-ray continuum increases with
rapidly at low overvoltage but then more slowly as the over- Z, as given by Eq. 4.9. The increased complexity of the charac-
voltage increases. teristic X-rays at higher Z is also readily apparent.
6 P/B
Relative X-ray intensity
Relative intensity and P/B
0
1.0 1.5 2.0 2.5 3.0 3.5 4.0
Overvoltage, U = E0/Ec
50 Chapter 4 · X-Rays
1e+5
4
1e+4
Intensity
1e+3
1e+2
1e+1
1e+0
0 5 10 15 20
X-ray photon energy (keV)
1e+6
1e+5
Intensity
1e+4
1e+3
1e+2 C
Cu
Au
1e+1
1e+0
0 5 10 15 20
X-ray photon energy (keV)
4.3.3 Range of X-ray Production crude estimate of the limiting range of X-ray generation,
can be obtained from a simple modification of the
As the beam electrons scatter inelastically within the tar- Kanaya–Okayama range equation (IV-5) to compensate
get and lose energy, inner shell ionization events can be for the portion of the electron range beyond which the
produced from U0 down to U = 1, so that depending on E0 energy of beam electrons has deceased below a specific
and the value(s) of Ec represented by the various elements value of Ec:
in the target, X-rays will be generated over a substantial
RK − O ( nm ) = 27.6 ( A / Z 0.89 ρ ) E01.67 − Ec1.67 (4.12)
portion of the interaction volume. The “X-ray range,” a
4.3 · X-Ray Continuum (bremsstrahlung)
51 4
. Table 4.2 lists calculations of the range of generation for in various matrices for the generation of various characteristic
copper K-shell X-rays (Ec = 8.98 keV) produced in various X-rays spanning a wide range of Ec is shown in . Fig. 4.14a–d.
host elements, for example, a situation in which copper is
present at a low level so it has a negligible effect on the over-
all Kanaya–Okayama range. As the incident beam energy 4.3.4 onte Carlo Simulation of X-Ray
M
decreases to E0 = 10 keV, the range of production of cop- Generation
per K-shell X-rays decreases to a few hundred nanometers
because of the very low overvoltage, U0 = 1.11. The X-ray range The X-ray range given by Eq. 4.12 provides a single value that
captures the limit of the X-ray production but gives no infor-
mation on the complex distribution of X-ray production
within the interaction volume. Monte Carlo electron simula-
.. Table 4.2 Range of Cu K-shell (Ec = 8.98 keV) X-ray
generation in various matrices tion can provide that level of detail (e.g., Drouin et al., 2007;
Joy, 2006; Ritchie, 2015), as shown in . Fig. 4.15a, where the
Matrix 25 keV 20 keV 15 keV 10 keV electron trajectories and the associated emitted photons of
C 6.3 μm 3.9 μm 1.9 μm 270 nm
Cu K-L3 are shown superimposed. For example, the limit of
production of Cu K-L3 that occurs when energy loss causes
Si 5.7 μm 3.5 μm 1.7 μm 250 nm the beam electron energy to fall below the Cu K-shell excita-
Fe 1.9 μm 1.2 μm 570 nm 83 nm tion energy (8.98 keV) can be seen in the electron trajecto-
Au 1.0 μm 630 nm 310 nm 44 nm
ries (green) that extend beyond the region of X-ray
production (red). The effects of the host element on the
Range (micrometers)
AIK X-rays
TiK X-rays TiK X-rays
6 FeK X-rays 6 FeK X-rays
CuK X-rays CuK X-rays
4 4
2 2
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Beam energy (keV) Beam energy (keV)
Range (micrometers)
1.0
0.5
0.5
0.0 0.0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Beam energy (keV) Beam energy (keV)
.. Fig. 4.14 a X-ray range as a function of E0 for generation of K-shell c X-ray range as a function of E0 for generation of K-shell X-rays of C, Al,
X-rays of C, Al, Ti, Fe, and Cu in a C matrix. b X-ray range as a function of Ti, Fe, and Cu in a Cu matrix. d X-ray range as a function of E0 for gen-
E0 for generation of K-shell X-rays of C, Al, Ti, Fe, and Cu in an Al matrix. eration of K-shell X-rays of C, Al, Ti, Fe, and Cu in an Au matrix
52 Chapter 4 · X-Rays
Cu K-L3 in Cu
E0 = 10 keV Cu L3-M5 in Cu
E0 = 10 keV
250 nm
250 nm
Cu K-L3 in Au Cu L3-M5 in Au
E0 = 10 keV E0 = 10 keV
250 nm 250 nm
Cu K-L3 in C Cu L3-M5 in C
E0 = 10 keV E0 = 10 keV
1 µm 1 µm
Number 5000
Select
Si
1 µm Repeat
Phiroz
f(chi)
Energy (keV) 20
Tilt/TOA 0
1 µm
Number 20000
Select
Si
Repeat
method to model ϕ(ρz) by dividing the target into layers of a peak value in ϕ(ρz) is reached, beyond which the X-ray
constant thickness parallel to the surface, counting the X-rays intensity decreases due to cumulative energy loss, which
produced in each layer, and then plotting the intensity as a his- reduces the overvoltage, and the relative number of backscat-
togram. The intensity in each layer is normalized by the inten- tering events decreases. The ϕ(ρz) distribution then steadily
sity produced in a single unsupported layer which is sufficiently decreases to a zero intensity when the electrons have sustained
thin so that no significant elastic scattering occurs: the electron sufficient energy loss to reach overvoltage U = 1. The limiting
trajectories pass through such a thin layer without deviation. X-ray production range is given by Eq. 4.12.
The ϕ(ρz) distribution has several important characteristics.
4 For a thick specimen, the intensity produced in the first layer
exceeds that of the unsupported reference layer because in 4.4 X-Ray Absorption
addition to the X-ray intensity produced by the passage of all of
the beam electrons through the first layer, elastic scattering The Monte Carlo simulations shown in . Fig. 4.15b–d are in
from deeper in the specimen creates backscattered electrons fact plots of the X-rays emitted from the sample. To escape the
which pass back through the surface layer to escape the target, sample, the X-rays must pass through the sample atoms where
producing additional X-ray generation. The intensity produced they can undergo the process of photoelectric absorption.
in the first layer, designated ϕ0, thus always exceeds unity An X-ray whose energy exceeds the binding energy (critical
because of this extra X-ray production due to backscattering. excitation energy) for an atomic shell can transfer its energy to
Below the surface layer, ϕ(ρz) increases as elastic scattering the bound electron, ejecting that electron from the atom with
increases the path length of the electrons that pass obliquely a kinetic energy equal to the X-ray energy minus the bind-
through each layer, compared to the relatively unscattered pas- ing energy, as shown in . Fig. 4.17, which initiates the same
sage of the incident electrons through the outermost layers processes of X-ray and Auger electron emission as shown
before elastic scattering causes significant deviation in the tra- in . Fig. 4.1 for inner shell ionization by energetic electrons.
jectories. The reverse passage of backscattered electrons also The major difference in the two processes is that the X-ray is
adds to the generation of X-rays in the shallow layers. Eventually annihilated in photoelectric absorption and its entire energy
K-shell
vacancy En= EK – EL = 277 eV
Ekin = EK – 2EL
4.4 · X-Ray Absorption
55 4
transferred to the ejected electron. Photoelectric absorption The dimensions of the mass absorption coefficient are cm2/g.
is quantified by the “mass absorption coefficient,” μ/ρ, which For a given material, mass absorption coefficients generally
determines the fraction of X-rays that pass through a thickness decrease with increasing photon energy, as shown for carbon
s of a material acting as the absorber: in . Fig. 4.18a. The exception is near the critical excitation
energy for the atomic shells of a material. The region near the
I / I 0 = exp − ( µ / ρ ) ρ s (4.13) C K-shell excitation energy of 284 eV is shown expanded in
. Fig. 4.18b, where an abrupt increase in μ/ρ occurs. An
where I0 is the initial X-ray intensity and I is the intensity after X-ray photon whose energy is just slightly greater than the
passing through a thickness, s, of a material with density ρ. critical excitation energy for an atomic shell can very
1e+4
1e+3
1e+2
1e+1
1e+0
1e-1
0 5 10 15 20
1e+5
1e+4
1e+3
1e+2
0.0 0.5 1.0 1.5 2.0
X-ray photon energy (keV)
56 Chapter 4 · X-Rays
1e+3
1e+2
1e+1
0 5 10 15 20
1e+5
1e+4
1e+3
0.0 0.5 1.0 1.5 2.0
X-ray photon energy (keV)
e fficiently couple its energy to the bound electron, resulting When a material consists of an atomic-scale mixture of
in a high value of μ/ρ. With further increases in photon two or more elements, the mass absorption for the mixture is
energy, the efficiency of the coupling of the photon energy to calculated as
the bound electron decreases so that μ/ρ also decreases. For
more complex atoms with more atomic shells, the mass
( µ / ρ ) mix = Σ j ( µ / ρ ) j Cj
i i
(4.14)
absorption coefficient behavior with photon energy becomes
more complicated, as shown for Cu in . Fig. 4.19a, b, which
shows the region of the three Cu L-edges. For Au, . Fig. 4.20a– where (μ/ρ)ji is the mass absorption coefficient for the X-rays
c shows the regions of the three Au L-edges and the five Au of element i by element j, and Cj is the mass concentration of
M-edges. element j in the mixture.
4.4 · X-Ray Absorption
57 4
.. Fig. 4.20 a Mass absorption coefficient for Au
as a function of photon energy. b Mass absorption X-ray mass absorption coefficient of gold
a 1e+6
coefficient for Au as a function of photon energy
near the Au L-shell critical excitation energies. c
Mass absorption coefficient for Au as a function of
1e+4
1e+3
1e+2
1e+1
0 5 10 15 20
X-ray photon energy (keV)
100
10 11 12 13 14 15
X-ray photon energy (keV)
X-ray mass absorption coefficient of gold
c 10000
Mass absorption coefficient (cm2/g)
1000
a
ED
S
4
S
b
Phiroz Phiroz
f[chi] f[chi]
Phiroz Phiroz
f[chi] f[chi]
.. Fig. 4.21 a Determination of the absorption path length from a layer of the ϕ(ρz) distribution located at depth, d, in the direction of the X-ray
detector. b Monte Carlo determination of the ϕ(ρz) distribution and absorption for carbon at various beam energies. (Joy Monte Carlo simulation)
Photoelectric absorption reduces the X-ray intensity that shown in . Fig. 4.21a from the depth, d, of the histogram
is emitted from the target at all photon energies. The absorp- slice, and the cosecant of the X-ray detector “take-off angle,”
tion loss from each layer of the ϕ(ρz) distribution is calcu- ψ, which is the elevation angle of the detector above the
lated using Eq. 4.15 with the absorption path, s, determined as target surface:
4.5 · X-Ray Fluorescence
59 4
in f(χ) are seen at the critical ionization energy of the K-shell
.. Table 4.3 Self-absorption of carbon K-shell X-rays as a
function of beam energy
in carbon, the K- and L-shells in copper, and the M- and
L-shells in gold, corresponding to the sharp increase in μ/ρ
E0 f(χ) just above the critical ionization energy. Because the X-ray
continuum is generated at all photon energies, the continuum
2 keV 0.974 is affected by every ionization edge represented by the atomic
5 0.867 species present, resulting in abrupt steps in the background.
An abrupt decrease in X-ray continuum intensity is observed
10 0.615
just above the absorption edge energy due to the increase in
20 0.237 the mass absorption coefficient. The characteristic peaks in
30 0.103 these spectra are also diminished by absorption, but because
a characteristic X-ray is always lower in energy than the ion-
ization edge energy from which it originated, the mass
absorption coefficient for characteristic X-rays is lower than
s = d cscψ (4.15) that for photons with energies just above the shell ionization
energies. Thus an element is relatively transparent to its own
Normalizing by the intensity generated in each layer, the characteristic X-rays because of the decrease in X-ray absorp-
ϕ(ρz) histogram gives the probability, with a value between 0 tion below the ionization edge energy.
and 1, that a photon generated in that layer and emitted into
the solid angle of the EDS detector will escape and reach the
detector, as shown in . Fig. 4.16b for each histogram bin of 4.5 X-Ray Fluorescence
the silicon ϕ(ρz) distribution. The escape probability of
X-rays integrated over the complete ϕ(ρz) histogram gives As a consequence of photoelectric absorption shown in
the parameter designated “f(χ),” which is the overall escape . Fig. 4.17, the atom will subsequently undergo de-excitation
probability, between 0 and 1, for an X-ray generated any- following the same paths as is the case for electron ionization
where in the ϕ(ρz) distribution. in . Fig. 4.1. Thus, the primary X-ray spectrum of character-
. Figure 4.21b shows a sequence of calculations of the istic and continuum X-rays generated by the beam electron
C K ϕ(ρz) distribution and subsequent absorption as a func- inelastic scattering events gives rise to a secondary X-ray
tion of incident beam energy. As the incident beam energy spectrum of characteristic X-rays generated as a result of tar-
increases, the depth of electron penetration increases so that get atoms absorbing those characteristic and continuum
carbon characteristic X-rays are produced deeper in the tar- X-rays and emitting lower energy characteristic X-rays.
get. For pure carbon with E0 = 5 keV, the cumulative value of Because continuum X-rays are produced up to E0, the
f(χ) = 0.867; that is, 86.7 % of all carbon X-rays that are gener- Duane–Hunt limit, all atomic shells present with Ec < E0 will
ated escape, while 13.3 % are absorbed. As the C X-rays are be involved in generating secondary X-rays, which is referred
produced deeper with increasing beam energy, the total to as “secondary X-ray fluorescence” by the X-ray micro-
X-ray absorption increases so that the value of f(χ) for C K analysis community. Generally, at any characteristic photon
decreases sharply with increasing beam energy, as shown in energy the contribution of secondary fluorescence is only a
. Fig. 4.21b and in . Table 4.3. few percent or less of the intensity produced by the direct
Thus, with E0 = 2 keV, 97.4 % of the carbon X-rays escape electron ionization events. However, there is a substantial
the specimen, while at E0 = 30 keV, nearly 90 % of the carbon difference in the spatial distribution of the primary and sec-
X-rays generated in pure carbon are absorbed before they ondary X-rays. The primary X-rays must be produced within
can exit the specimen. the interaction volume of the beam electrons, which gener-
When the parameter f(χ) is plotted at every photon energy ally has limiting dimensions of a few micrometers at most.
from the threshold of 100 eV up to the Duane–Hunt limit of The secondary X-rays can be produced over a much larger
the incident beam energy E0, X-ray absorption is seen to volume because the range of X-rays in a material is typically
sharply modify the X-ray spectrum that is emitted from the an order-of-magnitude (or more) greater than the range of
target, as illustrated for carbon (. Fig. 4.22), copper an electron beam with E0 from 5 to 30 keV. This effect is
(. Fig. 4.23), and gold (. Fig. 4.24). The high relative intensity shown in . Fig. 4.25 for an alloy of Ni-10 % Fe for the second-
of the X-ray continuum at low photon energies compared to ary fluorescence of Fe K-shell X-rays (EK = 7.07 keV) by the
higher photon energies in the generated spectrum is greatly electron-excited Ni K-L2,3 X-rays (7.47 keV). The hemispher-
diminished in the emitted spectrum because of the higher ical volume that contains 99 % of the secondary Fe K-L2,3
absorption suffered by low energy photons. Discontinuities X-rays has a radius of 30 μm.
60 Chapter 4 · X-Rays
4 0.6
f(χ)
0.4
0.2
0.0
0 5 10 15 20
Photon energy (keV)
1.0
0.8
0.6
f(χ)
0.4
0.2
0.0
0 1 2 3 4 5
Photon energy (keV)
1e+7 E0 = 20 keV
1e+6
1e+5
Intensity
1e+4
1e+3
1e+2
Generated
1e+1 Emitted
1e+0
0 5 10 15 20
X-ray photon energy (keV)
4.5 · X-Ray Fluorescence
61 4
.. Fig. 4.23 a Absorption parameter f(χ) as a function
of photon energy for copper and an incident beam a Absorption parameter f(χ) for copper, E0 = 20 keV
energy of E0 = 20 keV. Note the abrupt decrease just
above the ionization energies of the three L-shells near
1.0
0.930 keV and the K-shell ionization energy at 8.98 keV.
b Comparison of the generated (red) and emitted (blue)
X-ray spectra for copper with an incident beam energy 0.8
of E0 = 20 keV. c Comparison of the generated and
emitted X-ray spectra for copper with an incident beam
energy of E0 = 20 keV; expanded to show the region of
0.6
the Cu L-shell and Cu K-shell ionization e nergies
f(χ)
0.4
0.2
0.0
0 5 10 15 20
Photon energy (keV)
b Copper, E0 = 20 keV
1e+8
1e+7
1e+6
Emitted
X-ray intensity
1e+5 Generated
1e+4
1e+3
1e+2
1e+1
1e+0
0 5 10 15 20
Photon energy (keV)
Copper, E0 = 20 keV
c 1e+7
1e+6
Emitted
X-ray intensity
1e+5 Generated
1e+4
1e+3
1e+2
0 2 4 6 8 10
Photon energy (keV)
62 Chapter 4 · X-Rays
f(c)
0.4
0.2
0.0
0 5 10 15 20
Photon energy (keV)
b Gold, E0 = 20 keV
1e+8
1e+7 Emitted
Generated
1e+6
1e+5
X-ray intensity
1e+4
1e+3
1e+2
1e+1
1e+0
0 5 10 15 20
Photon energy (keV)
Gold, E0 = 20 keV
c
1e+8
1e+7 Emitted
Generated
1e+6
X-ray intensity
1e+5
1e+4
1e+3
1e+2
0 1 2 3 4 5
Photon energy (keV)
References
63 4
Fiori CE, Swyt CR, Myklebust RL (1992) Desktop Spectrum Analyzer (DTSA), a
Ni-10% Fe comprehensive software engine for electron-excited X-ray spectrome-
E0 = 20 keV try. National Institute of Standards and Technology (NIST) – National
Institutes of Health (NIH). United States Patent 5,299,138 (March 29,
1994) (Standard Reference Data Program, NIST, Gaithersburg, MD
50%
20899)
90% Heinrich KFJ, Fiori CE, Myklebust RL (1979) Relative transition probabili-
10 µm ties for the X-ray lines from the K level. J Appl Phys 50:5589
Joy D (2006) Monte Carlo simulation. Available at 7 http://www.lehigh.
99%
edu/~maw3/link/mssoft/mcsim.html
Kramers H (1923) On the theory of X-ray absorption and the continuous
X-ray spectrum. Phil Mag 46:836
.. Fig. 4.25 Range of secondary fluorescence of Fe K-L3 (EK = 7.07 keV) Krause M, Oliver J (1979) Natural widths of atomic K and L levels, Kα
by Ni K-L3 X-rays (7.47 keV). Red arc = extent of direct electron excita- X-ray lines and several KLL Auger lines. J Phys Chem Ref Data
tion of Ni K-L3 and Fe K-L3. Blue arc = range for 50 % of total secondary Monogr 8:329
fluorescence of Fe K-L3 by Ni K-L3; green arc = 90 %; magenta arc = 99 % Lifshin E, Ciccarelli MF, Bolon RB (1980) In: Beaman DR, Ogilvie RE, Wittry
DB (eds) Proceedings 8th international conference on X-ray optics
and microanalysis, vol 141. Pendell, Midland
Pouchou JL, Pichoir F (1991) Quantitative analysis of homogeneous or
References stratified microvolumes applying the model ‘PAP’. In: Heinrich KFJ,
Newbury DE (eds) Electron probe quantitation, vol 31. Plenum
Crawford J, Cohen D, Doherty G, Atanacio A (2011) Calculated K, L and Press, New York
M-shell X-ray line intensities for light ion impact on selected targets Powell C (1976) Use of Monte Carlo Calculations. In: Heinrich KFJ, Newbury
from Z = 6 to 100. ANSTO/E-774, Australian Nuclear Science and DE, Yakowitz H (eds) Electron probe microanalysis and scanning elec-
Technology Organisation (September, 2011) tron microscopy. National Bureau of standard Special Publication
Drouin D, Couture A, Joly D, Tastet X, Aimez V, Gauvin R (2007) CASINO 460. U.S. Government Printing Office, Washington, DC, 97
V2.42 – a fast and easy-to-use modeling tool for scanning electron Ritchie NWM (2015) NIST DTSA-II software, including tutorials, is available
microscopy and microanalysis users. Scanning 29:92–101 free at: 7 www.cstl.nist.gov/div837/837.02/epq/dtsa2/index.html
65 5
References – 91
a dapt b dapt
b density = 204 pA/nm2
1 nA
A B A B
1 nA
2α W α W
5 nm 2.5 nm
S S
.. Fig 5.2 a Current density in a circular electron beam; b current
density if the beam diameter is reduced by a factor of two with the .. Fig. 5.3 a Definition of beam cone opening angle 2α; b definition
same current of beam convergence (half ) angle α
5.2 · Electron Optical Parameters
69 5
beam may not be physically realistic, but it is simple to It is no coincidence that the arc tangent of 0.005 is almost
understand and works well for our purposes here of under- exactly equal to 0.005 radians, since a well-known approxi-
standing basic beam parameters. mation in trigonometry is that
In geometry, the opening angle of a cone is defined as the
vertex angle ASB at the point of the cone, as shown in tan −1 q = q. (5.5)
. Fig. 5.3a. When working with the electron optics of an
SEM, by convention we use the term convergence angle to Since in every practical case encountered in SEM imaging the
describe how quickly the electron beam narrows to its focus angle will be sufficiently small to justify this approximation, we
as it travels down the optic axis. This convergence angle is can write our estimate of the convergence angle in a much sim-
shown in . Fig. 5.3b as α, which is half of the cone opening pler form that does not require any trigonometric functions,
angle. In some cases, the beam convergence angle is referred
rapt d apt
to as the convergence half-angle to emphasize that only half of α= , or α = (5.6)
the opening angle is intended. W 2W
Generally the numerical value of the beam convergence
angle in the SEM is quite small, and the electron beam cones As mentioned earlier, this angle is quite small, approximately
are much sharper and narrower than the cones used for sche- equal to 0.25° or about 17 arc minutes.
matic purposes in . Fig. 5.3. In fact, if you ground down and
reshaped the sides of a sewing needle so that it was a true 5.2.6 Beam Solid Angle
cone instead of a cylinder sharpened only at the tip, you
would then have a cone whose size and shape is reasonably In the previous section we defined the beam convergence
close to the dimensions found in the SEM. angle in terms of 2D geometry and characterized it by a
Estimating the value of the convergence angle of an planar angle measured in the dimensionless units of radi-
electron beam is not difficult using the triangles drawn in ans. However, the electron beam forms a 3D cone, not a 2D
. Fig. 5.3. The length of the vertical dashed line along the triangle, so in reality it subtends a solid angle. This is a con-
optical axis is called the working distance, usually denoted by cept used in 3D geometry to describe the angular spread of
the symbol W. It is merely the distance from the bottom of a converging (or diverging) flux. The usual symbol for solid
the objective lens pole piece (taken here to be approximately angle is Ω, and its units of measure are called steradians,
the same plane as the final aperture) to the point at which abbreviated sr. Usually when solid angles are discussed in the
the beam converges, which is typically also the surface of the context of the SEM they are used to describe the acceptance
sample if the sample is in focus. In practical SEM configura- angle of an X-ray spectrometer, or sometimes a backscat-
tions this distance can be as small as a fraction of a millime- tered electron detector, but they are also important in fully
ter or as large as tens of millimeters or a few centimeters, but describing the electron optical parameters of the primary
in most situations W will be between 1 mm and 5 mm or beam in the SEM as well as the properties of electron guns.
so. The diameter of the wide end of the cone, line segment . Figure 5.4 shows the conical electron beam in 3D,
AB, is the aperture diameter, dapt. This can also vary widely emerging from the circular beam-defining aperture at the top
depending on the SEM model and the choices made by the
operator, but it is certainly no larger than a fraction of a mil-
limeter and can be much smaller, on the order of microm-
eters. For purposes of concreteness, let’s assume W is 5 mm dapt
(i.e., 5000 μm) and the aperture is 50 μm in diameter (25 μm aapt
in radius, denoted rapt).
From . Fig. 5.3b we can see that triangle ASB is com-
posed of two back-to-back right triangles. The rightmost of Ω
these has its vertex angle labeled α. The leg of that triangle
adjacent to α is the working distance W, the opposite leg is
the aperture radius rapt, and the hypotenuse of the right tri-
angle is the slant length of the beam cone, SB. From basic
trigonometry we know that the tangent of the angle is equal
to the length of its opposite leg divided by its adjacent leg, or
rapt
tan α = , (5.4) W
W
rapt −1 25 mm
a = tan −1 = tan .. Fig. 5.4 Definition of beam solid angle, Ω. The vertical dashed line
W 5000 mm represents the optical axis of the SEM, and the distance from the aperture
= tan −1 ( 0.005 ) = 0.005 radians = 5 mrad. plane to the beam impact point is the working distance, W. This is also
the radius of the imaginary hemisphere used to visualize the solid angle
70 Chapter 5 · Scanning Electron Microscope (SEM) Instrumentation
of the figure and converging on the sample surface. The Unless you work with solid angle calculations on a regular
diameter of that aperture is dapt, and the area of the aperture basis, this value probably has little physical meaning to you,
is aapt. As discussed earlier, the distance the beam travels ver- and you have no sense of how big or how small 78 microste-
tically from the final aperture to the point where it is focused radians are in real life. To provide some perspective, consider
to a spot is the working distance, W. Now imagine a complete that both the Moon and the Sun subtend about this same
sphere centered on the beam impact point, and with spheri- solid angle when viewed from the surface of the Earth using
cal radius equal to W. The upper hemisphere of this imagi- the naked eye. The exact angular diameters (and therefore
nary dome is depicted in . Fig. 5.4 as well, and since its also the solid angles) of both the Sun and the Moon vary
radius is W, the beam-defining aperture will lie on the sur- slightly during their orbits, depending on how far away they
face of this sphere. are at any given moment, but this variation is small and oscil-
5 The key to understanding the meaning of solid angles and lates around average values:
their numerical measure using units of steradians is to con-
2α Sun = 9.35 mrad, α Sun = 4.68 mrad
sider such a complete sphere and the fractional surface area
of that sphere that is occupied by the object of interest. Every 2α Moon = 9.22 mrad, α Moon = 4.61 mrad
complete sphere, regardless of diameter, subtends exactly 4π
steradians of solid angle. It follows that every hemisphere Ω Sun = 68.7 µ sr
represents a solid angle of 2π steradians, no matter how small
or how large the hemisphere might measure in meters. On Ω Moon = 66.7 µ sr.
the other hand, the surface area of a sphere Asphere most
certainly depends on the radius r, and can be calculated using Of course the Sun is much, much larger than the Moon in
the ancient formula diameter, but it is also much farther away, so the two celestial
bodies appear to be about the same angular size from the per-
Asphere = 4π r 2 . (5.7) spective of the Earthbound viewer. This similarity in angular
size is a coincidence, and it is the reason that during a solar
For the imaginary sphere and electron beam aperture shown eclipse that the Moon almost perfectly occludes the Sun for a
in . Fig. 5.4, we can assume realistic numbers for this calcu- short time. This analogy is instructive for the SEM operator
lation by adopting the values used in the beam convergence because it helps explain how a small final beam aperture
angle discussion above: W = 5000 μm, dapt = 50 μm, and combined with a short working distance can produce the
rapt = 25 μm. With these values we can calculate the surface same convergence angle (and therefore depth of field) as a
area of the complete sphere as configuration that uses a large aperture and a long working
distance. Likewise, an energy-dispersive X-ray spectrometer
Asphere = 4π W 2 = 4π ( 5000 µ m ) = 3.14 ×108 mm 2 . (5.8)
2
(EDS detector) with a small area of 10 mm2 can subtend the
same solid angle (and therefore collect the same number of
and we can calculate the area of the beam-defining aperture X-rays) as a much larger 100-mm2 detector sitting at a more
as distant detector-to-sample position.
length) causes the beam to converge to focus before the elec- Astigmatism
trons reach the surface of the sample. Since the electrons then The pointy cones drawn in . Fig. 5.5 are a useful fiction for
begin to diverge from this crossover point, the beam has representing the large-scale behavior of a focused electron
broadened beyond its narrowest waist and is wider than opti- beam, but if we consider the beam shape carefully near the
mal when it strikes the surface of the sample, thus producing beam crossover point this conical model of the beam breaks
an out-of-focus image. Conversely, the right-hand portion of down. . Figure 5.6 is a more realistic picture of the beam
. Fig. 5.5 shows the beam behavior in an underfocused con- shape as it converges to its narrowest point and then begins
dition. Here the magnetic field is too weak, and the beam is to diverge again below that plane. For a variety of reasons,
not fully brought to crossover before it strikes the surface of mostly the effects of lens aberrations and other imperfec-
the sample, and again the beam diameter is broader than tions, even at its narrowest point the beam retains a finite
optimal, resulting in an out-of-focus image. beam diameter. In other words, it can never be focused to a
Using these schematics as a guide, it is easier to under- perfect geometrical point. The left side of . Fig. 5.6 shows the
stand what is happening electron optically when the SEM beam narrowing gently but never reaching a sharp point,
user focuses the image. Changes in the focus control result reflecting this reality. Ideally, cross sections through the beam
in changes in the electrical current in the objective lens, at different heights will all be circles, as shown in the right of
which results in raising or lowering the crossover of the elec- . Fig. 5.6. If the beam is underfocused or overfocused, as
tron beam relative to the surface of the sample. The distance shown in . Fig. 5.5, the consequence is a blurry image caused
between the objective lens exit aperture and this beam cross- by the larger-diameter beam (larger blue circles in . Fig. 5.6
over point is displayed on the microscope as the working above and below the narrowest point).
distance, W. On most microscopes you can see the working In real SEMs the magnetic fields created in the electron
distance change numerically on the screen as you make optics are never perfectly symmetric. Although the
gross changes in the focus setting, reflecting this vertical manufacturers strive for ideal circular symmetry in round
motion of the beam crossover in the SEM chamber. It is lenses, invariably there are defects in the lens yoke, the elec-
important to note that the term working distance is also used trical windings, the machining of the pole pieces, or other
by some microscopists when referring to the distance problems that lead to asymmetries in the lens field and ulti-
between the objective lens pole piece and the surface of the mately to distortions in the electron beam. Dirt or contami-
sample. The value of W displayed on the microscope will nation buildup on the apertures in the microscope can also
accurately reflect this lens-to-sample distance if the sample be an important source of distorted beam shapes. Since the
is in focus. dirt on the aperture is electrically non-conductive, it can
5.2 · Electron Optical Parameters
73 5
accumulate an electric charge when the beam electrons strike cross section changes from an elongated ellipse with long axis
it, and the resulting electrostatic fields can warp and bend the in the y-direction, to a circle (albeit with a larger diameter
beam into odd and complex shapes that no longer have a cir- than the equivalent circle in . Fig. 5.6), and then to another
cular cross section. elongated ellipse, but this time with its long axis oriented in
By far the most important of these distortions is called the x-direction. This progression from a near-line-focus to a
two-fold astigmatism, which in practice is often referred to broader circular focus and then to a near-line-focus in an
just as astigmatism. In this specific distortion the magnetic orthogonal direction is the hallmark of a beam exhibiting
field that focuses the electrons is stronger in one direction astigmatism.
than in the orthogonal direction, resulting in a beam with an This behavior is also easily visible in the images produced
elliptical cross section instead of a circular one. In beams by rastering the beam on a sample. When the beam cross sec-
exhibiting astigmatism the electrons come to closest focus in tion is highly elongated at the surface of the sample, the
the x-direction at a different height than the y-direction, con- image resolution is degraded badly in one direction, produc-
sistent with the formation of elliptical cross sections. These ing a blurring effect with pronounced linear asymmetry. It
effects are shown schematically in . Fig. 5.7. Similar to appears as if the image detail is sheared or stretched in one
. Fig. 5.6, the focused beam is shown in perspective on the direction but not the other. If the focus knob is adjusted when
left side of the diagram, while a series of cross sections of the the beam is astigmatic, a point can be reached when this
beam are shown on the right of the figure. In the case shown image shearing or linear asymmetry is eliminated or at least
here, as the beam moves down the optical axis of the SEM, the greatly reduced. This is the best focus obtainable without
74 Chapter 5 · Scanning Electron Microscope (SEM) Instrumentation
a a
b b
c c
.. Fig. 5.8 Three SEM micrographs showing strong astigmatism in the X .. Fig. 5.9 Three additional SEM micrographs from the same field of
direction, when the sample is a overfocused, b underfocused, and c near view shown in Fig. 5.8 above. Here the beam shows strong astigmatism
focus. Note that the shearing or “tearing” appearance of fine detail in a in the Y-direction, when the sample is a overfocused, b underfocused,
appears to be in a direction perpendicular to the effect in b and c near focus
correcting the astigmatism, although it generally produces when the objective lens is overfocused, with the beam
an image that is far inferior to the in-focus images obtainable crossover occurring above the sample surface, corresponding
with a properly stigmated beam. If the focus is further to the left diagram in . Fig. 5.7. In panel . Fig. 5.8b, the same
adjusted, past this point of symmetry, the image will again sample with the same astigmatic beam is shown in underfo-
exhibit large amounts of shearing and stretching of the fine cus, the right side diagram in . Fig. 5.6. . Figure 5.8c shows
details, but in a different direction. This sequence of effects the best achievable focus; here, the shearing and stretching is
can be seen in . Fig. 5.8. In . Fig. 5.8a the sample is shown minimized (or at least balanced), suggesting the cross section
5.3 · SEM Imaging Modes
75 5
SEMs are very flexible instruments, and the SEM operator has
control over a large number of electron beam, detector, and
stage parameters. Consequently, the number of different imag-
ing conditions that may be employed to analyze any given sam-
ple is nearly infinite, and it is the job of the analyst to choose
these conditions wisely to obtain useful information to meet
the needs of the analysis. Fortunately, in many situations these
choices can be narrowed to using one of the four basic modes of
c
SEM operation: (1) High Depth-of-Field Mode, (2) Resolution
Mode, (3) High-Current Mode, and (4) Low-Voltage Mode.
Below you will find practical information on how to con-
trol the fundamental electron optical parameters described
earlier in the text and specific guidance for operating the
SEM in the four basic modes just mentioned. Experienced
SEM operators will have mastered these four modes and will
be comfortable moving between them as needed. Choosing
any one of these modes is a compromise, since each of them
sacrifices microscope performance in some areas to achieve
other imaging goals. Appreciating the strengths and weak-
nesses of each mode is essential to understanding when each
mode is warranted. Of course some analyses will demand
imaging conditions that do not fall neatly into one of these
four basic modes, and the expert SEM operator will use the
.. Fig. 5.10 Three additional SEM micrographs from the same field of
full flexibility of the instrument when required.
view seen in . Figs. 5.8 and 5.9 above. Here the sample is imaged with
a fully corrected beam, so neither the overfocused image a nor the
underfocused image b shows significant anisotropic fine detail. Fur- 5.3.1 High Depth-of-Field Mode
ther, the in-focus image in panel c is much sharper then the best-focus
images obtained in panel c of . Figs. 5.8 and 5.9 Anyone familiar with compound light microscopes (LMs)
understands that they have a very limited Depth-of-Field
of the electron beam is approximately circular. . Figure 5.9 (DoF), meaning there is a limited range of vertical heights on
shows the sample field of view seen in . Fig. 5.8, but imaged the sample surface that will all appear to be in focus simul-
using a beam with pronounced astigmatism in the Y direc- taneously. Parts of the sample that fall outside this range
tion. In general the SEM beam will be astigmatic in both X appear blurry. One of the advantages of the SEM over the
and Y, and the operator must correct this beam distortion light microscope is that it is capable of a much deeper depth-
along both axes using the X and Y stigmators. When this is of-field than the LM.
76 Chapter 5 · Scanning Electron Microscope (SEM) Instrumentation
Scan Scan
Plane of
optimum Df
focus
The shallow depth-of-field in the LM arises from the strikes the surface of the sample, at a height denoted by the
properties of its glass lenses, but SEMs don’t use lenses to upper red dotted line. This is equivalent to underfocusing
form direct images; instead they rely on lenses to focus the the beam, with the same effect: the diameter of the probe at
beam and then scan this beam from pixel to pixel to image the sample surface is larger than optimal. If this increase in
the sample.1 Nonetheless, they suffer from limited DoF probe size is large enough, it will degrade the sharpness of
because of the effect shown schematically in . Fig. 5.11. the image. The height at which this blurring becomes mea-
Here the electron beam is shown striking the sample in sureable, denoted by the upper red dotted line, is the upper
three different locations, producing three different pixels in limit of the DoF for this beam. Similarly, for pixel 3 the
the image. For all three pixels the vertical position of the sample surface is lower (i.e., further from the objective lens)
electron beam crossover is the same; this height is called the than the middle case. This is analogous to overfocus because
plane of optimum focus and is represented in . Fig. 5.11 as the beam comes to crossover and begins to diverge again
a horizontal green dot-dashed line. For the case of pixel 2, before it lands on the sample. As before, this can degrade the
this plane coincides with the surface of the sample. For pixel sharpness of the image, and the height at which this degra-
1, the electron beam has not yet reached crossover when it dation is noticeable is the lower limit of the height range that
defines the DoF. The distance between these two dotted red
lines is labeled Df in . Fig. 5.11, denoting the depth-of-field.
1 Glass lenses and transmission electron microscope lenses also Because the definition of DoF requires the resulting blur
have a related property known as depth-of-focus, a term that is
often confused with depth-of-field. Depth-of-field refers to the
to be noticeable (or at least measureable), it depends on many
range of heights in simultaneous focus on the sample (i.e., the factors and can be somewhat subjective. For example, since
observed field). In contrast, depth-of-focus refers to the range of in most cases sub-pixel blurring is not a concern to the SEM
positions near the imaging plane of the lens where the image is in analyst, the effective DoF will often improve as the magnifi-
focus. This determines, for example, how far away from the ideal cation decreases and the pixel size increases. However, as the
imaging plane of the lens you can place a piece of film, or a CCD
detector, and still capture an in-focus image. Because SEMs
magnification decreases, much more of the sample is visible
capture images via scanning action, the term depth-of-focus is not in the field of view, increasing the chances that pronounced
relevant. topography will lead to blurring. In these cases the DoF is
5.3 · SEM Imaging Modes
77 5
.. Fig. 5.12 a Diagram of the
electron beam emerging from the a b
final aperture in the objective
lens and striking the sample
under typical imaging conditions;
a relatively large aperture diam-
eter and short working distance
create a large convergence angle
and therefore a shallow depth-of-
field. b High Depth-of-Field Mode.
Here a small aperture diameter
W
and long working distance W
combine to create a small conver-
gence angle and therefore a large
depth-of-field W
still increased at low magnification, but parts of the sample joystick, or stand-alone stage controller, especially on older
are still blurred in the image because of the large range of microscopes. Recent models typically use a graphical user
height visible in the expanded field of view. Nonetheless, interface, requiring the operator to enter a destination height
operating the SEM in High Depth-of-Field Mode at medium (or “Z position”) in millimeters and then executing the move.
to low magnifications is perhaps the most often used imaging Some also allow the stage height to be changed continuously
condition for routine SEM work. using the mouse.
The basic idea behind High Depth-of-Field Mode is to Depth-of-Field Mode is also optimized by selecting a
create a set of imaging conditions where the convergence relatively small final beam aperture. The mechanisms used
angle of the beam is small, producing a narrow pencil-like to change the diameter of the final aperture, and to center it
electron beam that does not change diameter rapidly with on the optical axis of the microscope, vary widely from one
height above the sample. . Figure 5.12 shows what this looks SEM model to the next. In fact, some SEMs are designed to
like schematically. . Figure 5.12a represents typical imaging use a fixed or semi-fixed final aperture and do not provide an
conditions, with short working distance W and normal aper- easy method of altering the aperture size. Many microscopes
ture diameter. . Figure 5.12b shows the imaging conditions have manual aperture controls mounted on the outside of
used in High Depth-of-Field Mode, where the working dis- the SEM column (. Fig. 5.13). Other microscopes use a
tance has been increased significantly and a smaller diameter graphical user interface (GUI) to allow the operator to select
aperture is inserted. These two changes decrease the conver- one of several available apertures for insertion. Following
gence angle of the electron beam and therefore increase the this selection, motors driven by an X/Y- motion controller
DoF. The effects of the aperture and working distance are physically move the selected aperture into place and recall
independent of each other, meaning either one can improve from memory the X- and Y- positions needed to center it. In
the depth-of-field by itself. either case the apertures themselves are arrayed linearly as a
For best results in Depth-of-Field Mode, determine the series of circular holes in a long, thin aperture strip.
lowest stage position available (largest working distance), A few microscopes permit you to configure Depth-of-
and drive the sample to that location. Changing the working Field Mode directly by selecting this option in the instrument
distance is straightforward on most SEMs. Those micro- control software. . Figure 5.14 shows an example screenshot
scopes with a manual stage will often have a physical knob on from one manufacturer’s user interface where the operator
the chamber door for changing the height of the sample. can select a dedicated “DEPTH” setting, automatically opti-
Motorized stages are sometimes controlled by a hand panel, mizing the instrument for a small convergence angle.
78 Chapter 5 · Scanning Electron Microscope (SEM) Instrumentation
a a
PROBE CURRENT
COARSE/FINE
FINE
b
b
PROBE CURRENT
COARSE FINE
a b
Azimuthal angle, ζ
Elevation angle, ψ
5 Arbitrary fixed
reference line
within specimen
chamber, e.g.,
parallel to front
face
.. Fig. 5.19 a Detector take-off angle, ψ. b Detector azimuthal angle around beam, ζ
Objective Objective
Lens Lens
Semiconductor detector
Ω Ω
ψ A B
5
C D
Scintillator To photomultiplier
Quadrant detector,
Ω Ω bottom view
ψ
.. Fig. 5.23 Semiconductor annular detector, quadrant design with
four separately selectable sections
under the objective lens where it will not interfere with other
detectors. The size and proximity to the specimen provide
a large solid angle and a high take-off angle. As shown in
.. Fig. 5.22 Large solid-angle passive BSE detector . Fig. 5.23, the semiconductor detector can also be assem-
bled from segments, each of which can be used as a separate
detector that provides a selectable apparent illumination of
kAdjustable Controls the SEM image, or the signals from any combination of the
The Wells–Robinson scintillation BSE detector is often detectors can be added. Semiconductor detectors can also
mounted on an externally controlled, motorized retractable be placed at various locations around the specimen, simi-
arm. In typical use the detector would be fully inserted to lar to the arrangement shown for scintillator detectors in
maximize the solid angle. A partial insertion that does not . Fig. 5.21.
interrupt the beam access to the specimen can be used to The semiconductor BSE detector has an energy threshold
intentionally provide an asymmetric detector placement to typically in the range 1 keV to 3 keV because of energy loss
give an apparent illumination from one side. suffered by the BSE during penetration through the entrance
surface electrode. Above this threshold, the response of the
Semiconductor BSE Detectors detector increases linearly with increasing electron energy, thus
Certain semiconductor devices can detect energetic elec- providing a greater gain from the high energy fraction of BSE.
trons that penetrate into the active region of the device where
they undergo inelastic scattering. One product of this energy kAdjustable Controls
deposition in the semiconductor is the promotion of loosely The semiconductor BSE detector does not have any user-
bound valence shell electrons (each leaving behind a vacancy adjustable parameters, with the exception of the choice of
or positively-charged “hole”) into the empty conduction the individual components of a composite multi-detector.
band where they can freely move through the semiconduc- In some systems, the individual quadrants or halves can be
tor in response to an applied potential bias. By applying a selected in various combinations, or the sum of all detectors
suitable electrical field, these free electrons can be collected can be used. Some SEMs add an additional semiconductor
at a surface electrode and measured. For silicon, this process detector that is placed asymmetrically away from the electron
requires 3.6 eV of energy loss per free electron generated, so beam to enhance the effect of apparent oblique illumination.
that a 15-keV BSE will generate about 4000 free electrons.
Thus a BSE current of 1 nA entering the detector will cre-
ate a collected current of about 4 μA as input for the next 5.4.4 Secondary Electron Detectors
amplification stage. The collection electrodes are located on
the entrance and back surfaces of the planar wafer detector, Everhart–Thornley Detector
which is shown in a typical mounting as an annular detec- The most commonly used SEM detector is the Everhart–
tor in . Fig. 5.23. The semiconductor BSE detector has the Thornley (E–T) detector, almost universally referred to as
advantage of being thin, so that it can be readily mounted the “secondary electron detector.” Everhart and Thornley
5.4 · Electron Detectors
87 5
Scintillator Chamber
Light guide Wall
+10 kV
Faraday +10 kV
cage
-50 V to Ω SE3
+300 V SE3
+300 V SE3
ψ Indirect BSE
Direct BSE SE1
SE2 Indirect BSE
iB
iSE iBSE
TTL iB
E-T
BSE
SE1
SE2
5 BSE
SE3 iSE
iSC iBSE
iSC
Picoammeter
.. Fig. 5.27 Currents flowing in and out of the specimen and the
electrical junction equivalent
complete suite of SE1, SE2, SE3, and the direct BSE is collected,
creating a complex mix of BSE and true SE image contrast
effects. 5.4.5 pecimen Current: The Specimen as Its
S
Own Detector
Through-the-Lens (TTL) Electron Detectors
zz The Specimen Can Serve as a Perfect Detector for the
TTL SE Detector Total Number of BSE and SE Emitted
In SEMs where the magnetic field of the objective lens Consider the interaction of the beam electrons to produce
projects into the specimen chamber, a “through-the-lens” BSE and SE. For a 20-keV beam incident on copper, about 30
(TTL) secondary electron detector can be implemented, as out of 100 beam electrons are backscattered (η = 0.3). The
illustrated schematically in . Fig. 5.26. SE1 from the inci- remaining 70 beam electrons lose all their energy in the solid,
dent beam footprint and SE2 emitted within the BSE sur- are reduced to thermal energies, and are captured.
face distribution are captured by the magnetic field and Additionally, about 10 units of charge are ejected from cop-
spiral up through the lens. After emerging from the top of per as secondary electrons (δ = 0.1). This leaves a total of 60
the lens, the secondary electrons are then attracted to an excess electrons in the target. What is the fate of these elec-
Everhart–Thornley type biased scintillator detector. The trons? To understand this, an alternative view is to consider
advantage of the TTL SE detector is the near complete the electron currents, defined as charge per unit time, which
exclusion of direct BSE and the abundant SE3 class gener- flow in and out of the specimen. Viewed in this fashion, the
ated by BSE striking the chamber walls and pole piece. specimen can be treated as an electrical junction, as illus-
Since these remote SE3 are generated on surfaces far from trated schematically in . Fig. 5.27, and is subject to the fun-
the optic axis of the SEM, they are not efficiently captured damental rules which govern junctions in circuits. By
by the lens field. Because the SE3 class actually represents Thevinin’s junction theorem, the currents flowing in and out
low resolution BSE information, removing SE3 from the of the junction must exactly balance, or else there will be net
overall SE signal actually improves the sensitivity of the accumulation or loss of electrical charge, and the specimen
image to the true SE1 component, which is still diluted by will charge on a macroscopic scale. If the specimen is a con-
the BSE-related SE2 component. A further refinement of ductor or semiconductor and if there is a path to ground
the through-the-lens detector is the introduction of energy from the specimen, then electrical neutrality will be main-
filtering which allows the microscopist to select a band of tained by the flow of a current, designated the “specimen cur-
SE kinetic energy. rent” (also referred to as the “target current” or “absorbed
current”), either to or from ground, depending on the exact
TTL BSE Detector conditions of beam energy and specimen composition. What
For a flat specimen oriented normal to the beam, the cosine is the magnitude of the specimen current?
distribution of BSE creates a significant flux of BSE that pass Considering the specimen as a junction, the current flow-
up through the bore of the objective lens. A TTL BSE detec- ing into the junction is the beam current, iB, and the cur-
tor is created by providing either a direct scintillation BSE rents flowing out of the junction are the backscattered
detector or a separate surface above the lens for BSE-to-SE electron current, iBS, and the secondary electron current,
conversion and subsequent detection with another E-T type iSE. For charge balance to occur, the specimen current, iSC,
detector. is given by
5.4 · Electron Detectors
89 5
iSC = iB − iBS − iSC (5.15) et al. 1974). Such amplifiers can operate with currents as low
as 10 pA and still provide adequate bandwidth to view
For the copper target, the BSE current will be iBSE = η×iB = 0.3 acceptable images at slow visual scan rates (one 500-line
iB and the SE current will be iSE = δ×iB = 0.1 iB. Substituting frame/s).
these values in Eq. (5.15) gives the result that the specimen
current will be iSC = 0.6 iB, double the largest of the conven-
5.4.6 Useful, Practical Measure of a
A
tional emitted imaging currents, the BSE signal. If a path to
ground is not provided so that the specimen current can flow,
Detector: Detective Quantum
the specimen will rapidly charge. Efficiency
Note that in formulating Eq. (5.15) no consideration is
given to the large difference in energy carried by the BSE and The geometric efficiency is just one factor in the overall per-
SE. Since current is the passage of charge per unit time, the formance of a detector, and while this quantity is relatively
ejection of a 1 eV SE from the specimen carries the same straightforward to define in the case of a passive BSE detec-
weight as a 10 keV BSE in affecting the specimen current sig- tor, as shown in . Fig. 5.20, it is much more difficult to
nal. Moreover, the specimen current is not sensitive to the describe for an E–T detector because of the mix of BSE and
direction of emission of BSE and SE, or to their subsequent direct SE1 and SE2 signal components and the complex con-
fate in the SEM specimen chamber, as long as they do not version and collection of the remote SE3 component pro-
return to the specimen as a result of re-scattering. Thus, duced where BSE strike the objective lens, BSE detector, and
specimen current constitutes a signal that is sensitive only to chamber walls. A second important factor in detector perfor-
number effects, that is, the total numbers of BSE and SE leav- mance is the efficiency with which each collected electron is
ing the specimen. converted into useful signal. Thirdly, noise may be intro-
The specimen serves as its own collector for the specimen duced at various stages in the amplification process to the
current. As such, the specimen current signal is readily avail- digitization which creates the final intensity recorded in the
able just by insulating the specimen from electrical ground computer memory for the pixel at which the beam dwells.
and then measuring the specimen current flowing to ground All of these factors are taken into account by the detective
through a wire to ground. Knowledge of the actual specimen quantum efficiency (DQE). The DQE is a robust measure of
current is extremely useful for establishing consistent operat- detector performance that can be used in the calculation of
ing conditions, and is critical for dose-based X-ray micro- limitations imposed on imaging through the threshold cur-
analysis. The original beam current itself be measured by rent/contrast equation (Joy et al. 1996).
creating a “Faraday cup” in the specimen or specimen stage The DQE is defined as (Jones 1959)
by drilling a blind hole and directing the incident beam into
DQE = ( S / N ) / (S / N )
2 2
(5.16)
the hole: since no BSE or SE can escape the Faraday cup, the experimental theoretictal
Gray level
150
factorily during the sequence of measurements required. As
an alternative to silicon, a metallographically polished (but
not etched) pure metallic element (metallic) surface, such as 100
nickel, molybdenum, gold, etc., will be suitable. Because cal-
5 culation of the theoretical S/N ratio is required for the DQE 50
calculation with Eq. 5.16, the beam current must be accurately Y-Intercept = 36
measured. The SEM must thus be equipped with a picoam- 0
meter to measure the beam current, and if an in-column 0 2 4 6 8 10
Beam current (nA)
Faraday cup is not available, then a specimen stage Faraday
cup (e.g., a blind hole covered with a small [e.g.,
<100-μm-diameter] aperture) is required to completely cap- .. Fig. 5.28 Plot of measured gray level versus incident beam current
ture the beam without loss of BSE or SE so that a measure- for a BSE detector. E0 = 20 keV; Mo target
ment of the specimen current equals the beam current.
Because the detector will have a “dark current,” i.e., a
response with no beam current, it is necessary to make a
series of measurements with changing beam current. It is
also important to defeat any automatic image gain scaling
that some SEMs provide as a “convenience” feature for the
user that acts to automatically compensate for changes in the
incident beam current by adjusting the imaging amplifier
gain to maintain a steady mid-range gray level.
Measurement sequence
1. Choose a beam current which will serve as the high end
of the beam current range, and using the image histo-
gram function, adjust the imaging amplifier controls
(often designated “contrast” and “brightness”) to place
the average gray level of the specimen near the top of the 0 255
range, being careful that the upper tail of the gray level
Count: 757008 Min: 129
distribution of the image of the specimen does not
Mean: 134.117 Max: 137
saturate (“clip”) at the maximum gray level (255 for an
StdDev: 0.806 Mode: 134 (355350)
8-bit image, 65,535 for a 16-bit image).
2. Keeping the same values for the image amplifier
parameters (autoscaling of the imaging amplifier must .. Fig. 5.29 Output of image histogram from IMAGE-J for the 4 nA
be defeated before beginning the measurement process), image from Fig. 5.28
choose a beam current that places the average gray level
of the specimen near low end of the gray level range,
checking to see that the gray level distribution of the 5. Choose an image recorded within this range and
image is completely within the histogram range—that is, determine the mean gray level, Gmean, and the variance,
there is no clipping of the distribution at the bottom Svar (the square of the standard deviation) using the
(black) of the range. image histogram function, as shown in . Fig. 5.29.
3. With the minimum and maximum of the current range
established, record a sequence of images with different Calculation sequence:
beam currents between the low and high values and use
the image histogram tool to determine the average gray ( S / N )experimental = ( GLmean − GLDC ) / Svar (5.17)
level for each beam current.
4. A graphical plot of data measured with a semiconductor where is Svar the variance (the square of the standard devia-
BSE detector for a polished Mo target produces the tion) of the gray level distribution. For the values in . Fig. 5.29
result illustrated in . Fig. 5.28, where the y-axis intercept for the 4 nA data point obtained with ImageJ-Fiji,
value is a measure of the dark current gray level inten-
sity, GLDC (corresponding to zero beam current) of this ( S / N )experimental = (134.3 − 36 ) / 0.5752 = 297.3 (5.18)
particular BSE detector.
References
91 5
The corresponding theoretical S/N ratio is estimated from
.. Table 5.1 DQE of electron detectors from different
the number n of BSE produced, which depends on the inci- manufacturers (Joy et al. 1996)
dent beam current IB, the BSE coefficient η, and the dwell
time per pixel τ: SE detector DQE
Everhart–Thornley 0.17
where the coefficient 6.24 is appropriate for beam current
expressed in pA and the dwell time expressed in μs. Everhart–Thornley 0.12
Because the image pixels are independent and uncorre- Everhart–Thornley 0.017
lated, if a mean number n of BSE is produced at each pixel the
Everhart–Thornley 0.0008
expected variance is n1/2:
High performance SEM:
( S / N )theory = n / n = ( 6.24 I B ητ )
1/ 2
1/ 2
=n 1/ 2
(5.20) Everhart–Thornley (lower) 0.18
( S / N )theory = ( 6.24 I B ητ )
1/ 2
= 779.1 (5.21) BSE detector
DQE = ( S / N ) / (S / N )
2 2
E–T BSE mode (negative bias) 0.001
experimental theoretictal
.. Fig. 5.30 Average gray level versus beam current for an Everhart–
Thornley detector on an electron probe microanalyzer. Specimen: Si;
E0 = 10 keV
93 6
Image Formation
6.1 Image Construction by Scanning Action – 94
6.2 Magnification – 95
6.2.1 Magnification, Image Dimensions, and Scale Bars – 95
References – 110
6.1 I mage Construction by Scanning At any particular time, there is only one ray path (solid line)
Action through the scanning system and the beam reaches only one
location on the specimen, for example, position 3 in . Fig. 6.1.
After leaving the electron source, the beam follows the central The SEM image is a geometric construction created under
(optic) axis of the lens system and is sequentially defined by computer control by addressing the focused beam to a sequence
apertures and focused by the magnetic and/or electrostatic of discrete x-y locations on the specimen and measuring the
fields of the lens system. Within the final (objective) lens, a sys- effect of the interaction of the beam with the specimen at each
tem of scan coils acts to displace the beam off the optic axis so location. For a single gray-scale SEM image, this interaction
that it can be addressed to a location on the specimen, as illus- could be the output from a single electron detector, such as the
trated schematically for single deflection scanning in . Fig. 6.1. Everhart–Thornley detector. It is also possible to measure the
output from more than one detector simultaneously while the
beam is addressed to a single x-y location. When this is done,
multiple gray-scale SEM images are built up at the same time
6 during the scan. It is essential to realize that even when these
multiple signals are being collected simultaneously and multi-
ple images are produced, only a single scan is needed; the par-
allel nature of the acquisition arises from parallel detection, not
parallel scanning. Note that no “true image” actually exists
Scan coils within the SEM in the same sense as the image created in a light
optical microscope, where actual ray paths extend from each
point on the specimen through the lens system to a corre-
sponding point on the image recording medium, whether that
is the eye of a human viewer or the positionally sensitive detec-
tor of a digital camera. In the SEM, at each location sampled by
the incident electron beam, each signal is measured with an
appropriate detector and the analog measurement is converted
to an equivalent digital value (using an analog-to-digital con-
verter, ADC). The beam x-y location and the intensity(ies) Ij of
the signal(s) of interest generate a digital stream of data packets
(x, y, Ij), where the index j represents the various signals avail-
able: backscattered electrons (BSE), secondary electrons (SE),
1 2 3 4 5 6 7 8 9 absorbed current, X-rays, cathodoluminescence, etc.
A simple description of this “scanning action” to create an
image is shown schematically in . Fig. 6.2, where an area
with equal edge dimensions l being scanned on the specimen
.. Fig. 6.1 Scanning action to produce a sequence of discrete beam
locations on the specimen is effectively divided into an x-y grid of square picture ele-
NIST
10 μm
NIST
3
4 μm
4
++
++
1500 μm
1000 μm
750 μm
6 500 μm
250 μm
100 μm
50 μm
NIST
1
NIST
1500 μm
NIST
b
NIST
3
A: 200 nm pitch
G F E D C B A B: 280 nm pitch
4
C: 400 nm pitch
G
D: 500 nm pitch
F
E: 700 nm pitch
E
F: 1 µm pitch
C
G: 2 µm pitch
D
D
C
E
F
B
with 1 µm crosses on
G
A B C D E F G non-connected structures
and with a 1 µm grid on
2
11 connected structures
NIST
NIST
.. Fig. 6.4 a Scale calibration artifact Reference Material 8820 (National Institute of Standards and Technology, U.S.) (From Postek et al. 2014).
b Detail within the feature noted in . Fig. 6.4a (From Postek et al. 2014)
Technology [USA]), shown in . Fig. 6.4. This scale calibra- magnification of 10× on a 10 x 10-cm display). Scanned fields
tion artifact consists of an elaborate collection of linear fea- as small as 1 × 1 μm (e.g., a nominal magnification of
tures produced by lithography on a silicon substrate. It is 100,000×) can be calibrated with the series of structures with
important to calibrate the SEM over the full range of magnifi- pitches of various repeat distances shown in . Fig. 6.4b. The
cations to be used for subsequent work. RM 8820 contains structures present in RM 8820 enable simultaneous calibra-
large-scale structures suitable for low and intermediate mag- tion along the x- and y-axes of the image so that image distor-
nifications, for example, a span of 1500 μm (1.5 mm) as indi- tion can be minimized. Accurate calibration in orthogonal
cated by the red arrows in . Fig. 6.4a, that permit calibration directions is critical for establishing “square pixels” in order to
of scan fields ranging up to 1 × 1 cm (e.g., a nominal avoid introducing serious distortions into the scanned image.
6.3 · Making Dimensional Measurements With the SEM: How Big Is That Feature?
97 6
.. Fig. 6.5 a Careful calibration
a
of the x- and y-scans produces
square pixels, and a faithful repro-
duction of shapes lying in the
scan plane perpendicular to the
optic axis. b Distortion in the dis-
play of an object caused by non-
square pixels in the image scan
l
Beam locations on specimen
and specimen pixels L
l'
L
l
With square pixels, the shape of an object is faithfully trans- functions. As shown in . Fig. 6.6a, starting with an image
ferred, as shown in . Fig. 6.5a, while non-square pixels in the of a primary or secondary calibration artifact (i.e., where
specimen scan result in distortion in the displayed image, “secondary” refers to a commercial vendor artifact that is
. Fig. 6.5b. traceable to a primary national measurement calibration
Note that for all measurements the calibration artifact must artifact) that contains a set of defined distances, the user
be placed normal to the optic axis of the SEM to eliminate can specify a vector that spans a particular pitch to establish
image foreshortening effects (see further discussion below). the calibration at that magnification setting. This calibra-
tion procedure should then be repeated to cover the range
Using a Calibrated Structure in ImageJ-Fiji of magnification settings to be used for subsequent measure-
The image-processing software engine ImageJ-Fiji includes a ments of unknowns. Note that the calibration that has been
“Set Scale” function that enables a user to transfer the image performed is only strictly valid for the SEM working distance
calibration to subsequent measurements made with various at which the calibration artifact has been imaged. When a
98 Chapter 6 · Image Formation
A* B* C* D*
0 tilt
70 º tilt
Tilt Axis
100 Chapter 6 · Image Formation
successive scan rays of the SEM image have such a small angu- plane is tilted to θ = 45°, the grid appears to have rectangular
lar spread relative to the optic axis that they create a nearly openings, as shown in . Fig. 6.9b, with the shortened side of
parallel projection to create the geometric mapping of the the true squares running parallel to the y-axis, while the cor-
specimen three-dimensional space to the two-dimensional rectly sized side runs parallel to the x-axis, which is the axis of
image space. As shown in . Fig. 6.7, a linear feature of length tilt. Some SEMs are equipped with a “tilt correction” feature in
Ltrue lying in a plane tilted at an angle, θ, (where θ is defined which the y-scan perpendicular to the tilt axis is decreased to
relative to a plane perpendicular to the optic axis) is fore- compensate for the extended length (relative to the x-scan
shortened in the SEM image according to the relation along the tilt axis) of the scan excursion on the tilted speci-
men, as shown schematically in . Fig. 6.9c. Tilt correction
Limage = Ltrue * cos θ (6.3) creates the same magnification (i.e., the same pixel dimen-
sion) along orthogonal x- and y-axes, which restores the
For the situation shown in . Fig. 6.7a, all four linear objects proper shape of the squares, as seen in . Fig. 6.9c. However,
would have the same apparent size in the SEM image, but only this scan transformation is only correct for objects that lie in
6 one, object B, would be shown with the correct length since it the plane of the specimen. . Figure 6.9c also contains a spher-
lies in a plane perpendicular to the optic axis, while the true ical particle, which appears to be circular at θ = 0° and at
lengths of the other linear objects would be significantly θ = 45° without tilt correction, since the normal scan projects
underestimated. For the most severe case, object D, which lies the intersection of the plane of the scan sphere as a circle.
on the most highly tilted surface with θ = 75°, the object is a However, when tilt correction is applied at θ = 45°, the sphere
factor of 3.9 longer than it appears in the image. The effect of now appears to be a distorted ovoid. Thus, applying tilt cor-
foreshortening is dramatically illustrated in . Fig. 6.7b, where rection to the image of an object with three-dimensional fea-
familiar objects, paper clips, are seen in a wide area SEM tures of arbitrary orientation will result in image distortions
image at 0° tilt and 70° tilt. At high tilt, the length of the first that will increase in severity with the degree of local tilt.
paper clip parallel to the tilt axis remains the same, while the
second paper clip that is perpendicular to the tilt axis is highly
foreshortened (Note that the third paper clip, which also lies 6.4.2 Image Defocusing (Blurring)
parallel to the tilt axis, appears shorter than the first paper clip
because the effective magnification decreases down the tilted The act of focusing an SEM image involves adjusting the
surface). As shown schematically in . Fig. 6.8, foreshortening strength of the objective lens to bring the narrowest part of
causes a square to appear as a rectangle. The effect of fore- the focused beam cross section to be coincident with the
shortening is shown for an SEM image of a planar copper grid surface. If the specimen has a flat, planar surface placed nor-
in . Fig. 6.9, where the square openings of the grid are cor- mal to the beam, then the situation illustrated in . Fig. 6.10a
rectly imaged at θ = 0° in . Fig. 6.9a. When the specimen will exist at sufficiently low magnification. . Figure 6.10a
SEM image
True length
Tilt axis
6.4 · Image Defects
101 6
a b
.. Fig. 6.9 a SEM/E–T (positive) image of a copper grid with a poly- openings into rectangles. c Grid tilted to 45°; “tilt correction” applied,
styrene latex sphere; tilt = 0° (grid normal to electron beam). b Grid but note that while the square grid openings are restored to the proper
tilted to 45°; note the effect of foreshortening distorts the square grid shape, the sphere is highly distorted
shows the locations of the beam at the pixel centers in the small fraction of each pixel area so that there is no possibility
middle of the squares and the effective sampling footprint. of overlap, i.e., sampling of adjacent pixels. Now consider
The sampling footprint consists of the contribution of the what happens as the magnification is increased, i.e., the
incident beam diameter (in this case finely focused to a length l in Eq. (6.1) decreases while the pixel number, n,
diameter <10 nm) and the surface emergence area of the BSE remains the same: the distance between pixel centers
and SE, which is controlled by the interaction volume. decreases, but the beam sampling footprint remains the
. Figure 6.10a considers a situation of a low beam energy same size for this particular material and beam landing
(e.g., 5 keV) and a high atomic number (e.g., Au). For these energy. The situation shown in . Fig. 6.10b for the original
conditions, the beam sampling footprint only occupies a beam sampling footprint relative to the pixel spacing and
102 Chapter 6 · Image Formation
c
d
In-focus
Out-of-focus
.. Fig. 6.11 a Trivial example of optimal lens strength (focused at distances along the optic axis. c Effects similar to . Fig. 6.11b but for a
blue plane) and defocusing caused by selecting the objective lens three-dimensional specimen of arbitrary shape. d An imaging situation
strength too high (focused at green plane) and too low (focused at corresponding to . Fig. 6.11c: coated fragments of Mt. St. Helens ash
magenta plane) relative to the specimen surface. b Effect of a tilted pla- mounted on conducting tape and imaged under high vacuum at
nar surface. The beam is scanned with fixed objective lens strength, so E0 = 20 keV with an E–T (positive) detector
that different beam diameters encounter the specimen at different
6.5.1 Qualitative Stereomicroscopy human visual process and creating two angularly separated
views of the specimen with parallax information. In SEM
The human visual process creates the perception of depth imaging, the electron beam is the “eye of the observer” (see
and the three-dimensional character of objects by combin- the “Image Interpretation” module), so the two required
ing the separate two-dimensional views provided by the left images for stereo imaging must be obtained by either
eye and the right eye to create a fused image, as shown in changing the orientation of the beam relative to the speci-
. Fig. 6.13. The angular difference between the eyes creates men (beam rocking method), or by changing the orienta-
two distinct views containing parallax information, which tion of the specimen relative to the fixed electron beam
is the horizontal shift (relative to a vertical axis) of a feature (specimen tilting method). An appropriate image separa-
common to the two separate views. The parallax is the criti- tion method such as the anaglyph technique (e.g., using red
cal information that the brain uses to create the sensation of and cyan filters to view color-coded images) then presents
depth in the fused image: the larger the parallax, the closer the each member of the image pair to the left or right eye so
the object is to the viewer. To create a similar sensation of that the viewer’s natural imaging process will create a fused
depth, SEM stereomicroscopy operates by mimicking the image that reveals the third dimension. (Note that there is a
104 Chapter 6 · Image Formation
10 µm
along the vertical. In either case, note the location of the and blue (right eye) glasses. Note: The image-processing
Everhart–Thornley detector in the image, which will software may allow fine scale adjustments (shifts and/or
provide the general sense of illumination. Ideally, the rotations) to improve the registration of the images. This
position of the E–T detector should be at the top of the procedure is illustrated in . Fig. 6.14 for the “Anaglyph
image. However, after image rotation to orient the tilt Maker STEREOEYE” software (7 http://www.stereoeye.jp/
axis vertically, the effective position of the Everhart- index_e.html). Examples of “stereo pairs” created in this
Thornley detector is likely to be different from this ideal manner are shown in . Fig. 6.15 (a particle of ash from
12-o’clock position (top center of image). the Mt. St. Helens eruption) and . Fig. 6.16 (gypsum
2. Record an image of the area of interest at the low tilt crystals from cement).
angle, for example, stage tilt = 0°.
3. Using this image as a reference, increase the tilt angle to kNote
the desired value, e.g., stage tilt = 5°, while maintaining While usually successful, this SEM stereomicroscopy “recipe”
the location of the field of view. Depending on the may not produce the desired stereo effect on your particular
mechanical sophistication of the specimen stage, instrument. Because of differing conventions for labeling tilt
changing the tilt may cause the field of view to shift motions or due to unexpected image rotation applied in the
laterally, requiring continual relocation of the desired software, the sense of the topography may be reversed (e.g., a
field of view during the tilting process to avoid losing the topographic feature that is an “inner” falsely becomes an
area of interest, especially at high magnification on “outer” and vice versa). It is good practice when first imple-
specimens with complex topography. menting stereomicroscopy with an SEM to start with a simple
4. The vertical position of the specimen may also shift specimen with known topography such as a coin with raised
during tilting. To avoid introducing rotation in the lettering or a scratch on a flat surface. Apply the procedure
second (high tilt image) by changing the objective lens above and inspect the results to determine if the proper sense
strength to re-focus, the vertical stage motion (z-axis) of the topography has been achieved in the resulting stereo
should be used to refocus the image. After careful pair. If not, be sure the parallax shift is horizontal, that is,
adjustment of the x-y-z position of the stage using the across a vertical axis (if necessary, use software functions to
low tilt image to locate the area of interest, record this rotate the images to vertically orient the tilt axis). If the tilt axis
high tilt image. is vertical but the stereo pair still shows the wrong sense of the
5. Within the image-processing software, assign the low tilt topography, try reversing the images so the “high tilt” image is
image to the RED image channel and the high tilt image now viewed by the left eye and the “low tilt” image viewed by
to the CYAN (GREEN-BLUE channels combined, or the the right eye. Once the proper procedure has been discovered
individual GREEN or BLUE image channels, depending to give the correct sense of the topography on a known test
on the type of anaglyph viewing filters available). Apply structure, follow this convention for future stereomicroscopy
the image fusion function to create the stereo image, and work. (Note: A small but significant fraction of observers find
view this image display with appropriate red (left eye) it difficult to fuse the images to form a stereo image.)
106 Chapter 6 · Image Formation
Cement:
gypsum crystals
100 mm
q2 > q1 XR Parallax, P
XL
q1 Tilt Axis q2
Coordinates relative to the reference point (0, 0, 0) in both images
(Boyde 1973, 1974a,b; Wells 1974). This procedure can be Measuring a Simple Vertical Displacement
accomplished even if the operator is not personally able to The stereo pair in . Fig. 6.20a illustrates a typical three-
perceive the qualitative stereo effect using the anaglyph or dimensional measurement problem: for this screw thread,
other methods to present the two images. how far above or below is the feature circled in green relative
1. The first step is to record a stereo pair with tilt angles θ1 to the feature circled in yellow? The left image (low tilt, θ = 0°)
and θ2 and with the tilt axis placed in a vertical orienta- and right image (high tilt, θ = 5°) are prepared according to
tion in the images. The difference in tilt angle between the convention described above and oriented so that the tilt
the members of the stereo pair is a critical parameter: axis is vertical. It is good practice to inspect the stereo pair
with the anaglyph method shown in . Fig. 6.14 to ensure that
∆θ = θ 2 − θ1 the stereo pair is properly arranged, and to qualitatively
(6.4)
assess the nature of the topography, i.e., determine how fea-
2. A set of orthogonal axes is centered on a recognizable tures are arranged relative to each other, as shown for this
feature, as shown in the schematic example in image of the screw thread in . Fig. 6.20a. In . Fig. 6.20b, a
6 . Fig. 6.19. This point will then be arbitrarily assigned set of x- (horizontal) and y- (vertical) axes are established in
the X-, Y-, Z-coordinates (0, 0, 0) and all subsequent each image centered on the feature in the yellow circle, which
height measurements will be with respect to this point. is assigned the origin of coordinates (0, 0, 0). Using this coor-
The axes are selected so that the y-axis is parallel to the dinate system, measurements are made of the feature of
tilt axis and the x-axis is perpendicular to the tilt axis. interest (within the green circle) in the left (XL = 144 μm,
3. For the feature of interest, the (X, Y)-coordinates are YL = −118 μm) and right (XR = 198 μm, YR = −118 μm)
measured in the Left (XL, YL) and Right (XR, YR) images. The parallax P is then
members of the stereo pair using the calibrated distance
marker. The parallax, P, of a feature is given by P = X L − X R =144 µm −198 µm = − 54 µm (6.10)
P =( XL − XR ) (6.5) Note that the sign of the parallax is negative, which means
that the green circle feature is below the yellow circle feature,
With this convention, points lying above the tilt axis will a result that is confirmed by the qualitative inspection of the
have positive parallax values P. Note that as an internal stereo pair in . Fig. 6.20a. Inserting these values into Eq.
consistency check, YL = YR if the y-axis has been properly (6.6), the Z-coordinate of the end of the green circle feature
aligned with the tilt axis. relative to the yellow circle feature is calculated to be:
4. For SEM magnifications above a nominal value of 100×,
the scan angle will be sufficiently small that it can be Z 3 = P / 2 sin ( ∆θ / 2 )
assumed that the scan is effectively moving parallel to
the optic axis, which enables the use of simple formulas = − 54 µm / 2 sin ( 5° / 2 )
for quantification. With reference to the fixed point (0, 0, = − 619 µm (6.11)
0), the three-dimensional coordinates X3, Y3, Z3 of the
chosen feature are given by Thus, the feature in the green circle is 619 μm below the fea-
ture in the yellow circle at the origin of coordinates. The
Z 3 = P / 2 sin ( ∆θ / 2 ) (6.6) uncertainty budget for this measurement consists of the fol-
lowing components:
X 3 = ( P / 2) + X L = X R − ( P / 2) (6.7) 1. Scale calibration error: with the careful use of a primary
or secondary dimensional artifact, this uncertainty
contribution can be reduced to 1 % relative or less.
(Note that Eq. (6.7) provides a self-consistency check for
2. Measurement of the feature individual coordinates: The
the X3 coordinate.)
magnitude of this uncertainty contribution depends on
Y=
3 Y=
L YR (6.8) how well the position of a feature can be recognized and
on the separation of the features of interest. By selecting
Note that if the measured coordinates yL and yR are not a magnification such that the features whose vertical
the same then this implies that the tilt axis is not separation is to be measured span at least half of the
accurately parallel to Y and the axes must then be image field, the uncertainty in the individual coordinates
rotated to correct this error. should be approximately 1 % relative, and in the differ-
By measuring any two points with coordinates, (XM, YM, ence of X-coordinates (XL–XR) about 2 % relative. For
ZM) and (XN, YM, ZM), the length L of the straight line closely spaced features, the magnitude of this uncer-
connecting the points is given by tainty contribution will increase.
3. Uncertainty in the individual tilt settings: The magni-
L = SQRT ( X M − X N ) + (YM − YN ) + ( Z M − Z N ) (6.9)
2 2 2 tude of this uncertainty is dependent on the degree of
backlash in the mechanical stage motions. Backlash
6.5 · Making Measurements on Surfaces With Arbitrary Topography: Stereomicroscopy
109 6
.. Fig. 6.20 a Stereo pair of a
machined screw thread—SEM/E– a
T(positive) images; E0 = 20 keV. b
Stereo pair with superimposed
axes for measurement of coordi-
nates needed for quantitative ste-
reomicroscopy calculations
effects can be minimized by selecting the initial (low) tilt estimated to be approximately 2 % for Δθ = 50, with the
value to correspond to a well-defined detent position if relative uncertainty increasing for smaller values of Δθ.
the mechanical stage is so designed, such as a physical 4. Considering all of these sources of uncertainty, the
stop at 0° tilt. With a properly maintained mechanical measurement should be assigned an overall uncertainty
stage, the uncertainty in the tilt angle difference Δθ is of ±5 % relative.
110 Chapter 6 · Image Formation
6
111 7
References – 121
7.1 Information in SEM Images secondary electrons are likely to escape from a shallow
surface region of a specimen and convey surface
Information in SEM images about specimen properties is information.
conveyed when contrast in the backscattered and/or second-
ary electron signals is created by differences in the interac-
tion of the beam electrons between a specimen feature and its 7.2 I nterpretation of SEM Images
surroundings. The resulting differences in the backscattered of Compositional Microstructure
and secondary electron signals (S) convey information about
specimen properties through a variety of contrast mecha- 7.2.1 tomic Number Contrast
A
nisms. Contrast (Ctr) is defined as With Backscattered Electrons
Ctr = ( S max − S min ) / S max (7.1) The monotonic dependence of electron backscattering upon
atomic number (η vs. Z, shown in . Fig. 2.3) constitutes a
where is Smax is the larger of the signals. By this definition, number effect with predictable behavior that enables SEM
0 ≤ Ctr ≤ 1. imaging to reveal the compositional microstructure of a spec-
7 Contrast can be conveyed in the signal by one or more of imen through the contrast mechanism variously known as
three different mechanisms: “atomic number contrast,” “compositional contrast,” “material
1. Number effects. Number effects refer to contrast which contrast,” or “Z-contrast.” Ideally, to observe unobscured
arises as a result of different numbers of electrons leaving atomic number contrast, the specimen should be flat so that
the specimen at different beam locations in response to topography does not independently modify electron back-
changes in the specimen characteristics at those loca- scattering. An example of atomic number contrast observed
tions. in a polished cross section of Raney nickel alloy using signal
2. Trajectory effects. Trajectory effects refer to contrast collected with a semiconductor backscattered electron (BSE)
resulting from differences in the paths the electrons detector is shown in . Fig. 7.1, where four regions with pro-
travel after leaving the specimen. gressively higher gray levels can be identified. The systematic
3. Energy effects. Energy effects occur when the contrast is behavior of η versus Z allows the observer to confidently con-
carried by a certain portion of the backscattered clude that the average atomic number of these four regions
electron or secondary electron energy distribution. For increases as the average gray level increases. SEM/EDS micro-
example, the high-energy backscattered electrons are analysis of these regions presented in . Table 7.1 gives the
generally the most useful for imaging the specimen compositional results and calculated average atomic number,
using contrast mechanisms such as atomic number Zav, of each phase. The Zav values correspond to the trend of
contrast or crystallographic contrast. Low-energy the gray levels of the phases observed in . Fig. 7.1.
1
2
3
20 µm
7.2 · Interpretation of SEM Images of Compositional Microstructure
113 7
.. Table 7.1 Raney nickel alloy (measured composition, calculated average atomic number, backscatter coefficient, and atomic number
contrast across the boundary between adjacent phases)
Phase Al (mass frac) Fe (mass frac) Ni (mass frac) Zav Calculated, η Contrast
Ctr = ( S max − S min ) / S max = (hmax − hmin ) / hmax (7.2) Ctr = ( 0.487 − 0.484 ) / 0.487 = 0.0062 (7.6)
Values of the backscatter coefficient for E0 ≥ 10 keV can be . Figure 7.2 summarizes this behavior in a plot of the BSE
conveniently estimated using the fit to η versus Z (Eq. 2.2). atomic number contrast for a unit change in Z as a function of Z.
Note that for mixtures that are uniform at the atomic level
(e.g., alloy solid solutions, compounds, glasses, etc.), the 7.2.3 SE Atomic Number Contrast
B
backscattered electron coefficient can be calculated from the
With the Everhart–Thornley Detector
mass fraction average of the atomic number inserted into Eq.
2.2 (as illustrated for the Al-Fe-Ni phases listed in . Table 7.1),
The appearance of atomic number contrast for a polished cross
or alternatively, from the mass fraction average of the pure
section of Al-Cu aligned eutectic, which consists of an Al-2 %
element backscatter coefficients.
Cu solid solution and the intermetallic CuAl2, is shown as
The greater the difference in atomic number between two
viewed with a semiconductor BSE detector in . Fig. 7.3a and an
materials, the greater is the atomic number contrast. Consider
Everhart–Thornley detector (positively biased) in . Fig. 7.3b.
two elements with a significant difference in atomic number,
The E–T detector is usually thought of as a secondary electron
for example, Al (Z = 13, η = 0.152) and Cu (Z = 29, η = 0.302).
detector, and while it captures the SE1 signal, it also captures
From Eq. (7.1), the atomic number contrast between Al and
BSEs that are directly emitted into the solid angle defined by the
Cu is estimated to be
scintillator. Additionally, BSEs are also represented in the E–T
detector signal by the large contribution of SE2 and SE3, which
Ctr = (hmax − hmin ) / hmax are actually BSE-modulated signals. Thus, although the SE1 sig-
= ( 0.302 − 0.152 ) / 0.302 = 0.497 (7.3) nal of the E–T detector does not show predictable variation
with composition, the BSE components of the E–T signal reveal
When the contrast is calculated between elements separated the atomic number contrast seen in . Fig. 7.3b. It must be
by one unit of atomic number, much lower values are found, noted, however, that because of the sensitivity of the E–T detec-
which has an important consequence on establishing visibil- tor to edge effects and topography, these fine-scale features are
ity, as discussed below. Note that the slope of η versus Z much more visible in . Fig. 7.3b than in . Fig. 7.3a.
decreases as Z increases, so that the contrast (which is the For both the dedicated semiconductor BSE detector and
slope of η vs. Z) between adjacent elements (ΔZ = 1) also the E–T detector, the higher atomic number regions appear
decreases. For example, the contrast between Al (Z = 13, brighter than the lower atomic number regions, as indepen-
η = 0.152) and Si (Z = 14, η = 0.164) where the slope of η ver- dently confirmed by energy dispersive X-ray spectrometry of
sus Z is relatively high is both materials. However, the semiconductor BSE detector
114 Chapter 7 · SEM Image Interpretation
0.4
0.3
Contrast
0.2
7
0.1
0.0
0 20 40 60 80 100
Atomic number
a b
.. Fig. 7.3 Aligned Al-Cu eutectic; E0 = 20 keV: a semiconductor BSE detector (SUM mode); b Everhart–Thornley detector (positive bias)
actually enhances the atomic number contrast over that esti- 7.3 I nterpretation of SEM Images
mated from the composition (Al-0.02Cu, η = 0.155; CuAl2, of Specimen Topography
η = 0.232, which gives Ctr = 0.33). The semiconductor detec-
tor shows increased response from higher energy backscat- Imaging the topographic features of specimens is one of the
tered electrons, which are produced in greater relative most important applications of the SEM, enabling the
abundance from Cu compared to Al, thus enhancing the dif- microscopist to gain information on size and shape of fea-
ference in the measured signals. The response of the Everhart- tures. Topographic contrast has several components arising
Thornley detector (positive bias) to BSEs is more complex. from both backscattered electrons and secondary electrons:
The BSEs that directly strike the scintillator produce a greater 1. The backscattered electron coefficient shows a strong
response with increasing energy. However, this component is dependence on the surface inclination, η versus θ. This
small compared to the BSEs that strike the objective lens pole effect contributes a number component to the observed
piece and chamber walls, where they are converted to SE3 contrast.
and subsequently collected. For these remote BSEs, the lower 2. Backscattering from a surface perpendicular to the
energy fraction actually create SEs more efficiently. beam (i.e., 0° tilt) is directional and follows a cosine
7.3 · Interpretation of SEM Images of Specimen Topography
115 7
distribution η(φ) ≈ cos φ (where φ is an angle measured 7.3.1 Imaging Specimen Topography
from the surface normal) that is rotationally symmetric With the Everhart–Thornley Detector
around the beam. This effect contributes a trajectory
component of contrast. SEM images of specimen topography collected with the
3. Backscattering from a surface tilted to an angle θ Everhart–Thornley (positive bias) detector (Everhart and
becomes more highly directional and asymmetrical as θ Thornley 1960) are surprisingly easy to interpret, considering
increases, tending to peak in the forward scattering how drastically the imaging technique differs from ordinary
direction. This effect contributes a trajectory component human visual experience: A finely focused electron beam steps
of contrast. sequentially through a series of locations on the specimen and a
4. The secondary electron coefficient δ is strongly depen- mixture of the backscattered electron and secondary electron
dent on the surface inclination, δ(θ) ≈ sec θ, increasing signals, subject to the four number and trajectory effects noted
rapidly as the beam approaches grazing incidence. This above that result from complex beam–specimen interactions, is
effect contributes a number component of contrast. used to create the gray-scale image on the display. Nevertheless,
a completely untrained observer (even a young child) can be
Imaging of topography should be regarded as qualitative in reasonably expected to intuitively understand the general shape
nature because the details of the image such as shading of a three-dimensional object from the details of the pattern of
depend not only on the specimen characteristics but also highlights and shading in the SEM/E–T (positive bias) image.
upon the response of the particular electron detector as In fact, the appearance of a three-dimensional object viewed in
well as its location and solid angle of acceptance. an SEM/E–T (positive bias) image is strikingly similar to the
Nevertheless, the interpretation of all SEM images of view that would be obtained if that object were viewed with a
topography is based on two principles regardless of the conventional light source and the human eye, producing the so-
detector being used: called “light-optical analogy.” This situation is quite remarkable,
1. Observer’s Point-of-View: The microscopist views the and the relative ease with which SEM/E–T (positive bias)
specimen features as if looking along the electron beam. images can be utilized is a major source of the utility and popu-
2. Apparent Illumination of the Scene: larity of the SEM. It is important to understand the origin of this
a. The apparent major source of lighting of the SEM/E–T (positive bias) light-optical analogy and what patho-
scene comes from the position of the electron logical effects can occur to diminish or destroy the effect, pos-
detector. sibly leading to incorrect image interpretation of topography.
b. Depending on the detector used, there may appear to The E–T detector is mounted on the wall of the SEM
be minor illumination sources coming from other specimen chamber asymmetrically off the beam axis, as illus-
directions. trated schematically in . Fig. 7.4. The interaction of the beam
+10 kV +300 V
SE3
SE3
SE3
SE3
BSE
SE1,2 BSE
SE1,2
A B
C
116 Chapter 7 · SEM Image Interpretation
with the specimen results in backscattering of beam electrons 7.3.2 he Light-Optical Analogy to the SEM/
T
and secondary electron emission (type SE1 produced by the E–T (Positive Bias) Image
beam electrons entering the specimen and type SE2 produced
by the exiting BSEs). Energetic BSEs carrying at least a few The complex mix of direct BSEs, SE1 and SE2, and remote SE3
kilo-electronvolts of kinetic energy that directly strike the illustrated in . Fig. 7.4 effectively illuminates the specimen
E–T scintillator are always detected, even if the scintillator is in a way similar to the “real world” landscape scene illus-
passive with no positive accelerating potential applied. In trated schematically in . Fig. 7.5 (Oatley 1972). A viewer in
typical operation the E–T detector is operated with a large an airplane looks down on a hilly landscape that is direction-
positive accelerating potential (+10 kV or higher) on the ally illuminated by the Sun at a shallow (oblique) angle, high-
scintillator and a small positive bias (e.g., +300 V) on the lighting sloping hillsides such as “A,” while a general pattern
Faraday cage which surrounds the scintillator. The small of diffuse light originates from scattering of sunlight by
positive bias on the cage attracts SEs with high efficiency to clouds and the atmosphere that illuminates all features,
the detector. Once they pass inside the Faraday cage, the SEs including those not in the direct path of the sunlight, such
are accelerated to detectable kinetic energy by the high posi- as hillside “B,” while the cave “C” receives no illumination.
tive potential applied to the face of the scintillator. In addi- To establish this light-optical analogy, we must match com-
7 tion to the SE1 and SE2 signals produced at the specimen, the ponents with similar characteristics:
E–T (positive bias) detector also collects some of the remotely 1. The human observer’s eye, which has a very sharply
produced SE3 which are generated where the BSEs strike the defined line-of-sight, is matched in characteristic by the
objective lens and the walls of the specimen chamber. Thus, electron beam, which presents a very narrow cone angle
in . Fig. 7.4 a feature such as face “A,” which is tilted toward of rays: thus, the observer of an SEM image is effectively
the E–T detector, scatters some BSEs directly to the scintilla- looking along the beam, and what the beam can strike is
tor, which add to the SE1, SE2, and SE3 signals that are also what can be observed in an image.
collected, making “A” appear especially bright compared to 2. The illumination of an outdoor scene by the Sun consists
face “B.” Because “B” is tilted away from the E–T (positive of a direct component (direct rays that strongly light
bias) detector, it does not make a direct BSE contribution, but those surfaces that they strike) and an indirect compo-
some SE1 and SE2 signals will be collected from “B” by the nent (diffuse scattering of the Sun’s rays from clouds and
Faraday cage potential, which causes SEs to follow curving the atmosphere, weakly illuminating the scene from all
trajectories, while remote SE3 signals from face “B” will also angles). For the E–T detector (positive bias), there is a
be collected. Only features the electron beam does not direct signal component that acts like the Sun (BSEs
directly strike, such as the re-entrant feature “C,” will fail to emitted by the specimen into the solid angle defined by
generate any collectable signal and thus appear black. the scintillator, as well as SE1 and SE2 directly collected
A B
C
7.3 · Interpretation of SEM Images of Specimen Topography
117 7
from the specimen) and an indirect component that acts collection of the SE3 component. Thus, if we imagine the
like diffuse illumination (SE3 collected from all surfaces specimen scene to be illuminated by a primary light source,
struck by BSEs). then that light source occupies the position of the E–T detec-
tor and the viewer of that scene is looking along the electron
Though counterintuitive, in the SEM the detector is the beam. The SE3 component of the signal provides a general
apparent source of illumination while the observer looks diffuse secondary source of illumination that appears to
along the electron beam. come from all directions.
a b
10 mm
10 mm
7
c d
10 mm 10 mm
.. Fig. 7.7 a SEM image of a particle on a surface as prepared with on a surface as prepared with the E–T (negative bias) detector in the 0°
the E–T (negative bias) detector in the 90° clockwise position shown; clockwise (12 o’clock) position shown; E0 = 20 keV. Note strong shadow-
E0 = 20 keV. Note strong shadowing pointing away from E–T . b SEM ing pointing away from E–T . d SEM image of a particle on a surface as
image of a particle on a surface as prepared with the E–T (negative prepared with the E–T (positive bias) detector in the 0° clockwise (12
bias) detector in the 45° clockwise position shown; E0 = 20 keV. Note o’clock) position shown; E0 = 20 keV. Note lack of shadowing but bright
strong shadowing pointing away from E–T . c SEM image of a particle surface facing the E–T (positive bias) detector
down on the features of the specimen. When the top lighting apparent lighting. By comparison, the image of an undulat-
condition is violated and the observer is unaware of the ing surface of an unknown object may provide no clues that
alteration of the scene illumination, then the sense of the cause the proper sense of the topography to “click in” for the
topography can appear inverted. Arbitrary scan rotation can observer. Having top illumination is critical in such cases.
effectively place the E–T detector, or any other asymmetri- When a microscopist works in a multi-user facility, the pos-
cally placed (i.e., off-axis) detector, at the bottom or sides of sibility must always be considered that a previous user may
the image, and if the observer is unaware of this situation of have arbitrarily adjusted the scan rotation. As part of a per-
unfamiliar illumination, misinterpretation of the specimen sonal quality-assurance plan, the careful microscopist should
topography is likely to result. This is especially true in the confirm that the location of the E–T detector is at the top
case of specimens for which there are limited visual clues. center of the image. . Figure 7.7 demonstrates a procedure
For example, the SEM image of an insect contains many that enables unambiguous location of the E–T detector.
familiar features—e.g., head, eyes, legs, etc.—that make it Some (but not all) implementations of the E–T detector
almost impossible to invert the topography regardless of the enable the user to “deconstruct” the E–T detector image by
7.3 · Interpretation of SEM Images of Specimen Topography
119 7
selectively excluding the SE component of the total signal— . Fig. 7.8a demonstrates the efficiency of the E–T detector
either by changing the Faraday cage voltage to negative val- for collection of signal from virtually all surfaces of the spec-
ues to reject the very low energy SEs (e.g., −50 V cage bias) imen that the primary beam strikes.
or by eliminating the high potential on the scintillator so that
SEs cannot be accelerated to sufficient kinetic energy to
excite scintillation. Even without the high potential applied 7.3.3 Imaging Specimen Topography
to the scintillator, the E–T detector remains sensitive to the With a Semiconductor BSE Detector
high energy BSEs generated by a high energy primary beam,
for example., E0 ≥ 20 keV, which creates a large fraction of A segmented (A and B semicircular segments) semiconduc-
BSEs with energy >10 keV. As a passive scintillator or with tor BSE detector placed directly above the specimen is illus-
the negative Faraday cage potential applied, the E–T (nega- trated schematically in . Fig. 7.9. This BSE detector is
tive bias) detector only collects the small fraction of high mounted below the final lens and is placed symmetrically
energy BSEs scattered into the solid angle defined by the E–T around the beam, so that in the summation mode it acts as
scintillator. When the direct BSE mode of the E–T (negative an annular detector. A simple topographic specimen is illus-
bias) detector is selected, debris on a flat surface is found to trated, oriented so that the left face directs BSEs toward the
create distinct shadows that point away from the apparent A-segment, while the right face directs BSEs toward the
source of illumination, the E–T detector. By using the scan B-segment. This A and B detector pair is typically arranged
rotation, the effective position of the E–T detector can then so that one of the segments, “A,” is oriented so that it appears
be moved to the top of the image, as shown in the sequence to illuminate from the top of the image, while the “B” seg-
of . Fig. 7.7a–c, thus achieving the desired top-lighting situ- ment appears to illuminate from the bottom of the image.
ation. When the conventional E–T (positive bias) is used to The segmented detector enables selection of several modes
image this same field of view (. Fig. 7.7d), the strong shadow of operation: SUM mode (A + B), DIFFERENCE mode
of the particle disappears because of the efficient collection (A−B), and individual detectors A or B) (Kimoto and
of SEs, particularly the SE3 component, and now has a bright Hashimoto 1966).
edge along the top which reinforces the impression that it
rises above the general surface.
SUM Mode (A + B)
Note that physically rotating the specimen stage to change
the angular relation of the specimen relative to the E–T (or The two-segment semiconductor BSE detector operating in
any other) detector does not change the location of the appar- the summation (A + B) mode was used to image the same
ent source of illumination in the displayed image. Rotating pyrite specimen previously imaged with the E–T (positive
the specimen stage changes which specimen features are bias) and E–T (negative bias), as shown in . Fig. 7.8c. The
directed toward the detector, but the scan orientation on the placement of the large solid angle BSE is so close to the pri-
displayed image determines the relative position of the detec- mary electron beam that it creates the effect of apparent
tor in the image presented to the viewer and the apparent wide-angle illumination that is highly directional along the
direction of the illumination. line-of-sight of the observer, which would be the light-optical
equivalent of being inside a flashlight looking along the
beam. With such directional illumination along the observ-
econstructing the SEM/E–T Image
D er’s line-of-sight, the brightest topographic features are those
of Topography oriented perpendicular to the line-of-sight, while tilted sur-
It is often useful to examine the separate SE and BSE compo- faces appear darker, resulting in a substantially different
nents of the E–T detector image. An example of a blocky impression of the topography of the pyrite specimen com-
fragment of pyrite (FeS2) imaged with a positively-biased pared to the E–T (positive bias) image in . Fig. 7.8a. The
E–T detector is shown in . Fig. 7.8a. In this image, the effec- large solid angle of the detector acts to suppress topographic
tive position of the E–T detector relative to the presentation contrast, since local differences in the directionality of BSE
of the image is at the top center. . Figure 7.8b shows the same emission caused by differently inclined surfaces are effec-
field of view with the Faraday cage biased negatively to tively eliminated when the diverging BSEs are intercepted by
exclude SEs so that only direct BSEs contribute to the SEM another part of the large BSE detector.
image. The image contrast is now extremely harsh, since Another effect that is observed in the A + B image is the
topographic features facing toward the detector are illumi- class of very bright inclusions which were subsequently
nated, while those facing away are completely lost. Comparing determined to be galena (PbS) by X-ray microanalysis. The
. Fig. 7.8a, b, the features that appear bright in the BSE-only large difference in average atomic number between FeS2
image are also brighter in the full BSE + SE image obtained (Zav = 20.7) and PbS (Zav = 73.2) results in strong atomic
with the positively biased E–T detector, demonstrating the number (compositional) between the PbS inclusions and the
presence of the direct-BSE component. The much softer con- FeS2 matrix. Although there is a significant BSE signal com-
trast of nearly all surfaces seen in the BSE + SE image of ponent in the E–T (positive bias) image in . Fig. 7.8a, the
120 Chapter 7 · SEM Image Interpretation
Pbs
inclusions
50 mm 50 mm
50 mm 50 mm
BSE B segment BSE B segment
BSE DIFFERENCE B-A BSE A segment
g
50 mm
BSE B segment
topographic contrast is so strong that it overwhelms the detector. As illustrated in . Fig. 7.9 for a two-segment BSE
compositional contrast. detector, the individual segments effectively provide an off-
axis, asymmetric illumination of the specimen. Comparing
xamining Images Prepared
E the A-segment and B-segment images of the pyrite crystal in
With the Individual Detector Segments . Fig. 7.8d, e, the features facing each detector can be dis-
Some semiconductor BSE detector systems enable the cerned and a sense of the topography can be obtained by
microscopist to view BSE images prepared with the signal comparing the two images. But note the strong effect of the
derived from the individual components of a segmented apparent inversion of the sense of the topography in the
References
121 7
.. Fig. 7.9 Schematic illustration
of a segmented annular semicon- Bottom view of BSE detector
ductor BSE detector
A-segment
off
BSE: A+B
B
A B
B-segment
on
BSEs BSEs
B-segment image, where the illumination comes from the assumption of top lighting has the effect for most viewers
bottom of the field, compared to the A-segment image, where of . Fig. 7.8g to strongly invert the apparent sense of the
the illumination comes from the top of the field of view. topography, so that protuberances in the A–B image become
concavities in the B–A image. If BSE detector difference
DIFFERENCE Mode (A−B) images are to be at all useful and not misleading, it is critical
The signals from the individual BSE detector segments “A” and to determine the proper order of subtraction. A suitable test
“B” can be subtracted from each other, producing the image procedure is to image a specimen with known topography,
seen in . Fig. 7.8f. Because the detector segments “A” and “B” such as the raised lettering on a coin or a particle standing on
effectively illuminate the specimen from two different direc- top of a flat surface.
tions, as seen in . Fig. 7.8d, e, taking the difference A–B
between the detector signals tends to enhance these directional
differences, producing the strong contrast seen in . Fig. 7.8f. References
Note that when subtracting the signals the order of the
segments in the subtraction has a profound effect on appear- Everhart TE, Thornley RFM (1960) Wide-band detector for micro-
ance of the final image. . Figure 7.8g shows the image created microampere low-energy electron currents. J Sci Instr 37:246
Kimoto S, Hashimoto H (1966) Stereoscopic observation in scanning
with the order of subtraction reversed to give B–A. Because
microscopy using multiple detectors. In: McKinley T, Heinrich K,
the observer is so strongly biased toward interpreting an Wittry D (eds) The electron microprobe. Wiley, New York, p 480
image as if it must have top lighting, bright features are Oatley CW (1972) The scanning electron microscope, Part I, the instru-
automatically interpreted as facing upward. This automatic ment. Cambridge University Press, Cambridge
123 8
References – 131
The detection in SEM images of specimen features such as the same nominal beam current and dwell time will produce
compositional differences, topography (shape, inclination, a range of values distributed about a mean count n, with the
edges, etc.), and physical differences (crystal orientation, standard deviation of this distribution described by n 1/ 2 .
magnetic fields, electrical fields, etc.), depends on satisfying This natural variation in repeated samplings of the signal S is
two criteria: (1) establishing the minimum conditions neces- termed “noise,” N. The measure of the signal quality is termed
sary to ensure that the contrast created by the beam–speci- the “signal-to-noise ratio,” S/N, given by
men interaction responding to differences in specimen
features is statistically significant in the imaging signal (back- S
= n=
/ n 1/ 2 n 1/ 2 (8.1)
scattered electrons [BSE], secondary electrons [SE], or a N
combination) compared to the inevitable random signal fluc-
tuations (noise); and (2) applying appropriate signal process- Equation (8.1) shows that as the mean number of collected
ing and digital image processing to render the contrast signal counts increases, the signal quality S/N improves as
information that exists in the signal visible to the observer the random fluctuations become a progressively smaller frac-
viewing the final image display. tion of the total signal.
. Figure 8.1 shows schematically the result of repeated
scans over a series of pixels that cross a feature of interest. The
8.1 ignal Quality: Threshold Contrast
S signal value S changes in response to the change in the speci-
men property (composition, topography, etc.), but the
8 and Threshold Current
repeated scans do not produce exactly the same response due
to the inevitable noise in the signal generation processes.
An SEM image is constructed by addressing the beam to a
When an observer views a scanned image, this noise is super-
specific location on the specimen for a fixed dwell time, τ,
imposed on the legitimate changes in signal (contrast) of
during which a number of beam electrons are injected
features in the image, reducing the visibility. Rose (1948)
through the focused beam footprint into the specimen. The
made an extensive study of the ability of observers viewing
resulting beam–specimen interactions cause the emission of
scanned television images to detect the contrast between
BSE and SE, a fraction of which will be detected and mea-
objects of different size and the background in the presence
sured with appropriate electron detectors. This measured BSE
of various levels of noise. Rose found that for the average
and/or SE signal is then assigned to that pixel as it is digitally
observer to distinguish small objects with dimensions about
stored and subsequently displayed as a gray-level image. Both
5 % of the image width against the background, the change in
the incident electron beam current and the measured BSE
signal due to the contrast, ΔS, had to exceed the noise, N, by
and/or SE signals, Si, involve discrete numbers of electrons:
a factor of 5:
nB, nBSE, and nSE. The emission of the incident beam current
from the electron gun and the subsequent BSE/SE generation ∆S > 5 N (8.2)
due to elastic and inelastic scattering in the specimen are sto-
chastic processes; that is, the mechanisms are subject to ran- Synthesized digital images in . Figs. 8.2 and 8.3 demonstrate
dom variations over time. Thus, repeated sampling of any how the visibility is affected by noise and the relative size of
imaging signal, S, made at the same specimen location with objects. . Figure 8.2a shows a synthesized object from the
a b
100
80
60
Counts
40
20
0
0 20 40 60 80 100
Pixel number
.. Fig. 8.2 a Synthesized image from the template shown in . Fig. 8.3a. b Trace of the signal across the circular object
template shown in . Fig. 8.3a with a specified signal and by (5/C)2, must be collected per picture element. Considering
superimposed random noise, and . Fig. 8.2b shows a plot of electrons as signal carriers, the number of electrons which
the signal through one of the test objects. . Figure 8.3 shows must be collected per picture element in the dwell time, τ,
synthesized images for various levels of the S and ΔS relative can be converted into a signal current, is
to N. In . Fig. 8.3b ΔS = 5 = N; . Fig. 8.3c ΔS = 10 = 2 N; and
. Fig. 8.3d ΔS = 25 = 5 N, which just matches the Rose crite- ne
is = (8.7)
rion. While the large-scale features are visible in all three τ
images, the fine-scale objects are completely lost in image
. Fig. 8.3b, and only fully visible when the Rose criterion is where e is the electron charge (1.6 × 10–19 C). Substituting Eq.
satisfied in image . Fig. 8.3d. (8.6) into Eq. (8.7) gives the following result:
The Rose visibility criterion can be used as the basis to
develop the quantitative relation between the threshold con- 25e
i> (8.8)
trast, that is, the minimum level of contrast potentially visible C 2τ
in the signal, and the beam current. The noise can be consid-
ered in terms of the number of signal events, N = n1/2: The signal current, is, differs from the beam current, iB, by the
fractional signal generation per incident beam electron (η for
∆S > 5n 1/ 2 (8.3) BSE and δ for SE or a combination for a detector which is
simultaneously sensitive to both classes of electrons) and the
Equation (8.3) can be expressed in terms of contrast (defined efficiency with which the signal is converted to useful infor-
as C = ΔS/S) by dividing through by the signal: mation for the image. This factor is given by the detective
quantum efficiency (DQE) (Joy et al. 1996) and depends on
∆S 5n 1 / 2 5n 1 / 2 the solid angle of collection and the response of the detector
=C > = (8.4)
S S n (see the full DQE description in the Electron Detectors
5 module):
C> (8.5)
n 1/ 2 is = iB (η ,δ ) DQE (8.9)
2
5
n > (8.6) Combining Eqs. (8.8) and (8.9) yields
C
.. Fig. 8.3 Synthesized digital images: a template; b object S = 5 counts above background, ΔS = 5 = N, S/B = 1.2; c object S = 10 counts above
background, ΔS = 10 = 2 N, S/B = 1.4; d object S = 25 counts above background, ΔS = 25 = 5 N, S/B = 2
The picture element dwell time, τ, can be replaced by the time For an image with 1024 × 1024 picture elements, Eq. (8.12)
to scan a full frame, tF, from the relation can be stated as
τ=
tF
(8.11) iB >
( 4 ×10 −12
A)
(8.13)
N PE (η ,δ ) DQE C 2tF
where NPE is the number of pixels in the entire image. Equation (8.12) is referred to as the “Threshold Equation”
Substituting Eq. (8.11) into Eq. (8.10), (Oatley et al. 1965; Oatley 1972) because it defines the mini-
mum beam current, the “threshold current,” necessary to
( 4 ×10 ) N
−18
observe a specified level of contrast, C, with a signal produc-
( coulomb / s = amperes )
PE
iB > (8.12)
(η ,δ ) DQE C 2tF tion efficiency specified by η and/or δ and the detector
8.1 · Signal Quality: Threshold Contrast and Threshold Current
127 8
.. Fig. 8.4 Plot of the threshold
contrast vs. frame time for an Threshold contrast vs. frame time
image with 1024 by 1024 pixels 1
and an overall signal conversion
efficiency of 0.25. Contours of
constant current from 1 µA to
1 pA are shown 1 pA
10 pA 5 pA
0.1
100 pA 50 pA
1 nA
500 pA
10 nA 5 nA
Threshold contrast
100 nA
50 nA
1 µA 500 pA
0.01
0.001
0.0001
0.01 0.1 1 10 100 1000
Frame time (s)
1e-8 0.02
0.05
0.1
1e-9 0.2
0.5
1
1e-10
1e-11
1e-12
0.01 0.1 1 10 100 1000
Frame time (s)
performance described by the DQE (Joy et al. 1996). For any particular selection of operating parameters—
Alternatively, if we measure the current which is available in beam current, signal (backscattered electrons, secondary elec-
the beam that reaches the specimen (e.g., with a Faraday cup trons, or a combination), detector performance (DQE), image
and specimen current picoammeter), then we can calculate the pixel density and dwell time—there is always a level of contrast
minimum contrast, the so-called “threshold contrast,” which below which objects cannot be detected. Objects producing
can be observed in an image prepared under these conditions. contrast below this threshold contrast cannot be recovered by
Objects in the field of view that do not produce this threshold applying any post-collection image-processing algorithms.
contrast cannot be distinguished from the noise of random The graphical plots shown in . Fig. 8.4 (threshold contrast
background fluctuations. Equations (8.12 and 8.13) lead to the vs. frame time for various values of the beam current) and
following critical limitation on SEM imaging performance: . Fig. 8.5 (threshold current vs. frame time for various values
128 Chapter 8 · The Visibility of Features in SEM Images
.. Fig. 8.6 Al-Si eutectic alloy. BSE images (1024 by 784 pixels; 1-μs pixel dwell) at various beam currents
of the contrast) provide a useful way to understand the rela- trast less than approximately 0.05 (5 %) against the back-
tionships of the parameters of the Threshold Equation. These ground will be lost. Once the current required to image a
plots have been derived from Eq. (8.13) with the assumptions specific contrast level is known from the Threshold Equation,
that the image has 1024 by 1024 pixels and the overall signal the minimum beam size that contains this current can be esti-
generation/collection efficiency (the product of η and/or δ mated with the Brightness Equation. A severe penalty in
and the DQE) is 0.25; that is, one signal-carrying electron minimum probe size is incurred when the contrast is low
(backscattered and/or secondary) is registered in the final because of the requirement for high beam current needed to
image for every four beam electrons that strike the specimen. exceed the threshold current. Moreover, this ideal beam size
This collection efficiency is a reasonable assumption for a tar- will be increased due to the aberrations that degrade electron
get, such as gold, which has high backscattering and second- optical performance.
ary electron coefficients, when the electrons are detected with The Rose criterion is actually a conservative estimate of
an efficient positively-biased E–T detector. . Figure 8.4 visibility threshold conditions since it is appropriate for small
reveals that imaging a contrast level of C = 0.10 (10 %) with a discrete features with linear dimensions down to a few per-
frame time of 1 s (a pixel dwell time of ~1 μs for a 1024 × 1024- cent of the image width or small details on larger structures.
pixel scan) requires a beam current in excess of 1 nA, whereas For objects that constitute a large fraction of the image or
if 100 s is used for the frame time (pixel dwell time of ~100 μs), which have an extended linear nature, such as an edge or a
the required beam current falls to about 10 pA. If the speci- fiber, the ability of an observer’s visual process to effectively
men only produces a contrast level of 0.05 (5%), a beam combine information over many contiguous pixels actually
current above 5 nA must be used. Conversely, if a particular relaxes the visibility criterion, as illustrated in the synthesized
value of the beam current is selected, . Fig. 8.5 demonstrates images in . Fig. 8.3 (Bright et al. 1998). The effect of the size
that there will always be a level of contrast below which objects of a feature on visibility of real features can be seen in
will be effectively invisible. For example, if a beam current of . Figs. 8.6 and 8.7, which show BSE images (semiconductor
1 nA is used for a 10-s frame time, all objects producing con- detector) of a commercial aluminum–silicon eutectic casting
8.1 · Signal Quality: Threshold Contrast and Threshold Current
129 8
.. Fig. 8.7 Al-Si eutectic alloy. BSE images (1024 by 784 pixels; 100-pA beam current) pixel dwell at various frame times
alloy under various conditions. The two principal phases of example is shown in . Fig. 8.8, which shows a sequence of
this material are nearly pure Al and Si, which produce a con- images of a polished carbon planchet upon which a droplet
trast based on the respective BSE coefficients of C = Δη/ηmax = containing a dilute salt was deposited by inkjet printing. The
(0.14−0.13)/0.14 ≈ 0.07 or 7 % contrast. As the beam current images were prepared at constant beam current but with
is decreased with fixed frame time (. Fig. 8.6) or the frame increasing pixel dwell time, which represents a section
time is decreased with fixed beam current (. Fig. 8.7), the through the Threshold Equation plot shown in . Fig. 8.9.
visibility of the fine-scale features at the right-hand side of Even the largest-scale features are lost in the image prepared
the image diminishes and these small features are eventually with the shortest dwell time. Careful study of these images
lost, whereas the large-scale features on the left-hand side of reveals that new information is being added throughout the
the image remain visible over the range of experimental image sequence, and likely there would be additional infor-
parameters despite having the same compositional difference mation recovered with further increases in the pixel time or
and thus producing the same contrast. by increasing the beam current.
While the Threshold Equation provides “gray numbers” Finally, it must be recognized that there is a substantial
for the threshold parameters due to the variability of the “observer effect” for objects producing contrast near the
human observer and the relative size of objects, the impact of threshold of visibility: different observers may have substan-
the Threshold Equation must be considered in developing tially different success in detecting features in images (Bright
imaging strategy. Unfortunately, poor imaging strategy can et al. 1998). Thus the threshold current or threshold contrast
render the SEM completely ineffective in detecting the fea- calculated with Eq. (8.12) should be considered a “fuzzy
tures of interest. A careful imaging strategy will first estimate number” rather than an absolute threshold, since visibility
the likely level of contrast from the objects of interest (or depends on several factors, including the size and shape of
assume the worst possible case that the features being sought the features of interest as well as the visual acuity of the par-
produce very low contrast, e.g., <0.01) and then select instru- ticular observer and his/her experience in evaluating images.
ment parameters capable of detecting that contrast. An The limitations imposed by the threshold equation and the
130 Chapter 8 · The Visibility of Features in SEM Images
.. Fig. 8.8 Threshold imaging visibility; image sequence with increasing pixel dwell time at constant beam current. Inkjet deposited droplet on
carbon planchet; E0 = 10 keV; Everhart–Thornley (positive bias) detector. Post-collection processing with ImageJ (FIJI) CLAHE function
References
131 8
.. Fig. 8.9 Threshold current
plot showing time sequence at Threshold current vs. frame time
constant beam current. Contours 1e-6
of constant contrast from 1 to
0.001 are shown
0.001
1e-7 0.002
0.005
0.01
1e-11
1e-12
0.01 0.1 1 10 100 1000
Frame time (s)
Image Defects
9.1 Charging – 134
9.1.1 What Is Specimen Charging? – 134
9.1.2 Recognizing Charging Phenomena in SEM Images – 135
9.1.3 Techniques to Control Charging Artifacts
(High Vacuum Instruments) – 139
References – 146
SEM images are subject to defects that can arise from a vari- Faraday cage of
ety of mechanisms, including charging, radiation damage, EDS Everhart -Thornley
contamination, and moiré fringe effects, among others. detector
Image defects are very dependent on the specific nature of
the specimen, and often they are anecdotal, experienced but
not reported in the SEM literature. The examples described
below are not a complete catalog but are presented to alert
the microscopist to the possibility of such image defects so as Bore of
to avoid interpreting artifact as fact. objective lens
Annular
9.1 Charging BSE detector
Dark halo:
decreased SE collection
.. Fig. 9.2 Examples of charging artifacts observed in images of dust particles on a metallic substrate. E0 = 20 keV; Everhart–Thornley (positive
bias) detector
136 Chapter 9 · Image Defects
SE BSE
.. Fig. 9.3 Comparison of images of a dust particle on a metallic substrate: (left) Everhart–Thornley (positive bias) detector; (right) semiconduc-
tor BSE (sum) detector; E0 = 20 keV
100 µm
Extreme
charging:
1. Scan deflection
2. Fully saturated areas (gray level 255)
3. Completely dark areas (gray level 0)
.. Fig. 9.4 Comparison of images of an uncoated calcite crystal viewed at (left) E0 = 1.5 keV, showing topographic contrast; (right) E0 = 5 keV,
showing extreme charging effects; Everhart–Thornley (positive bias) detector
Charging phenomena are incompletely understood and beam conditions (beam energy, current, and scan rate), the
are often found to be dynamic with time, a result of the time- injected charge may only partially dissipate before the beam
dependent motion of the beam due to scanning action and returns in the scan cycle, leading to strong effects in SEM
due to the electrical breakdown properties of materials as images. Moreover, local specimen properties may cause
well as differences in surface and bulk resistivity. An insulat- charging effects to vary with position in the same image. A
ing specimen acts as a local capacitor, so that placing the time-dependent charging situation at a pixel is shown sche-
beam at a pixel causes a charge to build up with an RC time matically in . Fig. 9.5, where the surface potential at a par-
constant as a function of the dwell time, followed by a decay ticular pixel accumulates with the dwell time and then
of that charge when the beam moves away. Depending on the decays until the beam returns. In more extreme behavior, the
exact material properties, especially the surface resistivity accumulated charge may cause local electrical breakdown
which is often much lower than the bulk resistivity, and the and abruptly discharge to ground. The time dependence of
9.1 · Charging
137 9
shown in . Fig. 9.6, where an image of an uncoated mineral
fragment taken with E0 = 1 keV appears to be free of charging
artifacts with a pixel dwell time of 1.6 μs, but longer dwell
times lead to the in-growth of a bright region due to charg-
ing. Charging artifacts can often be minimized by avoiding
–V slow scanning through the use of rapid scanning and sum-
ming repeated scans to improve the signal-to-noise of the
final image.
Charging of some specimens can create contrast that can
easily be misinterpreted as specimen features. An example is
shown in . Fig. 9.7, where most of the polystyrene latex
0 spheres (PSLs) imaged at E0 = 1 keV with the Everhart–
0 Time Thornley (positive bias) detector show true topographic
details, but five of the PSLs have bright dots at the center,
.. Fig. 9.5 Schematic illustration of the potential developed at a pixel which might easily be mistaken for high atomic number
as a function of time showing repeated beam dwells inclusions or fine scale topographic features rising above the
spherical surfaces. Raising the beam energy to 1.5 keV and
charging can result in very different imaging results as the higher reveals progressively more extensive and obvious evi-
pixel dwell time is changed from rapid scanning for survey- dence of charging artifacts. The nature of this charging arti-
ing a specimen to slow scanning for recording images with fact is revealed in . Fig. 9.8, which compares an image of the
better signal-to-noise. An example of this phenomenon is PSLs at higher magnification and E0 = 5 keV with a low
1.6 µs 4 µs
8 µs 32 µs
.. Fig. 9.6 Sequence of images of an uncoated quartz fragment imaged at E0 = 1 keV with increasing pixel dwell times, showing development of
charging; Everhart–Thornley (positive bias) detector
138 Chapter 9 · Image Defects
Incipient
Charging
Artifacts
2 keV 5 keV
.. Fig. 9.7 Polystyrene latex spheres imaged over a range of beam energy, showing development of charging artifacts; Everhart–Thornley (posi-
tive bias) detector
5 keV 2 keV
.. Fig. 9.8 (left) Higher magnification image of PSLs at E0 = 5 keV; (right) reflection image from large plastic sphere that was charged at
E0 = 10 keV and then imaged at E0 = 2 keV; Everhart–Thornley (positive bias) detector
9.1 · Charging
139 9
magnification image of a large plastic sphere (5 mm in diam- 0.5–2 keV, depending on the material), the emission of SE can
eter) that was first subjected to bombardment at E0 = 10 keV, actually reach very large values for insulators with δmax rang-
followed by imaging at E0 = 2 keV where the deposited charge ing from 2 to 20 depending on the material. Thus, in this
acts to reflect the beam and produce a “fish-eye” lens view of beam energy region η + δ > 1, resulting in positive charging
the SEM chamber. Close examination of the higher magnifi- which increases the kinetic energy of the incoming beam
cation PSL images shows that each of these microscopic electrons until the E2 energy is reached and charge balance
spheres is acting like a tiny “fish-eye lens” and producing a occurs. This dynamic charge stability enables uncoated insu-
highly distorted view of the SEM chamber. lators to be imaged, as shown in the example of the uncoated
mineral particle shown in . Fig. 9.10, where a charge-free
image is obtained at E0 = 1 keV, but charging effects are
9.1.3 echniques to Control Charging
T observed at E0 ≥ 2 keV. Achieving effective “dynamic charge
Artifacts (High Vacuum Instruments) balance microscopy” is sensitive to material and specimen
shape (local tilt as it affects BSEs and particularly SE emis-
Observing Uncoated Specimens sion), and success depends on optimizing several instrument
To understand the basic charging behavior of an uncoated parameters: beam energy, beam current, and scan speed. Note
insulator imaged with different selections of the incident that the uncoated mineral specimen used in the beam energy
beam energy, consider . Fig. 9.9, which shows the behavior of sequence in . Fig. 9.10 is the same used for the pixel dwell
the processes of backscattering and secondary electron emis- time sequence at E0 = 1 keV in . Fig. 9.6 where charging is
sion as a function of beam energy. For beam energies above observed when longer dwell times are used, demonstrating
5 keV, generally η + δ < 1, so that more electrons are injected the complex response of charging to multiple variables.
into the specimen by the beam than leave as BSEs and SEs,
leading to an accumulation of negative charge in an insulator. oating an Insulating Specimen for Charge
C
For most materials, especially insulators, as the beam energy Dissipation
is lowered, the total emission of BSEs and SEs increases, even- Conductive coatings can be deposited by thermal evaporation
tually reaching an upper cross-over energy, E2 (which typi- with electron beam heating (metals, alloys) or with resistive
cally lies in the range 2–5 keV depending on the material) heating (carbon), by high energy ion beam sputtering (met-
where η + δ = 1, and the charge injected by the beam is just als, alloys), or by low energy plasma ion sputtering (alloys).
balanced by the charge leaving as BSEs and SEs. If a beam The coating must cover all of the specimen, including com-
energy is selected just above E2 where η + δ < 1, the local build- plex topographic shapes, to provide a continuous conducting
up of negative charge acts to repel the subsequent incoming path across the surface to dissipate the charge injected into
beam electrons while the beam remains at that pixel, lowering the specimen by the electron beam. It is important to coat
the effective kinetic energy with which the beam strikes the all surfaces that are directly exposed to the electron beam
surface eventually reaching the E2 energy and a dynamically or which might receive charge from BSEs, possibly after re-
stable charge balance. For beam energies below the E2 value scattering of those BSEs. Note that applying a conductive
and above the lower cross-over energy E1 (approximately coating alone may not be sufficient to achieve efficient charge
dissipation. Many specimens may be so thick that the sides
may not actually receive an adequate amount of the coating
material, as illustrated in . Fig. 9.11, even with rotation dur-
ing the coating process. It is necessary to complete the path
from the coating to the electrical ground with a conducting
1.0
material that exhibits a low vapor pressure material that is
compatible with the microscope’s vacuum requirement, such
as a metal wire, conducting tape, or metal foil.
η+δ
It is desirable to make the coating as thin as possible,
and for many samples an effective conducting film can be
2–10 nm in thickness. A beam with E0 > 5 keV will penetrate
through this coating and 10–100 times (or more) deeper
~ 0.5-1 keV ~ 2-5 keV depending on material and the incident beam energy, thus
depositing most of the charge in the insulator itself. However,
E1 E2 E0 the presence of a ground plane and conducting path nanome-
ters to micrometers away from the implanted charge creates
.. Fig. 9.9 Schematic illustration of the total emission of backscat-
a very high local field gradient, >106 V/m, apparently lead-
tered electrons and secondary electrons as a function of incident beam ing to continuous breakdown and discharging. The strongest
energy; note upper and lower cross-over energies where η + δ = 1 evidence that a continuous discharge situation is established
140 Chapter 9 · Image Defects
1 keV 2 keV
5 keV 10 keV
.. Fig. 9.10 Beam energy sequence showing development of charging as the energy is increased. Specimen: uncoated quartz fragment; 1.6 μs
per pixel dwell time; Everhart–Thornley (positive bias) detector
15 keV
Si
SiO2_15keV20nA_4ks
Si_15keV20nA_2ks
100000
10000
Counts
10
0 2 4 6 8 10 12 14 16 18 20
Photon Energy (keV) 15 keV
NIST DTSA II Duane Hunt fitted value:
Si
SiO2_15keV20nA_4ks
Na
1000000 BareGlass_15keV20nA_2ks
SiO2 (C, 8 nm) 15.08 keV
AIMg
C
100000
Ni
Counts
10000
1000
Ni
Ca
K
10
0 2 4 6 8 10 12 14 16 18 20
Photon Energy (keV)
.. Fig. 9.12 Effects of charging on the Duane–Hunt energy limit coated (C, 8 nm) SiO2 and uncoated glass showing significant depres-
of the X-ray continuum: (upper) comparison of silicon and coated (C, sion of the Duane–Hunt limit due to charging
8 nm) SiO2 showing almost identical values; (lower) comparison of
scales with the atomic number. Note that because of pulse there is no significant charging in the coated SiO2. When an
coincidence events, there will always be a small number of uncoated glass slide is bombarded at E0 = 15 keV, the charg-
photons measured above the true Duane–Hunt limit. The ing induced by the electron beam causes charging and thus
true limit can be estimated with good accuracy by fitting severely depresses the Duane–Hunt limit to 8 keV, as seen in
the upper energy range of the continuum intensity, prefer- . Fig. 9.12 (lower plot), as well as a sharp difference in the
ably over a region that is several kilo-electronvolts in width shape of the X-ray continuum at higher photon energy.
and that does not contain any characteristic X-ray peaks,
and then finding where the curve intersects zero intensity Choosing the Coating for Imaging Morphology
to define the Duane–Hunt limit. NIST DTSA II performs The ideal coating should be continuous and featureless so
such a fit and the result is recorded in the metadata reported that it does not interfere with imaging the true fine-scale fea-
for each spectrum processed. Once the beam energy is tures of the specimen. Since the SE1 signal is such an impor-
established on a conducting specimen, then the experiment tant source of high resolution information, a material that
consists of measuring a coated and uncoated insulator. In has a high SE coefficient should be chosen. Because the SE1
. Fig. 9.12 (upper plot) spectra are shown for Si (measured signal originates within a thin surface layer that is a few
Duane–Hunt limit = 15.11 keV) and coated (C, 8 nm) SiO2 nanometers in thickness, having this layer consist of a high
(measured Duane–Hunt limit = 15.08 keV), which indicates atomic number material such as gold that has a high SE
142 Chapter 9 · Image Defects
9
9.2 Radiation Damage a
)
(+
Certain materials are susceptible to radiation damage (“beam E-T
damage”) under energetic electron bombardment. “Soft”
materials such as organic compounds are especially vulnera-
ble to radiation damage, but damage can also be observed in
“hard” materials such as minerals and ceramics, especially if
water is present in the crystal structure, as is the case for
hydrated minerals. Radiation damage can occur at all length
scales, from macroscopic to nanoscale. Radiation damage
may manifest itself as material decomposition in which mass
is actually lost as a volatile gas, or the material may change
density, either collapsing or swelling. On an atomic scale,
atoms may be dislodged creating vacancies or interstitial
atoms in the host lattice. 100 mm
SE MAG: 200 x HV: 20.0 kV WD: 11.0 mm Px: 0.55 mm
An example of radiation damage on a coarse scale is
illustrated in . Fig. 9.13, which shows a conductive double- b
sided sticky polymer tab of the type that is often used as a Collapsed area
substrate for dispersing particles. This material was found
to be extremely sensitive to electron bombardment. As the
magnification was successively reduced in a series of 20-s
scans, radiation damage in the form of collapse of the
structure at the previous higher magnification scan was
readily apparent after a single 20-s scan (20 keV and 10 nA).
Note that when this tab is used as a direct support for par-
ticles, the susceptibility of the tab material to distortion due
radiation damage can lead to unacceptable image drift.
Instability in the position of the target particle occurs due
to changes in the support tape immediately adjacent to the
particle of interest where electrons strike, directly at the 100 mm
edges of the image raster and as a result of backscattering SE MAG: 200 x HV: 20.0 kV WD: 11.0 mm Px: 0.55 mm
off the particle. Other support materials are less susceptible .. Fig. 9.14 Conducting tape: a Initial image. b Image after a dose of
to radiation damage. . Figure 9.14 shows a detail on a dif- 15 min exposure at higher magnification (20 keV and 10 nA); Everhart–
ferent conductive sticky tape material. After a much higher Thornley (positive bias)
9.3 · Contamination
143 9
dose (15 min of bombardment at 20 keV and 10 nA), a bombardment. Contamination is a manifestation of radia-
much less significant collapse crater is seen to have formed. tion damage in which the material that undergoes radiation
It is prudent to examine the behavior of the support materi- damage is unintentionally present, usually as a result of the
als under electron bombardment prior to use in a particle original environment of the specimen or as a result of inad-
preparation. equate cleaning during preparation. Contamination typically
If radiation damage occurs and interferes with successful arises from hydrocarbons that have been previously depos-
imaging of the structures of interest, the microscopist has ited on the specimen surface, usually inadvertently. Such
several possible strategies: compounds are very vulnerable to radiation damage.
1. Follow a minimum dose microscopy strategy. Hydrocarbons may “crack” under electron irradiation into
a. Radiation damage scales with dose. Use the lowest gaseous components, leaving behind a deposit of elemental
possible beam current and frame time consistent carbon. While the beam can interact with hydrocarbons
with establishing the visibility of the features of present in the area being scanned, electron beam induced
interest. It may be necessary to determine these migration of hydrocarbons across the surface to actually
parameters for establishing visibility for the particu- increase the local contamination has been observed (Hren
lar specimen by operating initially on a portion of 1986). Sources of contamination can occur in the SEM itself.
the specimen that can be sacrificed. However, for a modern SEM that has been well maintained
b. Once optimum beam current and frame time have and for which scrupulous attention has been paid to degreas-
been established, the SEM can be focused and stig- ing and subsequently cleanly handling all specimens and
mated on an area adjacent to the features of interest, stage components, contamination from the instrument itself
and the stage then translated to bring the area of should be negligible. Ideally, an instrument should be
interest into position. After the image is recorded equipped with a vacuum airlock to minimize the exposure of
using the shortest possible frame time consistent the specimen chamber to laboratory air and possible con-
with establishing visibility, the beam should be tamination during sample exchange. A plasma cleaner that
blanked (ideally into a Faraday cup) to stop further operates in the specimen airlock during the pump down
electron bombardment while the stored image is cycle can greatly reduce specimen-related contamination by
examined before proceeding. decomposing the hydrocarbons, provided the specimen itself
2. Change the beam energy is not damaged by the active oxygen plasma that is produced.
Intuitively, it would seem logical to lower the beam A typical observation of contamination is illustrated in
energy to reduce radiation damage, and depending on . Fig. 9.15a, where the SEM was first used to image an area at
the particular material and the exact mechanism of certain magnification and the magnification was subsequently
radiation damage, a lower beam energy may be useful. reduced to scan a larger area. A “scan rectangle” is observed in
However, the energy deposited per unit volume actually the lower magnification image that corresponds to the area
increases significantly as the beam energy is lowered! previously scanned at higher magnification. Within this scan
From the Kanaya–Okayama range, the beam linear rectangle, the SE coefficient has changed because of the depo-
beam penetration scales approximately as E01.67 so that sition of a foreign material during electron bombardment,
the volume excited by the beam scales as (RK-O)3 or E05. most likely a carbon-rich material which has a lower SE coef-
The energy deposited per unit volume scales as E0/E05 or ficient. Note that the contamination is most pronounced at
1/E04. Thus, the volume density of energy deposition the edge of the scanned field, where the beam is briefly held
increases by a factor of 104 as the beam energy decreases stationary before starting the next scanned line so that the
from E0 = 10 keV to E0 = 1 keV. Raising the beam energy greatest electron dose is applied along this edge.
may actually be a better choice to minimize radiation “Etching,” the opposite of contamination, can also occur
damage. (Hren 1986). An example is shown in . Fig. 9.15b, where a
3. Lower the specimen temperature bright scan rectangle is observed in an image of an aluminum
Radiation damage mechanisms may be thermally sensi- stub after reducing the magnification following scanning for
tive. If a cold stage capable of achieving liquid nitrogen several minutes at higher magnification. In this case, the
temperature or lower is available, radiation damage may radiation damage has actually removed an overlayer of con-
be suppressed, especially if low temperature operation is tamination on the specimen, revealing the underlying alumi-
combined with a minimum dose microscopy strategy. num with its native oxide surface layer (~4 nm thick), which
has an increased SE coefficient compared to the carbon-rich
contamination layer.
9.3 Contamination Contamination is usually dose-dependent, so that the high
dose necessary for high resolution microscopy, for example,
“Contamination” broadly refers to a class of phenomena a small scanned area (i.e., high magnification) with a high
observed in SEM images in which a foreign material is depos- current density beam from a field emission gun, is likely to
ited on the specimen as a result of the electron beam encounter contamination effects. This situation is illustrated
144 Chapter 9 · Image Defects
a b
Cu
E0 = 10 keV
.. Fig. 9.15 a Contamination area observed after a higher magnifica- increasing the scanned area. b Etching of a surface contamination layer
tion scan; Everhart–Thornley (positive bias). The extent of the con- observed during imaging of an aluminum stub; Everhart–Thornley
tamination is visible upon lowering the magnification of the scan, thus (positive bias); 10 keV and 10 nA
.. Fig. 9.16 Contamination observed during dimensional measurements performed under high resolution conditions on a patterned silicon
substrate (Postek and Vladar 2014). Note broadening of the structure (right) due to contamination
in . Fig. 9.16, which shows contamination in scanned areas 9.4 oiré Effects: Imaging What Isn’t
M
on a patterned silicon sample used for dimensional metrol- Actually There
ogy (Postek and Vladar 2014). The c ontamination in this case
was so severe that it actually altered the apparent width of An SEM image appears to be continuous, but it is con-
the measured features. To perform successful measurements, structed as a regular repeating two-dimensional pattern of
the authors developed an aggressive cleaning procedure that pixels. Thus, the viewer is effectively looking at the speci-
minimized contamination effects for this class of specimens. men through a two-dimensional periodic grid, and if the
Their strategy may prove useful for other materials as well specimen itself has a structure that has a regularly repeating
(Postek and Vladar 2014). pattern, then a moiré pattern of fringes can form between
9.4 · Moiré Effects: Imaging What Isn’t Actually There
145 9
.. Fig. 9.17 Moiré fringe effects observed for the periodic structures in NIST RM 8820 (magnification calibration artifact). Note the different
moiré patterns in the different calibration regions; Everhart–Thornley (positive bias) detector
the two patterns. The form of the moiré interference fringes the moiré pattern to change. Finally, at sufficiently high
depends on the spacing and orientation of the specimen magnification, the specimen periodic structure becomes
periodic pattern and the scan pattern. Moiré patterns are sufficiently different from the scan pattern that the moiré
maximized when the spatial frequencies of the two patterns fringes are lost.
are similar or an integer multiple of each other (i.e., they are Moiré effects can be very subtle. The periodic bright flares
commensurate). The formation of moiré patterns is illus- at fine edges, as seen in . Fig. 9.18, are moiré patterns created
trated in . Fig. 9.17, which shows various etched patterns in when the fine scale structure approaches the periodicity of
the NIST RM 8820 magnification calibration artifact. The the scan grid.
structures have different spacings in each of the fields To avoid interpreting moiré effects as real structures, the
viewed at the lowest magnification so that different moiré relative position and/or rotation of the specimen and the
patterns are observed in each field. As the magnification is scan grid should be changed. A real structure will be pre-
increased the scan field decreases in size so that the SEM served by such an action, while the moiré pattern will
pattern changes its periodicity (spatial frequency), causing change.
146 Chapter 9 · Image Defects
References
Hren J (1986) Barriers to AEM: contamination and etching. In: Joy D,
Moire effects Romig A, Goldstein J (eds) Principles of analytical electron micros-
at fine edges copy. Plenum, New York, p 353
Postek M, Vladar A (2014) Does your SEM really tell the truth? How
would you know? Part 2. Scanning 36:347
.. Fig. 9.18 Moiré effects seen as periodic bright flares at the edge
of fine structures in NIST RM 8820 (magnification calibration artifact);
Everhart–Thornley (positive bias) detector
147 10
References – 164
100× 1000 1 μm
2000× 50 50 nm
4000× 25 25 nm
10,000× 10 10 nm b
20,000× 5 5 nm
100,000× 1 1 nm
biased Everhart–Thornley (E–T) detector collects a complex bright edges and uniform interiors and are due in part to the
mixture of BSE and SE signals, including a large BSE compo- dominance of the SE1 component that occurs at the edges of
nent (Oatley 1972). The BSE component consists of a rela- structures but which are lost in the pure BSE image of
tively small contribution from the BSEs that directly strike the . Fig. 10.1a.
scintillator (because of its small solid angle) but this direct
BSE component is augmented by a much larger contribution
of indirectly collected BSEs from the relatively abundant SE2 10.4 econdary Electron Contrast at High
S
(produced as all BSEs exit the specimen surface) and SE3 (cre- Spatial Resolution
ated when the BSEs strike the objective lens pole piece and
specimen chamber walls). For an intermediate atomic num- The secondary electron coefficient responds to changes in the
ber target such as copper, the SE2 class created as the BSEs local inclination (topography) of the specimen approximately
emerge constitutes about 45 % of the total SE signal collected following a secant function:
by the E–T(positive bias) detector (Peters 1984, 1985). The
SE3 class from BSE-to-SE conversion at the objective lens pole δ (θ ) = δ 0 secθ (10.1)
piece and specimen chamber walls constitutes about 40 % of
the total SE intensity. The SE2 and SE3, constituting 85 % of the where δ0 is the secondary electron coefficient at normal beam
total SE signal, respond to BSE number effects and create most incidence, i.e., θ = 0°. The contrast between two surfaces at
of the atomic number contrast seen in the E–T(positive bias) different tilts can be estimated by taking the derivative of
image. However, the SE2 and SE3 are subject to the same lat- Eq. 10.1:
eral delocalization suffered by the BSEs and result in a similar
loss of edge resolution. Fortunately for achieving useful high dδ (θ ) = δ 0 sec θ tan θ dθ (10.2)
resolution SEM, the E–T (positive bias) detector also collects
10 the SE1 component (about 15 % of the total SE signal for cop- The contrast for a small change in tilt angle dθ is then
per) which is emitted from the footprint of the incident beam.
The SE1 signal component thus retains high resolution spatial C ~ dδ (θ ) / δ (θ ) = δ 0 sec θ tan θ dθ / δ 0 sec θ
information on the scale of the beam, and that information is = tan θ dθ (10.3)
superimposed on the lower resolution spatial information
carried by the BSE, SE2, and SE3 signals. Careful inspection of As the local tilt angle increases, the contrast between two
. Fig. 10.1b reveals several examples of discrete fine particles adjacent planar surfaces with a small difference in tilt angle,
which appear in much sharper focus than the boundaries of dθ, increases as the average tilt angle, θ, increases, as shown
the Al-Cu eutectic phases. These particles are distinguished by in . Fig. 10.2 for surfaces with a difference in tilt of dθ = 1°, 5°
0.1
0.01
Dq = 10 degrees
Dq = 1 degree
0.001
Dq = 5 degrees
0.0001
0 20 40 60 80
SE1 SE1
SE1 SE2
SE2
SE2
SE signal
Scan position
152 Chapter 10 · High Resolution Imaging
a b
.. Fig. 10.5 a SEM image at E0 = 5 keV of TiO2 particles using a through-the-lens detector for SE1 and SE2 (Bar = 100 nm). b Note bright edge
effects and convergence of bright edges for the smallest particles (Example courtesy John Notte, Zeiss)
excess of SEs compared to the bulk interior. In addition, signal compared to bulk is a significant advantage. This is
there will be enhanced escape of BSEs near the edge, and especially true considering the limitations that are imposed
these BSEs will likely strike other nearby specimen and on high resolution performance by the demands of the
10 instrument surfaces, producing even more SEs. All of the Threshold Current/Contrast Equation, as discussed below.
signals collected when the beam is placed at a given scan
location are assigned to that location in the image no matter
where on the specimen or SEM chamber those signals are 10.4.2 ven More Localized Signal: Edges
E
generated. The apparent SE emission coefficient when the Which Are Thin Relative to the Beam
beam is placed near an edge is thus greatly increased over Range
the bulk interior value, often by a factor of two to ten
depending on the exact geometric circumstances. The edges The enhanced SE escape near an edge shown in . Fig. 10.4
of an object will appear very bright relative to the interior of is further increased when the beam approaches a feature
the object, as shown in . Fig. 10.5 (e.g., objects in yellow edge that is thin enough for penetration of the beam elec-
circles in . Fig. 10.5b) for particles of TiO2. Since the edges trons. As shown schematically in . Fig. 10.6, not only are
are often the most important factor in defining a feature, a additional SEs generated as the beam electrons emerge as
contrast mechanism that produces such an enhanced edge “BSEs” through the bottom and sides of the thin edge
SE1
SE1 SE1
SE2 SE1 SE2
SE2 SE2 SE2
SE1
BSE
BSE SE2
BSE SE2
SE2
SE1 BSE SE1
10.4 · Secondary Electron Contrast at High Spatial Resolution
153 10
structure, but these energetic BSEs will continue to travel, 10.4.3 oo Much of a Good Thing: The Bright
T
backscattering off other nearby specimen surfaces and the Edge Effect Can Hinder
SEM lens and chamber walls, producing additional genera- Distinguishing Shape
tions of SEs at each surface they strike. These additional SEs
will be collected with significant efficiency by the E–T (pos- As the dimensions of a free-standing object such as a particle
itive bias) detector and assigned to each pixel as the beam or the diameter of a fiber approach the secondary electron
approaches the edge, further increasing the signal at a thin escape length, the bright edge effects from two or more edges
edge relative to the interior and thus increasing the contrast will converge, as shown schematically in . Fig. 10.7 and in
of edges. the image of TiO2 particles (e.g., objects in magenta circles)
SE1 SE1
SE1
SE1 SE2 SE2
SE2 SE2
SE signal
Scan position
b
SE1 SE1 SE1
SE
SE1SE2 SE SE2 1
2 SE2
SE2 SE1
SE2
SE1
SE1
SE1
Scan position
154 Chapter 10 · High Resolution Imaging
shown in . Fig. 10.5b. While the object will appear in high view enabled a best fit estimate of the shape and dimensions
contrast as a very bright feature against the background, of the line. The structure was subsequently cross-sectioned
making it relatively easy to detect, as an object decreases in by ion beam milling to produce the SEM view shown in
size it becomes difficult and eventually impossible to discern . Fig. 10.10b. The best estimate of the structure obtained
the true shape of an equiaxed object and to accurately mea- from the top-down imaging and modeling (red trace) is
sure its dimensions. shown superimposed on the direct image of the cross-section
edges (blue trace), showing excellent correspondence with
this approach.
10.4.4 oo Much of a Good Thing: The Bright
T
Edge Effect Hinders Locating the True
Position of an Edge for Critical
Dimension Metrology
0.8
Intensity(arb. units)
0.6
0.4
0.2
0.0
255 260 265 270 275 280
X (nm)
10.4 · Secondary Electron Contrast at High Spatial Resolution
155 10
.. Fig. 10.10 a Top-down SEM
image of line-width test struc- a
tures; E0 = 15 keV. b Side view of 400
structures revealed by focused
ion beam milling showing esti-
mated shape from modeling of
the top-down image (red trace)
compared with the edges directly
found in the cross sectional
image (blue) (Villarrubia et al.
2015) 300
Linescan #
200
100
0
0 100 200 300 400 500
X (nm)
∼60 nm
156 Chapter 10 · High Resolution Imaging
SE1
SE2
10.5 · Achieving High Resolution with Secondary Electrons
157 10
.. Fig. 10.12 a High resolution achieved at low beam energy, courtesy John Notte, Zeiss). b High spatial resolution achieved at low
E0 = 1 keV: image of carbon nanofibers. Note broad fibers (cyan arrows) landing energy: SnO2 whisker imaged with a landing energy of 0.2 keV
with bright edges and darker interiors and thin fibers (yellow arrows) (left, Bar = 100 nm) (right, Bar = 10 nm) (Images courtesy V. Robertson,
for which the bright edge effects converge (Bar = 200 nm) (Example JEOL)
energy is reduced so that E0 ≤ 2 keV, the situation illustrated There are limitations of low beam energy operation that
in . Fig. 10.11b is reached for a finely focused beam (Joy must be acknowledged (Pawley 1984). An inevitable con-
1984; Pawley 1984). The BSE-SE2-SE3 distributions collapse sequence of low beam energy operation is the linear reduc-
onto the SE1 distribution, and all the signals now represent tion in source brightness, which reduces the current that is
high spatial resolution information. An example of carbon contained in the focused probe which in turn affects feature
nanofibers imaged at E0 = 1 keV to achieve high resolution is visibility. Low energy beams are also more susceptible to inter-
shown in . Fig. 10.12a. In . Fig. 10.12a, the edges of the ference from outside sources of electromagnetic radiation.
wider fibers appear bright (e.g., blue arrows) relative to the
interior, as shown schematically in . Fig. 10.7a. . Figure 10.12 High Beam Energy Strategy
also illustrates the convergence of the bright edges of the As the beam energy is increased, the electron range increases
narrow fibers (e.g., yellow arrows), as illustrated in rapidly as E01.67, broadening the spatial distribution of the
. Fig. 10.12b, to produce a very bright object against the BSE-SE2-SE3 signals while the SE1 distribution remains fixed
background. to the beam footprint. For example, when the beam energy
By applying a negative potential to the specimen, the is increased from 10 to 30 keV, the range increases by a fac-
landing energy can be reduced even further while preserving tor of 6.3. With sufficient broadening, the spatial distribu-
high spatial resolution, as shown in . Fig. 10.12b for tin tions of the BSE-SE2-SE3 signals do not significantly respond
oxide whiskers imaged with a TTL SE detector at a landing during beam scanning to small-scale features to which the
energy of E0 = 0.2 keV. SE1 are sensitive. The BSE-SE2-SE3 signals then represent
158 Chapter 10 · High Resolution Imaging
Grid bar,
e.g., Cu
Deep
blind
hole in
carbon
b
Everhart-Thornley
detector
Grid bar,
e.g., Cu
On-axis
bright field
detector
Annular
dark-field
detector
160 Chapter 10 · High Resolution Imaging
50 mm 5 mm
.. Fig. 10.16 SEM imaging glass shards deposited on a thin (~ 10-nm carbon) at E0 = 20 keV and placed over a deep blind hole in a carbon block
(K411 glass) in . Fig. 10.16. By selecting operation at the high- will appear dark. Portions of the specimen that do scatter suf-
est beam energy available, for example, 20–30 keV, backscat- ficiently will appear bright. Since elastic scattering depends
tering will be minimized along with the SE2 and SE3 signals. strongly on local atomic number, compositional effects can
be observed in the dark field STEM image. An example of a
high resolution STEM-in-SEM image created with an annu-
canning Transmission Electron Microscopy
S
lar off-axis detector is shown in . Fig. 10.17.
in the Scanning Electron Microscope
(STEM-in-SEM)
The “thin film” support method for nanoscale particles and
other thin specimens (either inherently thin or prepared as
thin sections by ion beam milling) can be further exploited
by collecting the beam electrons that transmit through the
specimen to create a scanning transmission electron micro-
scope (STEM) image, as illustrated in . Fig. 10.15b. To create
the STEM image, an appropriate detector, such as a passive
scintillator-photomultiplier, is placed below the specimen
grid on the optical axis. The size of this detector is such that
it accepts only electrons traveling close to the optical axis that
pass through the specimen unscattered. Those electrons that
experience even a small angle elastic scattering event that
causes an angular deviation of a few degrees will miss the
detector. Thus, the regions of the specimen with minimal
scattering will appear bright, while those with sufficient mass
to cause elastic scattering will appear dark, creating a “bright-
field” image. A more elaborate STEM detector array can
include an annular ring detector co-mounted with the cen-
tral on-axis bright-field detector to capture the elastically
scattered transmitted electrons from the specimen, as illus- .. Fig. 10.17 Dark-field annular detector STEM image of BaFe12O19
nanoparticles; E0 = 22 keV using oriented dark-field detector in the
trated in . Fig. 10.15b. This off-axis annular detector pro- Zeiss Gemini SEM. The 1.1-nm (002) lattice spacing is clearly evident
duces a “dark-field” image since the thin regions such as the (Image courtesy of John Notte, Carl Zeiss. Image processed with
support film that do not produce significant scattering events ImageJ-Fiji CLAHE function)
10.5 · Achieving High Resolution with Secondary Electrons
161 10
.. Fig. 10.18 Schematic
cross section of a STEM-in-SEM
detector that makes use of the
Everhart–Thornley(positive bias) Short
detector to form a bright-field
working Through the lens detector: SE image
STEM image
distance, ~ 3 mm
C-film and
support grid Scattered
beam
Aperture electrons
E-T detector: “STEM Image”
Electrons
scattered by Unscattered beam electrons
particle are
cut off by
aperture
Au surface
a b
500 nm
.. Fig. 10.19 Aerosol contamination particles deposited on lacey-carbon film and simultaneously imaged with a TTL detector for SE1 and the
STEM-in-SEM detector shown in . Fig. 10.18 (Example courtesy John Small, NIST)
A simple STEM-in-SEM bright-field detector can be STEM image collected with the E–T(positive bias) detector,
created as shown in . Fig. 10.18. The grid carrying the thin as shown for particles supported on a lacey-carbon film in
specimen is placed over an aperture that serves to stop elec- . Fig. 10.19.
trons that have suffered an elastic scattering event in the
specimen. The unscattered beam electrons pass through this
aperture and strike a gold-covered surface below, where they 10.5.3 liminate the Use of SEs Altogether:
E
generate strong SE emission, which is then attracted to the “Low Loss BSEs“
E–T(positive bias) detector, creating a bright-field image. If
the SEM is also equipped with a TTL detector, the nearly pure BSEs are usually considered a low resolution signal because
SE1 image that arises from a thin specimen can be collected of the substantial delocalization that results from multiple
with the TTL detector simultaneously with the bright-field elastic scattering of the beam electrons at conventional beam
162 Chapter 10 · High Resolution Imaging
energy, for example, E0 ≥ 10 keV. However, high resolution scattering event greater than 20°, which also has a suitable
SEM can be achieved by eliminating the use of SEs as the azimuthal angular component along the trajectory, can carry
imaging signal and instead relying on the BSEs, specifically the beam electron out of the specimen as a low loss BSE after
those that have lost very little of the initial beam energy. traveling along a short path within the specimen. The energy
Because of the energy loss due to inelastic scattering that filter with an applied potential V + ΔV then serves to deceler-
occurs for high energy beam electrons at a nearly constant ate and exclude BSEs that have lost more than a specified ΔE
rate, dE/ds, with distance traveled in the specimen, low loss of the incident energy. Since the electrons that pass through
BSEs represent beam electrons that have emerged from the the filter have been retarded to a low kinetic energy, the
specimen after traveling very short paths through the detector following the filter must include an acceleration
specimen. These low loss electrons are thus sensitive to spec- field, such as that of the Everhart–Thornley detector, to raise
imen scattering properties very close to the entrance beam the kinetic energy to a detectable level for detection.
footprint and from a very shallow surface region, thus consti- An example comparing TTL SE and LL BSE (10 % energy
tuting a high resolution signal. Wells (1974a, b) first demon- window) images of etched photoresist at low beam energy
strated the utility of this approach by using an energy filter to (E0 = 2 keV) is shown in . Fig. 10.21 (Postek et al. 2001). Note
select the “low loss” backscattered electrons (LL BSE) that the enhanced surface detail visible on the top of the resist
had lost less than a specified fraction, for example, 5 %, of the pattern in the LL BSE image compared to the SE image. The
initial beam energy. At normal beam incidence, the LL BSE extreme directionality of the LL BSE detector leads to loss of
fraction of the total BSE population is very low, and their tra- signal on surfaces not tilted toward the detector, resulting in
jectories are spread over a wide angular range, the 2π azi-
muth around the beam, making their efficient collection
difficult. The population of LL BSE can be increased, and
their angular spread greatly decreased, by tilting the speci-
10 men to a high angle, for example, 70° or higher. As shown
schematically in . Fig. 10.20, at this tilt angle a single elastic
E0
BSE
E0 - ∆E
Energy filter
-V + ∆V
.. Fig. 10.21 SE (upper) and low loss BSE (lower) images of photore-
.. Fig. 10.20 Schematic illustration of low loss BSE imaging from a sist at E0 = 2 keV. Note the enhanced detail visible on the surface of the
highly tilted specimen using an energy filter LL-BSE image compared to the SE image (Postek et al. 2001)
10.6 · Factors That Hinder Achieving High Resolution
163 10
poor signal collection on the sides of the steps, which are illu- appropriate to the signal selected, DQE is the detective quan-
minated in the TTL SE detector image. tum efficiency, which includes the solid angle of collection
While the example in . Fig. 10.21 illustrates the utility of for the electrons of interest and the conversion into detected
LL BSE imaging at low beam energy, LL BSE imaging also signal, C is the contrast that the feature produces, and tF is the
enables operation of the SEM at high beam energy (Wells frame time. Equation 10.4 reveals the constraints the micros-
1971), thus maximizing the electron gun brightness to enable copist faces: if the beam current is determined by the require-
a small beam with maximum current. Low loss images pro- ment to maintain a certain beam size and the detector has
vide both high lateral spatial resolution and a shallow sam- been optimized for the signal(s) that the features of interest
pling depth. are likely to produce, then the only factor remaining to
manipulate to lower the threshold contrast is to extend the
dwell time per pixel (tF/NPE). While using longer pixel dwell
10.6 actors That Hinder Achieving High
F times is certainly an important strategy that should be
Resolution exploited, other factors may limit its utility, including speci-
men drift, contamination, and damage due to increased dose.
10.6.1 Achieving Visibility: The Threshold Thus, performing high resolution SEM almost always a
Contrast dynamic tension when establishing the visibility of low con-
trast features between the electron dose needed to exceed the
High resolution SEM involves working with a finely focused threshold of visibility and the consequences of that electron
beam which even when optimized to minimize the effects dose to the specimen.
of aberrations inevitably carries a small current, often as
low as a few picoamperes, because of the restrictions
imposed by the Brightness Equation. The inevitable conse- 10.6.2 Pathological Specimen Behavior
quence of operating with low beam current is the problem
of establishing the visibility of the features of interest The electron dose needed for high resolution SEM even with
because of the restrictions imposed by the Threshold an optimized instrument can exceed the radiation damage
Equation. For a given selection of operating parameters, threshold for certain materials, especially “soft” materials
including beam current, detector solid angle, signal conver- such as biological materials and other weakly bonded organic
sion efficiency, and pixel dwell time, there is always a and inorganic substances. Damage may be readily apparent
threshold of detectable contrast. Features producing con- in repeated scans, especially when the magnification is low-
trast below this threshold contrast will not be visible at the ered after recording an image. If such specimen damage is
pixel density selected for the scan, even with post-process- severe, a “minimum-dose” strategy may be necessary, includ-
ing of the image with various advanced image manipulation ing such procedures as focusing and optimizing the image on
algorithms. It is important to understand that a major con- a nearby area, blanking the beam, translating the specimen to
sequence of the Threshold Equation is that the absence of a an unexposed area, and then exposing the specimen for a
feature in an SEM image is not a guarantee of the absence of single imaging frame.
that feature on the specimen: the feature may not be pro- Another possibility is to explore the sensitivity of the
ducing sufficient contrast to exceed the threshold contrast specimen to damage over a wide range of beam energy. It
for the particular operating conditions chosen. Because of may seem likely that operating at low beam energy should
the action of the “bright edge effect” in high resolution SE minimize specimen damage, but this may not be the case.
images to produce very high contrast, approaching unity, Because the electron range scales as E01.67 and the volume as
between the edges of a feature and its interior, the ready vis- (E01.67)3 while the energy deposited scales as E0, the energy
ibility of the edges of features, while obviously useful and deposited per unit volume scales roughly as
important, can give a false sense of security with regard to
Energy / unit volume = E0 / ( E01.67 ) =1 / E0 4
3
the absence of topographic details within the bulk of a fea- (10.5)
ture. In fact, those weaker topographic features may be pro-
ducing contrast that is below the threshold of visibility. To Thus, increasing the beam energy from 1 to 10 keV lowers
perform “due diligence” and explore the possibility of fea- the energy deposited per unit volume by a factor of approxi-
tures lurking below the threshold of visibility, the threshold mately 10,000. This simplistic argument obviously ignores
contrast must be lowered: the substantial variation in the energy density within the
interaction volume as well as the possibility that some dam-
( 4 ×10 ) N
−18
age mechanisms have an energy threshold for activation that
( coulomb / s = amperes )
PE
iB > (10.4)
(η ,δ ) DQEC 2tF may be avoided by lowering the beam energy. Nevertheless,
Eq. 10.5 suggests that examining the material susceptibility to
where NPE is the number of pixels in the image scan, η and δ damage over a wide range of beam energy may be a useful
are the backscatter or secondary electron coefficients as strategy.
164 Chapter 10 · High Resolution Imaging
Contamination
A modern SEM that is well maintained should not be the
source of any contamination that is observed. The first
requirement of avoiding contamination is a specimen prepa-
ration protocol that minimizes the incorporation of or reten-
tion of contaminating compounds when processing the
specimen. This caution includes the specimen as well as the
mounting materials such as sticky conductive tape. A speci-
men airlock that minimizes the volume brought to atmo- 100 nm
sphere for specimen exchange as well as providing the
important capability of pre-pumping the specimen to remove .. Fig. 10.22 SEM image of nanoparticles showing tearing at the
volatile compounds prior to insertion in the specimen cham- particle edges caused by some source of electromagnetic interference
ber is an important capability for high resolution SEM. The whose frequency is constant and apparently locked to the 60 Hz AC
specimen airlock can also be equipped with a “plasma power
cleaner” that generates a low energy oxygen ion stream for
destruction and removal of organic compounds that produce References
contamination. If contamination is still observed after a care-
ful preparation and insertion protocol has been followed, it is Joy DC (1984) Beam interactions, contrast and resolution in the SEM. J
much more likely that the source of contamination remains Microsc 136:241
10 the specimen itself and not the SEM vacuum system.
Oatley CW (1972) The scanning electron microscope, part 1, the instru-
ment. Cambridge University Press, Cambridge
Pawley JB (1984) Low voltage scanning electron microscopy. J Microsc
Instabilities 136:45
Unstable imaging conditions can arise from several sources. Peters K-R (1984) Generation, collection, and properties of an SE-1
(1) Drift and vibration: The specimen preparation, the enriched signal suitable for high resolution SEM on bulk specimens.
In: Kyser DF, Niedrig H, Newbury DE, Shimizu R (eds) Electron beam
method of attachment to the substrate, the attachment of the interactions with solids for microscopy, microanalysis, and micro-
specimen mount to the stage, and the stage itself must all lithography. SEM, Inc, AMF O’Hare, p 363
have high stability to avoid drift, which is most noticeable at Peters K-R (1985) “Working at higher magnifications in scanning elec-
high magnification, and isolation from sources of vibration. tron microscopy with secondary and backscattered electrons on
Note that some mounting materials such as sticky tape may metal coated biological specimens” SEM/1985. SEM, Inc, AMF
O’Hare, p 1519
be subject to beam damage and distortion when struck by the Postek MT, Vladar AE, Wells OC, Lowney JL (2001) Application of the low-
beam electrons, leading to significant drift. One of the most loss scanning electron microscope image to integrated circuit tech-
stable mechanical stage designs is to be mounted within the nology. Part 1 – applications to accurate dimension measurements.
bore of the objective lens, although such designs severely Scanning 23:298
limit the size of the specimen and the extent of lateral motion Villarrubia J, Vladar A, Ming B, Kline R, Sunday D, Chawla J, List S (2015)
Scanning electron microscope measurement of width and shape of
that can be achieved. (2) Electromagnetic radiation interfer- 10 nm patterned lines using a JMONSEL-modeled library. Ultrami-
ence: A periodic distortion is sometimes observed that is a croscopy 154:15–28
result of interference from various sources of electromag- Wells OC (1971) Low-loss image for surface scanning electron micros-
netic radiation, including emissions from 60-Hz AC sources, copy. Appl Phys Lett 19:232
including emissions from fluorescent lighting fixtures. Rather Wells OC (1974a) Scanning electron microscopy. McGraw-Hill, New York
Wells OC (1974b) Resolution of the topographic image in the SEM.
than being random, this type of interference can synchronize SEM/1974. IIT Research Inst, Chicago, p 1
with the scan and can be recorded. An example of this type of
image defect is shown in . Fig. 10.22. Eliminating this type of
interference and the resulting image defects can be extremely
challenging.
165 11
References – 172
The incident beam energy is one of the most useful parame- will be discussed below, useful SEM imaging can now be
ters over which the microscopist has control because it deter- accomplished down to remarkably low arrival energies at the
mines the lateral and depth sampling of the specimen specimen surface, less than 100 eV. The upper bound for E0 is
properties by the critical imaging signals. The Kanaya– arbitrary, but a reasonable limit is the value discussed in the
Okayama electron range varies strongly with the incident “Low Beam Energy X-Ray Microanalysis” module, where it is
beam energy: found that E0 = 5 keV is the lowest beam energy for which a
useful characteristic X-ray peak can be excited for all ele-
RK − O ( nm ) = ( 27.6 A / Z 0.89 ρ ) E01.67 (11.1) ments of the periodic table, excepting H and He, which do
not produce characteristic X-rays. Thus, the plot of the range
where A is the atomic weight (g/mol), Z is the atomic num- for E0 ≤ 5 keV shown in . Fig. 11.1b will be taken to define
ber, ρ is the density (g/cm3), and E0 (keV) is the incident the range for low beam energy SEM.
beam energy, which is shown graphically in . Fig. 11.1a–c.
The rapid but continuous decrease of the range with E0 shown The characteristics of the secondary electron (SE) and back-
in . Fig. 11.1a raises the question, Where does “low” beam scattered electron (BSE) signals observed in conventional
energy SEM imaging begin? That is, what value of E0 consti- SEM imaging performed at high beam energy (E0 ≥ 10 keV)
tutes the upper bound of “low” beam energy microscopy? As can be summarized as follows: (1) For most elements, η > δ.
6 0.3
Range (mm)
Gold Gold
4 0.2
2 0.1
0 0.0
0 5 10 15 20 25 30
0 1 2 3 4 5
Beam energy (keV)
Beam energy (keV)
Kanaya-Okayama range
c 0.04
Carbon
Aluminum
0.03 Titanium
Iron
Silver
Range (µm)
Gold
0.02
0.01
0.00
0.0 0.2 0.4 0.6 0.8 1.0
Beam energy (keV)
.. Fig. 11.1 Plot of the Kanaya–Okayama range for various elements: a 0–30 keV; b 0–5 keV; c 0–1 keV
11.2 · Secondary Electron and Backscattered Electron Signal Characteristics in the Low Beam Energy Range
167 11
.. Fig. 11.2 Secondary electron
coefficient, δ, as a function of Secondary electron coefficient vs. Beam energy
beam energy for C, Al, Cu, and Au, 3.5
taken from the data of Bongeler
et al. (1993)
3.0 Carbon
Aluminum
Copper
2.5 Gold
1.5
1.0
0.5
0.0
0 1 2 3 4 5
Beam energy (keV)
(2) Although the SE1 are sensitive to surface characteristics the surface characteristics of the material as the beam
within the escape depth of ~ 10 nm (metals), this surface energy is reduced. However, the surface of a material is
sensitivity is diluted by the more numerous SE2 and SE3, often unexpectedly complex. Upon exposure to the atmo-
which compose about 75–85 % of the total SE signal. SE2 and sphere, most “pure” elements form a thin surface oxide
SE3 carry BSE information since they are created by the exit- layer, for example, approximately 4 nm of Al2O3 forms on
ing BSEs at the specimen surface and on the chamber walls. Al. Moreover, this surface layer may incorporate water
Because the BSEs escape from approximately 15 % (high Z) chemically to form hydroxide and/or carbon dioxide to
to 30 % (low Z) of RK–O, BSE depth sensitivity in turn deter- form carbonate, or there may be physical adsorption of
mines the effective sampling of sub-surface information car- these and other compounds from the environment which
ried by the SE2 and the SE3, which is one to two orders of may not evaporate under vacuum. Additionally, there may
magnitude greater than the ~10 nm of the SE1. be unexpected contamination from hydrocarbons depos-
As E0 is reduced into the low beam energy range below ited on the specimen surface which generally arise from the
5 keV, the rapid reduction in the electron range given by environment to which the specimen was exposed prior to
equation 11.1, as shown in . Fig. 11.1 b, strongly influences the SEM. In some cases such contamination may be depos-
the secondary electron coefficient: (1) The fraction of the ited from the SEM vacuum system if sufficient care has not
incident energy lost by the beam electrons near the surface been previously taken to eliminate sources of volatile con-
increases, which in turn increases the production of SEs, so tamination by rigorous specimen cleaning and by pre-
that δ increases as the beam energy is reduced, as shown in pumping in an airlock prior to transferring into the
. Fig. 11.2 for several elements for measurements con- specimen chamber. Complex surface composition is the
ducted in one laboratory. Because of this significant increase likely reason for the wide range of δ values reported by
in SE production in the low beam energy range, generally various researchers measuring a nominally common target,
δ > η, as shown for Au in . Fig. 11.3. In low beam energy as illustrated in . Fig. 11.4 for aluminum, where reported
SEM, backscattering still occurs, but due to their much values of δ span a factor of 4 or more. This is a common
greater abundance SEs generally dominate the signal col- result across the periodic table, as seen in the SE database
lected by the Everhart–Thornley (E-T)(positive bias) detec- compiled by Joy (2012). The strong surface sensitivity of the
tor. (2) As the beam energy decreases, the collapse of the SE and BSE signals at low beam energy to the condition of
lateral and depth ranges increases the fraction of the SE2 the specimen surface means that SEM image interpretation
and SE3 that carry surface information equivalent to the of “real” as-received specimens will be challenging. In situ
SE1. This trend makes the SE image increasingly sensitive to cleaning by ion beam milling in a “dual beam” platform may
168 Chapter 11 · Low Beam Energy SEM
1.2
0.4
0.2
0.0
0 1 2 3 4 5
Beam energy (keV)
11
.. Fig. 11.4 Secondary electron
coefficient, δ, as a function of Aluminum
beam energy for Al (Taken from 4.0
the data of various authors)
Kanter (1961)
Bruining and De Boer (1938)
2.5
Bronstein and Fraiman (1969)
Hilleret et al. (2000)
2.0
1.5
1.0
0.5
0.0
0 1 2 3 4 5
Beam energy (keV)
remove such artifacts. However, even with ion beam clean- high that a monolayer of oxide will form on a reactive sur-
ing, it must be recognized that at the vacuum levels of the face such as Al in a matter of seconds. Thus, while ion beam
conventional “high vacuum” SEM, for example, 10−4 Pa milling may successfully remove contamination, oxide for-
(10−6 torr), the partial pressure of oxygen is sufficiently mation at least at the monolayer level may be unavoidable
11.3 · Selecting the Beam Energy to Control the Spatial Sampling of Imaging Signals
169 11
.. Fig. 11.5 SEM image of a
Ag filter, E0 = 500 eV
silver filter obtained at E0 = 0.5 keV
with a through-the-lens
secondary electron detector;
Bar = 5 µm (Image courtesy of
Keana Scott, NIST)
unless an ultrahigh vacuum instrument is used, where the most frequent choice to achieve high lateral spatial resolu-
chamber pressure is <10−8 Pa (10−10 torr). tion imaging of bulk specimens, as discussed in detail in
the “High Resolution SEM” module. An example of high
spatial resolution achieved at low beam energy is shown in
11.3 electing the Beam Energy to Control
S . Fig. 11.5 for a silver filter material imaged at E0 = 0.5 keV
the Spatial Sampling of Imaging with a “through-the-lens” secondary electron detector.
Signals Unfortunately, there is no simple rule like η vs. Z at high
beam energy for interpreting the contrast seen in this
11.3.1 ow Beam Energy for High Lateral
L image. For example, why does the population of nanoscale
particles appear extremely bright against the general mid-
Resolution SEM
level gray of the bulk background of the silver structure.
These features may appear bright because of local compo-
The electron range controls the lateral spatial distribution
sitional differences such as thicker oxides or there may be a
of the backscattered electrons: 90 % of BSEs escape radially
physical change such as increased surface area for SE emis-
from approximately 30 % RK-O (high Z) to 60 % RK-O (low
sion due to nanoscale roughening.
Z). The lateral spatial distribution of the SE2, which is cre-
ated as the BSE escape through the surface, and the SE3,
which is the BSE-to-SE conversion signal that results when
BSE strike the objective lens pole piece, the stage compo- 11.3.2 ow Beam Energy for High Depth
L
nents, and the chamber walls, effectively sample the same Resolution SEM
spatial range as the BSE. As the incident beam energy is
lowered, the BSE (SE3) and SE2 signal lateral distributions The strong exponential dependence of the beam penetration
collapse onto the SE1 distribution, which is restricted to on the incident energy controls the sampling of sub-surface
the beam footprint, so that at sufficiently low beam energy specimen properties by the BSEs and SEs, which can provide
all of these signals carry high spatial resolution informa- insight on the depth dimension. Observing a specimen as the
tion similar to the SE1. With a modern high performance beam energy is progressively lowered to record systematic
SEM equipped with a high brightness source, for example, changes can reveal lateral heterogeneities in surface compo-
a cold field emission gun or a Schottky thermally assisted sition. . Fig. 11.6 shows such a sequence of images from high
field emission gun, capable of delivering a useful beam beam energy (30 keV) to low beam energy (1 keV) prepared
current into a nanometer or sub- nanometer diameter with an E–T(positive bias) detector where the specimen is an
beam, low beam energy SEM operation has become the aluminum stub upon which was deposited approximately
170 Chapter 11 · Low Beam Energy SEM
30 keV a 20 keV b
10 keV c 5 keV d
11
3 keV e 2 keV f
1 keV g
.. Fig. 11.6 Beam energy series of images of a carbon film, nominally 7 nm thick, deposited on an aluminum SEM stub in the as-received
condition prepared with an E–T(positive bias) detector: a 30 keV; b 20 keV; c 10 keV; d 5 keV; e 3 keV; f 2 keV; g 1 keV; Bar = 800 μm
11.3 · Selecting the Beam Energy to Control the Spatial Sampling of Imaging Signals
171 11
7 nm of carbon shadowed through a grid. The contrast 11.3.3 Extremely Low Beam Energy Imaging
between the carbon and the aluminum behaves in a complex
fashion. The C-Al contrast is only weakly visible above High performance SEMs typically operate down to beam
E0 = 5 keV despite a high electron dose, long image integra- energies below 0.5 keV, with the lower limit depending on
tion, and post-acquisition image processing for contrast the vendor and the particular model. Ultralow beam energies
enhancement. The C-Al contrast increases sharply as the below 0.1 keV can be achieved through different electron-
beam energy decreases below 5 keV, reaching a maximum at optical techniques, including biasing the specimen to
E0 = 2 keV and then decreasing below this energy. The –V. Specimen biasing acts to decelerate the beam electrons
increase in contrast below 5 keV is consistent with the emitted at energy E0 from the column so that the landing
increasing separation between the values of δ for C and Al energy, that is, the kinetic energy remaining when the beam
seen in . Fig. 11.2. The eventual decrease in the C-Al contrast electrons reach the specimen surface, is E0 –eV, where e rep-
below E0 = 2 keV is not consistent with the measurements resents the electronic charge. Ultralow beam energy imaging
plotted in . Fig. 11.2, where the difference between δ for C is illustrated in . Fig. 11.7, where the surface of a silica (SiO2)
and Al actually increases below E0 = 2 keV, which should specimen is imaged at a landing energy of 0.030 keV (30 eV).
increase the contrast. Despite the difficulty in interpreting . Figure 11.8 shows gold islands on carbon imaged with a
these trends in contrast, this example demonstrates that lat- landing energy of 0.01 keV (10 eV). At such low incident
eral differences in the surface can be detected, provided care energy, only the outermost atomic/molecular layers are
is taken to fully explore the image response to changing the probed by the beam.
beam energy parameter.
11
References – 185
12
.. Fig. 12.2 a Uncoated polymer foam imaged (left) with high vacuum imaged with VPSEM, E0 = 20 keV, BSE detector; 60 Pa N2 (left, bar = 500 µm)
SEM, E0 = 4 keV, E-T(+) detector (bar = 200 µm); and (right) VPSEM, (right, bar = 50 µm) (Images courtesy J. Mershon, TESCAN; sample source:
E0 = 20 keV, off-axis BSE detector (bar = 500 µm) (Images courtesy Lehigh Microscopy School)
J. Mershan, TESCAN). b (left and right) Uncoated freshly popped popcorn
5000 35
12.3.2 ontrolling the Water Environment
C
of a Specimen 30
4000
Careful control and preservation of the water content of a Liquid 25
Vapor pressure (Pa)
Ice
Vapor pres (Torr)
Hydrophilic
monolayer on Si
(erythrocyte
membrane)
Hydrophobic
monolayer on Si
(octadecanethiol)
100 µm
with a cooling stage capable of reaching −5 °C to 5 °C. With 1 mm to 10 mm or more before reaching the specimen sur-
careful control of both the pressure of water vapor added face. As illustrated schematically in . Fig. 12.5, elastic scatter-
to the specimen chamber and of the specimen tempera- ing events that occur with the gas molecules along this path
ture, the microscopist can select the relative humidity in cause beam electrons to substantially deviate out of the
the sample chamber so that water can be e vaporated, con- focused beam to create a “skirt”. Even a small angle elastic
densed, or maintained in liquid–gas or solid-gas equilib- event with a 1-degree scattering angle that occurs 1 mm
rium. In addition to direct examination of water-containing above the specimen surface will cause the beam electron to
specimens, experiments can be performed in which the be displaced by 17 μm radially from the focused beam.
presence and quantity amount of water is controlled as a How large is the gas-scattering skirt? The extent of the
variable, enabling a wide range of chemical reactions to be beam skirt can be estimated from the ideal gas law (the density
observed. . Figure 12.4 shows an example of the condensa- of particles at a pressure p is given by n/V = p/RT, where n is
tion of water on a silicon wafer, one side of which was cov- the number of moles, V is the volume, R is the gas constant,
ered with a hydrophobic layer while the other was coated and T is the temperature) and by using the cross section for
with a hydrophilic layer, directly revealing the differences elastic scattering for a single event (Danilatos 1988):
in the wetting behavior on the two applied layers, as well as
the bare silicon exposed by fracturing the specimen.
Rs = ( 0.364 Z / E )( p / T )
1/ 2
L3/ 2 (12.1)
Unscattered
beam intensity
Intensity
Skirt intensity
Position
1 Pa
10 Pa
80 100 Pa
1000 Pa
2500 Pa
Skirt Radius (micrometers)
60
40
30
20
0
0 2 4 6 8 10
Gas Path Length (mm)
12.4 · Gas Scattering Modification of the Focused Electron Beam
179 12
Kanaya–Okayama range equation. For a copper specimen into the skirt. The VPSEM menu of DTSA-II allows selection
and E0 = 20 keV, the full range RK-O = 1.5 μm, which is also a of the critical variables: the gas path length, the gas pressure,
good estimate of the diameter of the interaction volume pro- and the gas species (He, N2, O2, H2O, or Ar). . Table 12.1
jected on the entrance surface. With a beam/interaction vol- gives an example of the Monte Carlo output for the electron
ume footprint radius of 0.75 μm, the gas scattering skirt of scattering out of the beam for a 5-mm gas path length
30-μm radius is thus a factor of 40 larger in linear dimension, through 100 Pa of water vapor. In addition to the radial dis-
and the skirt is a factor of 1600 larger in area than that due tribution, the DTSA II Monte Carlo reports the unscattered
the focused beam and beam specimen interactions. fraction that remains in the focused beam, a value that is
Considering just a 10-nm incident beam diameter (5-nm critical for estimating the likely success of VPSEM imaging,
radius), the gas scattering skirt is 6000 times larger. as described below.
While Eq. 12.1 is useful to estimate the extent of the gas . Figure 12.7a plots the gas scattering predicted by the
scattering skirt under VPSEM conditions, it provides no Monte Carlo simulation for a gas path length of 5 mm and
information on the relative fraction of the beam that remains 100 Pa of O2, presented as the cumulative electron inten-
unscattered or on the distribution of gas-scattered electrons sity as a function of radial distance out to 50 μm from the
within the skirt. The Monte Carlo simulation embedded in beam center. For these conditions the unscattered beam
NIST DTSA-II enables explicit treatment of gas scattering to retains about 0.70 of the beam intensity that enters the
provide detailed information on the unscattered beam elec- specimen chamber. The skirt out to a radius of 30 μm con-
trons as well as the spatial distribution of electrons scattered tains a cumulative intensity of 0.84 of the incident beam
.. Table 12.1 NIST DTSA-II Monte Carlo simulation for 20-keV electrons passing through 5 mm of water vapor at 100 Pa
Ring Inner Radius, Outer radius, Ring area, Electron count Electron Cumulative (%)
μm μm μm2 fraction
0.7
0.6
3 mm GPL
5 mm GPL
10 mm GPL
0.5
0.4
0 10 20 30 40 50
Radial distance from beam center (micrometers)
VPSEM 100 Pa O2
b
1.0
12 0.9
Cumulative electron intensity
0.8
0.7
0.6
3 mm GPL
5 mm GPL
10 mm GPL
0.5
0.4
0 200 400 600 800 1000
current. To capture 0.95 of the total beam current requires gas path length on the skirt radius, which follows a 3/2
a radial distance to approximately 190 μm, as shown in exponent in the scattering Eq. 12.1, can be seen in
. Fig. 12.7b, and the last 0.05 of the beam electrons are . Fig. 12.7 by comparing the plots for 3-, 5-, and 10-mm
distributed out to 1000 μm (1 mm). The strong effect of the gas path lengths.
12.5 · VPSEM Image Resolution
181 12
a b
.. Fig. 12.8 a High resolution SEM imaging of gold deposited on SEM imaging of gold deposited on carbon in VPSEM; 300 Pa N2;
carbon in conventional SEM; E0 = 30 keV; E–T (positive bias) detector E0 = 30 keV; BSE detector (bar = 200 nm) (Image courtesy J. Mershon,
(bar = 200 nm) (image courtesy J. Mershon, TESCAN). b High resolution TESCAN)
12.5 VPSEM Image Resolution of the original beam current reaches the specimen surface
unscattered and contained within the focused beam. The elec-
Remarkably, despite the strong gas scattering and the develop- trons that remain in the beam behave exactly as they would in
ment of the skirt around the focused beam, the image resolu- a high vacuum SEM, creating the same interaction volume and
tion that can be achieved in VPSEM operation is very similar generating secondary and backscattered electrons with exactly
to that for the same specimen imaged at the same incident the same spatial distributions. The electrons that land in the
beam energy in a conventional high vacuum SEM. A compari- scattering skirt also generate secondary and backscattered elec-
son of high vacuum SEM and VPSEM imaging performance trons in response to the local specimen characteristics they
for gold islands on carbon using a modern thermal field emis- encounter, for example, surface inclination, roughness, com-
sion gun SEM is shown in . Fig. 12.8, showing comparable spa- position, an so on, which may be different from the region
tial resolution, as originally demonstrated by Danilatos (1993). sampled by the focused beam. Because these skirt electron
This extraordinary imaging performance in the VPSEM can be interactions effectively arise from a broad, diffuse area rather
understood by recognizing that elastic scattering is a stochastic than a focused beam, they cannot respond to fine-scale spatial
process. As beam electrons encounter the gas molecules and details of the specimen as the beam is scanned. The skirt elec-
atoms in the elevated pressure region, elastic scattering events trons interact over such a broad area that effectively they only
occur, but not every electron suffers elastic scattering immedi- contribute increased noise to the measurement, degrading the
ately. There remains an unscattered fraction of electrons that signal-to-noise ratio of the useful high resolution signal gener-
follows the expected path defined by the objective lens field ated by the unscattered electrons that remain in the focused
and lands in the focused beam footprint identical to the situa- beam. This degraded signal-to-noise does degrade the visibility
tion at high vacuum but with reduced intensity due to the gas threshold, which can be compensated by increasing the beam
scattering events that rob the beam of some of the electrons. As current and/or by increasing the pixel dwell time. The degrada-
the gas scattering path, which is a product of working distance tion of feature visibility due to gas scattering has the most
and the gas pressure, increases, the unscattered fraction of the impact at the short pixel dwell times (high scan rates) that are
beam decreases and eventually reaches zero intensity. The frac- typically selected for rapidly surveying a specimen to search for
tion of unscattered electrons that remain in the beam can be features of interest. The prudent VPSEM microscopist will
calculated by the Monte Carlo simulation in DTSA-II, and an always use long pixel dwell times to reduce the contrast visibil-
example is plotted in . Fig. 12.9. For 20-keV electrons passing ity threshold to ensure that a low-contrast feature of interest
through 10 mm of water vapor at 200 Pa, approximately 20 % can be observed.
182 Chapter 12 · Variable Pressure Scanning Electron Microscopy (VPSEM)
0.8
0.4
0.2
0.0
0 200 400 600 800 1000
Pressure (Pa)
12.6.1 Backscattered Electrons—Passive To utilize the low energy secondary electrons in the VPSEM,
Scintillator Detector a special elevated pressure SE detector that utilizes ioniza-
tion of the environmental gas (gaseous secondary electron
As noted above, the E–T detector, or any other detector detector, GSED) has been developed (Danilatos 1990). As
which employs a high accelerating voltage post-specimen, shown schematically in . Fig. 12.12, an electrode (which
such as the channel plate multiplier, cannot be used at ele- may also serve as the final pressure limiting aperture) in
vated VPSEM pressures due to ionization of the gas atoms close proximity to the electrically grounded specimen is
leading to large-scale electrical breakdown. The passive back- maintained at a modest accelerating voltage of a few hun-
scattered electron detectors, including the semiconductor dred volts positive. The exact value of this applied voltage is
and scintillator detectors, are suitable for elevated pressure, selected so as not to exceed the breakdown voltage for the
since the backscattered electrons suffer negligible energy loss gas species and pressure being utilized. The SE emitted from
while transiting the environmental gas and thus retain suffi- the specimen are accelerated toward this electrode and
cient energy to activate the scintillator without post-specimen undergo collisions with the gas molecules, ionizing them
acceleration. In fact, an added advantage of elevated pressure and creating positive ions and more free electrons. The
VPSEM operation is that the gas discharging allows the bare mean free path for this process is a few tens of micrometers,
scintillator to be used without the metallic coating required depending on the gas pressure and the accelerating voltage,
for conventional high vacuum operation. An example of a so that multiple generations of ionizing collisions can occur
VPSEM image of polymer foam prepared with a large sym- between SE emission from the specimen and collection at
metric BSE detector placed symmetrically above the speci- the positive electrode. Moreover, the electrons ejected from
men is shown in . Fig. 12.10 (left). the gas atoms are also accelerated toward the wire and ion-
. Figure 12.11a shows an example of a BSE image of pol- ize other gas atoms, resulting in a cascade of increasing
ished Raney nickel alloy obtained with a passive scintillator charge carriers, progressively amplifying the current col-
detector in water vapor at a pressure of 500 Pa (3.8 torr) with lected at the electrode by a factor up to several hundred
a beam energy of 20 keV. This BSE image shows composi- compared to the SE current originally emitted from the
tional contrast similar to that observed under high vacuum specimen. While BSE can also contribute to the total signal
conventional SEM imaging. collected at the electrode by collisions with gas molecules,
12.6 · Detectors for Elevated Pressure Microscopy
183 12
.. Fig. 12.10 Uncoated polymer foam imaged under VPSEM conditions at E0 = 20 keV: (left) large solid angle symmetric BSE detector placed
above the specimen; (bar = 500 µm) (right) same area with induced field SE detector (bar = 500 µm) (Images courtesy J. Mershan, TESCAN)
BSE GSED
a 150 mm b 150 mm
.. Fig. 12.11 VPSEM imaging of a polished Raney nickel alloy surface. a backscattered electron detector (BSE). b gaseous secondary electron
detector (GSED). Note the details visible in the shrinkage cavity in the GSED image
the mean free path for gas collisions increases rapidly with obtained with the GSED is shown in . Fig. 12.11b, operat-
increasing electron energy thus decreasing the frequency of ing under VPSEM conditions with water vapor at a pressure
gas ionizations by the BSE. The contribution of the high of 600 Pa.
energy BSE to the current amplification cascade is much Other variants of the GSED have been developed that
less than that of the SE. To make simultaneous use of both make use of other physical phenomena that occur in the
the BSE and SE signals, a detector array such as that shown complex charged particle environment around the beam
in . Fig. 12.13 can be utilized, combining an annular scintil- impact on the specimen, including the magnetic field induced
lator BSE detector with the GSED. An example of the same by the motion of the charged particles and the cathodolumi-
area of the polished Raney nickel alloy simultaneously nescence of certain environmental gases induced by the SE
184 Chapter 12 · Variable Pressure Scanning Electron Microscopy (VPSEM)
150 mm 10 mm
50 mm 2 mm
.. Fig. 12.14 Sequence of VPSEM images showing details revealed at the bottom of a cavity in polished Raney nickel alloy
of the cavity, the SEs generated deep in the cavity create addi-
tional generations of cascade multiplication, increasing the
signal compared to the flat surface of the specimen, making
the cavities appear bright relative to the flat surface.
+V
+ + 30 to
References
+
+ + 600 V Danilatos GD (1988) Foundations of environmental scanning electron
+ 1 - 20 torr microscopy. Adv Electronics Electron Opt 71:109
Danilatos GD (1990) Mechanism of detection and imaging in the ESEM.
J Microsc 160:9
Danilatos GD (1991) Review and outline of environmental SEM at pres-
ent. J Microsc 162:391
Danilatos DD (1993) Introduction to the ESEM instrument. Microsc Res
Tech 25:354
Newbury D (1996) Imaging deep holes in structures with gaseous sec-
ondary electron detection in the environmental scanning electron
.. Fig. 12.15 Effect of extended gas path length created by a cavity microscope. Scanning 18:474
187 13
References – 193
Software is an essential tool for the scanning electron micros- becoming popular, Rasband undertook the development of a
copist and X-ray microanalyst (SEMXM). In the past, soft- Mac version of the tool called NIH Image. Largely to enable
ware was an important optional means of augmenting the cross-platform compatibility and to allow non-Macintosh
electron microscope and X-ray spectrometer, permitting users to run the program, it was again rewritten, this time
powerful additional analysis and enabling new characteriza- using the Java programming language. The result was the first
tion methods that were not possible with bare instrumenta- version of ImageJ in 1997 (Schneider et al. 2012, 2015).
tion. Today, however, it is simply not possible to function as The availability of ImageJ on the Microsoft PC and Unix
an SEMXM practitioner without using at least a minimal platforms as well as Macintosh undoubtedly added to its
amount of software. A graphical user interface (GUI) is an popularity, but just as important was the decision to create an
integral part of how the operator controls the hardware on open software architecture that encouraged contributions
most modern microscopes, and in some cases it is the only from a large community of interested software developers. As
interface. Even many seemingly analog controls such as focus a result, ImageJ benefitted from a prodigious number of code
knobs, magnification knobs, or stigmators are actually digital submissions in the form of macros and plugins as well as edits
interfaces mounted on hand-panel controllers that connect to to the core application itself. Partly to manage this organic
the microscope control computer via a USB interface. growth of the package, partly to reorganize the code base, and
In addition to its role in data acquisition, software is now in part to introduce improvements that could not be added
indispensable in the processing, exploration, and visualiza- incrementally, NIH funded the ImageJ2 project in 2009 to
tion of SEMXM data and analysis results. Fortunately, most overhaul this widely useful and very popular program, and to
manufacturers provide high-quality commercial software create a more robust and more capable foundation for future
packages to support the hardware they sell and to aid the enhancements (7 http://imagej.net/ImageJ2).
analyst in the most common materials characterization tasks. Both ImageJ and ImageJ2 have benefitted from indepen-
Usually this software has been carefully engineered, often at dent software development projects that interoperate with
great cost, and smart analysts will take advantage of this soft- these programs. The Bio-Formats file I/O library as well as
ware whenever it meets their needs. However, closed-source other related projects led by the Laboratory for Optical and
commercial software suffers from several limitations. Because Computational Instrumentation (LOCI) at the University of
the source code is not available for inspection, the procedures Wisconsin (7 https://loci.wisc.edu) are important resources
and algorithms used by the software cannot be checked for in the ImageJ universe and have added valuable functional-
accuracy or completeness, and must be accepted as a “black ity. The Bio-Formats project responded to the community’s
box.” Further, it is often very difficult to modify closed source need for software that would read and write the large num-
13 software, either to add missing features needed by the analyst ber of vendor-supplied image file formats, mostly for light
or to customize the workflow to meet specific job require- microscopy (LM). Today the Bio-Formats library goes well
ments. In this regard, open source software is more flexible beyond LM vendor formats and encompasses 140 different
and more extensible. The cost of commercial software pack- file types, including many useful for SEMXM, such FEI and
ages can also be a downside, especially in an academic or JEOL images, multi-image TIFFs (useful for EDS multi-
teaching environment or in any situation where many dupli- element maps), movie formats like AVI for SEM time-lapse
cate copies of the software are required. Clearly a no-cost, imaging, etc. A follow-on LOCI project called SCIFIO aims
open source solution is preferable to a high-cost commercial to extend the I/O library’s scope to include N-dimensional
application if you need to install 50 copies for instructional files (Hiner et al. 2016). Both projects are closely associated
purposes. with the Open Microscopy Environment (OME) project and
One of the most popular free and open source software the OME consortium (7 http://www.openmicroscopy.org).
packages for SEM image analysis is ImageJ, a Java program Similarly, the ImgLib2 project aims to provide a neutral, Java-
that has grown over the decades from a small application based computational library for processing N-dimensional
started at the National Institutes of Health (NIH) into a large scientific datasets of the kind targeted by SCIFIO (Pietzsch
international collaboration with hundreds of contributors et al. 2012).
and many, many thousands of users (7 http://imagej.net). Given the complexity of this rapidly evolving ecosystem
of interrelated and interoperable tools that support ImageJ, it
is not surprising that some users find it difficult to under-
13.1 The ImageJ Universe stand how all the pieces fit together and how to exploit all the
power available in this software suite. Fortunately, there is a
ImageJ has grown into a large and multifaceted suite of simple way to access much of this power: by installing Fiji.
related tools, and how all these parts fit together (and which
are useful for SEM and X-ray microanalysis) may not be
immediately obvious. The project began in the late 1970s 13.2 Fiji
when Wayne Rasband, working at NIH, authored a simple
image processing program in the Pascal programming lan- Fiji, which is a recursive acronym that stands for “Fiji Is Just
guage that he called Image. This original application ran only ImageJ,” is a coherent distribution of ImageJ2 that is easy to
on the PDP-11, but in 1987 when the Apple Macintosh II was install and comes pre-bundled with a large collection of u
seful
13.2 · Fiji
189 13
plugins and enhancements to the bare ImageJ2 application run from its own directory as a standalone application.
(Schindelin et al. 2012). It is often thought of as “ImageJ with Installation is as simple as downloading the distribution and
Batteries Included.” The Fiji website provides several conve- unpacking it; Fiji does not use an installer, does not copy
nient installation packages for both the 32-bit and 64-bit ver- shared libraries into destination directories scattered around
sions of Fiji for common operating systems such as Microsoft the file system, and it does not store configuration information
Windows (currently Windows XP, Vista, 7, 8, and 10) and in system databases (e.g., the Windows registry). Because of
Linux (on amd64 and x86 architectures). Pre-built and tested this design, once installed it can be moved or copied simply by
versions for Mac OS X 10.8 (Mountain Lion) and later are also moving the directory tree. This portability also means it runs
available. By default, these bundles include a version of the quite well from a USB flash drive or removable hard drive.
Java Runtime Environment (JRE) configured for Fiji’s use that After launching Fiji you will be presented with the Fiji
can coexist with other instances of Java on the host computer, main window (. Fig. 13.1a), which contains the Menu bar,
but “bare” distributions of Fiji are available that will attempt the Tools bar, and a Status bar for messages and other appli-
to utilize your computer’s existing JRE if that is preferred. Of cation feedback to the user. Selecting “Update…” or “Update
course, as an Open Source software project, all of the source Fiji” on the Help menu will trigger the updater, one of the
code for Fiji can be downloaded. most useful features of Fiji. Because Fiji is configured by
Installation of Fiji is straightforward because it has been default to start the updater immediately after program
configured as a portable application, meaning it is designed to launch, for many new users this is the first piece of Fiji
b
190 Chapter 13 · ImageJ and Fiji
13
functionality they encounter. Upon activation the updater c ommunity. Fiji comes with some of the most useful plugins
will scan your local Fiji installation and calculate checksums pre-installed, and these are accessible from the Plugins menu
for everything to see if any components are out-of-date, or if item. Hundreds of powerful features are accessible this way,
new features have been added since it was last run. It will exposed to the user in a series of cascading menus and sub-
then confer with the global Fiji code repositories to look for menus. Such a large set of choices can be overwhelming at
updated Java Archive files (.jar files) and offer to download first, but many of the plugins are meant for light microscopy,
and install them for the user. . Figure 13.1b shows an exam- so the SEM analyst may find it simpler to ignore some of
ple of this, where the updater has located numerous changes them. However, the Non-local means denoising plugin, the
in the ImageJ, Fiji, and Bio-Formats repositories. By select- Optic flow plugin, and the myriad of morphological opera-
ing the “Apply changes” button the software will fetch the tions under the Plugins|Process menu are all useful for SEM
latest code and apply all the patches to the user’s local Fiji microscopists, as are the dozens of features in the Registration,
installation. . Figure 13.2 shows a window listing a selection Segmentation, Stacks, Stitching, Transform, and Utilities
of available Fiji update sites illustrating the rich community submenus.
resources. Sometimes the appearance of a plugin as a single entry in
the Fiji menu structure belies the full power of that plugin.
Indeed, some of the most impressive plugins available for Fiji
13.3 Plugins might be considered entire image processing packages in their
own right. An example of this is the Trainable WEKA Classifier
One of the most powerful features of Fiji is the enormous plugin that appears as a single entry on the Segmentation sub-
collection of plugins, macros, and other extensions that have menu of the Plugins menu. WEKA is an acronym that stands
been developed by third-party contributors in the scientific for “Waikato Environment for Knowledge Analysis,” a tool
13.4 · Where to Learn More
191 13
developed by the Machine Learning Group at the University a much larger amount of supplemental code and capability.
of Waikato in New Zealand (Hall et al. 2009). WEKA is a full- For example, the “Cookbook” site listed in . Fig. 13.4a adds
featured and very popular open source software suite written a new top-level menu item to the Fiji main window, as
in Java for machine learning (ML) researchers. It provides an shown in . Fig. 13.4b. This new menu contains example
open, cross-platform workbench for common ML tasks such code to help new users follow along with a community-
as data mining, feature selection, clustering, classification, and written tutorial introduction to ImageJ, available on the
regression, going well beyond just image analysis. The Fiji plu- ImageJ website (7 http://imagej.net/Cookbook).
gin is a gateway into this large array of tools and provides a Occasionally a set of useful plugins will be written by a
convenient interface for processing SEM images using a mod- researcher or contributor who is unable or unwilling to make
ern machine learning framework (7 http://imagej.net/ them available as an update site. The ImageJ website offers
Trainable_Weka_Segmentation). free hosting of update sites for any author of plugins, and
Some of the most widely used and powerful plugins in organizations can run their own Fiji update sites if they wish.
Fiji have been back-ported into ImageJ itself, and are avail- If these are not already a selectable option on the Manage
able directly from the main application’s menu structure. An update sites list (. Fig. 13.4a), the “Add update site” button
example of this is the Process menu option known as allows the user to manually follow a third-party update site.
“Contrast Limited Adaptive Histogram Equalization,” or As a last resort, plugins may also be manually installed into
CLAHE (Zuiderveld 1994). First developed in 1994, this the Fiji plugins directory, but they will not be automatically
algorithm has been implemented in a wide variety of image updated so this is discouraged.
processing tools. It is designed to amplify local contrast by Thus, there are really four tiers of plugins across the
performing histogram equalization on small subsets (tiles) ImageJ universe: (1) core ImageJ plugins that are bundled
within the source image, but to limit the allowed amplifica- into the base ImageJ package (more than 1000 plugins in
tion to reduce the tendency to magnify the noise in relatively 2016); (2) core Fiji plugins, included by default in the
homogeneous patches. . Figure 13.3a shows a scanning elec- “Batteries Included” Fiji distributions (more than 1000 addi-
tron micrograph of microfabricated features on silicon, tional plugins in 2016); (3) plugins available from additional
acquired at 20 keV using an Everhart–Thornley detector. update sites; and (4) plugins that must be located, down-
Because of slight misalignment of the raster with the linear loaded, and installed manually. While this last category of
features, Moiré contrast is evident in the image as bright plugins is the most likely to be buggy and poorly supported,
edges on some features, and there are pure white and pure any plugin written by a co-worker or officemate will often fall
black horizontal lines that have been added to simulate con- into this category, so the code may be highly specific to your
trast artifacts. These extreme limits of intensity preclude the task or your organization—don’t overlook these!
usual brightness/contrast adjustments, but the CLAHE algo-
rithm recovers invisible details without loss of information,
. Fig. 13.3b. 13.4 Where to Learn More
While it is possible that the ideal software tool for your
project is available in Fiji itself (e.g., CLAHE) or in one of Learning ImageJ or Fiji can be a daunting task for the
the many plugins loaded into Fiji by default (e.g., Trainable beginner, and no attempt was made here to provide even a
Weka Segmentation), it is much more likely that the tool basic introduction to opening, exploring, manipulating,
you are looking for is not in the distribution you down- and saving SEM micrographs or X-ray data. However, there
loaded from the Fiji website. Only a small fraction of the are many excellent resources for learning Fiji on the web,
plugins available to the user have been installed in the and the community offers several support channels for
menu tree. A much larger collection awaits the user who is those who need additional help. Fiji itself has a built-in
willing to explore the many optional Fiji update sites. The Help menu with links to the ImageJ and Fiji websites, news-
Updater window shown in . Fig. 13.1b has a “Manage groups, online documentation, example code, developer
update sites” button at the lower left. If you press this but- tools, guidance documents, etc. The ImageJ Help page
ton you are presented with a list of optional plugin reposi- maintains links to the ImageJ Forum, Chat Room, and IRC
tories, as shown in . Fig. 13.4a. When checked, these channel as well as pointers to the ImageJ tag on Stack
additional update sites will be accessed and used by the Overflow and Reddit, popular online locations for ImageJ
Updater to find new functionality to add into the base dis- and Fiji questions and answers. Finally, there is a synoptic
tribution. Some of the sites shown in . Fig. 13.4a only add search engine for many of the above resources at 7 http://
one or two items to the Plugins menu, while others import search.imagej.net.
192 Chapter 13 · ImageJ and Fiji
13
References
193 13
.. Fig. 13.4 a Fiji’s Manage
a
Update Sites window, showing
some of the many optional plugin
repositories available for use. b A
new top-level menu item called
“Cookbook” imported from the
Cookbook update site
with surface tilt (topographic contrast). SE emission increases 14.4 electing the Beam Energy for SEM
S
as the beam energy decreases. Three classes of SEs are recog- Imaging
nized: (1) SE1 are produced as the beam electrons enter the
specimen surface within footprint of the beam, potentially The optimum beam energy depends on the nature of the
carrying high resolution information, and are sensitive to the imaging problem to be solved. The location of the feature (s)
first few nm below the surface. (2) SE2 are produced as beam of interest on the surface or within the specimen; the contrast
electrons exit as BSEs and are actually sensitive to BSE charac- generating mechanism (s), and the degree of spatial resolu-
teristics (lateral and depth sampling). (3) SE3 are produced as tion to be achieved are examples of factors to be considered.
the BSEs strike the objective lens and specimen chamber
walls, and are also sensitive to BSE characteristics (lateral and
depth sampling). SEs are sensitive to electrical and magnetic 14.4.1 Compositional Contrast
fields, and even a few volts of surface potential (“charging”) With Backscattered Electrons
can alter SE trajectories and eventual collection.
Choose E0 ≥ 10 keV: Above 5 keV, electron backscattering
follows a nearly monotonic increase with atomic number,
14.3 Selecting the Electron Detector resulting in easily interpretable compositional contrast (aka
“atomic number contrast”; “Z-contrast”). Because of the
14.3.1 Everhart–Thornley Detector energy threshold of the passive scintillator BSE detector and
(“Secondary Electron” Detector) semiconductor BSE detector (~1 keV to 5 keV), by selecting
E0 ≥ 10 keV the BSE detector will operate reliably with the
Virtually all SEMs are equipped with an Everhart–Thornley energy spectrum of BSEs produced by the specimen. For
detector, often referred to as the “secondary electron (SE)” maximum compositional contrast, a flat polished specimen
detector. While SEs constitute a large fraction of the E–T sig- should be placed at 00 tilt (i.e., perpendicular to the beam).
nal, the E–T detector is also sensitive to BSEs directly and indi-
rectly through the collection of SE2 and SE3. The E–T detector
is the usual choice for imaging problems involving fine spatial 14.4.2 Topographic Contrast
details. The effective collection angle for SEs is nearly 2π sr. With Backscattered Electrons
Some E–T detectors allow user selection of the potential
applied to the SE-collecting Faraday cage so that the SE signal Choose E0 ≥ 10 keV: BSE detectors can respond strongly to
can be minimized or eliminated leaving a BSE signal. This BSE variations in specimen topography, so the same beam energy
signal is collected over a very small solid angle, ~ 0.01 sr. conditions apply as for compositional contrast (7 Sect. 14.4.1)
14 to assure efficient BSE detector response. Local variations in
14.3.2 Backscattered Electron Detectors the specimen surface tilt cause BSEs to travel in different
directions. BSE topographic contrast is maximized by a small
Most SEMs are also equipped with a “dedicated” backscattered BSE detector placed on one side of the beam (e.g., Everhart–
electron detector which has no sensitivity to SEs. Passive scin- Thornley detector with zero or negative Faraday cage bias)
tillator BSE detectors and semiconductor BSE detectors are and minimized by large BSE detectors placed symmetrically
typically placed on the bottom of the objective lens above the around the beam (e.g., large passive scintillator or semicon-
specimen, giving a large solid angle of collection approaching ductor detector).
2π sr. Both types have an energy threshold below which there is
no response, the value of which depends on the particular
detector in use and is typically in the range 1 keV to 5 keV. Above 14.4.3 Topographic Contrast
this threshold, the detector response increases nearly linearly With Secondary Electrons
with BSE energy, creating a modest energy selectivity.
Choose any E0 within the operating range: Topographic con-
trast is usually viewed in “secondary electron” images pre-
14.3.3 “Through-the-Lens” Detectors pared with the E–T detector, positively biased for SE
collection. The E–T detector is designed to efficiently collect
Some high performance SEMs include “through-the-lens” and detect SEs, which are produced at all incident beam
(TTL) detectors which use the strong magnetic field of the energies and are maximized at low beam energy.
objective lens to capture SEs. The collection is restricted to the
SE1 and SE2 signals, with the SE3 component excluded. Since
SE3 actually carry lower resolution BSE information, exclud- 14.4.4 High Resolution SEM Imaging
ing SE3 benefits high resolution imaging. TTL BSE detectors
capture the portion of the BSEs emitted into the bore of the Two beam energy strategies optimize imaging fine-scale
lens. Some TTL SE and TTL BSE detectors can energy filter details by maximizing the contribution of the SE that are pro-
the signal-carrying electrons according to their energy. duced within the footprint of the focused beam:
14.7 · Image Interpretation
199 14
Strategy 1 specimen. The visibility of this contrast depends on satisfying
Choose the highest available beam energy, E0 ≥ 25 keV. The SE1 the Threshold Current Equation:
component of the total SE signal retains the high resolution
information at the scale of the beam entrance footprint. Due to (
I th > 4 pA / δ DQE C2tr t F ) (14.3a)
lateral spreading of the interaction volume, the BSE and their or in terms of the contrast threshold as
associated SE2 and SE3 signals actually degrade spatial resolu-
tion at intermediate beam energy (e.g., 5 keV to 20 keV). As the (
C th > SQRT 4 pA / IB δ DQE t F )
(14.3b)
beam energy increases, the electron range increases as E01.67,
where δ is the secondary electron coefficient (η if imaging
causing the lateral spreading of BSEs to increase. When these
with backscattered electrons), DQE is the detective quantum
signal components are spread out as much as possible by using
efficiency (effectively the fraction of the collected electrons—
the maximum beam energy, their contribution diminishes
detector solid angle and detection—that contribute to the
toward random noise, while the high resolution SE1 contribu-
measured signal), and tF is the frame time (s) for a 1024 by
tion remains. Degraded signal-to-noise means that longer pixel
1024-pixel image. Lower values of Cth can be obtained with
dwell will be necessary to establish visibility of weak contrast.
higher beam current and/or longer frame times. For any selec-
An additional advantage is the improvement in gun brightness,
tion of beam current and frame time, there is always a threshold
which increases linearly with E0, so that more beam current can
contrast below which features will not be visible.
be obtained in the focused beam of a given size.
Contrast (Ctr), Ctr = (S2 – S1)/S2, where S2 > S1, arises when the
properties of a feature (e.g., composition, mass thickness, and/ 14.7.2 Direction of Illumination
or surface tilt) cause a difference in the BSE (η) and/or SE (δ)
thus altering the measured signal, Sfeature = S2, compared to the The apparent source of illumination is from the position of
background signal, Sbackground = S1, from adjacent parts of the the detector. The detector is the apparent flashlight!
200 Chapter 14 · SEM Imaging Checklist
100 µm
15.1 · Case Study: How High Is That Feature Relative to Another?
203 15
100 µm
.. Fig. 15.3 “3D Viewer” rotation of the rendered surface
.. Fig. 15.4 Step height to be measured (yellow arrow)
c
204 Chapter 15 · SEM Case Studies
a b
.. Fig. 15.6 a Use of the single pixel measurement feature in ImageJ-Fiji select the reference pixel (center of the blue circle) on the upper surface
to select the reference pixel (center of the red circle) on the lower surface feature
feature. b Use of the single pixel measurement feature in ImageJ-Fiji to
Parallax = XLeft – Xright = 77.1 μm – 66.9 μm = 10.2 μm (Note BSE rather than in the numbers of the BSE or SE signals, so
that the parallax has a positive sign, so the feature is above the that an appropriate detector should be chosen that empha-
reference point.) sizes the BSE trajectory component of topographic contrast.
(3) Because the shallow relief is likely to provide very few
“clues” as to the sense of the topography, it is critical to estab-
Z = P / 2 sin ( ∆θ ) / 2 = 10.2 µm / 2 sin ( 40 / 2 ) lish a condition of top lighting so that the sense of the local
= 146.1 µm ± 6 % (15.3) topography can be more easily determined. (4) Establishing
15 the visibility of low contrast requires exceeding a high thresh-
old current, so that careful control of beam current will be
Thus, the step represented by the yellow arrow in . Fig. 15.4 necessary. (5) The displayed image must be contrast manipu-
is 146.1 μm ± 6 % above the origin of the yellow arrow. lated to render the low contrast visible to the observer, which
The estimated uncertainty has two major components: an may be challenging if other sources of contrast are present.
uncertainty of 0.10 in the tilt angle difference contributes an An example of the shallow surface relief imaging problem
uncertainty of ± 5 % to the calculated step height. A ± is illustrated by a highly polished specimen with a micro-
1 pixel uncertainty in selecting the same reference pixels for structure consisting of large islands of Fe3C (cementite) in
the lower and upper features in both images contributes ± pearlite (interpenetrating lath-like structures of Fe3C and an
4 % to the calculated step height. iron-carbon solid solution). The strategy for obtaining a use-
ful image of this complex specimen is based on the realiza-
tion that the weak contrast from the shallow topography will
15.2 Revealing Shallow Surface Relief be maximized with the BSE signal detected with a detector
with a small solid angle of collection placed asymmetrically
Surfaces with topographic structures that create shallow sur- relative to the specimen and with a shallow detector elevation
face relief a few tens to hundreds of nanometers above the angle above the surface to produce the effect of oblique illu-
general surface provide special challenges to SEM imaging: mination. The small solid angle means that most BSE trajec-
(1) Shallow topography creates only small changes in the elec- tories not directed into the detector will be lost, which
tron interaction volume and in the resulting emitted second- actually increases the contrast. The asymmetric placement
ary electron (SE) and backscattered (BSE) signals as the beam and shallow elevation angle ensures that the apparent illumi-
is scanned across a feature, resulting in low contrast. (2) The nation will come from a source that skims the surface, creat-
strongest changes in the emitted signals from the weak topo- ing the effect of oblique illumination which creates strong
graphic features will be found in the trajectory effects of the shadows. The Everhart–Thornley (E–T) detector when
15.2 · Revealing Shallow Surface Relief
205 15
biased negatively to reject SEs becomes a small solid angle intensity levels over a larger gray-scale output range. The
BSE detector with these characteristics. The E–T detector is contrast can be interpreted as follows: With the apparent
typically mounted so as to produce a shallow elevation angle illumination established as coming from the top of the
relative to a specimen plane that is oriented perpendicular to image, bright edges must therefore be facing upward, and
the incident beam (0° tilt for a planar specimen). Before pro- conversely, dark edges must be facing away. Thus, the topog-
ceeding with the imaging campaign, the relative position of raphy of the Fe3C islands can be seen to project slightly
the E–T detector is confirmed to be at the 12-o’clock position above the general surface. This situation occurs because the
in the image by using the “scan rotation” function and a Fe3C is harder than the iron-carbon solid solution, so that
specimen with known topography. when this material is polished, the softer iron-carbon solid
. Figure 15.7 shows an image of the iron-carbon micro- solution erodes slightly faster than the harder Fe3C phase,
structure with the negatively biased E–T detector placed at which then stands in slight relief above the iron-carbon solid
the top of the image. A high beam current (10 nA) and a solution.
long dwell time (256 μs per pixel) were used to establish the When this same field of view is imaged with the E–T
visibility of low contrast. The displayed contrast was detector positively biased, . Fig. 15.8a, the same general con-
expanded by first ensuring that the histogram of gray levels trast is seen, but there are significant differences in the fine-
in the raw image was centered at mid-range and did not clip scale details. Several of these differences are highlighted in
at the black or white ends. The “brightness” and “contrast” . Fig. 15.8b. (1) It is much easier to discern the numerous
functions in ImageJ-Fiji were used to spread the input BSE small pits (e.g., yellow circles) in the E–T(positive bias) image
because of the strong “bright edge” effects that manifest along
the lip of each hole. (2) There are small objects (e.g., blue
E-T (-bias)
circles) which appear in the E–T(negative bias) image but
which appear anomalously dark in the E–T(positive bias)
image. These objects are likely to be non-conducting oxide
inclusions that are charging positively, which decreases SE
collection.
When this same field of view is imaged with the annular
semiconductor BSE detector (sum mode, A + B) which pro-
vides apparent uniform, symmetric illumination along the
beam, as shown in . Fig. 15.9, the contrast from the shallow
topography of the edges of the Fe3C islands is entirely lost,
whereas the compositional contrast (atomic number con-
trast) between the Fe3C islands and the iron-carbon solid
solution and Fe3C lamellae is much more prominent.
Finally, when the BSE difference mode (A − B) mode is
used, . Fig. 15.10, the atomic number contrast is suppressed
.. Fig. 15.7 Highly polished iron-carbon specimen imaged at
and the topographic contrast is enhanced. Note that the fea-
E0 = 20 keV and Ip = 10 nA with the negatively biased E–T detector. tures highlighted in the blue circles in . Fig. 15.8b are almost
Image dimensions: 140 × 105 μm (Bar = 30 µm) completely lost.
.. Fig. 15.8 a Same area imaged with a positively biased E–T detector. b Selected features highlighted for comparison
206 Chapter 15 · SEM Case Studies
Two populations
of deposits
500 µm
.. Fig. 15.9 Same area imaged with an annular semiconductor BSE .. Fig. 15.11 SEM-ET (positive bias) image of ink-jet deposits on a
detector (sum mode, A + B) polished carbon substrate with E0 = 5 keV; 32 μs/pixel = 25 s frame time
c d
e f
g h
.. Fig. 15.12 a SEM-ET (positive bias) image of ink-jet deposits on a pol- frame time. e 16 μs/pixel = 12.8 s frame time. f 32 μs/pixel = 25 s frame time.
ished carbon substrate with E0 = 10 keV: 1 μs/pixel = 0.8 s frame time. b 2 μs/ g 64 μs/pixel = 50 s frame time. h 64 μs/pixel = 100 s frame time
pixel = 1.6 s frame time. c 4 μs/pixel = 3.2 s frame time. d 8 μs/pixel = 6.4 s
209 16
References – 234
X-rays
K
L Photo-
electron
Electrons
hγ-Eκ
Hole-electrons
Si K-L3 (3.8eV / Pair)
holes
Si 'dead' layer
Window:
Si3N4 or
polymer supported
on etched Si grid
-V
b
X-rays
hγ K
L Photo-
electron
hγ-Eκ
Hole-electrons
SDD backsurface SiKα Si K-L3 (3.8eV / Pair)
Ring electrodes
SDDs have a complex
back surface electrode
structure. Applied
potential creates
internal radial Resistor bridge
collection channel
Central anode,
80 mm diameter
.. Fig. 16.1 a Basic principle of photon measurement with a semiconductor-based energy dispersive X-ray spectrometer. b Schematic of silicon
drift detector (SDD) design, showing the complex patterned back surface electrode with a small central anode
212 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
Number of photons
Repeated measurements of the same
photon energy produce a range of
values: this is the spectrometer
resolution effect.
Photon energy
100 000
Mn K-L2+3
Mn L2-M4+N1, Mn L3-M4+5 Mn L2-M2+3
Mn_20keV
80 000 Mn
E0 = 20 keV
60 000
Counts
Mn L2-M1+3-M1
40 000
Mn K-M2+3
Si K-L2+3
C K-L2+3
Si K-M2+3
Mn esc
20 000
0
0 2 4 6 8 10
16
Photon energy (keV)
100 000
Mn_20keV
Mn K-L3
80 000
Mn K-L2
60 000
Counts
Mn K-M2+3
40 000
20 000
0
5 000 5 200 5 400 5 600 5 800 6 000 6 200 6 400 6 600 6 800 7 000
Photon energy (keV)
.. Fig. 16.3 SDD-EDS spectrum of Mn with E0 = 20 keV and a time constant that produces a resolution of FWHM = 129 eV at Mn K-L2,3
16.1 · The Energy Dispersive Spectrometry (EDS) Process
213 16
.. Fig. 16.4 EDS “black box”
representation of the Si-escape
peak artifact Artifacts: escape peak
Si K-L3 = 1.74 keV
EDS black box
Input:
X-ray photon
Output:
The EDS estimate of
photon energy is robbed
by the amount of the
Si K-L3 photon energy
(1.74 keV) that is lost.
Number of photons
Actually
placed here!
Photon energy
subject to the EDS resolution function, including the X-ray another photoelectron and further contributing to the charge
bremsstrahlung (continuum) background, but because the generation. However, in a small number of events, as illus-
continuum is created at all energies up to E0 and because of trated schematically in . Fig. 16.4, the Si K-shell X-ray will
its slow variation with photon energy, the distortions intro- escape from the detector, carrying with it 1.740 keV (for a Si
duced into the continuum background by the EDS resolution K-L3 X-ray) and robbing the original photon being captured
function are more difficult to discern. of this amount of energy, which creates an artifact peak at an
The major impact of peak broadening is the frequent energy corresponding to:
occurrence in practical analytical situations of mutually
interfering peaks that arise even with pure elements, for Escape peak energy = Parent peak energy − 1.740 keV (16.3)
example, Si K-L3 and Si K-M3 and the Fe L-family. When
mixtures of elements are analyzed, Interferences are espe- Si-escape peaks are illustrated for tin and gold in . Fig. 16.5.
cially frequent when elements with atomic numbers above 20 The intensity ratio of the Si-escape peak/parent peak depends
are present since these elements have increasingly complex on the energy of the parent photon, with a maximum value
spectra of L- and M- shell X-rays that have a wide energy for this ratio occurring for photon energies just above the Si
span. A secondary impact of peak broadening occurs when K-shell ionization energy (1.838 keV) and decreasing as the
trace elements are to be measured. Peak broadening has the photon energy increases. It is important to identify Si-escape
effect of spreading the characteristic X-rays over a wide range peaks so that they are not mistaken for elements present at
of the X-ray continuum background. Variance in the back- minor or trace levels.
ground sets the ultimate limit of detection.
Sn L3-M4+5
12 000
Sn
10 000 E0 = 20 keV
SnL2-M4 escape
8 000
Sn L2-M4
Counts
C K-L2+3
6 000
Sn M4-N2+5-N3
Sn M3-N5
Sn L1-M3
Sn L3-N5
4 000
Sn L1-N2+3
Sn L1-M2
Sn M2-N1
Sn L2-N4
Sn L3-M1
Sn M3-N1
Sn M2-N4
Sn L2-M1
2 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
16 000 Au_20kV10nA50s
Au M5-N6+7 Au
14 000
E0 = 20 keV
12 000
AuM5-N6,7 escape
Au N4-N6+5-N6
AuM4-N6 escape
10 000
Au M2-N1+3-N4+3-N5
Counts
Au M4-N6
8 000
Au M4-N2+5-N3
6 000
Au M3-O4+5
Au M3-N1
Au M2-N4
4 000
2 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
.. Fig. 16.5 Si-escape peaks observed with an SDD-EDS for Sn and Au (E0 = 20 keV)
count.
Coincidence count
Photon energy
16.1 · The Energy Dispersive Spectrometry (EDS) Process
215 16
Al_20keV14%DT
1,000,000
Al
E0 = 20 keV
800,000 Deadtime = 14%
Counts
600,000
400,000
200,000
0
0 2 4 6 8 10
Photon energy (keV)
18,000
Al_20keV14%DT
coincidence
Al+Al
16,000
14,000
12,000
Counts
10,000
8000
6000
4000
2000
0
0 1 2 3 4 5
Photon energy (keV)
.. Fig. 16.7 Al at E0 = 20 keV. Coincidence peak (Al + Al) observed at dead-time = 14 %. Note proximity of this artifact peak to Ar K-L2,3 and Ag L-M
family
chain to suppress this effect by rejecting the measurement of (which corresponds to the incident beam energy, E0) and
both photons, but as the flux of X-rays increases, an increasing should not be mistaken for the true limit.
frequency of events will occur in which the time separation
between the two events is too short for the anti-coincidence
function to recognize and reject the separate photons, result- 16.1.4 inor Artifacts: Si Absorption Edge
M
ing in an artifact sum photon. This coincidence phenomenon and Si Internal Fluorescence Peak
can occur between any two photons, for example, two charac-
teristic X-rays, a characteristic X-ray plus a continuum X-ray, X-rays entering the EDS must pass through a window, typi-
or two continuum X-rays. Coincidence produces a readily cally a thin polymer, which is often supported on an etched
recognizable artifact peak when coincidence occurs between silicon grid. Some X-rays will be absorbed in this grid silicon,
two photon energies that are particularly abundant, which is especially those whose photon energy is just above the Si
the case for high intensity characteristic X-ray peaks. An K-ionization energy (1.839 keV). In addition, there is a thin
example is shown in . Fig. 16.7, where two Al K-L3 photons inactive Si layer (“dead-layer”) just below the entrance elec-
(1.487 keV) combine to produce a coincidence peak at trode of the EDS that also acts to absorb X-rays. The X-ray
2.972 keV. Coincidence events can be formed from any two mass absorption coefficient of silicon increases abruptly at the
characteristic peaks, for example, O K-L + Si K-L2. It is impor- K-shell ionization energy, and this has the effect of increasing
tant to identify coincidence peaks so that they are not mis- the absorption of the X-ray continuum, producing an abrupt
taken for characteristic peaks of elements present at minor or step. However, the EDS resolution function acts to broaden all
trace levels. Coincidence events involving lower energy pho- photon energies so that this sharp feature is also broadened, as
tons will occur above the Duane–Hunt high energy limit seen in . Fig. 16.8 (after peak fitting for Si) and made into a
216 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
18 000
Cu_20kV9.95nA300s
16 000
14 000
12 000
Counts
10 000
8 000 Cu
E0 = 20 keV
6 000
4 000
2 000
0
0 500 1 000 1 500 2 000 2 500
Photon energy (eV)
18 000
Cu_20kV9.95nA300s
Residual[Cu_20kV9.95nA300s]
16 000
14 000
12 000
Counts
10 000
8 000
6 000
4 000
2 000
0
0 500 1 000 1 500 2 000 2 500
Photon energy (eV)
.. Fig. 16.8 Cu at E0 = 20 keV. The artifact Si peak is a combination of support grid and the Si detector dead-layer, and subsequent Si X-ray
the Si K-absorption edge and the Si internal fluorescence peak (peak emission
16 fitting in lower spectrum) created by absorption of X-rays in the Si
peak-like structure. The absorption of X-rays by the Si grid 16.2.1 Operation of the EDS System
and Si dead-layer ionizes Si atoms and subsequently results in
the emission of Si K-shell X-rays, which contribute a false Si Before commencing any EDS microanalysis campaign, the
peak to the spectrum. In the example for a copper target analyst should follow an established checklist with careful
shown in . Fig. 16.8, the apparent level of Si contributed by attention to the measurement science of EDS operation. To
the internal fluorescence artifact is approximately 0.002 mass establish the basis for quantitative analysis, the EDS param-
fraction. eters must be chosen consistently, especially if the analyst
wishes to use archived spectra to serve as standards.
16.2 “ Best Practices” for Electron-Excited hoosing the EDS Time Constant (Resolution
C
EDS Operation and Throughput)
The EDS amplifier time constant (a generic term which may be
While modern EDS systems are well supported by computer locally known as “shaping time,” “processing time,” “resolu-
automation, there remain parameters whose selection is the tion,” “count rate range,” “1–6,” etc.) should be checked. There
responsibility of the user. are usually at least two settings, one that optimizes resolution
16.2 · “Best Practices” for Electron-Excited EDS Operation
217 16
(at the cost of X-ray throughput) and one that optimizes If the EDS is mounted on a translatable slide that can alter the
throughput (at the cost of resolution). Confirming the desired detector-to-specimen distance, then the user must select a
choice of the time constant is critical for consistent recording specific value for this distance for consistency with archived
of spectra, especially if the analyst is using archived spectra to standard spectra if these are to be used in quantitative analy-
serve as standards for quantitative analysis. This is especially sis procedures. Because of the exponent on the distance
important when the EDS system is in a multi-user environ- parameter r in Eq. (16.4), a small error in r propagates to a
ment, since the previous user may have altered this parameter. much larger error in the solid angle and a proportional devia-
tion in the measured intensity.
Channel Width and Number
The energy width of the histogram bins is typically chosen as electing a Beam Current for an Acceptable
S
5, 10, or 20 eV. The bin energy width determines how many Level of System Dead-Time
bins will define an X-ray peak. Since the peak width is a func- X-rays are generated randomly in time with an average rate
tion of photon energy, as described by Eq. (16.2), decreasing determined by the flux of electrons striking the specimen,
from approximately 129 eV at Mn K-L3 (5.895 keV) to thus scaling with the incident beam current. As discussed
approximately 50 eV FWHM for C K-L2 (0.282 keV), a selec- above, the EDS system can measure only one X-ray photon at
tion of a 5-eV bin width is a useful choice to optimize peak a time, so that it is effectively unavailable if another photon
fitting since this choice will provide 10 channels across C arrives while the system is “busy” measuring the first photon.
K-L2. The number of bins that comprise the spectrum multi- Depending on the separation in the time of arrival of the sec-
plied by the bin width gives the energy span. It is useful to ond photon, the anti-coincidence function will exclude the
capture the complete energy spectrum from a threshold of second photon, but if the measurement of the first photon is
0.1 keV to the incident beam energy, E0. Thus, to span not sufficiently advanced, both photons will be excluded
0–20 keV with 5 eV bins requires 4096 channels. from the measurement and effectively lost. Due to this pho-
ton loss, the output count rate (OCR) in counts/second of the
Choosing the Solid Angle of the EDS detector will always be less than the input count rate (ICR).
The solid angle Ω of a detector with an active area A at a dis- The relation between the OCR and ICR is shown in . Fig. 16.9
tance r from the specimen is for a four-detector SDD-EDS. An automatic correction func-
tion measures the time increments when the detector is busy
W = A / r2 (16.4) processing photons, and to compensate for possible photon
loss during this “dead-time,” additional time is added at the
conclusion of the user-specified measurement time so that all
Deadtime
40
127.5 eV at MnKα
400,000
ICR (c/clock-s) 30
OCR (Medium)
20
200,000
Deadtime
10
Single SDD chip: maximum output ~ 130 kHz
0 0
0 500 kHz 1MHz 1.5 MHz 2 MHz 2.5 MHz
Input count rate
218 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
measurements are made on the basis of the same “live-time” time over the full dead-time range to 80 % or higher. (Note
so as to achieve constant dose for quantitative measurements. that as a component of a quality measurement system, the
The level of activity of the EDS is reported to the user as a dead-time correction function should be periodically
percentage “dead-time”: checked by systematically changing the beam current and
comparing the measured X-ray intensity with predicted.)
Deadtime ( % ) = ( ICR − OCR ) / ICR ×100% (16.5) However, as the dead-time increases and the arrival rate of
X-rays at the EDS increases, coincidence events become
Dead-time increases as the beam current increases. The progressively more prominent. This effect is illustrated in
dead time correction circuit can correct the measurement . Fig. 16.10 for a sequence of spectra from a glass with six
80 000 Ca 0.109
Fe 0.0774
60 000
40 000
20 000
0
0 2 4 6 8 10
Photon energy (keV)
Coincidence Coincidence
14 000
peaks peaks
16 12 000
SiK+O K
2MgK
2AlK
10 000
2SiK
8 000
20DT O 0.428
Mg 0.117
6 000 16DT
Al 0.0491
12DT Si 0.212
4 000
10DT Ca 0.109
Fe 0.0774
2 000 7DT
0
0 2 4 6 8 10
Energy (keV)
.. Fig. 16.10 Development of coincidence peaks as a function of dead- peaks for O, Mg, Al, Si, Ca, and Fe. (lower) SDD-EDS spectra recorded over
time. NIST Standard Reference Material SRM 470 (Mineral Glasses) K412. arrange of dead-times showing in-growth of coincidence peaks. Note
(upper) SDD-EDS spectrum at 7 % dead-time showing the characteristic elemental misidentifications that are possible
16.3 · Practical Aspects of Ensuring EDS Performance for a Quality Measurement Environment
219 16
major constituents (O, Mg, Al, Si, Ca, and Fe) measured at setting it too high may make trace element analysis
increasing dead-time. The spectra show the in-growth of a challenging.
series of coincidence peaks as the dead-time increases. With 9. If the ratio is too large, decrease the probe current and
the long pulses of the Si(Li) EDS technology, the pulse re-measure the probe current and the Al spectrum.
inspection function was effective in minimizing coincidence Re-measure the ratio I(Al K + Al K)/I(Al K).
effects to dead-times in the range 20–30 %. There is more 10. Repeat steps 5–10 until a suitable probe current has
vendor-to-vendor variability in SDD-EDS technology. Some been determine.
vendors provide coincidence detection that will permit 11. Finally, note the suitable probe current and use it
dead-times of up to 50 %, while others are restricted to 10 % consistently at the beam energy for which it was
dead-time. Since there is variability among vendors’ SDD determined.
performance, it is useful to perform a measurement to deter-
mine the performance characteristic of each detector. See
the sidebar for a procedure implementing such a procedure. 16.3 ractical Aspects of Ensuring EDS
P
Regardless of dead-time restrictions, an SDD-EDS is still a Performance for a Quality
factor of 10 or more faster than an Si(Li)-EDS for the same
Measurement Environment
resolution. In summary, as a critical step in establishing a
quality measurement strategy, the beam current (for a spe-
The modern energy dispersive X-ray spectrometer is an
cific EDS solid angle) should be selected to produce an
amazing device capable of measuring the energy of tens of
acceptable rate of coincidence events in the worst-case sce-
thousands of X-ray events per second. The spectra can be
nario. This beam current can then be used for all measure-
processed to extract measures of composition with a preci-
ments with reasonable expectation that the dead-time will
sion of a fraction of a weight-percent. However, this poten-
be within acceptable limits.
tial will not be realized if the detector is not performing
kSidebar: Protocol for Determining the Optimal Probe
optimally. It is important to ensure that the detector is
Current and Dead-Time
mounted and configured optimally each time it is used.
Some parameters change infrequently and need only be
Aluminum produces one of the highest fluxes of X-rays per
checked when a significant modification is made to the
unit probe current: With the Al K-shell ionization energy of
detector or the SEM. Other parameters and performance
1.559 keV, a modest beam energy of 15 keV provides an over-
metrics can change from day-to-day and need to be verified
voltage of 9.6 for strong excitation. Al K-L2 is of sufficient
more frequently. The following sections will step through a
energy (1.487 keV) that it has low self-absorption, and at this
series of tests in a rationally ordered progression. The initial
energy the SDD efficiency is also relatively high. The Al K-L2
tests and configuration steps need only be performed occa-
energy is low enough that this peak is also quite susceptible
sionally, for example, when the detector is first commis-
to coincidence events. Pure aluminum thus makes an ideal
sioned or when a significant service event has occurred.
sample for testing the coincidence detection performance of
Later steps, like ensuring proper calibration, should be per-
a detector and for determining the maximum practical probe
formed regularly and a archival record of the results
current for a given beam energy.
maintained.
1. Place a mounted, flat, polished sample of pure Al in the
SEM chamber at optimal analytical working distance.
2. Mount a Faraday cup with a picoammeter in the SEM
chamber. 16.3.1 Detector Geometry
3. Configure the detector at the desired process time.
4. Configure the SEM at the desired beam energy and an In most electron-beam instruments, the EDS detector is
initial probe current. Measure the probe current using mounted on a fixed flange to ensure a consistent sample/
the Faraday cup/picoammeter. detector geometry with a fixed elevation angle. Almost all
5. Collect a spectrum from the pure Al sample with at modern EDS detectors are mounted in a tubular snout with
least 10,000 counts in the Al K peak. the crystal mounted at the end of the snout and the face of
6. Use your vendor’s software (or NIST DTSA-II) to the active detector element perpendicular to the principle
integrate the background-corrected intensity in the Al axis of the snout. The principal axis of the snout is oriented in
K peak (E = 1.486 keV). the instrument such that it intersects with the electron beam
7. Use your vendor’s software (or NIST DTSA-II) to look axis at the “optimal working distance.” This geometry is illus-
for and integrate the background-corrected intensity in trated in . Fig. 16.11.
the Al K + Al K coincidence peak (E = 2.972 keV). Often the detector is mounted on the flange on a sliding
8. Determine the ratio of the integrated intensity I(Al mechanism that allows the position of the detector to trans-
K + Al K)/I(Al K). We desire this ratio to be smaller late (move in and out) along the axis of the snout. The eleva-
than 0.01 (1 %). In some trace analysis situations, it may tion angle is nominally held fixed during the translation but
desirable to have this ratio less than 0.001 (0.1 %). the distance from the detector crystal changes and along with
Setting this limit too low will limit throughput but it the solid angle (Ω) subtended by the detector. The solid
220 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
Objective
lens
t
nou
rs
cto
e te
Sd
Axes X ED
Y
Optimal
Z working ψ:
distance Elevation angle
Sample
Objective
lens
16
Detector
W crystal
Sample Sample-to
-detector
distance
angle is illustrated in the . Fig. 16.12. Moving the detector It is important to be able to maintain a reproducible solid
away from the sample is designed to decrease the solid angle angle through consistent repositioning of the detector. Some
but does not change the elevation angle or the optimal work- slide mechanisms are motorized. Motorized mechanisms
ing distance. should define an “inserted” position and a “retracted”
16.3 · Practical Aspects of Ensuring EDS Performance for a Quality Measurement Environment
221 16
position, which you should test to ensure that the positioning 4. Test the reproducibility of the insertion point by
is reproducible between insertions. Manual mechanisms usu- extracting and inserting the detector and collecting a
ally provide a threaded screw with a manual crank to pull the series of spectra. If the characteristic peak intensities
detector in and out. The threaded screw will usually have a are reproducible (to much better than a fraction of a
pair of interlocking nuts which can be positioned to define a percent) between insertions, the precision is adequate.
consistent insertion position. The procedure in the sidebar
below will allow you to set and maintain a constant solid The take-off angle is the angle at which X-rays exit a flat sam-
angle and thus also a consistent detector collection efficiency. ple in the direction of the detector. For a flat sample mounted
perpendicular to the electron beam at the optimal working
distance, the take-off angle equals the elevation angle. If the
zz Sidebar: Setting a Constant Detector-to-Sample sample is tilted or the sample surface is at a slightly different
Distance to Maintain Solid Angle working distance, then the take-off angle can be computed
1. Locate the pair of lock nuts on the screw mechanism. from the sample tilt, the working distance and the sample-to-
Move the lock nuts to the inner most position on the detector distance. This is shown in . Fig. 16.13. Often you
treaded rod. will hear the terms elevation angle and take-off angle used
2. Insert the detector as close to the sample as possible. interchangeably. It is more precise however to think of the
Ensure that the detector does not touch the interior elevation angle as being a fixed property of the instrument/
of the microscope. The detector snout must be detector geometry and the take-off angle being dependent
electrically isolated (no conductive path) from the upon instrument-specimen configuration.
interior of the chamber to eliminate noise caused by
electrical ground loops. zz Check 1: Verify the Elevation Angle
3. Twist the upper lock nut to limit the motion of the It is critical that your quantitative analysis software has the
detector towards the sample. Tighten the lower nut correct elevation/take-off angle. Matrix correction a lgorithms
to lock the upper nut into position. use the take-off angle to calculate the correct absorption
Objective
lens
ut
sno
tor
Op tec
tim
al w S de
ork ED
Axes X ingd
ista
Y nce
ψ:
Actu Elevation angle
al w
o
Z Sam rking
ple dist
anc
e
Sample tilt
222 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
AP3 AP5
Thickness (µm) 380 265
Rib width (µm) 59 45
Opening width (µm) 190 190 Thickness
.. Fig. 16.15 Window support grid dimensions for two common window types (Source: MOXTEK)
224 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
6. Plot the data to demonstrate the variation of the nished intensities leading to low analytical totals
intensity as a function of position. . Figure 16.16 and sub-optimal quantitative results.
shows the map from a well-oriented detector plotted 3. Typically, the Cu L-family peak is more sensitive
using a thermal color scheme in which red due to absorption by the vacuum window support
represents the highest intensity and blue represents grid’s Si ribs. . Figure 16.15 (source: Moxtek)
zero intensity. . Figure 16.17 shows traverses shows the design of two recent Moxtek support
extracted from the . Fig. 16.16 data on diagonals grids. The vertical sensitivity is usually minimized
representing parallel to the detector axis and by orienting the grid ribs vertically.
perpendicular to the detector axis. Verify that the
most intense region in the intensity plots is in the
center of the image area. zz Sidebar: Processing a “RAW” Spectrum Image with
7. Note the extent of the region of uniform efficiency. ImageJ-Fiji
Variation from ideal uniform sensitivity has 1. Convert the X-ray spectrum image data into a RAW
consequences. file. A RAW file is large binary representation of the
1. Low magnification X-ray spectrum images will data in the spectrum image. Each pixel in the
suffer from reduced intensity towards the edges. spectrum image consists of a spectrum encode in an
2. Point mode X-ray spectrum acquisitions collected integer binary format. The pixels are organized in a
off the optical axis will also suffer from dimi continuous array row-by-row. The size of the file is
typically equal to (channel depth) × (row
dimension) × (column dimensions) × (2 or 4 bytes
per integer value).
De
255.0 2. Import the RAW data file into ImageJ using the
te
cto “Import → Raw” tools to create a “stack” as shown in
r . Fig. 16.18.
3. As imported, the orientation of the stack will depend
upon how the data in the RAW file is organized.
Regardless of the original orientation, you will need to
pivot the data a couple times using the “Image →
0.0 Stack → Reslice” tool. First, to identify the range of
channels that represent the Cu L-family and Cu K-L2,3
intensities. Second, to align the spectrum data with
the Z dimension so that the “Image → Stack →
4m
m
150
100
50
0
0 1 2 3 4 5 6
mm
16.3 · Practical Aspects of Ensuring EDS Performance for a Quality Measurement Environment
225 16
channels to sum together for the Cu L-family and Cu 16.3.5 Count Rate Linearity
K-L2,3 lines. The Cu K-L2,3 lines are often quite dim.
5. Using the “Image → Stack → Reslice” tool (see One of the most important circuits in an X-ray pulse proces-
. Fig. 16.20) rotate the stack until it looks like sor accounts for the time during which the pulse processor is
. Fig. 16.21 busy processing X-ray events. When the pulse processor is
6. Use the “Image → Stack → Z-project” twice processing an X-ray, it is unavailable to process new incoming
(. Fig. 16.22) to extract the range of channels X-rays. This time is called “dead-time.” In contrast, the time
identified in . Fig. 16.19 as associated with the Cu during which the processor is not busy and is available is called
L-family and Cu K-L2,3 lines. “live-time.” The sum of “live-time” and “dead-time” is called
7. Convert the gray-scale image to a thermal scale using “real-time” – the time you would measure using a wall clock.
“Image → Lookup Tables → Thermal.” For calculating the effective probe dose, live-time is the
8. Convert the image to a plot using “Analyze → critical parameter. The effective probe dose consists of those
Surface Plot” electrons which could produce measurable X-rays. So the
9. Extract traverses from the image using the straight effective probe dose equals the live-time times the probe cur-
line tool to define the traverse and then the “Analyze rent. The effective probe dose is always less than the real
→ Plot Profile” tool to extract and plot the data. probe dose, which is the product of the real time and the
Cu L-family Cu K-L2,3
.. Fig. 16.19 The spectrum image perspective as imported into ImageJ-Fiji. The bright strip is the Cu L-family while the much fainter band is
Cu K-L2,3
226 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
2,000,000
Counts
1,500,000
Det 2
1,000,000 Linear (Det 2)
500,000
0
0 10 20 30 40 50 60 70 80 90
Probe dose (nA.s)
36,050
36,000
35,950
35,900
Counts / (nA.s)
35,850
35,800
35,750 Det 2
35,700
35,650
35,600
35,550
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Probe current (nA)
use this value for all your data acquisition. Before each day’s energy. Two points are sufficient to unambiguously calibrate
measurements ensure that the detector is calibrated consis- a linear function. The calibration (peak position) and resolu-
tently by following the same protocol to check and, if neces- tion (peak shape) should be constant with input count rate
sary, recalibrate your detector. (or dead-time), as shown in . Fig. 16.24.
On a modern pulse processor, calibration is usually per- To a very high degree, modern EDS detectors are linear.
formed using the EDS vendor’s software. The software will However if you look carefully in the mid-range of energies,
prompt you to collect a spectrum from an established mate- you many notice the KLM markers may be misaligned by a
rial. The software will examine the spectrum and extract the channel or two. This is evidence that your detector is not per-
positions of various characteristic X-ray features. The soft- fectly linear but this need not represent a true performance
ware will then perform an internal adjustment to center these problem.
features in the correct channels. Most modern pulse proces- Since energy calibration is so critical but is also one of
sors perform a continuous zero offset calibration using a many parameters that should be measured as part of a com-
“zero strobe pulse” the pulse processor adds to the signal plete EDS Quality Control (QC) program, the validation will
stream for diagnostic purposes. Thus the only parameter be discussed in a later section.
they usually adjust when performing a calibration is an elec-
tronic gain. Usually, this involves identifying a single high
energy characteristic line (like the Mn K-L2,3 or the Cu K-L2,3) 16.3.7 Other Items
and adjusting the gain until this feature is centered on the
appropriate channel. The calibration is thus a two-point cali- zz Light Transparency and IR Cameras
bration—either a low energy characteristic line or the zero Most (but not all) modern EDS detectors have a vacuum
strobe at low energy and a second characteristic line at high tight window that is opaque to infrared and visible light.
228 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
Counts
10 000
5 000
0
7.7 7.9 8.1 8.3 8.5 8.7 8.9 9.1
Energy (keV)
150 000
Counts
100 000
50 000
0
0.68 0.78 0.88 0.98 1.08
Energy (keV)
Usually, the window is coated with a thin layer of aluminum 16.3.8 Setting Up a Quality Control Program
(or other metal) to keep light in the chamber from creating a
spurious signal on the EDS detector. An ongoing QC program is a valuable way to demonstrate
Regardless, it is worthwhile to test whether your detector that your data and results can be trusted——yesterday, today,
is sensitive to light. You may be surprised by a pinhole light and tomorrow. If a client ever questions some data, it is useful
leak or a window without an adequate opaque layer. to be able to go back to the day that data was collected and
16 show that your instrument and detector were performing
zz Check 6: Check for SEM Light Sources adequately. A well-designed QC program need not take
The windows on some EDS detectors are not opaque to light much time. A single spectrum from a consistent sample col-
and light in the chamber will produce noise counts particu- lected under consistent conditions is sufficient to identify
larly in low channels. most common failure modes and to document the long-term
1. Enumerate the potential sources of light inside your performance of your detector. A well-designed QC program
SEM. Sources to consider: is likely to save time by eliminating the possibility of collect-
1. An IR camera ing data when the detector is miscalibrated or otherwise
2. Stage position sensors misbehaving.
3. The tungsten filament (essentially a light bulb)
4. Chamber windows zz Check 7: Implement a Quick QC Program
5. Cathodoluminescence from samples like zinc 1. Maintain a sample consisting of a Faraday cup and a
selenide or benitoite. piece of Cu or Mn. Make use of this sample each day
2. Collect a series of spectra and examine these spectra for on which you intend on collecting quantitative EDS
anomalies. data to ensure that the detector is calibrated.
1. When possible collect the spectra without an electron 2. Image the sample at a consistent working distance
beam so there should be no source of X-rays. (the “optimal working distance”), a consistent beam
2. Collect a spectrum with the light source turned off energy, and a consistent probe current.
and a spectrum with the source on. There should be 3. Collect a spectrum from a sample for a consistent
no difference. live-time.
16.3 · Practical Aspects of Ensuring EDS Performance for a Quality Measurement Environment
229 16
4. Process the spectrum to extract the raw intensities in ideal because it is readily available, stable and has
the K-L2,3 and L-family lines, the resolution, the actual both K-L2,3 (~8.04 keV) and L-family (~930 eV)
zero offset and gain, and the total number of counts. lines.
Record and plot these values on a control chart. 55The beam energy should be sufficient to adequately
excite (overvoltage > 2) all the lines of interest in the
Using the QC Tools Within DTSA-II sample.
While it is possible to implement a QC program by combining 55The nominal working distance should be the
an EDS vendor’s software with careful record-keeping, detector’s optimal working distance. The sample
DTSA-II provides tools specifically designed to implement a should be brought into focus (using the stage Z-axis if
basic EDS detector QC program. The tools import spectra, necessary) at this working distance before collecting a
make some basic sanity checks, process the spectrum to extract spectrum.
QC metrics, archive the metrics and report the metrics. 55The nominal probe current is the probe current at
which the spectra will be collected (to within a few
Creating a QC Project percent). Usually, the “use probe current
55 A QC project is an archive of spectra collected from a normalization” will be selected so that all intensities
specified sample under similar conditions on a specified are scaled relative to the probe current measured with
detector/instrument. The spectra are fitted with a a in-lens cup or a Faraday cup.
modeled spectrum and the resulting quantities 55 After creating the project, you will need to go
recorded to track these values over the lifetime of the through the QC tool additional times to add measured
detector. spectra.
55 Creating a QC Project involves specifying a detector
and the conditions which are to be held consistent
(. Fig. 16.25). If you use a single detector with different zz Adding Spectra to a QC Project
process times, either select a constant process time or, 55 To add spectra to a QC project, you first need to
better yet, create individual projects for each process specify which project you will be adding the spectra.
time. Once you select the project, the “material,” “beam
55The material may be simple or complex but energy,” “nominal working distance,” and “nominal
should be robust, durable and provide characteristic probe current” boxes will fill with the associated
lines over a large range of energies. Copper is information (see . Fig. 16.26).
55 Once the project has been selected, you will need to You will specify the detector and QC project along with
specify a spectrum to add to the project. Needless to the fit values that you wish to report. The report will be gen-
say, the spectrum should have been collected on the erated into a new HTML document and the result displayed
correct detector under the conditions specified (see in your system’s default web browser. The report will look
. Fig. 16.27). like . Fig. 16.30, with header information and a series of
55 After selecting the spectrum, the spectrum will be control charts. At the bottom of the report is a table contain-
processed by fitting it to a modeled spectrum shape. ing all the data values that went into creating the control
The resulting fit parameters will be reported and charts. All the values computed when the spectrum was
compared with the fit parameters from previous fits. added to the QC project will be available to display in the
16 The results are organized into columns associated QC report.
with the current fit, the average of all fits, average of
the first (up to) 10 fits and the average of the last (up
to) 10 fits. Review these values to determine whether 16.3.9 Purchasing an SDD
there has been short or long term drift in any of the
fit parameters. The same information is shown in the If you were to survey EDS vendor’s advertisements, you’d
DTSA-II Report table, as shown in . Fig. 16.28. come to conclusion that two hardware characteristics deter-
55 You can also generate reports containing these mine the “best” EDS detector—resolution and detector area.
quantities as tabular values, as shown in . Fig. 16.29, Over the last decade, detector areas have become larger and
and plotted on control charts, as shown in . Fig. 16.30. larger and detector resolutions have improved significantly
too. The performance of Si(Li) detectors scaled poorly with
size because the detector capacitance scaled with size. SDD,
zz Generating QC Reports on the other hand, perform only slightly worse (throughput
A QC Report is a quick way to track the long-term perfor- and resolution) as the detector area increases. As a result,
mance of your detector. QC Reports are also generated using even a basic modern SDD-EDS detector is larger and per-
the QC tool accessible through the “Tools → QC Alien” forms better than the best Si(Li)-EDS detector of a decade
menu item. ago.
16.3 · Practical Aspects of Ensuring EDS Performance for a Quality Measurement Environment
231 16
.. Fig. 16.27 Panel for selection
of measured spectrum for archive
QC Measurement Recorded
Index 920
.. Fig. 16.29 Report of current QC spectrum measurement parameters compared to archival values
Regardless of what the EDS vendors literature tells us, esolution and Peak Position Stability
R
while both of these performance characteristics are impor- with Count Rate
tant, neither is the basis of a well-considered choice of detec- The detector resolution and peak position must not change
tor. Resolution and area are indirect proxies for the appreciably with a variation of a factor of ten or more in
16 performance characteristics that should really drive the deci- X-ray flux.
sion process—good throughput at an adequate resolution The same spectra used to demonstrate linearity can be used
and a large solid angle of detection. to demonstrate peak position and resolution stability. Use
First, a word or two about two characteristics which are DTSA-II’s calibration tool to fit the spectra and extract full
absolutely required for good quantitative analysis. width at half-maximum (FWHM) and channel width values for
Fortunately, almost all modern SDD meet these two impor- each spectrum. Plot the spectra and results as shown in
tant requirements—linearity and stability. . Fig. 16.31.
Having ensured these two basic characteristics, the
inearity of Output Count Rate
L choice of next most important characteristic depends
with Live-Time Dose upon how your detector will be used. If signal quantity is a
The number of X-rays measured must be proportional to the problem because you are limited to low probe currents,
number of X-rays generated. If you generated ten times as STEM mode analysis of microparticles, low beam energy
many X-rays, you should measure ten times as many X-rays. analysis, or another reason why the flux of X-rays is lim-
Otherwise, the k-ratio, the basis of all quantitative analysis, ited, then a detector that maximizes solid angle is impor-
would depend not only upon the composition of the material tant. If on the other hand, you can produce a lot of X-rays,
but also the probe current. then throughput at an adequate resolution is more impor-
Perform the check in section Count Rate Linearity to tant. Regardless, both criteria should be part of your eval-
evaluate a candidate detector’s linearity performance. uation process.
16.3 · Practical Aspects of Ensuring EDS Performance for a Quality Measurement Environment
233 16
Channel width
10,06
10,05 Channel width
10,04
10,03
10,02
10,01
10,00
9,99
9,98
9,97
10/13/11 6/1/12 1/18/13 9/7/13 4/26/14 12/14/14 8/2/15 3/21/16 6/14/16
Cu Kα
131.5 9.999
9.998
131.0
9.997
9.996
130.5
9.995
130.0 9.994
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Probe current (nA)
234 Chapter 16 · Energy Dispersive X-ray Spectrometry: Physical Principles and User-Selected Parameters
Solid Angle for Low X-ray Flux performance of the detector is three-way trade-off between
The fraction of X-rays emitted by the sample that strike the throughput, resolution, and coincidence rate. Typically, this is
detector is proportional to the solid angle. The solid angle is a accomplished by defining an acceptable coincidence rate as dis-
function of both the active area of the detector and the dis- cussed in the section on process time and determining the pro-
tance from the sample to the detector. The detected fraction is cess time that maximizes the throughput at this coincidence
linearly proportional to the active area of the detector but rate. This process time will typically be a slight compromise
inversely proportional to the square of the distance from the from the one that produces the optimal resolution but typically
sample to the detector so the position of the detector is critical. not by more than a few eV FWHM at Mn K-L2,3 (Kα). A few eV
It is not reasonable to assume that a larger area detector will of resolution degradation is usually an acceptable compromise
always produce a larger solid angle. Larger area detectors may as throughput is far more important than resolution for accu-
require larger diameter snouts which may not be able to be rate quantitative EDS microanalysis.
positioned as close to the sample. Larger area detectors may
also produce slightly poorer resolution and/or slightly lower zz Special Case: Low Energy Sensitivity
ultimate throughput due to increases in “ballistic deficit” — If measuring low energy X-rays in the sub-200-eV range is
the spreading of electron packets in the active detector area. particularly important to you, then you should focus your
The only way to know ahead of time what solid angle you criteria on this energy region and understand that to opti-
can expect is to ask they vendor to provide schematics show- mize this regime will likely require compromises to through-
ing how your detector will be positioned in your instrument. put. Find samples similar to the ones you will commonly
The critical parameters are sample-to-detector distance at the measure and use these samples to evaluate the performance
maximum insertion position, the optimal working distance, of the candidate detectors.
the detector area, and the elevation angle. These parameters
can be used within DTSA-II to model the X-ray signal you
can expect to measure from the types of samples and the
probe currents you use.
References
Fiori C, Newbury D (1978) Artifacts in energy dispersive X-ray spectrom-
aximizing Throughput at Moderate
M etry. Scan Electron Microsc 1:401
Resolution Fitzgerald R, Keil K, Heinrich KFJ (1968) Solid-state energy-dispersive
Modern detectors are capable of extraordinary resolutions and spectrometer for electron-microprobe X-ray analysis. Science
159:528
high throughput, though not both at the same time. The best Gatti E, Rehak P (1984) Semiconductor drift chamber – an application of
resolutions are achieved at long pulse process times, which pro- a novel charge transport scheme. Nucl Instr Meth A 225:608
duce poor ultimate throughput. The highest throughputs are Struder L, Fiorini C, Gatti E, Hartmann R, Holl P, Krause N, Lechner P,
achieved at short pulse process times; but, while it may be pos- Longini A, Lutz G, Kemmer J, Meidinger N, Popp M, Soltau H, Van
sible to measure many X-rays per unit time, coincidence events Zanthier C (1998) High-resolution high-count-rate X-ray spectros-
copy with state-of-the-art silicon detectors. Mikrochim Acta Suppl
(pulse pile up) limit the quantitative accuracy. The quantitative 15:11
16
235 17
References – 264
17.1.1 Motivation DTSA-II was designed around common user interface meta-
phors and should feel consistent with other programs on
Reading about a new subject is good but there is nothing like your operating system. It has a main menu which provides a
doing to reinforce understanding. With this in mind, the handful of high-level interactions such as file access, process-
authors of this textbook have designed a number of practical ing, simulation, reporting, and help. Many of these menu
exercises that reinforce the book’s subject matter. Some of items lead to “wizard-style” dialogs which take you step-by-
these exercises can be performed with software you have step through some more complex operation like experiment
available to you—either instrument vendor software or a design, spectrum quantification, or spectrum simulation.
spreadsheet like MS Excel or LibreOffice/OpenOffice Calc. The goal is to make common operations as simple as possible.
Other exercises require functionality which may not be pres- Additional tools are often available through context sen-
ent in all instrument vendor’s software. Regardless, it is much sitive menus. Each region on the DTSA-II main window pro-
easier to explain an exercise when everyone is working with vides different functionality. Within these regions, you are
the same tools. likely to want to perform various context sensitive opera-
To ensure that everyone has the tools necessary to perform tions. These operations are accesses through menus that are
the exercises, we have developed the National Institute of accessed by placing the mouse over the region and issuing a
Standards and Technology software called DTSA-II (Ritchie, “right-button click” on Windows/Linux/UNIX or an Apple-
2009, 2010, 2011a, b, c, 2012, 2017a, b). NIST DTSA-II is Command key + mouse click on OS X. The context-sensitive
quantitative X-ray microanalysis software designed with edu- menu items may perform operations immediately, or they
cation and best practices in mind. Furthermore, as an output may request additional information through dialog boxes.
of the US Federal Government, it is not subject to copyright On the main screen, the bottom half of window is three
restrictions and freely available to all regardless of affiliation tabs—the Spectrum tab, the Report tab, and the Command
or nationality. From a practical perspective, this means that tab. The spectrum tab is useful for investigating and manipu-
you can install and use DTSA-II on any suitable computer. lating spectra. The report tab provides a record of the work
You can give it to colleagues or students. You can use it at completed during this invocation of the program. Since the
home, in your office, or in the lab. If you are so inclined, the report is in HTML (Hypertext Markup Language) and stored
source code is available to allow you to review the implemen- by date, it is possible to review old reports either in DTSA-II
tation or to enhance the tool for your own special purposes. or in a standard web browser. Some operating systems will
Exercises in the textbook will be designed around the index HTML documents making it easy to find old designs,
capabilities of DTSA-II. Many will take advantage of the analyses or simulations.
graphical user interface to manipulate and interrogate spec- The final tab contains a command line interface. The
tra. A number will take advantage of the command line command line interface implements a Python syntax script-
scripting interface to access low-level data or to perform ing environment. Python is a popular, powerful and com-
advanced operations. plete scripting language. Through Python it is possible
(though not necessarily easy) to perform anything that can
be done through the GUI. It is also possible to do a lot more.
For example, the GUI makes available some common, useful
17.1.2 Platform geometries for performing Monte Carlo simulations of spec-
17 trum generation. Through scripting, it is possible to simulate
NIST DTSA-II is written in the multi-platform run-time arbitrary sample and detector geometries. Some of the exam-
environment Java. This means that the same program runs ples in the text will involve scripting. These scripts will be
on Microsoft Windows (XP, Vista, 7, 8.X, 10.X), Apple OS X installed with the software so they are readily available and so
10.6+, and many flavors of Linux and UNIX. The run-time you can use these as the basis for your own custom scripts.
environment adjusts the look-and-feel of the application to An important foundational concept in DTSA-II is the
be consistent with the standards for each operating system. definition of an X-ray detector. The software comes with a
For Windows, Linux, and UNIX, the main menu is part of “default detector” which represents a typical Si(Li) detector
the main application window. In OS X, the main menu is at on a typical SEM. This detector will produce adequate results
the top of the primary screen. for many purposes. However, it is better and more useful if
To make installation as easy as possible on each environ- you create your own detector definition or definitions to
ment, an installer has been developed which works on reflect the design and performance of your detector(s) and
Windows, OS X, and Linux/UNIX. The installer verifies that SEM. You define your own detector using the “Preferences”
an appropriate version of Java is available and place the exe- dialog which is access through the “File → Preferences” main
cutable and data files in a location that is consistent with menu item. To select and activate your detector, you select it
operating system guidelines. Detailed installation instruc- in the “Default Detector” drop down lists on the middle, left
tions are available on the download site. side of the main DTSA-II window on the “Spectrum Tab.”
17.1 · Getting Started With NIST DTSA-II
237 17
17.1.4 Design Simulation
Spectrum generation can be modeled either using analytical
DTSA-II is, in many ways, much more like vendor software models or using Monte Carlo models. The difference is that
used to be. This has advantages and disadvantages. Over the analytical models are deterministic, they always produce the
years, vendors have simplified their software. They have same output for the equivalent input, and they are less com-
removed many more advanced spectrum manipulation tools putationally intensive. They are limited, however, in the
and they have streamlined their software to make getting an geometries for which we know how to perform the analytical
answer as straightforward as possible. If your goal is simply calculation. Monte Carlo models are based on pseudo-
to collect a spectrum, press a button, and report a result, the random simulation of the physics of electron interactions
vendor software is ideal. However, if you want to develop a and X-ray production. Individual electron trajectories are
more deep understanding of how spectrum analysis works, traced as they meander through the sample. Interactions like
many vendors have buried the tools or removed them elastic scattering off the electrons and nucleus in the sample
entirely. DTSA-II retains many of the advanced spectrum are modeled. Inelastic interactions like core-shell ionization
manipulation and interrogation tools. are also modeled. Each core shell ionization is followed by
DTSA-II is designed with Einstein’s suggestion about either an Auger electron or an X-ray photon. The trajectories
simplicity in mind: “Everything should be made as simple as of these can also be modeled. The resulting X-rays can be
possible, but not simpler.” DTSA-II was designed with the collected in a modeled detector and the result presented as a
goal of making the most reliable and accurate means of quan- dose-correct spectrum.
tification, standards-based quantification, as simple as pos- So, in summary, analytical models are quicker, but Montel
sible, but not simpler. When there is a choice that might Carlo models are more flexible. Regardless, in domains where
compromise reliability or accuracy for simplicity, reliability they are both applicable, they produce similar but not identi-
and accuracy wins out. cal results.
One such example is “auto-quant.” Most microanalysis
software will automatically place peak markers on spectra. Quantification
Unfortunately, these markers have time and time again Accurate, reliable quantification is the goal. Turning mea-
been demonstrated to be reliable in many but far from all sured spectra into reliable estimates of material composition
cases. Users grow dependent on auto-quant and when it can be a challenge. Our techniques work well when we are
fails they often don’t have the experience or confidence to careful to prepare our samples, collect our spectra, and pro-
identify the failures. The consequence is that the qualitative cess the data. However, there are many pitfalls and potential
and then since the qualitative results are used to produce sources of error for the novice or the overconfident.
quantitative results, the quantitative results are just plain DTSA-II implements some of the most reliable algo-
wrong. rithms for spectrum quantification. First, DTSA-II assumes
Rather than risking being wrong, DTSA-II requires the that you will be comparing your unknown spectrum to spec-
user to perform manual peak identification. The process is tra collected from standard materials. Standards-based
more tedious and requires more understanding by the user. quantification is the most accurate and reliable technique
But no more understanding than is necessary to judge known. Second, DTSA-II implements robust algorithms for
whether the vendor’s auto-qual has worked correctly. If you comparing peak intensities between standards and unknown.
as a user can’t perform manual qualitative analysis reliably, DTSA-II uses linear least squares fitting of background fil-
you should not be using the vendor’s auto-qual. tered spectra. This algorithm is robust, accurate, and makes
very good use of the all the information present in each
peak. It also provides mechanism called the residual to
determine whether the correct elements have been identified
17.1.5 he Three -Leg Stool: Simulation,
T and fit.
Quantification and Experiment Fitting produces k-ratios which are the first-order esti-
Design mates of composition. To extract the true composition, the
k-ratios must be scaled to account for differences in absorp-
NIST DTSA-II is designed to tie together three tools which tion, atomic number, and secondary fluorescence. DTSA-II
are integral to the process of performing high-quality X-ray implements a handful of different matrix correction algo-
microanalysis—simulation, quantification, and experiment rithms although users are encouraged to use the default algo-
design. Simulation allows you to understand the measure- rithm (‘XPP’ by Pouchou and Pichoir, 1991) unless they have
ment process for both simple measurements and more com- a compelling reason to do otherwise.
plex materials and geometries. Quantification allows you to
turn spectra into estimates of composition. Experiment Experiment Design
design ties together simulation and quantification to allow One thing that has long hindered people from performing
you to develop the most accurate and reliable measurement standards-based quantification is the complexity of design-
protocols. ing an optimal standards-based measurement. The choices
238 Chapter 17 · DTSA-II EDS Software
that go into designing a good measurement are subtle. How DTSA-II was designed around the idea of being able
does one select the optimal beam energy? How does one simulate what you measure. With DTSA-II, it is possible to
select the best materials to use as standards? How does one simulate the full measurement process for both simple and
determine when a reference spectrum1 is needed in addition complex samples. You can simulate the spectrum from an
to the standard spectra? How long an acquisition is required unknown material and from the standard materials neces-
to produce the desired measurement precision? What limits sary to quantify the unknown spectrum. You can quantify
of detection can I expect to achieve? Do I want to optimize the simulated spectra just like you can quantify measured
accuracy or simply precision? Am I interested in minimizing spectra. This ability allows you to understand the measure-
the total error budget or am I interested in optimizing the ment process in ways that are simply not possible otherwise.
measurement of one (or a couple of) elements? It is possible to investigate how changes in sample geometry
In fact, many of these decisions are interrelated in subtle or contamination or coatings will influence the results. It is
ways. Increasing the beam energy will often improve preci- possible to visualize the electron trajectories and X-ray pro-
sion (more counts) but will reduce accuracy (more absorp- duction and absorption.
tion). The best standard for a precision measurement is likely However to do this, it is necessary to be able to model the
a pure element while the best standard for an accurate mea- sample, the physics of electron transport, atomic ionization
surement is likely a material similar to the unknown. and X-ray production and transport, and the detection of
Then we must also consider subtle interactions between X-rays. The physics of electrons and X-rays is not perfectly
elements. If the emission from one element falls near in known, but at least it doesn’t change between one instrument
energy to the absorption edges of another element, accu- and another. The biggest change between instruments is the
racy may be reduced due to complex near edge absorption X-ray detection process. Not all detectors are created equal.
effects. All the different considerations make the mind reel To compensate for the detection process, DTSA-II builds
and intimidate all but the most confident practitioners of algorithmic models of X-ray detectors based on the proper-
the art. ties of the detector. These models are then used to convert the
DTSA-II addresses these problems through an experi- simulated X-ray flux into a simulated measured spectrum.
ment design tool. The experiment design tool calculates the The better these models, the better DTSA-II is able to simu-
uncertainty budget for an ensemble of different alternative late and quantify spectra.
measurement protocols. It then suggests the experiment pro-
tocol which optimizes the uncertainty budget. It outlines Modeled Detectors (. Fig. 17.1)
which spectra need to be acquired and the doses necessary to To make optimal use of DTSA-II, you will need to create a
achieve the user’s desired measurement precision. This is detector model to describe each of your X-ray detectors.
then presented in the report page as a recipe that the analyst Each detector model reflects the performance of a specific
can taking into the laboratory. detector in a specific instrument at a specific resolution/
Experiment design builds upon an expert’s understand- throughput setting. Each physical detector should be associ-
ing of the quantification process and makes extensive use of ated with at least one detector model. A single physical detec-
spectrum simulation. Through spectrum simulation, tor may have more than one detector model if the detector is
DTSA-II can understand how peak interferences and detec- regularly operated at different resolution / throughput set-
tor performance will influence the measurement process. tings.
Through spectrum simulation carefully calibrated to the per- Some of the information necessary to build the detector
formance of your detectors, the experiment optimizer can model is readily available from product literature or from a
17 predict how much dose (probe current x time) is required. call to the vendor. Unfortunately, some pieces of information
Often the result is good news. We often spend much too are less easy to discover. Some require access to very special-
much time on some spectra and too little on others. ized samples or equipment, but fortunately, accepting the
default values won’t overly affect utility of the simulated
results. . Table 17.1 identifies which values are critical and
17.1.6 Introduction to Fundamental which are less critical.
Concepts Detector models are created in the “Preferences” dialog
which is accessed through the “File → Preferences” main
For the most part, the functionality of DTSA-II will be intro- menu item. The tree view on the left side of the dialog allows
duced along with the relevant microanalytical concept. you to navigate through various preference pages. By default,
However, there are a handful of concepts which provide a a root node labeled “Instruments and Detectors” is created
skeleton around which the rest of the program is built. It is with a branch called “Probe” and a leaf node called “Si(Li).”
necessary to understand these concepts to use the program The branch “Probe” reflects a very basic traditional SEM/
effectively. microprobe. The leaf “Si(Li)” reflects a typical lithium-drifted
silicon detector with a ultra-thin window and a resolution of
132 eV at Mn Kα. You can examine the definition of this
1 Don’t worry if you don’t understand the difference between a detector to determine which pieces of information are neces-
reference and a standard spectrum. This will be explained later. sary to fully describe a detector.
17.1 · Getting Started With NIST DTSA-II
239 17
.. Fig. 17.1 The preferences dialog showing a panel containing properties of a detector
.. Table 17.1 This table identifies parameters that have a critical influence on simulated spectra and those that have a less critical
influence. You should be able to determine the correct value of the critical parameters from vendor literature or a call to the vendor.
The vendor may be able to provide the less critical values too but if they can’t just accept the defaults
Crystal thickness
Number of channels
Resolution at Mn Kα (approximate)
Azimuthal angle
240 Chapter 17 · DTSA-II EDS Software
Some pieces of information are specific to your instru- 1 keV. Here the absorption edges in the elements making up
ment and the way the detector is mounted in the instrument. the windows can lead to large jumps in efficiency over nar-
Window-type, detector area, crystal thickness, resolution, row energy ranges. Diamond represents an extreme example
gold layer thickness, aluminum layer thickness, nickel layer in which the absorption edge at 0.283 keV leads to a three
thickness, and dead layer thickness are model specific prop- order-of-magnitude change in efficiency.
erties of the detector. Elevation angle, optimal working dis- Your vendor should be able to tell you the make and
tance, sample-to-detector distance, and azimuthal angle are model of the window on your detector.
determined by how the detector is mounted in your instru-
ment. The energy scale, zero offset, and the resolution are The Optimal Working Distance (. Figs. 17.3
dependent upon hardware settings that are usually config- and 17.4)
ured within the vendor’s acquisition software. The oldest sys- The position and orientation of your EDS detector is optimized
tems may have physical hardware switches. for a certain sample position. Typically, the optimal sample
position is located on the electron-beam axis at an optimal
Window Type (. Fig. 17.2) working distance. At this distance, the effective elevation angle
As is discussed elsewhere, most detectors are protected from equals the nominal elevation angle. Sometimes, the optimal
contamination by an X-ray transparent window. Older win- working distance will be specified in the drawings the EDS ven-
dows were made of ultrathin beryllium foils or occasionally dor used to design the detector mounting hardware (. Fig. 17.4).
boron-nitride or diamond films. Almost all modern detectors Other times, it is necessary to estimate the optimal working
use ultrathin polymer windows although the recently intro- distance finding the sample position that produces the largest
duced silicon nitride (Si3N4) windows show great promise. X-ray flux. The optimal working distance is measured on the
Each type of window has a different efficiency as a func- same scale as the focal distance since the working distance
tion of energy. The largest variation in efficiency is seen below value recorded in spectrum files is typically this value.
a b
1.0 1.0
0.9
0.9
0.8
0.7 0.8
0.6
0.7
0.5
0.4 0.6
0.3
0.5
0.2
0.1 0.4
0.0
10 100 eV 1000 10000 0.3
Beryllium (5 mm) Beryllium (8 mm) 0.2
Beryllium (12 mm) Beryllium (25 mm)
Moxtek Ap 1.3 Moxtek Ap 1.7 0.1
Moxtek Ap 3.3 (model) Moxtek Ap 3.3 (manufacturer’s table) 0.0
Diamond (0.45 mm) Boron nitride (0.25 mm) 10 100 eV 1000 10000
No window Beryllium (5 mm) Beryllium (12 mm)
Beryllium (8 mm)
17 c d
Beryllium (25 mm)
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
10 100 1000 10000 10 100 eV 1000 10000
eV
Moxtek AP 1.3 Moxtek AP 1.7 Beryllium (8 mm) Moxtek Ap 3.3 (model)
Moxtek AP 3.3 (model) Moxtek AP 3.3 (manufacturer’s table) Diamond (0.45 mm) Boron nitride (0.25 mm)
Optimal
working
Optic axis
distance
Nominal elevation
De
te
ct
Effective elevation or
ax
is Actual
working
distance
0.5
69
17
34 16
,01 ,43
3
242 Chapter 17 · DTSA-II EDS Software
The difference between the effective elevation defined by Number of Channels, Energy Scale, and Zero
the actual working distance and the nominal elevation as Offset
defined by the intersection of the detector axis and the optic A detector’s energy calibration is described by three quanti-
axis. The effective elevation angle can be calculated from the ties—the number of bins or channels, the width of each bin
actual working distance given the optimal working distance, (energy scale), and the offset of the zero-th bin (zero offset).
the nominal elevation angle, and the nominal sample-to- The number of bins is often a power-of-two, most often 2,048
detector distance. but sometimes 1,024 or 4,096. This number represents the
number of individual, adjacent energy bins in the spectrum.
Elevation Angle The width of each bin is assumed to be a nice constant—typ-
The elevation angle is defined as the angle between the detec- ically 10 eV, 5 eV, 2.5, or occasionally 20 eV. The detector
tor axis and the plane perpendicular to the optic axis. The electronics are then adjusted (in older systems through phys-
elevation angle is a fixed property of the detector as it is ical potentiometers or in modern systems through digital
mounted in an instrument. The elevation angle is closely calibration) to produce this width.
related to the take-off angle. For a sample whose top surface The zero offset allows the vendor to offset (‘shift’) the
is perpendicular to the optic axis, the take-off angle at the energy scale for the entire spectrum by a fixed energy or to
optimal working distance equals the elevation angle. compensate for a slight offset in the electronics. Some ven-
Elevation angles typically range between 30° and 50° with dors don’t make use of ability and the zero offset is fixed at
between 35° and 40° being the most common. The correct zero. Other vendors use a negative zero offset to measure
detector elevation is important for accurate quantification as the full width of an artificial peak they intentionally insert
the matrix correction has a strong dependence on this into the data stream at 0 eV called the zero-strobe peak.
parameter. The zero-strobe peak is often used to automatically correct
for electronic drift. Often, DTSA-II can read these values
Sample-to-Detector Distance from a vendor’s spectrum file using the “Import from spec-
The sample-to-detector distance is the distance from front face trum” tool.
of the detector crystal to the intersection of the optic and detec- You don’t need to enter the exact energy scale and zero
tor axes. The sample-to-detector distance helps to define the offset when you create the detector as the calibration tool can
solid angle of acceptance for the detector. The sample-to-detec- be used to refine these values.
tor distance can often be extracted from the drawings used to
design the detector mounting hardware (See . Fig. 17.3). Resolution at Mn Kα (Approximate)
Alternatively, you can estimate the distance and adjust the
The resolution is a measure of the performance of an EDS
value by comparing the total integrated counts in a simulated
detector. Since the resolution depends upon X-ray energy in
spectrum with the total integrated counts in an equivalent
a predictable manner, the resolution is by long established
measured spectrum. The simulated counts will decrease as
standard reported as the “full width half maximum” (FWHM)
the square in the increase of the sample-to-detector distance.
of the Mn Kα peak (5.899 keV).
Detector Area The full width at half-maximum is defined as the width of
a peak as measured half way from the base to the peak. This
The detector area is the nominal surface area of the detector
is illustrated in the . Fig. 17.5. The full height is measured
crystal visible (unobstructed) from the perspective of the opti-
from the level of the continuum background to the top of the
mal analysis point. The detector area is one of the values that
17 detector vendors explicitly specify when describing a detector.
peak. A line is drawn across the peak at half the full height.
To account for the finite bin width, a line is drawn on each
Typical values of detector area are 5, 10, 30, 50, or 80 mm2. The
side of the peak from the center of the bin above the line to
detector area does not account for area obstructed by grid bars
the center of the bin below the line. The intersection of this
on the window but does account for area obstructed by a col-
diagonal line is assumed to be the true peak edge position.
limator or other permanent pieces of hardware.
The width is then measured from these intersection points
Crystal Thickness and calibrated relative to the energy scale.
The graphical method for estimating the FWHM is not as
The detection efficiency for hard (higher-energy) X-rays
accurate as numerical fitting of Gaussian line shapes. The
depends upon the thickness of the active detector crystal
calibration tool uses the numerical method and is the pre-
area. Si(Li) detectors tend to have much thicker crystals and
ferred method (. Fig. 17.6).
thus measure X-rays with energies above 10 keV more effi-
ciently. Silicon drift detectors (SDD) tend to be about an
order of magnitude thinner and become increasingly trans- Azimuthal Angle
parent to X-rays above about 10 keV. DTSA-II defaults to a The azimuthal angle describes the angular position of the
thickness of 5 mm for Si(Li) detectors and 0.45 mm for detector rotated around the optic axis. The azimuthal angle is
SDD. These values will work adequately for most purposes if particularly important when modeling samples that are tilted
a vendor specified value is unavailable. or have complex morphology.
17.1 · Getting Started With NIST DTSA-II
243 17
35 000
Mn K-L3
Mn 1
30 000
20 000
Counts
15 000
Mn K-L2
10 000
5 000
0
5 600 5 700 5 800 5 900 6 000 6 100 6 200
Energy (eV)
.. Fig. 17.5 Estimating the full width at half-maximum peak width. This peak is approximately 139 eV FWHM which you can confirm with a ruler
Electron
column
De
te
De
cto
te
rs
ct
no
or
ut
Electron
Ele
column
DD
WD
Elevation X
X
Azimuth
Y Sample
Z Y
second effect is a low energy tail, called incomplete charge criminator should be set to an energy just above the high
collection, on low energy X-ray peaks. energy tail of the zero strobe.
The dead layer in modern detectors is very thin and typi-
cally produces very little incomplete charge collection. Older Material Editor Dialog (. Figs. 17.9, 17.10, 17.11,
Si(Li) detectors had thicker layers and worse incomplete 17.12, 17.13, and 17.14)
charge collection. The material editor dialog is used to enter compositional and
density information throughout DTSA-II. This dialog allows
you to enter compositional information either as mass frac-
Zero Strobe Discriminator (. Figs. 17.7 and 17.8) tions or atomic fractions. It also provides shortcut mecha-
The zero strobe discriminator is an energy below which all nisms for looking up definitions in a database or entering
spectrum counts will be set to zero before many spectrum compositions using the chemical formula.
processing operations are performed. Method 1: Mass fractions (see . Fig. 17.10)
The zero strobe is an artificial peak inserted by the detec- Method 2: Atomic fractions (see . Fig. 17.11)
tor electronics at 0 eV. The zero strobe is used to automati- Method 3: Chemical formula (see . Fig. 17.12)
cally determine the noise performance of the detector and to Method 4: Database lookup (see . Fig. 17.13)
automatically adjust the offset of the detector to compensate Method 5: Advanced chemical formulas (see . Fig. 17.16)
for shifts in calibration. The zero strobe does not interfere
with real X-ray events because it is located below the energies So if your database contains a definition for “Albite” and
at which the detector is sensitive. you press the search button , the table will be filled
Some vendors automatically strip out the zero strobe out with the mass and atomic fractions and the density as
before presenting the spectrum. Others leave it in because it recorded in the database for “Albite.” The database is updated
can provide useful information. When it does appear, it can each time you select the “Ok” button. Over time, it is possible
negatively impact processing low energy peaks. To mitigate to fill the database with every material that you commonly
this problem, the zero strobe discriminator can be used to see in your laboratory. The name “unknown” is special and is
strip the zero strobe from the spectrum. The zero strobe dis- never saved to the database.
600 000 Cu
400 000
Counts
200 000
0
-500 0 500 1 000 1500
Energy (eV)
17
.. Fig. 17.7 A raw Cu spectrum showing the zero strobe peak centered at 0 eV and the Cu L peaks centered near 940 eV
68 eV Cu
6 000 Cu
166
Counts
4 000
2 000
0
-500 0 500 1 000 1500
Energy (eV)
.. Fig. 17.8 The blue line shows an appropriate placement of the zero strobe discriminator between the high energy edge of the zero strobe
and the start of the real X-ray data
17.2 · Simulation in DTSA-II
245 17
.. Fig. 17.9 The material editor dialog. Materials are defined by a .. Fig. 17.11 The relative amount of each element in atomic fractions
name (“Albite”), a density (“3.61 g/cm3”), and a mapping between ele- may be entered manually using the “Element” and “Quantity” edit boxes
ments and quantities. Albite is defined as NaAlSi3O8 which is equivalent and the “Add” button. Note that the mode radio button is set to “Atomic
to the mass fractions and atomic fractions displayed in the table Proportions” and that the quantity is entered as a number of atoms in a
unit cell. The element may be specified by the common abbreviation
(“Si”), the full name (“silicon”) or the atomic number (“14”)
17.2.1 Introduction
.. Fig. 17.12 It is possible to enter the chemical formula directly into .. Fig. 17.13 To assist the user, DTSA-II maintains a database of
the “Name” edit box. When the search button is pressed the chemical materials. Each time the user enters a new material or redefines an
formula will be parsed and the appropriate mass and atomic fractions old material, the database is updated. The database is indexed by
entered into the table. Capitalization of the element abbreviations is “Name”
important as “CO” is very different from “Co”—one is a gas and the other
a metal. More complex formulas like fluorapatite (“Ca5(PO4)3F”) can be
entered using parenthesis to group terms. It is important that the for-
mula is unambiguous. Once the formula has been parsed you may
specify a new operator friendly name for the material like “Albite” of “Flu-
orapatite” in the “Name” edit box
55 Monte Carlo model of a block sample at which the beam intersects (. Figs. 17.29, 17.30,
55The user can specify the block base (square) and the and 17.31).
height. All the models require you to specify at least one material.
55 Monte Carlo model of an equilateral prism The material editor (described elsewhere) allows you to spec-
55The user can specify the edge of the triangle and the ify the material. Since the density is a critical parameter, you
length of the prism (. Figs. 17.16, 17.17, 17.18, 17.19, must specify it (. Fig. 17.32).
17.20, 17.21, 17.22, 17.23, 17.24, 17.25, 17.26, 17.27, Simulations are designed to model the spectra you could
and 17.28). collect on your instrument with your detector. By default, the
simulation “instrument configuration” page assumes that
Each simulation mode takes different parameters to config- you want to simulate the “default detector” as specified on the
ure the sample geometry. This page is for the simulation of a main “Spectrum” tab. However, you can specify a different
cube which requires two materials (substrate and cube), the instrument, detector and calibration if you desire.
dimensions of the cube and the sample rotation. You also need to specify an incident beam energy. This is
The sample rotation parameter is available for all modes the kinetic energy with which the electrons strike the sample
for which are not rotationally invariant. The best way to and is specified in kilo-electronvolts (keV).
understand the sample rotation parameter is to imagine The probe dose determines the relative intensity in the
rotating the sample about the optic axis at the point on the spectrum. Probe dose is specified in nano-amp seconds
17
.. Fig. 17.16 Bulk, homogeneous material .. Fig. 17.17 Thin film on substrate. Parameters: film thickness
17.2 · Simulation in DTSA-II
249 17
.. Fig. 17.18 Spherical particle on a substrate. Parameters: Sphere’s .. Fig. 17.19 Cubic particle on a substrate. Parameters: Cube
radius dimension
Zn K-L3 4 µm × 4 µm
Zn K-L3 2.19 µm × 2.19 µm
.. Fig. 17.20 Block inclusion in substrate. Parameters: Thickness and .. Fig. 17.21 Interface between two materials. Parameters: Distance
edge length from interface
250 Chapter 17 · DTSA-II EDS Software
.. Fig. 17.22 Square pyramid on a substrate. Parameters: Height and .. Fig. 17.23 Cylinder (fiber) on substrate. Parameters: Fiber diameter
base edge length and length
17
Zn K-L3 2.19 µm × 2.19 µm
Zn K-L3 2.19 µm × 2.19 µm
.. Fig. 17.24 Cylinder (can) on end on substrate. Parameters: Height .. Fig. 17.25 Hemispherical cap on substrate. Parameters: Cap radius
and fiber diameter
17.2 · Simulation in DTSA-II
251 17
.. Fig. 17.26 Rectangular block on substrate. Parameters: Block .. Fig. 17.27 Triangular prism on substrate. Parameters: Triangle
height and base edge length edge and prism lengths
y y
x
x
y
y
17
(nA · s = nC). This quantity is a product of the probe current tistics,” you may also select to output multiple spectra based
(nA) and the spectrum acquisition live-time (seconds.) on the simulated spectrum but differ by pseudorandom
Remember the probe current is a measure of the actual count statistics. You may also select to run additional simu-
number of electrons striking the sample per unit time, so the lated electron trajectories. The number of simulated elec-
probe dose is equivalent to a number of electrons striking tron trajectories determines the simulation to simulation
the sample during the measurement. Doubling the dose variance in characteristic X-ray intensities. The default
doubles the number of electrons striking the sample and number of electron trajectories typically produces about 1 %
thus also doubles the average number of X-rays generated in variance. The variance decreases as the square-root of the
the sample. number of simulated trajectories. You may also specify
The incidence angle (nominally 0°) allows you to simu- which X-ray generation modes to simulate including both
late a tilted sample (. Fig. 17.33). characteristic and bremsstrahlung primary emission and
The incident angle is defined relative to the optic axis. The secondary emission due to characteristic primary emission
pivot occurs at the surface of the sample, which is placed at or bremsstrahlung primary emission (. Figs. 17.37, 17.38,
the detector’s optimal working distance. A positive rotation is and 17.39).
towards the X-axis (an azimuth of 0°) and a negative rotation The “configure VP” page provides an advanced option to
towards the –X axis (an azimuth of 180°.) The detector dis- simulate the beam scatter in a variable-pressure or environ-
played is at an azimuth of 180°. A negative incidence angle mental SEM. If the check box is selected, you may select a gas
would tilt the sample toward the detector. Arbitrary tilts may (“water,” “helium,” “nitrogen,” “oxygen” or “argon”), a gas path
be simulated by moving the detector around the azimuth length and a nominal pressure. The gas path length is the dis-
(. Figs. 17.33, 17.34, 17.35, and 17.36). tance from the final pressure limiting aperture to the sample.
The “other options” page allows you to specify whether 1 Torr is equivalent to 133 Pa (. Figs. 17.40 and 17.41).
the spectrum is simulated with or without variance due to The primary output of a simulation is a spectrum. The
count statistics. If you select to “apply simulated count sta- simulated spectrum looks and acts to the best of our ability
Objective Lens
X-
ra
y
De
te
ct
or
Optimal
working
Distance
17 Incident
Angle
X axis
Sa
mp
le
Z axis
17.2 · Simulation in DTSA-II
255 17
Cu K-L3 1.81 µm × 1.81 µm Cu K-L3 1.81 µm × 1.81 µm Cu K-L3 1.81 µm × 1.81 µm
to simulate measured spectra like a spectrum collected from generated within the sample and emitted into a milli-stera-
the specified material under the specified conditions on the dian. The emitted column tabulates the X-rays that are
specified detector. You can treat simulated spectra like generated and also escape the sample in the direction of
measured spectra in the sense that you can simulate an the detector. The ratio is the fraction of generated X-rays
“unknown” standards and reference spectra and quantify the that escape the sample in the direction of the detector. The
“unknown” as though it had been measured (. Fig. 17.42). final rows compare the relative amount of characteristic
The report tab contains additional details about the simu- fluorescence to the amount of primary characteristic and
lation including configuration and results information. The bremsstrahlung emission (. Fig. 17.44).
first table summarizes the simulation configuration The emission images show where the measured X-rays
parameters. This table is available for all simulation modes were generated. Because the images only display X-rays
17 (. Fig. 17.43). that escape the sample, the distinction between the strongly
The “simulation results” table contains additional absorbed X-rays like the O K-L3 (Kα) and the less strongly
details derived from the simulation. The first two lines absorbed like the Si K-L3 (Kα) is evidenced by the flattened
contain links to the simulated spectrum and a Virtual emission profile in the Si K-L3 image. The last image shows
Reality Markup Language (VRML) representation of the the first 100 simulated electron trajectories (down to a
sample and electron trajectories. The subsequent rows kinetic energy of 50 eV.) The color of the trajectory seg-
contain raw data detailing the generation and emission of ment varies as the electron passes through the different
various different kinds of characteristic X-rays. Only those materials present in the sample. The gray lines exiting the
modes which were selected will be available. The generated top of the image are backscattered electrons (trajectories in
X-rays column tabulates the relative number of X-rays vacuum).
17.2 · Simulation in DTSA-II
257 17
10000
1000
Counts (Log)
100
10
1
0 2 4 6 8 10 12 14
Energy (keV)
2000
1500
Counts
1000
500
0
0 2 4 6 8 10 12 14
Energy (keV)
.. Fig. 17.37 Simulated admiralty brass (69 % Cu, 30 % Zn, and 1 % Tin) for various different selections of generation modes
60
Counts
40
20
0
0 2 4 6 8 10 12 14
Energy (keV)
.. Fig. 17.38 Simulated admiralty brass (69 % Cu, 30 % Zn, and 1 % Tin) with (blue) and without (red) simulated variation due to count statistics
258 Chapter 17 · DTSA-II EDS Software
17
17.2 · Simulation in DTSA-II
259 17
Generated Emitted
Transition (ratio) (ratio)
O AII 0.0011 0.0007
AI Kα 0.0135 0.0096
Comparing Characteristic to AI Kβ 0.0000 0.0000
Characteristic Fluorescence
Si Kα 0.0035 0.0013
Si Kβ 0.0023 0.0009
Ca Kα 0.0000 0.0000
17 Ca Kβ 0.0000 0.0000
Generated Emitted
Transition (ratio) (ratio)
O AII 0.0006 0.00004
AI Kα 0.0039 0.00022
Comparing Characteristic to AI Kβ 0.0000 0.00000
Bremsstrahulung Fluorescence
Si Kα 0.0051 0.00026
Si Kβ 0.0034 0.00018
Ca Kα 0.0236 0.0125
Ca Kβ 0.0236 0.0137
17.2 · Simulation in DTSA-II
261 17
Al K-L3 5.14 µm × 5.14 µm Al K-L3 5.14 µm × 5.14 µm Cl K-L3 5.14 µm × 5.14 µm
4.28E-2 Emission
Si K-M3 Trajectories
44 µm × 44 µm
17.2.4 Optional Tables One way to understand the scatter is to consider a series of
concentric rings on the surface of the sample centered at the
The “Fractional Emission Depths and Volumes” Table beam axis. The “VP Scatter Data” table (. Fig. 17.47), sum-
(. Fig. 17.45) marizes the number and number fraction of the incident elec-
When simulating a bulk sample, an additional report trons which intersect the various rings. In this simulation,
17 table shows the depth and the volumes from which 50 % or 87 % of the initial electrons are undeflected. However, at least
90 % of the measured X-rays are emitted. The depth and vol- one electron (0.1 %) is scattered further than 700 μm and 1.2 %
ume are largely determined by the ionization edge energy are scattered more than 50 μm. This qualitative information is
and X-ray absorption. Low ionization edge energies emit useful because it gives us a sense of how significant beam scat-
from larger volumes but lower energy X-rays also tend to be ter will be in variable pressure mode. It gives a sense of whether
more strongly absorbed. true quantitative analysis is possible and how much of an error
will be introduced by the beam scatter. The consequences are
The “VP Scatter Data” Table evident in the spectrum from an inclusion of admiralty brass
In variable pressure mode, the gas in the chamber can scatter in an aluminum. The aluminum is present in significant quan-
the incident electrons before they strike the sample. Despite tities in the variable pressure mode acquisition.
the fact that these scatters tend to be small angle events, the . Figure 17.48 shows EDS spectra calculated for a brass
path length is relatively long and electrons can scatter hun- inclusion in an aluminum matrix under VPSEM (red) and
dreds of microns to millimeters. The consequence can be conventional vacuum (blue) operation. The large peak for Al
demonstrated by simulating a moderate sized inclusion, under VPSEM conditions reveals the extent of gas scattering
shown in . Fig. 17.46. In simulated high-vacuum mode, the outside the focused beam. Interestingly, the Al is not zero in
excitation volume remains entirely within the inclusion. In the “high-vacuum” spectrum because of continuum gener-
simulated variable-pressure mode, the electrons can scatter ated secondary fluorescence. Increasing the size of the inclu-
out of the beam, entirely missing the inclusion and striking sion does not eliminate the slight Al peak but turning off the
the surrounding matrix. simulation of continuum fluorescence does.
17.2 · Simulation in DTSA-II
263 17
.. Fig. 17.47 Distribution of
gas-scattered electrons into a VP Scatter Data
series of concentric rings Electron trajectory count = 1000
Inner Radius Outer Radius Ring area
Ring µm2 Electron Count Electron Fraction Cumulative
µm µm
Undeflected - - - 869 0.869 -
1 0.0 2.5 19.6 921 0.921 92.1%
2 2.5 5.0 58.9 15 0.015 93.6%
3 5.0 7.5 98.2 8 0.008 94.4%
4 7.5 10.0 137.4 8 0.008 95.2%
5 10.0 12.5 176.7 8 0.008 96.0%
6 12.5 15.0 216.0 7 0.007 96.7%
7 15.0 17.5 255.3 4 0.004 97.1%
8 17.5 20.0 294.5 2 0.002 97.3%
9 20.0 22.5 333.8 3 0.003 97.6%
10 22.5 25.0 373.1 3 0.003 97.9%
11 25.0 27.5 412.3 2 0.002 98.1%
12 27.5 30.0 451.6 2 0.002 98.3%
13 30.0 32.5 490.9 0 0.000 98.3%
14 32.5 35.0 530.1 0 0.000 98.3%
15 35.0 37.5 569.4 1 0.001 98.4%
16 37.5 40.0 608.7 1 0.001 98.5%
17 40.0 42.5 648.0 1 0.001 98.6%
18 42.5 45.0 687.2 0 0.000 98.6%
19 45.0 47.5 726.5 1 0.001 98.7%
20 47.5 50.0 765.8 1 0.001 98.8%
21 50.0 100.0 23561.9 8 0.008 99.6%
22 100.0 150.0 39269.9 0 0.000 99.6%
23 150.0 2000.0 54977.9 1 0.001 99.7%
24 200.0 250.0 70685.8 0 0.000 99.7%
25 250.0 300.0 86393.8 1 0.001 99.8%
26 300.0 350.0 102101.8 0 0.000 99.8%
27 350.0 400.0 117809.7 1 0.001 99.9%
28 400.0 450.0 133517.7 0 0.000 99.9%
29 450.0 500.0 149225.7 0 0.000 99.9%
30 500.0 550.0 164933.6 0 0.000 99.9%
31 550.0 600.0 180641.6 0 0.000 99.9%
32 600.0 650.0 196349.5 0 0.000 99.9%
33 650.0 700.0 212057.5 0 0.000 99.9%
34 700.0 750.0 227765.5 1 0.001 100.0%
35 750.0 800.0 243473.4 0 0.000 100.0%
36 800.0 850.0 259181.4 0 0.000 100.0%
37 850.0 900.0 274889.4 0 0.000 100.0%
38 900.0 950.0 290597.3 0 0.000 100.0%
39 950.0 1000.0 306305.3 0 0.000 100.0%
264 Chapter 17 · DTSA-II EDS Software
0 2 4 6 8 10 12 14
Energy (keV)
.. Fig. 17.48 Comparison of spectra calculated for a brass inclusion in aluminum: VPSEM (red) and conventional vacuum (blue) operation. 1-mm
gas path length through water vapor at 133 Pa
References Ritchie N (2011b) “Manipulating Spectra with DTSA-II.” In: Lyman C (ed)
Microscopy Today. Cambridge Univ. Press. 19(3):34
Ritchie N (2011c) “Standards-Based Quantification in DTSA-II – Part I.”
Pouchou J-L, Pichoir P (1991) “Quantitative analysis of homogeneous or
In: Lyman C (ed) Microscopy Today. Cambridge Univ. Press. 19(5):30
stratified microvolumes applying the model PAP”. In: Heinrich K,
Ritchie N (2012) “Standards-Based Quantification in DTSA-II – Part II.”
Newbury D (eds) Electron Probe Quantitation. New York: Plenum,
In: Lyman C (ed) Microscopy Today. Cambridge Univ. Press. 20(1):14
p163
Ritchie N (2017a) “DTSA-II, a multiplatform software package for quantita-
Ritchie N (2009) Spectrum simulation in DTSA-II. Microsc Microanal
tive x-ray microanalysis.” Available free at the National Institute of
15:454
Standards and Technology website (search “DTSA-II”): 7 http://www.
Ritchie N (2010) Using DTSA-II to simulate and interpret energy disper-
cstl.nist.gov/div837/837.02/epq/dtsa2/index.html
sive spectra from particles. Microsc Microanal 16:248
Ritchie N (2017b) “Efficient simulation of secondary fluorescence Via
Ritchie N (2011a) “Getting Started with NIST DTSA-II.” In: Lyman C (ed)
NIST DTSA-II Monte Carlo” Microsc Microanal 23:618
Microscopy Today. Cambridge Univ. Press. 19(1):26
17
265 18
References – 287
120 000
O K-L
CaK-L2,3
CaCO3
100 000
E0 = 10 keV
80 000
C K-L
Counts
60 000
CaK-M3
40 000
20 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
120 000
100 000
80 000
C K-L
O K-L
Counts
60 000
40 000
20 000
0
0.0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Photon energy (keV)
.. Fig. 18.1 Spectrum of calcium carbonate. Note non-linear behavior at low photon energy, e.g., the C K-shell peak is significantly shifted below
the true energy value given by the marker
The characteristic peak to continuum background, P/B, depending on the shells involved, with the fluorescence yields
which determines the visibility of peaks above the back- for a particular element generally trending K > L > > M. An
ground, is found as the ratio of equations (18.2) and example for barium L-shell and M-shell X-rays is shown in
(18.3b): . Fig. 18.3, where the Ba M-family X-rays are seen to have a
much lower P/B than the Ba L-family X-rays, making Ba dif-
P / B = I ch / I cm ~ (U 0 − 1) / (U 0 − 1)
n
ficult to identify with high confidence if only the Ba M-family
= (U 0 − 1)
n −1
(18.4) is excited, a condition that will exist for Ba if E0 is chosen
below the 5.25 keV ionization energy for the Ba L3-shell.
Since the exponent n ~ 1.5, in the expression for P/B the value X-ray Absorption
of n − 1 ~ 0.5, so that as U0 is lowered, the P/B decreases dra-
A third factor which can strongly influence the visibility and
matically, reducing the visibility of peaks, as shown in
detection of peaks is absorption of characteristic X-rays as
. Fig. 18.2 for the K-shell peaks of silicon.
they travel through the specimen and the window and sur-
Fluorescence Yield face layers of the EDS detector. X-ray absorption along a path
of length s through the specimen is a non-linear process:
A second factor that affects the detectability of characteristic
peaks is the fluorescence yield, the fraction of ionizations that
I / I 0 = exp − ( µ / ρ ) ρs (18.5)
leads to photon emission. The fluorescence yield varies sharply
268 Chapter 18 · Qualitative Elemental Analysis by Energy Dispersive X-Ray Spectrometry
10 000
1 000
Counts
100
10
1
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
Photon energy (keV)
14 000
12 000
10 000
8 000
Counts
6 000
4 000
2 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
Photon energy (keV)
.. Fig. 18.2 Si at various overvoltages, showing diminishing peak visibility as the excitation decreases
where I0 is the original intensity and I is the intensity that relative detector efficiency, εC < εSi, as described in the “EDS”
remains after path s through a material of density ρ having a module.) Because the absorption path length, s, depends
mass absorption coefficient μ/ρ for the photon energy of strongly on the electron range, RK-O, which scales
interest. The mass absorption coefficient depends strongly on approximately as the 1.7 power of the incident beam energy,
the photon energy and the specific elements that the photons decreasing the beam energy reduces the absorption path of
are passing through. Generally, a photon will be strongly the C K X-rays, making the C K-peak more prominent rela-
18 absorbed, i.e., there will be a large mass absorption coeffi- tive to the Si K-peak, as shown in . Fig. 18.4 for a series of
cient, if the energy of the photon lies in a range of approxi- progressively lower beam energies.
mately 1 keV above the critical ionization energy for another The possibility of a high absorption situation for elements
element that is present in the analyzed volume. An extreme that must be measured with a low photon energy requires an
case is illustrated in . Fig. 18.4 for SiC, where at E0 = 20 keV analytical strategy such that when analyzing an unknown,
with the spectrum scaled to the Si K-peak, the C K peak is the analyst should start at high beam energy, E0 ≥ 20 keV, and
barely visible despite C’s making up half of the composition work down in beam energy. The analyst must be prepared to
on an atomic basis. Strong absorption of the C K X-ray at utilize low beam energies, E0 ≤ 5 keV, to evaluate the possibil-
0.282 keV occurs because this energy lies just above the Si L3 ity of high absorption situations, such as those encountered
critical ionization energy at 0.110 keV, resulting in an for low atomic number elements (Z < 10). For these elements,
extremely large value for the mass absorption coefficient. the only detectable peaks have low photon energies (<1 keV)
(There are also other factors that apply to this case, including and are thus subject to high absorption when high incident
the relative fluorescence yields, for which ωC < ωSi, and the beam energy is used.
18.2 · Principles of Qualitative EDS Analysis
269 18
140 000
80 000
Counts
60 000
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
35 000
30 000
25 000
Counts
20 000
15 000
10 000
5 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
.. Fig. 18.3 EDS spectrum of BaCl2 showing Ba L-family and Ba M-family peaks
350 000
300 000
SiC
250 000 E0 = 20 keV
SiK-L2,3
200 000
Counts
100 000
50 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
C K-L
5 keV
SiK-L2,3
10 keV
20 keV
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
.. Fig. 18.4 EDS spectra of SiC: (upper) at E0 = 20 keV, the C K-L3 peak is barely visible; (lower) as the beam energy is lowered to reduce absorp-
tion, the C K-L3 peak becomes more prominent (all spectra scaled to the Si K-L3 region.)
.. Table 18.1 Comprehensive listing of all X-ray transitions for gold (DTSA-II database)
(continued)
272 Chapter 18 · Qualitative Elemental Analysis by Energy Dispersive X-Ray Spectrometry
Si Escape Peak detected because of the high count spectrum. The EDS sys-
The Si escape peak results when a Si K–L3 X-ray tem should mark all possible escape peaks and not subse-
(E = 1.740 keV), which is created following the photoion- quently misidentify them as other elements. For example, if
ization of a silicon atom and is usually reabsorbed within not properly assigned, the escape peak for Ti K-L2,3 could
the detector volume, escapes. This results in an energy loss be mistaken for Cl K-L2,3. Note that some vendor software
18 of 1.740 keV from the parent X-ray, creating a “silicon removes the escape peaks in the final processed spectrum
escape peak,” as shown in . Fig. 18.7 (upper spectrum) for that is displayed to the user, as shown in . Fig. 18.7 (lower
the titanium K-family X-rays. Escape peaks can only be spectrum).
created for parent photon energies above 1.740 keV and are
formed at a fixed fraction of the parent peak, with that frac- Coincidence Peaks
tion rapidly decreasing as the parent X-ray energy increases Although the EDS spectrum may appear to an observer to be
above 1.740 keV. For parent peaks with photon energies collected simultaneously at all photon energies, in reality
above 6 keV, escape peaks are so small that they are difficult only one photon can be measured at a time. Because X-rays
to detect. Escape peaks can occur from any parent peak, but are created randomly in time, as the rate of production (input
as a practical matter only the major members of a family count rate) increases, the possibility of two photons entering
are likely to produce detectable escape peaks. Note that in the detector and creating an artifact coincidence event
the example shown for the titanium K-family in . Fig. 18.7, increases in probability. An inspection function continu-
the escape peaks for Ti K-L2,3 and Ti K-M2,3 are both ously monitors the detector to reject such events, but at
18.2 · Principles of Qualitative EDS Analysis
273 18
200 000
BaF2
150 000 E0 = 10 keV
Counts
100 000
50 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
50 000
40 000
30 000
Counts
20 000
10 000
0
2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5
Photon energy (keV)
.. Fig. 18.5 EDS spectrum of BaF2 excited with a beam energy of 10 keV
60 000
50 000
40 000 Au
Counts
E0 = 10 keV
30 000
20 000
10 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
5 000
4 000 Ti
E0 = 20 keV
3 000
Counts
2 000
1 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
5 000
Ti coincidence
Counts
3 000
Ti coincidence
2 000
1 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
s ufficiently high input count rates its finite time resolution in . Fig. 18.10. As the dead-time increases, numerous coinci-
will be overwhelmed and coincidence events will begin to dence peaks are observed, several of which could be mis-
populate the spectrum. This phenomenon is illustrated in identified as elements present as trace to minor constituents.
. Fig. 18.7 for pure Ti, where coincidence peaks are seen for At the highest dead-times, the coincidence peaks are seen to
Ti K-L2,3 + Ti K-L2,3 and Ti K-L2,3 + Ti K-M2,3. The height of a occupy much of the useful spectral energy range where legit-
coincidence peak relative to the parent peak depends on the imate minor and trace constituents might be measured.
18 arrival rate of X-rays at the detector, and thus upon the dead- Coincidence can involve any two photons of any energies,
time. Even at low dead-time coincidence events are likely to but the noticeable effects in EDS spectra consist of two char-
be observed for highly excited, low photon energy peaks acteristic peak photons that originate from major constitu-
such as Al and Si in pure element or high concentration tar- ents that are highly excited, for exmple, A + A, A + B, B + B,
gets, as shown for Si in . Fig. 18.8 at 12 % dead-time and in an so on. Since coincidence depends on the arrival rate of
. Fig. 18.9 at 1 % dead-time. Operating at low dead-time can photons, the analyst can exert some control by operating at
reduce the height of the coincidence peak relative to the par- low dead-time (10 % or less) to minimize but not eliminate
ent peak, but coincidence can never entirely be eliminated by the effect. Some vendor EDS systems use statistical models of
reducing the input count rate. For complex compositions, a coincidence to post process the spectrum, removing the
wide array of coincidence peaks involving many combina- coincidence events and restoring the events to their proper
tions of highly excited peaks can be encountered, as shown parent peaks.
18.3 · Performing Manual Qualitative Analysis
275 18
10 00 000
Si
E0 = 10 keV
100 000 12% deadtime
Counts
10 000
1 000
100
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
12 000
Si-coincidence
10 000
8 000
Counts
6 000
4 000
2 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
10 00 000
Si
E0 = 20 keV
100 000
1% deadtime
Counts
10 000
1 000
100
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
4 000
Si-coincidence
3 000
Counts
2 000
1 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
CaKα
OK
AlK
140,000
Coincidence 66% DT
130,000
peaks 56%
120,000 43%
Coincidence 36%
110,000 29%
peaks 21%
100,000
11%
18 90,000 6%
SiK+CaKα , not CrKα
80,000 3%
SiK+O K
2%
2O K not NaK
70,000 1% DT
FeL
FeKα
O 0.428
AlK+CaKα , not VKα
CaKβ
60,000 Mg 0.117
2AlK
Al 0.0491
CK
2SiK
50,000 Si 0.212
2MgK
40,000 Ca 0.109
Fe 0.0774
30,000
FeKβ
20,000
10,000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 18.10 EDS spectra of NIST glass K412 over a range of dead-times from 1 % to 66 %
18.3 · Performing Manual Qualitative Analysis
277 18
peaks that arise from the silicon escape peak and from coin- it is important to the analyst to identify the presence of a
cidence peaks. A particularly insidious problem occurs when trace element (s) with a high degree of confidence, manual
automatic peak identification software delivers identifica- peak identification will be necessary.
tions of peaks with low peak-to-background too early in the
EDS accumulation before adequate counts have been
recorded. Statistical fluctuations in the continuum back- 18.3.2 erforming Manual Qualitative
P
ground create “false peaks” that may appear to correspond to Analysis: Choosing the Instrument
minor or trace constituents. This problem can be recognized Operating Conditions
when an apparent peak identification solution for these low
level peaks subsequently changes as more counts are accu- Beam Energy
mulated. The danger is that the analyst may choose to stop Equation 18.1 reveals that one selection of the beam energy
the accumulation prematurely and be misled by the low level may not be sufficient to solve a particular problem, and the
“peaks” that do not actually exist. analyst must be prepared to explore a range of beam energies
When the analyst must operate only at low beam energy to access desired atomic shells. The peak height relative to the
(E0 ≤ 5 keV), the peak misidentification problem is exacer- spectral background increases rapidly as U0 is increased,
bated by the loss of the higher photon energies where X-ray enabling better detection of the characteristic peak (s). Having
family members are more widely spread and more easily adequate overvoltage is especially important as the concen-
identified, as well as the confidence-increasing redundancy tration of an element decreases from major to minor to trace.
provided by having K-L and L-M family pairs for identifica- As a general rule, it is desirable to have U0 > 2 for the analyzed
tion of intermediate and high atomic number elements shells of all elements that occur in a particular analysis. For
(Newbury 2009). initial surveying of an unknown specimen, it is useful to select
Even well-implemented automatic peak identification a beam energy of 20 keV or higher to provide an overvoltage
software is likely to ignore peaks with low peak-to- of at least 2 for ionization edges up to 10 keV. Elements with
background that may correspond to trace constituents intermediate atomic numbers (e.g., 22, Ti ≤ Z ≤ 42, Mo) and
because the likelihood of a mistake becomes so large. Thus, if high atomic number (e.g., Z ≥ 56, Ba) elements have complex
278 Chapter 18 · Qualitative Elemental Analysis by Energy Dispersive X-Ray Spectrometry
atomic shell structures that produce two families of detectable measurements. It is often desirable to maximize the recorded
characteristic X-rays (with 20 keV ≤ E0 ≤ 30 keV), for exam- counts per unit of real (clock) time. Higher beam current lead-
ple, the Cu K-family and L-family; the Au L-family and ing to higher dead-time, for example, 30–40 %, can be utilized,
M-family. A second advantage of selecting the beam energy to but the spectrum is likely to have coincidence peaks like those
excite the higher energy X-ray family for an element is that it shown in . Fig. 18.10, which can greatly complicate the recog-
enables a high confidence identification since the peaks that nition and measurement of the peaks of minor and trace con-
form the family are more widely separated in photon energy stituents. Note that some vendor software systems effectively
and thus more likely to be resolved with EDS. Note that the block coincidence peaks or else remove them from the spec-
physics of X-ray generation requires that all members of the trum by post-processing with a stochastic model that predicts
X-ray family of a tentative elemental assignment must be coincidence peaks based on the parent peak count rates.
present. Identifying all family members in the correct relative
intensity ratios gives high confidence that the element assign- Obtaining Adequate Counts
ment is correct as well as avoiding subsequent misidentifica- The analyst must accumulate adequate X-ray counts to dis-
tion of these minor family members. tinguish a peak against the random fluctuations of the back-
ground (X-ray continuum). While it is relatively easy to
hoosing the EDS Resolution (Detector Time
C record sufficient counts to recognize the principal peak for
Constant) a major constituent, detection of the minor family member
EDS systems provide two or more choices for the detector (s) to increase confidence in the elemental assignment may
time constant. The user has a choice of a short detector time require recording a substantially greater total count. For
constant that gives higher throughput (photons recorded per minor or trace constituents, an even greater dose is likely to
unit time) at the expense of poorer peak resolution or a long be required just to detect the principal family members, and
time constant that improves the resolution at the cost of to obtain minor family members to increase confidence in
throughput. The analyst thus has a critical choice to make: an elemental identification will require a dose greater by
more counts per unit time or better resolution. Statham (1995) another factor of ten or more. A peak is considered detect-
analyzed these throughput-resolution trade-offs with respect able if it satisfies the following criterion (Currie 1968):
to various analytical situations and concluded that a strategy
nP > 3 nB1/ 2 (18.6)
that emphasizes maximizing the number of X-ray counts
rather than resolution produces the most robust results.
where nP is the number of peak counts and nB is the number
of background counts under the peak. Note that “detectable”
hoosing the Count Rate (Detector
C does not imply optimally measureable, for example, obtain-
Dead-Time) ing accurate peak energy. While Eq. 18.6 defines the mini-
A closely related consideration is the problem of pulse coinci- mum counts to detect a peak, accurate measurement of the
dence creating artifact peaks, which are reduced (but not peak position to identify the peak may require higher counts.
eliminated) by using lower dead-time. Note that a specific level The effect of increasing the total spectral intensity to “develop”
of dead-time, for example, 10 %, corresponds to a higher low relative intensity peaks from trace constituents is shown
throughput when a shorter time constant is chosen. With the in . Fig. 18.11.
beam energy and detector time constant selected, the rate at
which X-rays arrive at the EDS and subsequent output depends kGolden Rule
on two factors: (1) the detector solid angle and (2) the beam If it is difficult to recognize a peak above fluctuations in the
current. If the EDS detector is movable relative to the speci- background, accumulate more counts. Patience is a virtue!
18 men, the specimen-to-detector distance should be chosen in a
consistent fashion to enable subsequent return to the same
operating conditions for robust standards-based quantitative 18.4 Identifying the Peaks
analysis. A typical choice is to move the detector as close to the
specimen as possible to maximize the detector solid angle, After a suitable spectrum has been accumulated, the analyst
Ω = A/r2, by minimizing r, the detector-to-specimen distance, can proceed to perform manual qualitative analysis.
where A is the active area of the detector. Always ensure that
any possible stage motions will not cause the specimen to
strike the EDS. With the EDS solid angle fixed, the input count 18.4.1 Employ the Available Software Tools
rate will then be controlled by the beam current. A useful strat-
egy is to choose a beam current that creates an EDS dead-time Manual qualitative analysis is performed using the support of
of approximately 10 % on a highly excited characteristic X-ray, available software tools such as KLM markers that show the
such as Al K-L2,3 from pure aluminum. To establish dose-cor- energy positions and relative heights of X-ray family mem-
rected standards-based quantitative analysis, this same detec- bers to assign peaks recognized in the spectrum to specific
tor solid angle and beam current should be used for all elements. Before using this important software tool, the user
18.4 · Identifying the Peaks
279 18
1 00 000
10 000
Counts
1 000
100
0
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
14 000 NIST K491
E0 = 20 keV
12 000 Al: 0.00085 (850 ppm)
Ti: 0.0015 (1500 ppm)
10 000 Ce: 0.0046(4600 ppm)
200 s
Fe: 0.0018 (1800 ppm)
8 000 Ta: 0.0059 (5900 ppm)
Counts
6 000
100 s
4 000
50 s
2 000
20 s
0
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
.. Fig. 18.11 Detection of trace constituent peaks in NIST microanalysis glass K491 as the dose is increased. Integrated spectrum counts:
1 s = 0.14 million; 2 s = 0.28 million; 5 s = 0.70 million; 10 s = 1.4 million; 20 s = 2.8 million; 50 s = 7.1 million; 100 s = 14.3 million; 200 s = 28.5 million
should confirm that all elements in the periodic table are efinitely are present. It isn’t the natural perspective and it
d
enabled and all X-ray family members are shown in the KLM takes more thought and effort but it is much less prone to
markers. errors of omission.
As each element is tentatively identified from its major Imagine that you are at a zoo. You have a list of 20 animals
family peak, a systematic search must then be made to locate of various sizes and shapes and you are asked to answer the
all possible peaks that must be associated with that element: question, what animals in this list could possibly be in this cage.
(1) all minor family members; (2) a second X-ray family at You look around and see a rhinoceros laying down near the
lower energy (e.g., K and L or L and M), and for the highest back of the cage and no other animal. You might be tempted to
atomic number elements, the N-family can also be observed, say that the only animal that could possibly be in the cage is
as shown in . Fig. 18.6; and (3) any associated EDS artifact the one you see – the rhinoceros. However, the list also con-
peaks (escape peaks and coincidence peaks). This careful tains snakes, mice, fish and elephants. You can rule out ele-
inspection regimen and meticulous bookkeeping raises the phants because they are too big to hide behind a rhino. You
confidence in the tentative assignment. Properly assigning can rule out fish because the environment is inappropriate.
the minor family members and the artifact peaks to the You can’t rule out the possibility that a mouse or snake is in the
proper element will diminish the possibility of subsequent cage hidden behind the rhino. It is only by eliminating those
incorrect assignment of those peaks to other constituents animals that are too large (elephant) or can’t survive (fish)
that might appear to be present at minor or trace levels. behind the rhino that you can come to the full list of animals
It can be helpful to think of qualitative analysis as the pro- that could potentially be present in the cage – the rhino and
cess of eliminating those elements which cannot possibly be any animals which could be hiding behind the rhino. If you
present rather than the process of including those that want to be certain that you haven’t missed an animal that
280 Chapter 18 · Qualitative Elemental Analysis by Energy Dispersive X-Ray Spectrometry
could possibly be present in the cage, the process of culling ani- energy region (<4 keV). The logic for this strategy is that
mals that couldn’t possibly be present is far more robust than K-shell and L-shell characteristic X-rays above 4 keV are
the process of including animals that definitely are present. produced in families that provide two or more peaks with
Spectra are similar. Not only is it possible that the obvious distinctive relative abundances for which the energy resolu-
elements are present but also those that could be hidden by the tion of EDS is sufficient to easily separate these peaks. Having
ones that are readily identified. Fortunately, there is a tool to help two or more peaks to identify greatly increases the confi-
us to see through spectra and expose the hidden components – dence with which an elemental identification can be made,
the residual spectrum. The residual spectrum is the intensity enabling the analyst to achieve an unambiguous result. For
that remains in each channel after peak fitting has been per- each peak that is recognized, first test whether its energy cor-
formed for the specified elements. It is like being able to ask the responds closely to a particular K-L3 (Kα) peak. The physics
rhino to move and then being able to see what is hidden behind – of X-ray generation demands that the corresponding K-M3
maybe a mouse or a snake or maybe nothing. An example of the (Kβ) peak must also be present in roughly a 10:1 ratio. If
utility of the residual spectrum is shown in Figure 20.8. K-family peaks do not match the peak in question, examine
L-family possibilities, noting that three or more L-peaks are
likely to be detectable: L3–M4,5 (Lα), L2-M4 (Lβ), and L2-N4
18.4.2 I dentifying the Peaks: Major (Lγ). Locate and mark all minor family members such as L3-
Constituents M1 (Ll). Locate and mark the escape peaks, if any, associated
with the major family members. Locate and mark, if any, the
Start with peaks located in the higher photon energy coincidence peaks associated with the major family mem-
(>4 keV) region of the spectrum and work downward in bers, which may be located at very high energy, for example,
energy, even if there are higher peaks in the lower photon as shown for Cu K-L3 coincidence in . Fig. 18.12.
1 00 000
CuK coincidence
Cu
E0 = 20 keV
10 000
Counts
1 000
100
10
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
6 000
18 5 000
4 000
CuL coincidence
Counts
3 000
2 000
1 000
0
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
.. Fig. 18.12 EDS spectrum of Cu at E0 = 20 keV showing a coincidence peak for CuK-L3 at 16.08 keV
18.5 · A Worked Example of Manual Peak Identification
281 18
18.4.3 Lower Photon Energy Region 18.5 Worked Example of Manual Peak
A
Identification
As major spectral peaks located at lower photon energy
(<4 keV) are considered, the energy separation diminishes Alloy IN100 is a complex mixture of transition and heavy ele-
and the relative peak heights decrease for the members of each ments that provides several challenges to manual peak iden-
X-ray family. EDS is no longer able to resolve these peaks, tification:
leading to a situation where only one peak is available for iden- 1. . Figure 18.14a shows the spectrum from 0 to 20 keV
tification for K-family X-rays below 2 keV in energy. The K-L3 excited with E0 = 20 keV. Using the KLM marker tools in
peak appears symmetric since the K-M3 peak has low relative DTSA II, starting at high photon energy and working
intensity, as shown for Al K-L3 in . Fig. 18.13a. For L- and M- downward, the first high peak encountered shows a
family X-rays in the low photon energy range, the composite good match to Ni K-L3 and the corresponding Ni K-M3
peak appears asymmetric. As shown for Br in . Fig. 18.13b, is also found at the correct ratio, as well as the Ni
the major peaks L3-M5(Lα) and L2-M4(Lβ) occur with a ratio L-family at low photon energy. The position of the Ni
of approximately 2:1 and the low abundance but separated L3- K-L3 escape peak is marked. Inspection for possible
M1 (Ll) and L2-M1 (Lη) can also aid in the identification pro- coincidence peaks does not reveal a significant popula-
viding the spectrum contains adequate counts. Similarly, the tion due to the low dead-time (8 %) used to accumulate
M5-N6,7(Mα) and M4-N6(Mβ) peaks occur with a ratio of 1/0.6 the spectrum and the large number of peaks over which
and the well separated minor family members W M5,4-N3,2 the input count rate is partitioned so that even the most
(Mζ) and W M3-N5 (Mγ) can be detected in a high count spec- intense peak has a relatively low count rate and does not
trum, as shown for W in . Fig. 18.13c. produce significant coincidence.
2. Working down in energy (. Fig. 18.14b), the next
peak is seen to correspond to Co K-L3, but the Co
18.4.4 I dentifying the Peaks: Minor K-M3 suffers interference from Ni K-L3 and only
and Trace Constituents appears as an asymmetric deviation on the high
energy side. Likewise, the Co L-family is unresolved
After all major peaks and their associated minor family mem- from the Ni L-family.
bers and artifact peaks have been located and identified with 3. The next set of peaks match Cr, as shown in . Fig. 18.14c.
high confidence as belonging to particular elements, the ana- 4. Continuing, . Fig. 18.14d shows a match for the peaks of
lyst can proceed to identify any remaining peaks which are Ti, but the apparent ratio of Ti K-L3/Ti K-M3 is approxi-
now likely to be associated with minor and trace level constitu- mately 5:1, whereas the true ratio is about 10:1, which
ents. Achieving the same degree of high confidence in the iden- suggests that another element must be present.
tification of lower concentration constituents is more difficult Expansion of this region in . Fig. 18.14e reveals that V is
since the lower concentrations reduce all X-ray intensities so likely to be present but with severe interference between
that minor family members are more difficult to detect. The V K-L3 and Ti K-M3. While the anomalous peak ratio
situation is likely to require accumulating additional X-ray observed for TiK-L3/TiK-M3 is a strong clue that another
counts to improve the detectability of minor X-ray family element must be present, this example shows one of the
members and increase the confidence of the assignment of limitations of manual peak identification, namely, that
elemental identification. In general, establishing the presence peaks representing minor and trace constituents can be
of a constituent at trace level is a significant challenge that lost under the higher intensity peaks of higher concen-
requires not only collecting a high count spectrum that satisfies tration constituents as the concentration ratio becomes
the limit of detection criterion but also scrupulous attention to large. Detecting such interferences of constituents with
identifying all possible minor family members and artifacts large concentration ratios requires the careful peak-
from the X-ray families of the major and minor constituents. fitting procedure that is embedded in the quantitative
analysis procedures described in module 19.
5. In . Fig. 18.14f, the next peak group best matches the
18.4.5 Checking Your Work Mo L-family. This photon energy range involves possible
interferences from the S K-family, the Mo L-family, and
The only way to be confident that the qualitative analysis is cor- the Pb M-family. The possibility of identifying the peak
rect to quantify the spectrum and examine the residual spec- group as the Pb M-family which occurs this energy
trum. When every element has been correctly identified and range, can be rejected because of the absence of the Pb
quantified, the analytical total should be approximately unity L-family, as shown in . Fig. 18.14g. The possible
and there should be no obvious structure in the residual spec- presence of the S K-family (. Fig. 18.14h) is much more
trum that cannot be explained through chemistry or minor difficult to exclude because S cannot be effectively
chemical peak shifts. This iterative qualitative – q uantitative measured by an alternate X-ray family such as the S
analysis scheme to discover minor and trace elements hidden L-family due to the low fluorescence yield. While the
under the high intensity peaks of major constituents will shape of the peak cluster does not match S K-L3 and S
covered in 7 Chapter 19. K-M3, the presence of S can only be confidently
282 Chapter 18 · Qualitative Elemental Analysis by Energy Dispersive X-Ray Spectrometry
10 00 000
800 000
Al
E0 = 20 keV
Counts
600 000
400 000
200 000
0
1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00
Photon energy (keV)
200 000
KBr
150 000 E0 = 20 keV
Counts
100 000
50 000
0
1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00
Photon energy (keV)
c
350 000
300 000
W
E0 = 20 keV
250 000
18 200 000
Counts
150 000
100 000
50 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
.. Fig. 18.13 a EDS spectrum of Al at E0 = 20 keV; note symmetry of Al K-family peaks. b EDS spectrum of KBr at E0 = 20 keV; note asymmetry of Br
L-family peaks. c EDS spectrum of W at E0 = 20 keV; note asymmetry of W M-family peaks
18.5 · A Worked Example of Manual Peak Identification
283 18
a 40 000
IN100
E0 = 20 keV
30 000
8% deadtime
Counts
20 000
10 000
0
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
b 40 000
30 000
Counts
20 000
10 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
c 40 000
IN100
E0 = 20 keV
8% deadtime
30 000
Counts
20 000
10 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 18.14 a Alloy IN100 recorded with E0 = 20 keV and at 8 % of Mo. g Rejection of Pb. h Possible presence of S. i Identification of Al. j
dead-time showing identification of Ni. b Identification of Co. c Identi- Rejection of Br. k Identification of C. l Identification of Si
fication of Cr. d Identification of Ti. e Identification of V. f Identification
284 Chapter 18 · Qualitative Elemental Analysis by Energy Dispersive X-Ray Spectrometry
d 40 000
30 000
Counts
20 000
10 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
e
20 000
IN100
E0 = 20 keV
8% deadtime
15 000
Counts
10 000
5 000
0
4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0
Photon energy (keV)
f 20 000
15 000
Counts
10 000
5 000
18
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
g 10 000
IN100
E0 = 20 keV
8 000 8% deadtime
6 000
Counts
4 000
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0
Photon energy (keV)
h 20 000
15 000
Counts
10 000
5 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
i
40 000
IN100
E0 = 20 keV
8% deadtime
30 000
Counts
20 000
10 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
j
10 000
1 000
Counts
100
10
1
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
k
40 000
IN100
E0 = 20 keV
8% deadtime
30 000
Counts
20 000
10 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
l
10 000
8 000
Counts
6 000
4 000
18 2 000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
c onfirmed by peak fitting procedures during quantitative 7. The last significant peak is found to correspond to C K
analysis. (. Fig. 18.14k) noting that due to the non-linearity of the
6. The next peak matches the Al K-family (. Fig. 18.14i) photon energy scale for this detector below 400 eV, the
but in this photon energy range only one peak is peak is displaced to a lower energy from the ideal position.
available for identification. The Br L-family also fits this 8. Finally, inspection of the remaining low peak-to-
peak (. Fig. 18.14j) but Br can be dismissed because of background peaks reveals just one candidate, which cor-
the absence of the Br K-family. responds to the Si K-family (. Fig. 18.14l).
References
287 18
References Newbury D (2007) Mistakes encountered during automatic peak identi-
fication in low beam energy X-ray microanalysis. Scanning 29:137
Newbury D (2009) Mistakes encountered during automatic peak identi-
Currie LA (1968) Limits for qualitative detection and quantitative deter-
fication of minor and trace constituents in electron-excited energy
mination. Anal Chem 40:586
dispersive X-ray microanalysis. Scanning 31:1
Newbury D (2005) Misidentification of major constituents by automatic
Statham P (1995) Quantifying Benefits of resolution and count rate in
qualitative energy dispersive X-ray microanalysis: a problem that
EDX microanalysis. In: Williams D, Goldstein J, Newbury D (eds)
threatens the credibility of the analytical community. Microsc
X-ray spectrometry in electron beam instruments. Plenum,
Microanal 11:545
New York, pp 101–126
289 19
Quantitative Analysis:
From k-ratio to Composition
19.1 What Is a k-ratio? – 290
References – 307
19.1 What Is a k-ratio? known physical parameters are the same for an element
in the standard and the unknown.
A k-ratio is the ratio of a pair of characteristic X-ray line
intensities, I, measured under similar experimental condi- In the history of the development of quantitative electron-
tions for the unknown (unk) and standard (std): excited X-ray microanalysis, the X-ray intensities for the
unknown and standards were measured sequentially with a
k = I unk / I std (19.1) wavelength spectrometer in terms of X-ray counts. The raw
measurement contains counts that can be attributed to both
The measured intensities can be associated with a single the continuum background (bremsstrahlung) and character-
characteristic X-ray line (as is typically the case for wave- istic X-rays. Since k-ratio is a function of only the character-
length spectrometers) or associated with a family of charac- istic X-rays, the contribution of the continuum must be
teristic X-ray lines (as is typically the case for energy estimated. Usually, this is accomplished by measuring two
dispersive spectrometers.) The numerator of the k-ratio is off-peak measurements bounding the peak and using inter-
typically the intensity measured from an unknown sample polation to estimate the intensity of the continuum back-
and the denominator is typically the intensity measured from ground at the peak position. The estimated continuum is
a standard material—a material of known composition. subtracted from the measured on peak intensity to give the
Both the numerator and the denominator of the k-ratio characteristic X-ray line intensity.
must be measured under similar, well-controlled instrument Extracting the k-ratio with an energy dispersive spec-
conditions. The electron beam energy must be the same. The trometer can be done in a similar manner for isolated peaks.
probe dose, the number of electrons striking the sample dur- However, to deal with the peak interferences frequently
ing the measurement, should be the same (or the intensity encountered in EDS spectra, it is necessary to simultane-
scaled to equivalent dose.) The position of the sample relative ously consider all of the spectrum channels that span the
to the beam and to the detector should be fixed. Both the mutually interfering peaks. Through a process called linear
sample and the standard(s) should be prepared to a high least squares fitting, a scale factor is computed which repre-
degree of surface polish, ideally to reduce surface relief below sents the multiplicative factor by which the integrated area
50 nm, and the surface should not be chemically etched. If under the characteristic peak from the standard must be
the unknown is non-conducting, the same thickness of con- multiplied by to equal the integrated area under the charac-
ducting coating, usually carbon with a thickness below teristic peak from the unknown. This scale factor is the
10 nm, should be applied to both the unknown and the k-ratio, and the fitting process separates the intensity com-
standard(s). Ideally, the only aspect that should differ ponents of the interfering peaks and the continuum back-
between the measurement of the unknown and the standard ground. The integrated counts measured for the unknown
are the compositions of the materials. and for the standard for element Z enable an estimate of the
The k-ratio is the first estimate of material composition. precision of the measurement for that element. Linear least
From a set of k-ratios, we can estimate the unknown material squares fitting is employed in NIST DTSA-II to recover char-
composition. In many cases, to a good approximation: acteristic X-ray intensities, even in situations with extreme
peak overlaps.
CZ , unk ∼ k Z CZ ,std = I unk / I std CZ ,std (19.2) A measured k-ratio of zero suggests that there is none of
the associated element in the unknown. A measurement on a
where CZ,unk and CZ,std are the mass fraction of element Z in standard with exactly the same composition as the unknown
the unknown and standard, respectively, and kZ is the k-ratio will nominally produce a k-ratio of unity for all elements
measured for element Z. This relationship is called “Castaing’s present. Typically, k-ratios will fall in a range from 0 to 10
first approximation” after the seminal figure in X-ray micro- depending on the relative concentration of element Z in the
analysis, who established the k-ratio as the basis for quantita- unknown and the standard. A k-ratio less than zero can
tive analysis (Castaing 1951). occur when count statistics and the fitting estimate of the
19 By taking a ratio of intensities collected under similar background intensity conspire to produce a slightly negative
conditions, the k-ratio is independent of various pieces of characteristic intensity. Of course, there is no such thing as
poorly known information. negative X-ray counts, and negative k-ratios should be set to
1. The k-ratio eliminates the need to know the efficiency of zero before the matrix correction is applied. A k-ratio larger
the detector since both sample and unknown are mea- than unity happens when the standard generates fewer X-rays
sured on the same detector at the same relative position. than the unknown. This can happen if the standard contains
Since the efficiency as a multiplier of the intensity is less of the element and/or if the X-ray is strongly absorbed by
identical in the numerator and denominator of the the standard. Usually, a well-designed measurement strategy
k-ratio, the efficiency cancels quantitatively in the ratio. won’t result in a k-ratio much larger than unity. We desire to
2. The k-ratio mitigates the need to know the physics of the use a standard where the concentration of element Z is high
X-ray generation process if the same elements are excited so as to minimize the contribution of the uncertainty in the
under essentially the same conditions. The ionization amount of Z in the standard to the overall uncertainty budget
cross section, the relaxation rates, and other poorly of the measurement, as well as to minimize the uncertainty
19.3 · Sets of k-ratios
291 19
contribution of the count statistics in the standard spectrum uncertainties. For EDS, each channel in the spectrum has an
to the overall measurement precision. The ideal standard to associated Poissonian statistical uncertainty. In both cases,
maximize concentration is a pure element, but for those ele- the statistical uncertainties must be taken into account care-
ments whose physical and chemical properties prohibit them fully so that an estimate of the measurement precision can be
from being used in the pure state, for example, gaseous ele- associated with the k-ratio.
ments such as Cl, F, I, low melting elements such as Ga or In, The best practices for calculating and reporting measure-
or pure elements which deteriorate under electron bombard- ment uncertainties are described in the ISO Guide to
ment, for example, P, S, a binary compound can be used, for Uncertainty in Measurement (ISO 2008). Marinenko and
example, GaP, FeS2, CuS, KCl, etc. Leigh (2010) applied the ISO GUM to the problem of k-ratios
and quantitative corrections in X-ray microanalytical mea-
surements. In the case of WDS measurements, the applica-
TIP
tion of ISO GUM is relatively straightforward and the details
To get the most accurate trace and minor constituent are in Marinenko and Leigh (2010). For EDS measurements,
measurements, it is best to average together many the process is more complicated. If uncertainties are associ-
k-ratios from distinct measurements before applying the ated with each channel in the standard and unknown spec-
matrix correction. Don’t truncate negative k-ratios before tra, the k-ratio uncertainties are obtained as part of the
you average or you’ll bias your results in the positive process of weighted linear squares fitting.
direction.
ZAF ( Cstd ;P ) This procedure should be performed with care and the ana-
Cunk = k z Cstd (19.5b)
ZAF ( Cunk ;P ) lytic total reported along with the normalized mass fractions.
Normalization is not guaranteed to improve results and can
However, there is a problem. Our ability to calculate kZ cover up for some measurement errors like missing an ele-
depends upon knowledge of the composition of the unknown, ment or inappropriately accounting for sample morphology.
Cunk. Unfortunately, we don’t know the composition of the The analytical total is important information and the nor-
unknown. That is what we are trying to measure. malized mass fractions should never be reported without
19 Fortunately, we can use a trick called “iteration” or suc- also reporting the analytical total. Any analysis which sums
cessive approximation to solve this dilemma. The strategy is exactly to unity should be viewed with some skepticism.
as follows: Careful inspection of the raw analytical total is a critical
1. Estimate the composition of the unknown. Castaing’s First step in the analytical process. If all constituents present are
Approximation is a good place to start. measured with a standards-based/matrix correction proce-
2. Calculate an improved estimate of CZ,unk based on the dure, including oxygen (or another element) determined by
previous estimated composition. the method of assumed stoichiometry, then the analytical
3. Update the composition estimate based on the new calcu- total can be expected to fall in the range 0.98 to 1.02 (98
lation. weight percent to 102 weight percent). Deviations outside
4. Test whether the resulting computed k-ratios are suffi- this range should raise the analyst’s concern. The reasons for
ciently similar to the measured k-ratios. such deviations above and below this range may include
5. Repeat steps 2–5 until step 4 is satisfied. unexpected changes in the measurement conditions, such as
19.7 · Other Ways to Estimate C Z
293 19
variations in the beam current, or problems with the speci- occur in natural minerals. The choice of oxidation state can
men, such as local topography such as a pit or other excur- be selected by the user, often relying upon independent
sion from an ideal flat polished surface. For a deviation below information such as a crystallographic determination or
the expected range, an important additional possibility is that based upon the most common oxidation state that is encoun-
there is at least one unmeasured constituent. For example, if tered in nature.
a local region of oxidation is encountered while analyzing a The same basic concept can be applied to other elements
metallic sample, the analytical total will drop to approxi- which combine in predicable ratios.
mately 0.7 (70 weight percent) because of the significant frac-
tion of oxygen in a metal oxide. Note that “standardless
analysis” (see below) may automatically force the analytical 19.7.2 Waters of Crystallization
total to unity (100 weight percent) because of the loss of
knowledge of the local electron dose used in the measure- Water of crystallization (also known as water of hydration or
ment. Some vendor software uses a locally measured spec- crystallization water) is water that occurs within crystals.
trum on a known material, e.g., Cu, to transfer the local Typically, water of crystallization is annotated by adding
measurement conditions to the conditions used to measure “·nH2O” to the end of the base chemical formula. For exam-
the vendor spectrum database. Another approach is to use ple, CuSO4 · 5H2O is copper(II) sulfate pentahydrate. This
the peak-to-background to provide an internal normaliza- expression indicates that five molecules of water have been
tion. Even with these approaches, the analytical total may not added to copper sulfate. Crystals may be fully hydrated or
have as narrow a range as standards-based analysis. The ana- partially hydrated depending upon whether the maximum
lyst must be aware of what normalization scheme may be achievable number of water molecules are associated with
applied to the results. An analytical total of exactly unity (100 each base molecule. CuSO4 is partially hydrated if there are
weight percent) should be regarded with suspicion. fewer than five water molecules per CuSO4 molecule. Some
crystals hydrate in a humid environment. Hydration mole-
cules (water) can often be driven off by strong heating, and
19.7 Other Ways to Estimate CZ some hydrated materials undergo loss of water molecules
due to electron beam damage.
k-ratios are not the only information we can use to estimate Measuring water of crystallization involves measuring O
the amount of an element Z, CZ. Sometimes it is not possible directly and comparing this measurement with the amount
or not desirable to measure kZ. For example, low Z elements, of water predicted by performing a stoichiometric calcula-
like H or He, don’t produce X-rays or low Z elements like Li, tion on the base molecule. Any surplus oxygen (oxygen mea-
B and Be produce X-rays which are so strongly absorbed that sured but not accounted for by stoichiometry) is assumed to
few escape to be measured. In other cases, we might know be in the form of water and two additional hydrogen atoms
the composition of the matrix material and all we really care are added to each surplus oxygen atom. The resulting com-
about is a trace contaminant. Alternatively, we might know position can be reported as the base molecule + “·nH2O”
that certain elements like O often combine with other ele- where n is the relative number waters per base molecule.
ments following predictable stoichiometric relationships. In
these cases, it may be better to inject other sources of infor-
mation into our composition calculation algorithm. 19.7.3 Element by Difference
19.8 Ways of Reporting Composition Starting with the atomic fractions, the mass fractions are
calculated according to the formula
19.8.1 Mass Fraction N
Cw,i = ( Ca,i ∗ Ai ) Σ (C a,i ∗ Ai ) (19.11)
The most common way to report the composition of a mate- 1
Ca,Al = 26.9815
N 0.1166 0.04910.2120 0.1090 0.0774 0.4275
Ca,i = ( Cw,i / Ai ) Σ (C w,i / Ai ) (19.10) + + + + +
1 24.305 26.9815 28.085 40.078 55.845 15.999
(19.13)
Where N is the number of elements involved in the mixture.
19.9 · The Accuracy of Quantitative Electron-E xcited X-ray Microanalysis
295 19
.. Table 19.1 Three different ways to report the composition of NIST SRM 470 glass K412.
K412 Glass
Element Mg Al Si Ca Fe O Sum
Valence 2 3 1 2 2 −2 –
Oxide fraction 0.1933 ± 0.0020 0.0927 ± 0.0020 0.4535 ± 0.0020 0.1525 ± 0.0020 0.0996 ± 0.0020 – 0.9916
MgO Al2O3 Si02 CaO FeO
Note the analytic total is less than 1, indicating an imprecision in the certified value
6
15 kev
4 20 keV
0
4%
0%
4%
8%
2%
6%
0%
%
%
%
%
%
%
%
%
%
4%
10
16
22
28
34
40
46
-8
-2
-1
-5
-4
-3
-3
-2
-2
RDEV (%)
b 10
102 measurements
9
O by stoichiometry
8
7
6
Count
5
4
15 keV
3 20 keV
2
1
0
0%
4%
8%
2%
6%
0%
4%
%
%
%
%
%
%
%
%
%
4%
10
16
22
28
34
40
46
-8
-2
-5
-4
-3
-3
-2
-2
-1
RDEV (%)
298 Chapter 19 · Quantitative Analysis: From k-ratio to Composition
Standardless Analysis (two different vendors): If we solve each equation for k/Cunk and equate the resulting
expression, we discover that
M1 0.173 (+30 %) 0.400 (−3 %) 0.267 (−7 %) Cu-K
Y2Ba3Cu4O10 ZAFA = 1 ZAFB (19.15c)
M1 0.158 (+19 %) 0.362 (−12 %) 0.316 (+10 %) Cu-L
Y2Ba3Cu6O12Needless to say, these inconsistent definitions can cause sig-
M2 0.165 (+24 %) 0.387 (−6 %) 0.287 (+0.4 %) Cu-K nificant confusion. Whenever interpreting matrix correc-
Y2Ba3Cu5O11 tions in the literature, it is important to identify which
convention the author is using.
M2 0.168 (+26 %) 0.395 (−4 %) 0.276 (−3.5 %) Cu-L
Y4Ba6Cu9O21
The confusion extends to this book. Most of this book has
been written using the first convention (ZAFA) however, the
previous (third) edition of this book used the second conven-
tion (ZAFB). The following section which has been pulled
standards-based analysis, the formulae calculated from the
from the third edition continues to use the ZAFB convention
standardless results do not match the proper formula.
as this was the definition favored by the writer. NIST DTSA-II
Another shortcoming of standardless analysis is the loss
and CITZAF uses the k = Cunk ZAFA convention.
of the information on the dose and the absolute spectrometer
(Contribution of the late Prof. Joseph Goldstein taken from
efficiency that is automatically embedded in the standards-
SEMXM-3, 7 Chapter 9)
based k-ratio/matrix corrections protocol. Without the dose
Upon initial examination, it would seem that quantitative
and absolute spectrometer efficiency information, standard-
analysis should be extremely simple. Just form the ratio of the
less analysis results must inevitably be internally normalized
characteristic X-ray intensity for a given element measured
to unity (100 %) so that the calculated concentrations have
from the specimen to that measured from the standard, and
realistic meaning, thereby losing the very useful information
that ratio should be equal to the ratio of concentrations for a
present in the raw analytical total that is available in the
given element between the specimen and the standard. As
standards-based k-ratio/matrix corrections protocol. It must
was first noted by Castaing (1951), the primary generated
be noted that standardless analysis results will always sum to
intensities are roughly proportional to the respective mass
unity, even if one or more constituents are not recognized
fractions of the emitting element. If other contributions to
during qualitative analysis or are inadvertently lost from the
X-ray generation are very small, the measured intensity ratios
suite of elements being analyzed. If the local dose and spec-
between specimen and standard are roughly equal to the
trometer efficiency can be accurately scaled to the conditions
ratios of the mass or weight fractions of the emitting element.
used to record remote standards, then standardless analysis
This assumption is often applied to X-ray quantitation and is
can determine a meaningful analytical total, but this is not
called Castaing’s “first approximation to quantitative analy-
commonly implemented in vendor software.
sis” and is given by
19 Ci ,= =
unk / Ci,std I i,unk / I i ,std k (19.16)
19.10 Appendix
The terms Ci,unk and Ci,std are the composition in weight (mass)
19.10.1 he Need for Matrix Corrections
T concentration of element i in the unknown and in the stan-
To Achieve Quantitative Analysis dard, respectively. The ratio of the measured unknown-to-
standard intensities after continuum background is subtracted
There has long been confusion around the definition of the and peak overlaps are accounted for, Ii,unk/Ii,std, is the basic
expression ‘ZAF’ used to compensate for material differences experimental measurement which underlies all quantitative
in X-ray microanalysis measurements. There are two compet- X-ray microanalysis and is given the special designation as the
ing definitions. Neither is wrong and both exist in the literature “k-ratio.”
and implemented in microanalysis software. However, the two Careful measurements performed on homogeneous sub-
definitions lead to numerical values of the matrix corrections stances of known multi-element composition compared to
that are related by being numerical inverses of each other. pure element standards reveal that there are significant sys-
19.10 · Appendix
299 19
factors of ten or more in certain systems. Recognition of the
1.0 1.0 complexity of the problem of the analysis of solid samples has
Measured Fe
led numerous investigators to develop the theoretical treat-
ment of the quantitative analysis scheme, first proposed by
0.8 0.8 Castaing (1951).
0.6 0.6
19.10.2 The Physical Origin of Matrix Effects
0.4 0.4 What is the origin of these matrix effects? The X-ray intensity
generated for each element in the specimen is proportional to
the concentration of that element, the probability of X-ray
0.2 0.2
production (ionization cross section) for that element, the
Measured Ni path length of the electrons in the specimen, and the fraction
0.0 0.0 of incident electrons which remain in the specimen and are
0.0 0.2 0.4 0.6 0.8 1.0 not backscattered. It is very difficult to calculate the absolute
Weight fraction Ni, CNi generated intensity for the elements present in a specimen
directly. Moreover, the intensity that the analyst must deal
.. Fig. 19.5 Measured Fe K-L3 and Ni K-L3 k-ratios versus the weight with is the measured intensity. The measured intensity is even
fraction of Ni at E0 = 30 keV. Curves are measured k-ratio data, while more difficult to calculate, particularly because absorption
straight lines represent ideal behavior (i.e., no matrix effects) and fluorescence of the generated X-rays may occur in the
specimen, thus further modifying the measured X-ray inten-
tematic deviations between the ratio of measured intensities sity from that predicted on the basis of the ionization cross
and the ratio of concentrations. An example of these devia- section alone. Instrumental factors such as differing spec-
tions is shown in . Fig. 19.5, which depicts the deviations of trometer efficiency as a function of X-ray energy must also be
measured X-ray intensities in the iron-nickel binary system considered. Many of these factors are dependent on the
from the linear behavior predicted by the first approximation atomic species involved. Thus, in mixtures of elements,
to quantitative analysis, Eq. (19.16). . Figure 19.5 shows the matrix effects arise because of differences in elastic and
measurement of Ii,unk/Ii,std = k for Ni K-L3 and Fe K-L3 in nine inelastic scattering processes and in the propagation of
well-characterized homogeneous Fe-Ni standards (Goldstein X-rays through the specimen to reach the detector. For con-
et al. 1965). The data were taken at an initial electron beam ceptual as well as calculational reasons, it is convenient to
energy of 30 keV and a take-off angle ψ = 52.5°. The intensity divide the matrix effects into atomic number, Zi; X-ray
ratio kNi or kFe is the Ii,unk/Ii,std measurement for Ni and Fe, absorption, Ai; and X-ray fluorescence, Fi, effects.
respectively, relative to pure element standards. The straight Using these matrix effects, the most common form of the
lines plotted between pure Fe and pure Ni indicate the rela- correction equation is
tionship between composition and intensity ratio given in
Eq. (19.16). For Ni K-L3, the actual data fall below the linear Ci , unk / Ci ,std = = [ ZAF]i [ Ii,unk /Ii,std ] = [ ZAF ]i .ki (19.17)
first approximation and indicate that there is an X-ray
absorption effect taking place, that is, more absorption in the where Ci,unk is the weight fraction of the element of interest in
sample than in the standard. For Fe K-L3, the measured data the unknown and Ci,std is the weight fraction of i in the stan-
fall above the first approximation and indicate that there is a dard. This equation must be applied separately for each ele-
fluorescence effect taking place in the sample. In this alloy ment present in the sample. Equation (19.17) is used to
the Ni K-L3 radiation is heavily absorbed by the iron and the express the matrix effects and is the common basis for X-ray
Fe K-L3 radiation is increased due to X-ray fluorescence by microanalysis in the SEM/EPMA.
the Ni K-L3 radiation over that generated by the bombarding It is important for the analyst to develop a good idea of
electrons. the origin and the importance of each of the three major
These effects that cause deviations from the simple linear non-linear effects on X-ray measurement for quantitative
behavior given by Eq. (19.16) are referred to as matrix or analysis of a large range of specimens.
inter-element effects. As described in the following sections,
the measured intensities from specimen and standard need
to be corrected for differences in electron backscatter and 19.10.3 ZAF Factors in Microanalysis
energy loss, X-ray absorption along the path through the
solid to reach the detector, and secondary X-ray generation The matrix effects Z, A, and F all contribute to the correction
and emission that follows absorption, in order to arrive at the for X-ray analysis as given in Eq. (19.17). This section dis-
ratio of generated intensities and hence the value of Ci,unk. The cusses each of the matrix effects individually. The combined
magnitude of the matrix effects can be quite large, exceeding effect of ZAF determines the total matrix correction.
300 Chapter 19 · Quantitative Analysis: From k-ratio to Composition
Intensity ratio k
0.6 0.6
ing term, R, for the unknown and standard yields the atomic
number matrix factor, Zi, for each element, i, in the unknown.
A discussion of the R and S factors follows. 0.4
0.4
Backscattering, R: The process of elastic scattering in a
solid sample leads to backscattering which results in the
premature loss of a significant fraction of the beam elec- 0.2 0.2
trons from the target before all of the ionizing power of
Measured Au
those electrons has been expended generating X-rays of the
various elemental constituents. From . Fig. 2.3a, which 0.0 0.0
depicts the backscattering coefficient as a function of atomic 0.0 0.2 0.4 0.6 0.8 1.0
number, this effect is seen to be strong, particularly if the Weight fraction Au, CAu
elements involved in the unknown and standard have widely
differing atomic numbers. For example, consider the analy-
.. Fig. 19.6 Measured Au L3-M5 and Cu K-L3 k-ratios versus the weight
sis of a minor constituent, for example, 1 weight %, of alu- fraction of Au at E0 = 25 keV. Curves are measured k-ratio data, while
minum in gold, against a pure aluminum standard. In the straight lines represent ideal behavior (i.e., no matrix effects)
aluminum standard, the backscattering coefficient is about
15 % at a beam energy of 20 keV, while for gold the value is K-L3 for four well-characterized homogeneous Au-Cu stan-
about 50 %. When aluminum is measured as a standard, dards (Heinrich et al. 1971). The data were taken at an initial
about 85 % of the beam electrons completely expend their electron beam energy of 15 keV and a take-off angle of 52.5°,
energy in the target, making the maximum amount of Al and pure Au and pure Cu were used as standards. The atomic
K-L3 X-rays. In gold, only 50 % are stopped in the target, so number difference between these two elements is 50. The
by this effect, aluminum dispersed in gold is actually under straight lines plotted on . Fig. 19.6 between pure Au and
represented in the X-rays generated in the specimen relative pure Cu indicate the relationship between composition and
to the pure aluminum standard. The energy distribution of intensity ratio given in Eq. (19.17). For both Au L-M and Cu
backscattered electrons further exacerbates this effect. Not K-L3, the absorption matrix effect, Ai, is less than 1 %, and the
only are more electrons backscattered from high atomic fluorescence matrix effect, Fi, is less than 2 %. For Cu K-L3,
number targets, but as shown in . Fig. 2.16a, b, the back- the measured data fall above the first approximation and
scattered electrons from high atomic number targets carry almost all the deviation is due to the atomic number effect,
off a higher fraction of their incident energy, further reduc- the difference in atomic number between the Au-Cu alloy
ing the energy available for ionization of inner shells. The and the Cu standard. As an example, for the 40.1 wt% Au
integrated effects of backscattering and the backscattered specimen, the atomic number matrix factor, ZCu, is 1.12, an
electron energy distribution form the basis of the “R-factor” increase in the Cu K-L3 intensity by 12 %. For Au L-M, the
in the atomic number correction of the “ZAF” formulation measured data fall below Castaing‘s first approximation and
of matrix corrections. almost all the deviation is due to the atomic number effect.
Stopping power, S: The rate of energy loss due to inelastic As an example, for the 40.1 wt % Au specimen, the atomic
scattering also depends strongly on the atomic number. For number effect, ZAu, is 0.806, a decrease in the Au L-M
19 quantitative X-ray calculations, the concept of the stopping intensity by 20 %. In this example, the S factor is larger and
power, S, of the target is used. S is the rate of energy loss given the R factor is smaller for the Cu K-L3 X-rays leading to a
by the Bethe continuous energy loss approximation, Eq. (1.1), larger S/R ratio and hence a larger ZCu effect. Just the oppo-
divided by the density, ρ, giving S = − (1/ρ)(dE/ds). Using the site is true for the Au L-M X-rays leading to a smaller ZAu
Bethe formulation for the rate of energy loss (dE/ds), one effect. The effects of R and S tend to go in opposite directions
observes that the stopping power is a decreasing function of and to cancel.
atomic number. The low atomic number targets actually
remove energy from the beam electron more rapidly with X-ray Generation With Depth, φ(ρz)
mass depth (ρz), the product of the density of the sample (ρ), A second approach to calculating the atomic number effect is
and the depth dimension (z) than high atomic number tar- to determine the X-ray generation in depth as a function of
gets. atomic number and electron beam energy. As shown in
An example of the importance of the atomic number 7 Chapters 1, 2, and 4, the paths of beam electrons within the
effect is shown in . Fig. 19.6. This figure shows the measure- specimen can be represented by Monte Carlo simulations of
ment of the intensity ratio kAu and kCu for Au L-M and Cu electron trajectories. In the Monte Carlo simulation tech-
19.10 · Appendix
301 19
E0 = 15 keV
K-L3 = generation
ϕ(ρz) = distribution
AI Ti Cu
1 µm 1 µm 1 µm
.. Fig. 19.7 Monte Carlo simulations (Joy Monte Carlo) of X-ray gen- the resulting φ(ρz) distribution. (lower) the φ(ρz) distribution is plotted
eration at E0 = 15 keV for Al K-L3, Ti K-L3, and Cu K-L3, showing (upper) with the associated f(χ) distribution showing the escape of X-rays fol-
the sites of X-ray generation (red dots) projected on the x-z plane, and lowing absorption
nique, the detailed history of an electron trajectory is calcu- characteristic X-rays. A decrease in overvoltage, U, decreases
lated in a stepwise manner. At each point along the trajectory, the energy range over which X-rays can be produced.
both elastic and inelastic scattering events can occur. The One can observe from . Fig. 19.7 that there is a non-even
production of characteristic X-rays, an inelastic scattering distribution of X-ray generation with depth, z, for specimens
process, can occur along the path of an electron as long as the with various atomic numbers and initial electron beam ener-
energy E of the electron is above the critical excitation energy, gies. This variation is illustrated by the histograms on the left
Ec, of the characteristic X-ray of interest. side of the Monte Carlo simulations. These histograms plot the
. Figure 19.7 displays Monte Carlo simulations of the number of X-rays generated with depth into the specimen. In
positions where K-shell X-ray interactions occur for three detail the X-ray generation for most specimens is somewhat
elements, Al, Ti, and Cu, using an initial electron energy, E0, higher just below the surface of the specimen and decreases to
of 15 keV. The incoming electron beam is assumed to have a zero when the electron energy, E, falls below the critical excita-
zero width and to impact normal to the sample surface. X-ray tion energy, Ec, of the characteristic X-ray of interest.
generation occurs in the lateral directions, x and y, and in As illustrated from the Monte Carlo simulations, the
depth dimension, z. The micrometer marker gives the dis- atomic number of the specimen strongly affects the distribu-
tance in both the x and z dimensions. Each dot indicates the tion of X-rays generated in specimens. These effects are even
generation of an X-ray; the dense regions indicate that a large more complex when considering more interesting multi-
number of X-rays are generated. This figure shows that the element samples as well as the generation of L and M shell
X-ray generation volume decreases with increasing atomic X-ray radiation.
number (Al, Z = 13; Ti, Z = 22; Cu, Z = 29) for the same initial . Figure 19.7 clearly shows that X-ray generation varies
electron energy. The decrease in X-ray generation volume is with depth as well as with specimen atomic number. In prac-
due to (1) an increase in elastic scattering with atomic num- tice it is very difficult to measure or calculate an absolute
ber, which deviates the electron path from the initial beam value for the X-ray intensity generated with depth. Therefore,
direction; and (2) an increase in critical excitation energy, Ec, we follow the practice first suggested by Castaing (1951) of
that gives a corresponding decrease in overvoltage U using a relative or a normalized generated intensity which
(U = E0/Ec) with atomic number. This decreases the fraction varies with depth, called φ (ρz). The term ρz is called the
of the initial electron energy available for the production of mass depth and is the product of the density ρ of the sample
302 Chapter 19 · Quantitative Analysis: From k-ratio to Composition
2.0
2.5 Eo = 15 keV
Eo = 15 keV
ϕm
2.0 AI Kα in AI
1.5 ϕ0
Ai Kα in Ti
1.5
f (rz)
Cu Kα in Cu
ϕ(ρ z)
1.0 1.0
0.5
0.5
0.0
0 100 200 300 400 500 600 700
Mass-depth (ρ z) (10-6g/cm2)
0.0
0.0 0.5 1.0
ρ Rm ρRx
Mass-depth (ρz) (10 g/cm )
-3 2 .. Fig. 19.9 Calculated φ(ρz) curves for Al K-L3 in Al; Ti K-L3 in Ti; and
Cu K-L3 in Cu at E0 = 15 keV; calculated using PROZA
0.0 0.5 1.0
Rm Rx dent beam electrons that passed through the surface layer.
Depth (z) (10-4 cm, µm) After the depth ρRm, X-ray production begins to decrease
with depth because the backscattering of the beam electrons
reduces the number of electrons available at increasing depth
.. Fig. 19.8 Schematic illustration of the φ(ρz) depth distribution of ρz and the remaining electrons lose energy and therefore ion-
X-ray generation, with the definitions of specific terms: φ0, φm, ρRm, Rm,
ρRX, and RX
izing power as they scatter at increasing depths. Finally X-ray
production goes to zero at ρz = ρRx where the energy of the
beam electrons no longer exceeds Ec.
(g/cm3) and the linear depth dimension, z (cm), so that the Now that we have discussed and described the depth dis-
product ρz has units of g/cm2. The mass depth, ρz, is more tribution of the production of X-rays using the φ(ρz) versus
commonly used than the depth term, z. The use of the mass ρz curves, it is important to understand how these curves dif-
depth removes the strong variable of density when compar- fer with the type of specimen that is analyzed and the operat-
ing specimens of different atomic number. Therefore it is ing conditions of the instrument. The specimen and operating
important to recognize the difference between the two terms conditions that are most important in this regard are the aver-
as the discussion of X-ray generation proceeds. age atomic number, Z, of the specimen and the initial electron
The general shape of the depth distribution of the gener- beam energy, E0 chosen for the analysis. Calculations of φ(ρz)
ated X-rays, the φ (ρz) versus ρz curve, is shown in . Fig. 19.8. versus ρz curves have been made for this Appendix using the
The amount of X-ray production in any layer of the histogram PROZA program (Bastin and Heijligers 1990). In . Fig. 19.9,
is related to the amount of elastic scattering, the initial elec- the φ(ρz) versus ρz curves for the K-L3 X-rays of pure Al, Ti,
tron beam energy, and the energy of the characteristic X-ray and Cu specimens at 15 keV are displayed. The shapes of the
of interest. The intensity in any layer of the φ (ρz) versus ρz φ(ρz) versus ρz curves are quite different. The φ0 values, rela-
curve is normalized to the intensity generated in an ideal thin tive to the value of φm for each curve, increase from Al to Cu
layer, where “thin” is a thickness such that effectively no due to increased backscattering which produces additional
significant elastic scattering occurs and the incident electrons X-ray radiation. On an absolute basis, the φ0 value for Cu is
pass through perpendicular to the layer. As the incident beam smaller than the value for Ti because the overvoltage, U0, for
penetrates the layers of material in depth, the length of the the Cu K-L3 X-ray at E0 = 15 keV is low (U0 = 1.67) and the
trajectory in each successive layer increases because (1) elastic energy of many of the backscattered electrons is not sufficient
19 scattering deviates the beam electrons out of the straight line to excite Cu K-L3 X-rays near the surface. The values of ρRm
path, which was initially parallel to the surface normal, thus and ρRx decrease with increasing Z and a smaller X-ray exci-
requiring a longer path to cross the layer and (2) backscatter- tation volume is produced. This decrease would be much
ing results in electrons, which were scattered deeper in the more evident if we plotted φ(ρz) versus z, the linear depth of
specimen, crossing the layer in the opposite direction follow- X-ray excitation, since the use of mass depth includes the
ing a continuous range of angles relative to the surface nor- density, which changes significantly from Al to Cu.
mal. Due to these factors, X-ray production increases with . Figure 19.10 shows calculated φ(ρz) versus ρz curves,
depth from the surface, ρz = 0, and goes through a peak, φm, at using the PROZA program (Bastin and Heijligers 1990, 1991)
a certain depth ρRm (see . Fig. 19.8). Another consequence of at an initial beam energy of 15 keV for Al K-L3 and Cu K-L3
backscattering is that surface layer production, φ0, is larger radiation for the pure elements Al and Cu. These curves are
than 1.0 in solid samples because the backscattered electrons compared in . Fig. 19.10 with calculated φ(ρz) versus ρz
excite X-rays as they pass through the surface layer and leave curves at 15 keV for Al K-L3 and Cu K-L3 in a binary sample
the sample, adding to the intensity created by all of the inci- containing Al with 3 wt % Cu. The φ0 value of the Cu K-L3
19.10 · Appendix
303 19
0.5 µm 1 µm 1 µm
Cu
K-L3 = generation
ϕ(ρz) = distribution
.. Fig. 19.11 Monte Carlo simulations (Joy Monte Carlo) of the X-ray generation volume for Cu K-L3at E0 = 10 keV, 20 keV and 30 keV. The sites of
X-ray generation (red dots) are projected on the x-z plane, and the resulting φ(ρz) distribution is shown
2.0 AI Kα in Cu
(generated)
f (rz)
1.0
AI Kα in Cu
(emitted)
0.0
0 100 200 300 400 500 600 700
Mass-depth (ρz) (10-6g/cm2)
the φ(ρz) curve, a new curve results, which gives the depth the measurement of the light elements, for example, Be, B, C,
distribution of emitted X-rays. An example of the generated N, O, and so on. For example, the mass absorption coeffi-
and emitted depth distribution curves for Al K-L3 at an ini- cient for C K-L radiation in Ni is 17,270 cm2/g, so large that
tial electron beam energy of 15 keV (calculated using the in most practical analyses, no C K-L radiation can be mea-
PROZA program (Bastin and Heijligers 1990, 1991)) is sured if the absorption path length is 1 μm. Significant
shown in . Fig. 19.14 for a trace amount (0.1 wt%) of Al in a amounts of C K-L radiation can only be measured in a Ni
pure copper matrix. The area under the φ(ρz) curve repre- sample within 0.1 μm of the surface. In such an analysis situ-
sents the X-ray intensity. The difference in the integrated ation, the initial electron beam energy should be held below
area between the generated and emitted φ(ρz) curves repre- 10 keV so that the C K-L X-ray source is produced close to
sents the total X-ray loss due to absorption. The absorption the sample surface.
correction factor in quantitative matrix corrections is calcu- As shown in . Fig. 19.12, X-rays are generated up to sev-
lated on the basis of the φ(ρz) distribution. . Figure 19.5, for eral micrometers into the specimen. Therefore the X-ray
example, illustrates the large amount of Ni K-L3 absorbed in path length (PL = t) and the relative amount of X-rays avail-
the Fe-Ni alloy series as a function of composition. able to the X-ray detection system after absorption (I/I0)
X-ray absorption is usually the largest correction factor vary with the depth at which each X-ray is generated in the
that must be considered in the measurement of elemental specimen. In addition to the position, ρz or z, at which a
composition by electron-excited X-ray microanalysis. For a given X-ray is generated within the specimen, the relation
given X-ray path length, the mass absorption coefficient, of that depth to the X-ray detector is also important since a
(μ/ρ), for each measured characteristic X-ray peak controls combination of both factors determine the X-ray path
the amount of absorption. The value of (μ/ρ) varies greatly length for absorption. . Figure 19.15 shows the geometrical
from one X-ray to another and is dependent on the matrix relationship between the position at which an X-ray is gen-
elements of the specimen (see 7 Chapter 4, “X-rays”). For erated and the position of the collimator which allows
example, the mass absorption coefficient for Fe K-L3 radia- X-rays into the EDS detector. If the specimen is normal to
tion in Ni is 90.0 cm2/g, while the mass absorption coeffi- the electron beam (. Fig. 19.15), the angle between the
cient for Al K-L3 radiation in Ni is 4837 cm2/g. Using Eq. specimen surface and the direction of the X-rays into the
(19.18) and a nominal path length of 1 μm in a Ni sample detector is the take-off angle ψ. The path length, t = PL, over
containing small amounts of Fe and Al, the ratio of X-rays which X-rays can be absorbed in the sample is calculated by
emitted at the sample surface to the X-rays generated in the multiplying the depth in the specimen, z, where the X-ray is
sample, I/I0, is 0.923 for Fe K-L3 radiation but only 0.0135 for generated, by the cosecant (the reciprocal of the sine), of the
Al K-L3 radiation. In this example, Al K-L3 radiation is very take-off angle, ψ. A larger take-off angle will yield a shorter
heavily absorbed with respect to Fe K-L3 radiation in the Ni path length in the specimen and will minimize absorption.
sample. Such a large amount of absorption must be taken The path length can be further minimized by decreasing the
account of in any quantitative X-ray analysis scheme. Even depth of X-ray generation, Rx, that is by using the minimum
more serious effects of absorption occur when considering electron beam energy, E0, consistent with the excitation of
306 Chapter 19 · Quantitative Analysis: From k-ratio to Composition
X-ray Fluorescence, F
Photoelectric absorption results in the ionization of inner
atomic shells, and those ionizations can also cause the emis-
sion of characteristic X-rays. For fluorescence to occur, an
atom species must be present in the target which has a critical
Ψ excitation energy less than the energy of the characteristic
X-rays being absorbed. In such a case, the measured X-ray
intensity from this second element will include both the
Z t direct electron-excited intensity as well as the additional
=
PL intensity generated by the fluorescence effect. Generally, the
fluorescence effect can be ignored unless the photon energy is
Specimen less than 5 keV greater than the critical excitation energy, Ec.
The significance of the fluorescence correction, Fi, can be
illustrated by considering the binary system Fe-Ni. In this sys-
tem, the Ni K-L3 characteristic energy at 7.478 keV is greater
PL = Z COSEC Ψ
than the energy for excitation of Fe K radiation,
Ec = 7.11 keV. Therefore, an additional amount of Fe K-L3 radi-
.. Fig. 19.15 Schematic diagram showing the X-ray absorption path ation is produced beyond that due to the direct beam on Fe.
length in a thick, flat-polished sample: PL = absorption path length; . Figure 19.5 shows the effect of fluorescence in the Fe-Ni sys-
ψ = X-ray take-off angle (detector elevation angle above surface)
tem at an initial electron beam energy of 30 keV and a take-off
angle, ψ, of 52.5°. Under these conditions, the atomic number
effect, ZFe, and the absorption effect, AFe, for Fe K-L3 are very
.. Table 19.4 Path Length, PL, for Al K-L3 X-rays in Al close to 1.0. The measured kFe ratio lies well above the first
approximation straight line relationship. The additional inten-
E0 Take-Off Angle, ψ Rx (μm) Path Length, PL,
(μm)
sity is given by the effect of fluorescence. As an example, for a
10 wt% Fe – 90 wt% Ni alloy, the amount of iron fluorescence
10 15 0.3 1.16 is about 25 %.
10 60 0.3 0.35
The quantitative calculation of the fluorescence effect
requires a knowledge of the depth distribution over which the
30 15 2.0 7.7 characteristic X-rays are absorbed. The φ(ρz) curve of elec-
30 60 2.0 2.3 tron-generated X-rays is the starting point for the fluorescence
calculation, and a new φ(ρz) curve for X-ray-generated X-rays
19 is determined. The electron-generated X-rays are emitted iso-
the X-ray lines used for analysis. . Table 19.4 shows the tropically. From the X-ray intensity generated in each of the
variation of the path length that can occur if one varies the layers Δ(ρz) of the φ(ρz) distribution, the calculation next
initial electron beam energy for Al K-L3 X-rays in Al from considers the propagation of that radiation over a spherical
10 to 30 keV and the take-off angle, ψ = 15° and ψ = 60°. volume centered on the depth ρz of that layer, calculating the
The variation in PL is larger than a factor of 20, from absorption based on the radial distance from the starting layer
0.35 μm at the lowest keV and highest take-off angle to and determining the contributions of absorption to each layer
7.7 μm at the highest keV and lowest take-off angle. Clearly (ρz) in the X-ray-induced φ(ρz) distribution. Because of the
the analyst’s choices of the initial electron beam energy and longer range of X-rays than electrons in materials, the X-ray-
the X-ray take-off angle have a major effect on the path length induced φ(ρz) distribution covers a much greater depth, gen-
and therefore the amount of absorption that occurs. erally an order of magnitude or more than the electron-induced
References
307 19
φ(ρz) distribution. Once the X-ray-induced φ(ρz) generated Heinrich KFJ, Myklebust RL, Rasberry SD, Michaelis RE (1971) Preparation and
distribution is determined, the absorption of the outgoing Evaluation of SRM’s 481 and 482 Gold-Silver and Gold-Copper Alloys for
Microanalysis. U.S. Government Printing Office, Washington, DC. NBS
X-ray-induced fluorescence X-rays must be calculated with Spec. Publ., pp 260–28
the absorption path length calculated as above. ISO GUM (2008). Guide to the Expression of Uncertainty in Measurement;
The fluorescence factor, Fi, is usually the least important Guide 98–3:2008: Geneva, Switzerland, Joint Group for Guides in
factor in the calculation of composition by evaluating the Metrology, Working Group 1. ISO Central. Secretariat, Vernier,
[ZAF] term in Eq. (19.17). In most analytical cases secondary Geneva, Switzerland: International Organization for Standards
Marinenko R, Leigh S (2010) Uncertainties in electron probe microanaly-
fluorescence may not occur or the concentration of the ele- sis. Proc Eur Microbeam Anal Soc 7(1):012017
ment which causes fluorescence may be small. Of the three Marinenko R, Blackburn D, Bodkin J (1990) Glasses for Microanalysis:
effects, Z, A, and F, which control X-ray microanalysis calcu- SRMs 1871-1875” National Institute of Standards and Technology
lations, the fluorescence effect, Fi, can be calculated (Reed (U.S.) Special Publication 260–112 (U.S. Gov. Printing Office,
1965), with sufficient accuracy so that it rarely limits the Washington) available at: 7 http://www.nist.gov/srm/upload/
SP260-112.PDF
development of an accurate analysis. Newbury DE, Swyt CR, Myklebust RL (1995) ‘Standardless’ quantitative
electron probe microanalysis with energy-dispersive X-ray spec-
trometry: is it worth the risk? Anal Chem 67:1866
References Pouchou J-L, Pichoir F (1991) Quantitative analysis of homogeneous
or stratified microvolumes applying the model “PAP”. In: Heinrich
Bastin GF, Heijligers HJM (1990) Quantitative electron probe microanal- KFJ, Newbury DE (eds) Electron probe quantitation. Plenum,
ysis of ultralight elements (boron – oxygen). Scanning 12:225–236 New York, p 31
Bastin GF, Heijligers HJM (1991) Quantitative analysis of homogeneous Reed SJB (1965) Characteristic fluorescence corrections in electron-
or stratified microvolumes applying the model “PAP”. In: Heinrich probe microanalysis. Brit J Appl Phys 16:913
KFJ, Newbury DE (eds) Electron probe quantitation. Plenum Press, Saunders S, Karduck P, Sloof W (2004) Certified reference materials for
New York City, p 163 micro-analysis of carbon and nitrogen. Microchim Acta 145:209
Castaing R (1951) “Application of electron probes to local chemical and Yakowitz H (1975) Methods of quantitative analysis. In: Goldstein JI,
crystallographic analysis” Ph.D. thesis, University of Paris. (English Yakowitz H, Newbury DE, Lifshin E, Colby JW, Coleman JR (eds)
translation available from the Microanalysis Society at: 7 http:// Practical scanning electron microscopy. Plenum, New York, p 338
www.microanalysissociety.org/
309 20
Quantitative Analysis:
The SEM/EDS Elemental
Microanalysis k-ratio Procedure
for Bulk Specimens,
Step-by-Step
20.1 Requirements Imposed on the Specimen
and Standards – 311
References – 339
20.2 · Instrumentation Requirements
311 20
This chapter discusses the procedure used to perform a rig- grinding and polishing. “Chemical polishing” should
orous quantitative elemental microanalysis by SEM/EDS be avoided since chemical reactions may induce shal-
following the k-ratio/matrix correction protocol using the low, near-surface compositional changes that affect
NIST DTSA-II software engine for bulk specimens. Bulk the very shallow region that is excited and sampled
specimens have dimensions that are sufficiently large to by the electron beam. Ion beam milling can be used
contain the full range of the direct electron-excited X-ray to shape and finish the specimen, but it must be
production (typically 0.5–10 μm) as well as the range of recognized that implantation of the primary ion and
secondary X-ray fluorescence induced by the propagation differential material removal caused by differences in
of the characteristic and continuum X-rays (typically the sputtering rates of the elements can modify the
10–100 μm). composition of a shallow surface layer.
100 000
Mn_20kv10nA110kHz10DT
80 000
Mn
E0 = 20 keV
60 000
Counts
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
100 000
Mn_20kv10nA110kHz10DT
80 000
5 eV bin width
60 000
Counts
20 000
0
5.60 5.65 5.70 5.75 5.80 5.85 5.90 5.95 6.00 6.05 6.10 6.15 6.20
Photon energy (keV)
energy, E0. The energy span is given by the channel width photoelectron to determine the photon energy, and incre-
multiplied by the number of channels. With a 5-eV chan- menting the appropriate energy bin in the EDS histogram
nel width, a choice of 4096 channels will provide access to by one count. The EDS time constant (also known as the
photon energies as high as 20.48 keV. For beam energy shaping time, the processing time, the maximum through-
above 20 keV, the number of channels should be increased put, or other terms in different vendor EDS systems) effec-
to retain the 5-eV channel width, or alternatively the chan- tively determines the amount of time spent on the
nel width can be increased to 10 eV, but with the conse- measurement cycle. A short time constant enables more
20 quence that fewer channels will describe each characteristic photons to be processed per unit of real (clock) time, but
peak. the trade-off of faster processing is poorer accuracy in
assigning the photon energy. While the characteristic X-ray
DS Time Constant (Resolution
E peak has a sharply defined energy, with a natural peak width
and Throughput) of a few eV or less, the EDS measurement process inevitably
The EDS is only capable of processing one photon at a time. substantially broadens the measured peak. For example,
The basic measurement cycle is the photoelectric absorp- Mn K-L2,3 has a natural width of approximately 7 eV (deter-
tion of the photon in the detector active volume, measure- mined as the FWHM) but as displayed in the EDS histo-
ment of the charge deposited by scattering of the gram, the Mn K-L2,3 peak is broadened to 122–150 eV
20.2 · Instrumentation Requirements
313 20
FWHM or more, depending on the choice of time constant. reproducible choice must be made for r since this value has
For the particular silicon drift detector (SDD)-EDS and such a strong impact on Ω and thus on the number of pho-
time constant shown in . Fig. 20.1, the broadened EDS tons detected per unit of dose.
peak has a FWHM = 126 eV. The peak width increases (i.e.,
resolution becomes poorer) as the time constant decreases.
The shortest time constant, which gives the highest through- 20.2.2 Choosing the Beam Energy, E0
put but the broadest peaks (poorest resolution), is typically
chosen for analysis situations where it is important to max- The choice of beam energy depends on the particular
imize the total number of X-ray counts per unit of clock aspects of the analysis that the analyst wishes to optimize.
(real) time, such as elemental X-ray mapping. For quantita- As a starting point, a useful general analysis strategy is to
tive analysis, better peak resolution is desirable, and thus a optimize the excitation of photon energies up to 12 keV
longer time constant should be chosen. Whichever time by choosing an incident beam energy of 20 keV, which
constant strategy is selected, it is important for standards- provides sufficient overvoltage (E0/Ec > 1.5) for K-shell
based quantitative analysis that this same time constant be (to Br) and L-shell (elements to Bi) for reasonable excita-
used for all measurements of unknowns and standards, tion. The characteristic peaks of X-ray families that occur
especially if archived standards are used. in the photon energy range from 4 keV to 12 keV are gen-
erally sufficiently separated in energy to be resolved by
EDS Calibration EDS. When it is important to measure those elements
Assigning the proper energy bin for a photon measure- whose characteristic peaks occur below 4 keV, and espe-
ment depends on the EDS being calibrated. The vendor for cially for the low atomic number elements Be, B, C, N, O
a particular EDS system will have a recommended calibra- and F, for which the characteristic peaks occur below
tion procedure that should be followed on a regular basis 1 keV and suffer high absorption, then analysis with
as part of establishing a quality measurement environ- lower beam energy, 10 keV or lower, will be necessary to
ment, with full documentation of the measurements to optimize the results.
establish the on-going calibration record. A typical cali-
bration strategy is to choose a material such as Cu that
provides (with E0 ≥ 15 keV) strongly excited peaks in the 20.2.3 Measuring the Beam Current
low photon energy range (Cu L3M5 = 0.93 keV) and the
high photon energy range (Cu K-L2,3 = 8.04 keV). The SEM should be equipped for beam current measure-
Alternatively, some EDS systems that provide a “zero ment, ideally with an in-column Faraday cup which can be
energy reference” signal will use this value with a single selected periodically during the analysis procedure to
high photon energy peak such as Cu K-L2,3 or Mn K-L2,3 to determine the beam current. As an alternative, a picoam-
perform calibration. A good quality assurance practice is meter can be installed between the electrically isolated
to begin each measurement campaign by measuring a specimen stage and the electrical ground to measure the
spectrum of Cu (or another element, e.g., Mn, Ni, etc., or a absorbed (specimen) current that must flow to ground to
compound, e.g., CuS, FeS2, etc.) under the user-defined avoid specimen charging. The specimen current is the dif-
conditions. This Cu spectrum can be compared to the Cu ference between the beam current and the loss of charge
spectrum that is stored in the archive of standards to con- due to BSE and SE emission, both of which vary with com-
firm that the current measurement conditions are identi- position. To measure the true beam current, BSE and SE
cal to those used to create the archive. This starting Cu emission must be recaptured, which is accomplished by
spectrum should always be saved as part of the quality placing the beam within a Faraday cup, which is con-
assurance plan. structed as a blind hole in a conducting material (e.g.,
metal or carbon) covered with a small entrance aperture
EDS Solid Angle (e.g., an electron microscope aperture of 50 μm diameter
The solid angle of collection, Ω, is given by or less). This Faraday cup is then placed at a suitable loca-
tion on the electrically isolated specimen stage. By locat-
Ω = A / r2 (20.1) ing the beam in the center of the Faraday cup aperture
opening, the primary beam electrons as well as all BSEs
where A is the active area of the detector and r is the distance and SEs generated at the inner surfaces are collected with
from the X-ray source on the specimen to the detector. Some very little loss through the small aperture, so that the cur-
EDS systems are mounted on a retractable arm that rent flowing to the electrical ground is the total incident
enables the analyst to choose the value of r. A consistent and beam current.
314 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
20.2.4 Choosing the Beam Current SDD-EDS, a more conservative counting strategy is sug-
gested, such that the beam current is chosen so that the
After the analyst has chosen the EDS time constant, the dead-time on the most highly excited standard of interest,
detector solid angle (for a retractable detector), and the beam for example, Al or Si, is less than 10 %. Despite the opera-
energy, the beam current should be chosen so as to give a tion of the anti-coincidence function, SDD-EDS systems
reasonable detector throughput, as expressed by the system typically show evidence of coincidence peaks above a dead-
dead-time. . Figure 20.2 shows the relationship between the time of 10 % from highly excited parent peaks, as illustrated
input count rate (ICR) of X-rays that arrive at the detector in . Fig. 20.3, which shows the in-growth of an extensive
and the output count rate (OCR) of photons that are actually set of coincidence peaks from several parent peaks. If it is
stored in the measured spectrum. The OCR initially rises lin- important to measure low intensity X-ray peaks that cor-
early with the ICR, but as photons arrive at a progressively respond to minor or trace constituents that occur in spec-
greater rate at the detector, photon coincidence begins to tral regions affected by coincidence peaks, then choosing
occur and the anti-coincidence function begins to reject the low dead-time to minimize coincidence will be an
these coincidence events, reducing the OCR. Eventually a important issue in selecting the general analytical condi-
maximum OCR value is reached beyond which the OCR tions. If there is no interest in measuring X-ray peaks of
decreases with increasing ICR, eventually falling to zero possible constituents that occur in the region of coinci-
(“paralyzable dead-time”). A useful measure of the activity dence peaks, then these regions can be ignored and a
state of the EDS detector is the system “dead-time” which is counting strategy that involves higher dead-time operation
defined as can be used.
Once the analytical conditions (EDS time constant, solid
Dead-time ( % ) = ( ICR − OCR ) / ICR ∗100 (20.2) angle, beam energy, and beam current appropriate to the
complete suite of standards) have been chosen, these condi-
A classic strategy with the low throughput Si(Li)-EDS is to tions should be used for all standards and unknowns to
select a beam current on a highly excited pure element such achieve the basic measurement consistency required for the
as Al or Si that produces a dead-time of 30 % or less. With k-ratio/matrix corrections protocol.
160,000
Output Count Rate (counts/s)
120,000
80,000
40,000
20
0
0 40,000 80,000 120,000 160,000 200,000
Input Count Rate (counts/s)
20.2 · Instrumentation Requirements
315 20
a
3 500 000 K412_20kV5nAMED5eV40kHz3DT_5ks
2 000 000
1 500 000
1 000 000
5 000 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b
K412_20kV5nAMED5eV40kHz3DT_5ks
100 000
80 000
Counts
60 000
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
c K412_20kV5nA40kHz_3DT
K412_20kV50nA387kHz_29DT
Si+Mg; Al+Al
Si+O
E0 = 20 keV
Deadtime = 3%
Counts
Deadtime = 29%
Mg+Ca
Si+Ca
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 20.3 SDD-EDS spectra of NIST SRM (glass K412, E0 = 20 keV: a, b at 3 % dead-time (red); c 3 % (red) and 29 % dead-time (blue), showing
in-growth of coincidence peaks
316 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
20.3 Examples of the k-ratio/Matrix the error budget. In this analysis and the analyses reported
Correction Protocol with DTSA II below, the relative deviation from the expected value (RDEV)
(Newbury and Ritchie 2015b) (also referred to as “relative error”) is calculated with the
“expected” value taken as the stoichiometric formula value or
20.3.1 nalysis of Major Constituents
A the value obtained from an “absolute” analytical method, just
(C > 0.1 Mass Fraction) with as in gravimetric analysis:
Well-Resolved Peaks ( Analyzed value − expected value )
RDEV = × 100% (20.3)
The EDS spectra of the minerals pyrite (FeS2) and troilite expected value
(FeS) measured at E0 = 20 keV with a dead-time of ~ 10 % are
shown in . Fig. 20.4 and feature well separated peaks for the Optimizing Analysis Strategy
Fe K- and L- families and the S K-family. These spectra were The DTSA II analysis report includes for each analyzed ele-
analyzed with Fe and CuS serving both as peak-fitting refer- ment the ZAF factors and the estimated uncertainties in
ences and as standards. CuS is chosen for the S reference and these factors as well as uncertainties due to the counting sta-
standard rather than elemental S since CuS is stable under tistics associated with the measurements of the unknown and
electron bombardment while elemental S is not stable. The of the standard (Ritchie and Newbury 2012). Careful exami-
spectrum for FeS and the residual spectrum after peak-fitting nation of these factors can be used to refine the analytical
are also shown in . Fig. 20.4. The results for seven replicate strategy to optimize the measurement. Reducing the uncer-
analyses are listed in . Table 20.1 (FeS) and . Table 20.2 (FeS2) tainty due to the counting statistics requires increasing the
along with the ZAF correction factors and the components of dose. The absorption factor A is strongly influenced by the
a 180 000
FeS_20kV10n9%DT_50s
FeS_20kV10n9%DT_50s
160 000
E0 = 20 KeV
140 000 FeS
120 000 FeS2
100 000
Counts
80 000
60 000
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b
14 000 FeS_20kV10n9%DT_50s
Residual_FeS_20kV10n9%DT_50s
12 000
E0 = 20 KeV
10 000 FeS
Fitting residual
Counts
8 000
20 6 000
4 000
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 20.4 a SDD-EDS spectra of Pyrite (FeS2) (blue) and meteoritic Troilite (FeS (red) at E0 = 20 keV. b NIST DTSA-II analysis of FeS using Fe and
CuS as peak-fitting references and as standards. The original spectrum (red) and the residual spectrum after peak-fitting (blue) are shown
20.3 · Examples of the k-ratio/Matrix Correction Protocol with DTSA II
317 20
.. Table 20.1 Analysis of FeS (meteoritic troilite) at E0 = 20 keV .. Table 20.3 Analysis of FeS at E0 = 10 keV with CuS and Fe as
with CuS and Fe as fitting references and standards Integrated fitting references and standards Integrated spectrum count,
spectrum count, 0.1–20 keV = 7,048,000; uncertainties expressed 0.1–10 keV = 5,630,000; uncertainties expressed in mass
in mass fraction. Analysis performed with Fe K-L2,3 and S K-L2,3 fraction. Analysis performed with Fe K-L2,3 and S K-L2,3
S Fe S Fe
Cav (atom frac) 0.5052 0.4948 Cav (atom frac) 0.503 0.497
C (mass frac, single analysis) 0.3699 0.6305 C (mass frac, single analysis) 0.3627 0.6257
Counting error, std 0.00020 0.0003 Counting error, std 0.0002 0.0008
Counting error, unk 0.00020 0.0007 Counting error, unk 0.0003 0.0018
.. Table 20.2 Analysis of FeS2 (pyrite) at E0 = 20 keV with CuS .. Table 20.4 Analysis of FeS2 at E0 = 10 keV with CuS and Fe as
and Fe as fitting references and standards Integrated spectrum fitting references and standards Integrated spectrum count,
count, 0.1.–20 keV = 7,765,000; uncertainties expressed in mass 0.1–10 keV = 6,253,000); uncertainties expressed in mass
fraction. Analysis performed with Fe K-L2,3 and S K-L2,3 fraction. Analysis performed with Fe K-L2,3 and S K-L2,3
S Fe S Fe
Cav (atom frac) 0.6726 0.3274 Cav (atom frac) 0.671 0.329
C (mass frac, single analysis) 0.5485 0.4657 C (mass frac, single analysis) 0.537 0.4618
Counting error, std 0.0003 0.0002 Counting error, std 0.0003 0.0006
Counting error, unk 0.0003 0.0006 Counting error, unk 0.0003 0.0016
choice of beam energy. If the beam energy can be decreased, 10 keV gives the results shown in . Tables 20.3 and 20.4. The
considering also the constraints imposed by having sufficient absorption factor A from is reduced from 1.118 to 1.04 for S
overvoltage for all elements to be analyzed, the absorption in FeS and from 1.18 to 1.06 for S in FeS2, and the relative
correction factor and its uncertainty can also be reduced. For errors are also reduced slightly, from 1 to 0.59 % for S in FeS
the Fe-S examples, lowering the beam energy from 20 to and from 0.88 to 0.65 % for S in FeS2.
318 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
60 000
40 000
20 000
0
1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3
Photon energy (keV)
b PbS_10kV20nA
14 000 Residual_PbS_10kV20nA
12 000
10 000
Counts
8 000
20
6 000
4 000
2 000
0
1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3
Photon energy (keV)
.. Fig. 20.5 a SDD-EDS spectrum of PbS (red) and residual (blue) after DTSA II analysis using CuS and PbSe as fitting references and standards.
b Expanded view
20.4 · The Need for an Iterative Qualitative and Quantitative Analysis Strategy
319 20
analysis, O has been directly analyzed with the k-ratio/matrix
.. Table 20.5 Analysis of PbS at E0 = 10 keV with CuS and PbSe
as fitting references and standards; Integrated spectrum count,
corrections protocol and not by the method of assumed stoi-
0.1–10 keV = 5,482,000; uncertainties expressed in mass fraction. chiometry. The analytical results are seen to closely match the
Analysis performed with Pb M5-N6,7 and S K-L2,3 stoichiometry of the ideal mineral formula.
S Pb
300 000
E0 = 10 KeV
250 000 MoS2
Fitting residual
200 000
Counts
150 000
100 000
50 000
0
1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3
Photon energy (keV)
b MoS2_10kV20nA11%DT100s
Residual_MoS2_10kV20nA
20 000
15 000
Counts
10 000
5 000
0
1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3
Photon energy (keV)
.. Fig. 20.6 a SDD-EDS spectrum of MoS2 (red) at E0 = 10 keV (7,326,000 counts) and residual (blue) after DTSA II analysis using CuS and Mo as
fitting references and standards. b Expanded view
a Benitoite_10kV20nA7%DT
Residual_Benitoite_10kV20nA7%DT
20 000
15 000
Counts
10 000
5 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b Benitoite_10kV20nA7%DT
Residual_Benitoite_10kV20nA7%DT
5 000
4 000
Counts
3 000
2 000
1 000
0
3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7.0
Photon energy (keV)
.. Fig. 20.7 a SDD-EDS spectrum of BaTiSi3O9 (benitoite) (red) at E0 = 10 keV (11,137,000 counts) and residual (blue) after DTSA II analysis using
BaSi2O5 (sanbornite) and Ti as fitting references and standards. b Expanded view
.. Table 20.7 Analysis of BaTiSi3O9 (benitoite) at E0 = 10 keV with Ti and sanbornite (BaSi2O5) as fitting references and standards;
integrated spectrum count = 11,366,000. Analysis performed with O K- L2,3, Si K-L2,3, Ti K-L2,3 and Ba L3-M4,5
O Si Ti Ba
F-correction 1 1 1.007 1
a 24 000 K2496_1_10kV20nAMED73kHz8DT_100s
K2496 glass Residual_K2496_1_10kV20nAMED73kHz8DT_100s]
22 000 E0 = 10 kev
K2496
20 000 1000 nA-s
O 0.323
18 000 0.1-10keV integral = 12,175,000 counts
Si 0.229
16 000 SiO2
Ti 0.018
BaSi2O5 (Sanbornite) for Ba
Counts
14 000 Ba 0.430
Ti
12 000 Ba/Ti = 23.9
10 000
8 000
6 000
4 000
2 000
0
3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
Photon energy (keV)
b 6 000 K2496_10kV20nA8%DT
Residual_K2496_10kV20nA8%DT
K2496 glass
5 000
SiO2
BaSi2O5 (Sanbornite) for Ba
4 000 No peak fitting for Ti
Counts
3 000
2 000
1 000
0
3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7.0
Photon energy (keV)
.. Fig. 20.8 a SDD-EDS spectrum of NIST microanalysis glass K2496 dards. b Same analysis protocol, but not including Ti in the peak-fitting.
(red) at E0 = 10 keV (12,175,000 counts) and residual (blue) after DTSA II Note low level peaks for Ti K-L2,3 and Ti K-M3 (Ba L-family peaks marked
analysis using BaSi2O5 (sanbornite) and Ti as fitting references and stan- as green lines)
.. Table 20.8 Analysis of NIST microanalysis glass K2496 at E0 = 10 keV with Ti and sanbornite (BaSi2O5) as fitting references and
standards; integrated spectrum count = 12,175,000. Analysis performed with O K- L2,3, Si K-L2,3, Ti K-L2,3 and Ba L3-M4,5
O Si Ti Ba
F-correction 1 1 1.01 1
a IN100_20kV_residual
80 000 IN100 IN100_20kV
E0 = 20 kev
60 000
Counts
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b 10 000 IN100_20kV_residual
IN100_20kV
8 000
Counts
6 000
4 000
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV) Note anomaly
in background at TiK-M2,3
corresponding to V K-L2,3
.. Fig. 20.9 Analysis of IN100 alloy fitting for Al, Ti, Cr, Co, Ni and Mo: K-M4,5 (4.931 keV), which closely corresponds to the energy of V K-L2,3
a full spectrum (blue) and residual spectrum after peak-fitting (red); (4.952 keV)
b expanded view, note anomaly in background at the energy of Ti
Al 0.0559 ± 0.0007 0.0562 ± 0.0007 When is a Standard Not Suitable as a Peak-Fitting Reference?
Ti 0.0473 ± 0.0001 0.0474 ± 0.0001 One of the great strengths of the k-ratio/matrix correc-
tion protocol is simplicity of the required standards. Pure
V 0.0110 ± 0.0002 elements can be used for most of the periodic table, and for
Cr 0.0981 ± 0.0002 0.0949 ± 0.0006 those elements that are not in a suitable solid form at ambient
temperature and at the low chamber pressure, stoichiometric
Co 0.1551 ± 0.0003 0.1553 ± 0.0003
binary compounds that are stable under the beam can be
Ni 0.6065 ± 0.0008 0.6069 ± 0.0008 used. This is an excellent situation for the analyst, since it is
Mo 0.0315 ± 0.0002 0.0315 ± 0.0002 generally not possible to have a multi-element standard that
is homogeneous on the microscopic scale and similar in
324 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
a IN100_20kV
10 000 IN100 Residual[IN100_20kV]
E0 = 20 kev
8 000
6 000
Counts
4 000
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b IN100_20kV_residual
IN100 Residual[IN100_20kV]_withV-fit
4 000 Residual without V
Residual with V
3 000
Counts
2 000
1 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 20.10 a IN100 superalloy, fitting for Al, Ti, V, Cr, Co, Ni, and Mo. b Comparison of fitting residuals with and without inclusion of V; the
background anomaly is eliminated
composition to a particular unknown specimen. For many the results given in . Table 20.10 (column 2) are obtained.
simple binary as well as more complex mixtures of elements, Close examination of the residual spectrum reveals the peaks
Nature favors heterogeneity on the microscale, and many of the Mn K-family. When the analysis is repeated including
combinations of elements tend to phase separate to produce Mn in the suite of fitted elements, the results given in
chemically heterogeneous microstructures. However, there . Table 20.10 (column 3) are obtained with a concentration
are important cases where microscopically homogeneous, of Mn = 0.0154, and the residual spectrum no longer contains
multi-element compositions are available, such as minerals, anomalous peaks, as shown in . Fig. 20.11c, d.
glasses, and a few metal alloys. An example is NIST Standard
Reference Material 479, an Fe-Cr-Ni alloy which is certified
to be homogeneous on a microscopic scale. SRM 479 can 20.4.3 rogressive Discovery: Repeated
P
serve as a standard for the analysis of another more complex Qualitative–Quantitative Analysis
stainless steel. . Figure 20.11 shows the spectrum of a type Sequences
316 stainless steel which the initial qualitative analysis shows
20 that in addition to Cr, Fe, and Ni also contains peaks for Si Complex unknowns may require several iterations of qualita-
and Mo. While SRM 479 is an ideal standard for this analysis tive and quantitative analysis to discover all of the constitu-
of Cr, Fe, and Ni, it is not suitable to provide peak-fitting ref- ents. For such situations, the analytical total as well as the
erences for Cr, Fe, and Ni because of the mutual interference residual spectrum serve as powerful guides to reach a suc-
of these peaks. Thus, pure elements for Cr, Fe, and Ni are cessful result. As an example, consider the spectrum of a
used for the peak-fitting references, while SRM 479 is used as monazite (a lanthanum-cerium phosphate mineral) shown
the standard, reducing the magnitude of the matrix correc- in . Fig. 20.12a, b. The elements recognized in the first quali-
tions because of the close similarity of the unknown and tative analysis stage are O, P, La, Ce (major) and Al, Si, Ca,
standard compositions. When the analysis is performed, and Th (minor). The first quantitative analysis round for
including elemental Si and Mo as references and standards, these elements, with O calculated by stoichiometry, yielded
20.4 · The Need for an Iterative Qualitative and Quantitative Analysis Strategy
325 20
a
StainlessSteel_RR_20kV
100 000 StainlessSteel_RR_residual
Stainless steel
E0 = 20 kev
80 000
Counts
60 000
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b
StainlessSteel_RR_20kV
StainlessSteel_RR_residual
6 000
Counts
4 000
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
c StainlessSteel_RR_20kV
Residual[StainlessSteel_RR_20kV]
6 000
4 000
Counts
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
d Residual[StainlessSteel_RR_20kV]
StainlessSteel_RR_residual
4 000
3 000
Counts
Stainless steel
2 000 Residual without Mn
Residual with Mn
1 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 20.11 a Stainless steel, fitting for Si, Cr, Fe, Ni, and Mo; residual in blue. b Expanded vertical scale, note detection of Mn. c Stainless steel,
fitting for Si, Cr, Mn, Fe, Ni, and Mo; residual in blue. d Comparison of residuals with fitting for Mn (red) and without (blue)
326 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
(20.5)
20.5 Is the Specimen Homogeneous?
where N represents the mean of the set of measurements for
For the most part, Nature seems to prefer heterogeneity on a each parameter, for example:
microscopic scale. That is, many combinations of two or
20 more elements spontaneously form two or more phases,
n
N sam = ∑N i / nsam (20.6)
where a phase is defined as matter that is distinct in chemical i
composition and physical state, thus creating a chemical
20.5 · Is the Specimen Homogeneous?
327 20
a 350 000
kalimantan_Monazite
300 000
Monazite (Kalimantan, Indonesia)
E0 = 20 keV
250 000
200 000
Counts
150 000
100 000
50 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b
kalimantan_Monazite
40 000
Analyze for
Major: P, La, Ce (O by stoichiometry)
Minor: Al, Si, Ca, Th
30 000
Counts
20 000
10 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
c
kalimantan_Mona3-1_20kV9p95nAMED95kHz8DT_400s
40 000 Residual_Kalimantan_Mona3-1_La-Ce
30 000
Counts
20 000
10 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 20.12 Monazite (lanthanum-cerium phosphate mineral) at quantitative analysis; d P-region, first residuals; e La-Ce-region, first
E0 = 20 keV; 0.1 keV to 30 keV = 48.7 million counts: a original spectrum; residuals; Round 2: f P-region, second residuals; g La-Ce-region, second
b vertical expansion; c Round 1: full spectrum residuals after first residuals; Round 3; h P-region, all residuals; i La-Ce-region, all residuals
328 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
d
Kalimantan_Monazite
40 000 Kal_Mona_Si-P-FIT
30 000
Counts
20 000
10 000
0
1.40 1.50 1.60 1.70 1.80 1.90 2.00 2.10 2.20 2.30 2.40 2.50
Photon energy (keV)
e
Kalimantan_Monazite
40 000 Kal_Monazite_La-Ce-FIT
30 000
Counts
20 000
10 000
0
4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7.0
Photon energy (keV)
f
Kalimantan_Monazite
40 000 Kal_Mona_Si-P-Sr-Zr_FIT
30 000
Counts
20 000
10 000
0
1.40 1.50 1.60 1.70 1.80 1.90 2.00 2.10 2.20 2.30 2.40 2.50
Photon energy (keV)
20
.. Fig. 20.12 (continued)
20.5 · Is the Specimen Homogeneous?
329 20
g
Kalimantan_Monazite
40 000 Kal_Monazite_Ti-La-Ce-Pr-Nd_FIT
30 000
Counts
20 000
10 000
0
4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7.0
Photon energy (keV)
h
Kalimantan_Monazite
40 000 Kal_Mona_Si-P-Sr-Y-Zr-Nb_FIT
Kal_Mona_Si-P-Sr-Zr_FIT
Kal_Mona_Si-P-FIT
30 000
Counts
20 000
10 000
0
1.40 1.50 1.60 1.70 1.80 1.90 2.00 2.10 2.20 2.30 2.40 2.50
Photon energy (keV)
i
Kalimantan_Monazite
40 000 Kal_Monazite_Ti-La-Ce-Pr-Nd_FIT
Kal_Mona_Ti-La-Ce-Pr-Nd_Sm_Fe-FIT
Kal_Monazite_La-Ce_FIT
30 000
Counts
20 000
10 000
0
4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7.0
Photon energy (keV)
First analysis 0.2363 ± 0.0008 0.0013 ± 0.0000 0.0049 ± 0.0000 0.1114 ± 0.0006 0.0007 ± 0.0000
Second analysis 0.2828 ± 0.0009 0.0016 ± 0.0000 0.0059 ± 0.0001 0.1240 ± 0.0007 0.0007 ± 0.0000 0.0071 ± 0.0001
Third analysis 0.2908 ± 0.0010 0.0015 ± 0.0000 0.0061 ± 0.0001 0.1263 ± 0.0007 0.0007 ± 0.0000 0.0072 ± 0.0001
Element Fe Sr Y Zr Nb La
Second analysis 0.0016 ± 0.0001 0.0098 ± 0.0002 0.0113 ± 0.0002 0.1585 ± 0.0003
Third analysis 0.0022 ± 0.0001 0.0028 ± 0.0001 0.0030 ± 0.0002 0.0117 ± 0.0002 0.0006 ± 0.0001 0.1591 ± 0.0003
Second analysis 0.2699 ± 0.0004 0.0221 ± 0.0003 0.0750 ± 0.0003 0.0040 ± 0.0001 0.9629 ± 0.0013
Third analysis 0.2709 ± 0.0004 0.0221 ± 0.0003 0.0751 ± 0.0003 0.0073 ± 0.0005 0.0040 ± 0.0001 0.9960 ± 0.0030
and nsam and nstan are the numbers of measurements of the The use of Eq. (20.7) to calculate σc for an alloy with a
sample and standard. The corresponding precision in the composition of 0.215-Mo_0.785-W (21.5 wt % Mo and
measurement of the concentration is given by 78.5 wt % W) and the spectrum shown in . Fig. 20.13 is as
follows:
( N + N ( B ) ) / n ( N − N ( B ) ) 2 + First determine the number of Mo L3-M5 and W L3-M5
= C
sam sam sam sam sam
σC 2 2
counts measured on the sample and standard as well as the
( N stan + N stan ( B ) ) / nstan ( N stan − N stan ( B ) )
2
corresponding background counts for each:
At E0 = 20 keV and iB =10 nA for an SDD-EDS of
{1− ( a − 1) C / a }
2
Ω = 0.0077 sr, the spectrum of the alloy and the residual after
(20.7)
peak-fitting, as shown in . Fig. 20.13, gives the following
intensities for a single measurement:
where the parameter “a” is the constant in the hyperbolic
relation (Ziebold and Ogilvie 1964): Mo L3-M5 bkg W L3-M5 bkg
884416 195092 868516 111279
(1 − k ) / k = a (1 − C ) / C (20.8)
The pure element standards gave the following values for
a single measurement:
The parameter “a” can be calculated using Eq. (20.8) with the
measured value of k and the calculated value of C from the 7016889 211262 1147787 134382
quantitative analysis software results. These intensities yield the following mean k-values:
Equation 20.4 makes it possible to assess statistical uncer-
tainty in an estimate of composition. For example, it can be 0.1260 0.7567
used to construct a confidence interval (e.g., ± 1.96σC gives
From the NIST DTSA-II results and Eq. (20.6):
the 95 % confidence interval) for the difference of two sam-
20 ples or to plan how many counts must be collected to be able
Mo k = 0.1235 C = 0.2132 (normalized C = 0.2148) a = 1.92
to estimate differences between two samples at the desired
W k = 0.7540 C = 0.7792 (normalized C = 0.7852) a = 1.15
level of precision. The calculation of the confidence interval
is based on the normal distribution of the estimate of C for
Substituting these values in Eq. (20.4) gives
large samples. This confidence interval is only based on the
statistical uncertainty inherent in the X-ray counts. The full Mo W
error budget requires also estimating the uncertainty in the σC = 0.0003 σC = 0.0012
principal matrix corrections for absorption (A) and scatter-
Thus, from the statistics of the X-ray counts measured for
ing/energy loss (Z) (Ritchie and Newbury 2012). NIST
the alloy and the pure element standards, the 95 % confidence
DTSA-II provides these error estimates in addition to the
limit for reproducibility is given by ± 1.96σC
error in the measurement of the k-ratio.
20.6 · Beam-Sensitive Specimens
331 20
21.5-Mo_78.5-W
60 000 21.5-Mo_78.5-W_residual
40 000
Counts
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon Energy (keV)
.. Fig. 20.13 EDS spectrum of 0.215-Mo, 0.785-W alloy showing the residuals after peak fitting; E0 = 20 keV
0.2363 0.7637
8 000
Corning Glass A CorningA_poi
p nt_10s_15kV15nA
Si ka1+2
CorningA_poi
p nt_20s_15kV15nA
E0 = 15 keV CorningA_poi
p nt_50s_15kV15nA
p nt_100s_15kV15nA
CorningA_poi
6 000 Fixed beam
150 nA-s
300 nA-s
Counts
2 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
2 000
CorningA_point_10s_15kV15nA
Corning Glass A CorningA_point_20s_15kV15nA
CorningA_point_50s_15kV15nA
E0 = 15 keV CorningA_point_100s_15kV15nA
1 500 Fixed beam
150 nA-s
300 nA-s
Counts
1 000
750 nA-s
1500 nA-s
500
0
2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3
Photon energy (keV)
.. Fig. 20.14 Corning glass A, showing Na and K migration as a function of dose for a fixed beam (15 keV, 15 nA)
Comparing the first (150-nA-s dose) and the second spectra Methods to reduce alkali element migration are based on
(300-nA-s dose), the Na intensity is seen to fall by more than a modifying the total dose, the dose per unit area (and volume),
factor of two as the dose increases, while the K intensity and/or the dose rate. Reducing the dose per unit area is often
diminishes by approximately 20 %. After a dose of 1500 nA-s, one of the most effective ways to control migration. By defo-
the Na peak is reduced to approximately 10 % of its intensity cusing the fixed beam or by scanning the focused beam rap-
after 10 s, while the K peak decreases to approximately 25 % of idly over a large area, the dose per unit area can be greatly
its original value, whereas other non-alkali elements—e.g., reduced, often by several orders of magnitude, compared to a
Mg, Al, Ca, Si, etc.—remain nearly constant with dose. Even fixed, focused beam. Because of the basic assumption of the
this time series is somewhat misleading. If the initial dose is k-ratio/matrix correction protocol that the material being
reduced by a factor of 10, the Na intensity observed is higher analyzed must have the same composition over the entire vol-
by approximately 30 %, as shown in . Fig. 20.15, while the K ume excited by the electron beam, this increased-area strategy
20 intensity is higher by approximately 5 %. At the extremely high is only valid providing the region of analytical interest is
volumetric dose created by the fixed point beam in these homogeneous over a sufficiently large to accommodate the
experiments, significant alkali migration occurs even with the defocused or rapidly scanned beam. The effect of increasing
initial short beam dwell (e.g., 1 s, 15 nA). The effects of the the scanned area is shown in . Fig. 20.16 for Corning glass A,
dose on the results obtained by quantitative analysis with where the measured Na intensity increases rapidly as the
DTSA-II are given in . Table 20.13. Even in the first analysis scanned area is increased. . Table 20.14 compares DTSA-II
(150 nA-s dose), the measured Na concentration is a factor of quantitative analyses of spectra with the same dose (15 keV,
2 lower than the synthesized glass composition, and after the 1500 nA-s) obtained with a point beam and with that beam
maximum dose utilized for this series (1500 nA-s dose), the rapidly scanning over an area 100 μm square. The scanned
Na concentration has decreased by a factor of 11. area results correspond very closely to the as-synthesized
20.6 · Beam-Sensitive Specimens
333 20
CorningA_point_10s_15kV15nA
CorningA_point_1s_15kV15nA
Corning Glass A
E0 = 15 keV
15 nA-s
Counts
150 nA-s
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
CorningA_point_10s_15kV15nA
CorningA_point_1s_15kV15nA
Corning Glass A
E0 = 15 keV
15 nA-s
Counts
150 nA-s
2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3
Photon energy (keV)
.. Fig. 20.15 Corning glass A, showing Na and K migration compared as a function of dose with a reduction of a factor of 10 difference for a
fixed beam (15 keV, 15 nA); spectra normalized to the Si peak
.. Table 20.13 DTSA-II analysis of Corning Glass A (E0 = 15 keV), oxygen by assumed stoichiometry, fixed beam
Element As-synthesized 150 nA-s raw mass 300 nA-s raw mass 750 nA-s raw mass 1500 nA-s raw mass
mass conc conc conc conc conc
5 µm square
80 000
2 µm square
60 000 1 µm square
spot
40 000
20 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
20 000
Corning Glass A CorningA_1mu_15kV15nA
CorningA_2mu_15kV15nA
E0 = 15 keV CorningA_5mu_15kV15nA
CorningA_20mu_15kV15nA
100 µm square CorningA_50mu_15kV15nA
15 000 CorningA_100mu_15kV15nA
50 µm square CorningA_spot_15kV15nA
20 µm square
Counts
5 µm square
10 000 2 µm square
1 µm square
spot
5 000
0
2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3
Photon energy (keV)
.. Fig. 20.16 Corning glass A, showing Na and K migration as a function of dose for scanning beams covering various areas (20 keV, 10 nA)
v alues for the glass, including the alkali elements Na and K, and at the same dose with a fixed beam and two different
whereas the point beam results show reductions in the Na and sizes of scanned areas (lower). Thus, the use of a scanned area
K concentrations. . Figure 20.17 shows that the measured rather than a fixed beam may be necessary when collecting a
sodium and potassium concentrations increase to reach the standard spectrum, even on a crystalline material.
synthesized values as the scanned area dimensions are
increased to cover areas above 20 x 20-μm (nominal magnifi-
cation 5 kX) for the particular dose utilized (15 keV, 1500 nA-s). 20.6.2 aterials Subject to Mass Loss
M
Thus, while scanning a large homogeneous area obviously During Electron Bombardment—the
concedes the spatial resolution capability of electron-excited Marshall-Hall Method
X-ray microanalysis, this approach may be the most expedient
technique to control and minimize alkali element migration. Thin Section Analysis
20 Materials that can serve as useful standards for sodium The X-ray microanalysis of biological and polymeric speci-
include certain crystalline minerals such as albite (NaAlSi3O8) mens is made difficult, and sometimes impossible, by several
in which the sodium is much more stable under electron forms of radiation damage that are directly caused by the elec-
bombardment. However, even for albite the use of a station- tron beam. At the beam energies used in the SEM (0.1–
ary high intensity point beam may produce significant migra- 30 keV), it is possible for the kinetic energy of individual
tion effects, as shown in . Fig. 20.18 for spectra collected beam electrons to break and/or rearrange chemical bonds.
with a stationary point beam as a function of dose (upper) The radiation damage can release smaller molecules such as
20.6 · Beam-Sensitive Specimens
335 20
.. Table 20.14 DTSA-II quantitative analysis of Corning glass A: Comparison of results with a fixed beam and scanned beam (100 μm
square) (15 keV/15 nA); oxygen by assumed stoichiometry
Element As-synthesized mass conc 1500 nA-s (fixed beam) raw mass conc 1500 nA-s (100 -μm2 scan) raw mass conc
0.08 Na
Concentration (mass)
K
Ca
0.06 20 kX
0.04
50 kX
Fixed 100 kX
0.02 500 kX
Beam
200 kX
0.00
1e-5 1e-4 1e-3 1e-2 1e-1 1e+0 1e+1 1e+2 1e+3 1e+4
Area bombarded (mm2)
336 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
5th (50 s)
3 000 10th (100 s)
2 000
1 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
E0 = 15 keV
100 000 100 s spectra
1 µm x 1 µm
80 000
Counts
10 µm x 10 µm
60 000 Fixed beam
40 000
20 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
.. Fig. 20.18 Albite (NaAlSi3O8); E0 = 15 keV, 15 nA: (upper) effect of increasing dose on the Na peak; (lower) effect of fixed beam versus scanned
beam on the Na peak
CO, CO2, and H2O that evaporate into the vacuum, causing Marshall and Hall (1966) and Hall (1968) made the original
substantial mass loss from the interaction volume. At the suggestion that the X-ray continuum could serve as an inter-
highest beam currents, typically 10–100 nA, used with a nal standard to monitor specimen changes. This assumption
focused beam at a static location, it is also possible to cause permitted development of the key procedure for beam-sensi-
highly damaging temperature elevations, which further exac- tive specimens that is used extensively in the biological com-
erbate mass loss. Indeed, when analyzing this specimen class munity and that is also applicable in many types of polymer
it should be assumed that significant mass loss will occur dur- analysis. This application marks the earliest use of the X-ray
ing the measurement at each point of the specimen. If all con- continuum as a tool (rather than simply a hindrance) for
stituents were lost at the same rate, then simply normalizing analysis, and that work forms the basis for the development of
20 the result would compensate for the mass loss that occurs the peak-to-local background method applied to challenging
during the accumulation of the X-ray spectrum. Unfortunately, geometric forms such as particles and rough surfaces. The
the matrix constituents (principally carbon compounds and technique was initially developed for applications in the high
water) can be selectively lost, while the heavy elements of beam current EPMA, but the procedure works well in the
interest in biological microanalysis (e. g., Mg, P, S, K, Ca, Fe, SEM environment.
etc.) remain in the bombarded region of the specimen and The Marshall–Hall method (Marshall and Hall 1966)
appear to be present at effectively higher concentration than requires that several key conditions:
existed in the original specimen. What is then required of any 1. The specimen must be in the form of a thin section,
analytical procedure for biological and polymeric specimens where the condition of “thin” is satisfied when the
is a mechanism to provide a meaningful analysis under these incident beam penetrates with negligible energy loss. For
conditions of a specimen that undergoes continuous change. an analytical beam energy of 10–30 keV, the energy loss
20.6 · Beam-Sensitive Specimens
337 20
passing through a section consisting of carbon approxi- concentration, Zi is the atomic number, and Ai is the atomic
mately 100–200 nm in thickness will be less than weight. The subscript “A” identifies a specific trace or minor
500 eV. This condition permits the beam energy to be analyte of interest (e.g., Mg, P, S, Cl, Ca, Fe, etc.) in the organic
treated as a constant, which is critical for the develop- matrix, while the subscript “i” represents all elements in the
ment of the correction formula. Biological specimens are electron-excited region. E0 is the incident beam energy and J
thus usually analyzed in the form of thin sections cut to is the mean ionization energy, a function only of atomic
approximately 100-nm thickness by microtome. number as used in the Bethe continuous energy loss equation
Polymers may also be analyzed when similarly prepared Assumption 2 provides that the quantity ∑(Ci•Zi2/Ai) in
as thin sections by microtoming or by ion beam milling. Eq. (20.9) for the biological or polymeric specimen to be ana-
Such a specimen configuration also has a distinct lyzed is dominated by the low-Z constituents of the matrix.
advantage for improving the spatial resolution of the (Some representative values of ∑(Ci•Zi2/Ai) are 3.67 (water),
analysis compared to a bulk specimen. The analytical 3.01 (nylon), 3.08 (polycarbonate) and 3.28 (protein with S).
volume in such thin specimens is approximately the Typically the range is between 2.8 and 3.8 for most biological
cylinder defined by the incident beam diameter and the and many polymeric materials.) The unknown contribution
section thickness, which is at least a factor of 10–100 of the analyte, CA, to the sum may be neglected when consid-
smaller in linear dimensions than the equivalent bulk ering the specimen because CA is low when the analytes are
specimen case at the same energy, as shown in the trace constituents.
polymer etching experiment in the Interaction Volume To perform a quantitative analysis, Eq. (20.9) is used in
module. the following manner: A standard for which all elemental
2. The matrix composition must be dominated by light concentrations are known and which contains the analyte(s)
elements, for example, C, H, N, O, whose contributions of interest “A” is prepared as a thin cross section (satisfying
will form nearly all of the X-ray continuum and whose assumption 3). This standard is measured under defined
concentrations are reasonably well known for the beam and spectrometer parameters to yield a characteristic-
specimen. Elements of analytical interest such as Mg, P, to-continuum ratio, IA/Icm. This measured ratio IA/Icm is set
S, Cl, K, Ca, and so on, the concentrations of which are equal to the right side of Eq. (20.9). Since the target being
unknown in the specimen, must only be present irradiated is a reference standard, the atomic numbers Zi,
preferably as trace constituents (<0.01 mass fraction) so atomic weights Ai and weight fractions Ci are known for all
that their effect on the X-ray continuum can be constituents, and the Ji values can be calculated as needed.
neglected. When the concentration rises above the low The only unknown term is then the constant “c” in Eq. (20.9),
end of the minor constituent range (e.g., 0.01 to 0.05 which can now be determined by dividing the measured
mass fraction or more), the analyte contribution to the intensity ratio, IA/Icm, by the calculated term. Next, under the
continuum can no longer be ignored. same measurement conditions, the characteristic “A” inten-
3. A standard must be available with a known concentra- sity and the continuum intensity at the chosen energy are
tion of the trace/minor analyte of interest and for which determined for the specimen location(s). Providing that the
the complete composition of low-atomic-number low-Z elements that form the matrix of the specimen are
elements is also known and which is stable under similar to the standard, or in the optimum case these
electron beam bombardment. Glasses synthesized with concentrations are actually known for the specimen (or can
low atomic number oxides such as boron oxide are be estimated from other information about the actual, local-
suitable for this role. The closer the low–atomic-number ized, material being irradiated by the electrons, and not some
element composition of the standard is to that of the bulk property), then this value of “c” determined from the
unknown, the more accurate will be the results. standard can be used to calculate the weight fraction of the
analyte, CA, for the specimen.
The detailed derivation yields the following general expres- This basic theme can be extended and several analytes—
sion for the Marshall–Hall method: “A,” “B,” “C,” etc.—can be analyzed simultaneously if a suit-
able standard or suite of standards containing the analytes is
CA available. The method can be extended to higher concentra-
I ch AA tions, but the details of this extension are beyond the scope of
=c (20.9) this book; a full description and derivation can be found in
I cm Z2 E
∑i Ci Ai log e 1.166 J 0 Kitazawa et al. (1983). Commercial computer X-ray analyzer
i i systems may have the Marshall–Hall procedure included in
their suite of analysis tools. The Marshall–Hall procedure
In this equation, Ich is the characteristic intensity of the peak works well for thin specimens in the “conventional” analyti-
of interest, for example, S K-L2,3 or Ca K-L2,3, and Icm is the cal energy regime (E0 ≥ 10 keV) of the SEM. The method will
continuum intensity of a continuum window of width ΔE not work for specimens where the average atomic number is
placed somewhere in the high energy portion of the spec- expected to vary significantly from one analysis point to
trum, typically above 8 keV, so that absorption effects are another, or relative to that of the standard. A bulk specimen
negligible and only mass effects are important. Ci is the mass where the beam-damaged region is not constrained by the
338 Chapter 20 · Quantitative Analysis: The SEM/EDS Elemental Microanalysis k-ratio Procedure for Bulk Specimens, Step-by-Step
dimensions of the thin section, so that the region continues ratios are obtained, they can be used for quantitation in a
to change during electron bombardment also violates the number of ways. The P/B value for one element can be com-
fundamental assumptions. Consequently, many materials pared with the P/B value for another element in the same
science applications for “soft” materials cannot be accommo- sample and to a first order:
dated by the classic Marshall–Hall procedure.
( C / C ) = h ( P / B ) / ( P / B )
i j ij
(20.10)
Bulk Biological and Organic Specimens
i j
The quantitative procedures devised by Statham and Pawley where Ci and Cj are the percentage concentrations of ele-
(1978) and Small et al. (1979) for the analysis of particles and ments i and j and hij is a correction factor which can be
rough specimens have been adapted to the analysis of bulk obtained from measurements on a standard(s) of known
biological and organic samples (Roomans 1981, 1988; Echlin composition very similar in composition to the unknown.
1998). The method is based on the use of the ratio between Once hij has been empirically for the element(s) of interest,
the intensity of the characteristic and background X-rays measurements of the P/B ratio(s) from the unknown can be
defined as P/B, where P and B are measured over the range of immediately converted into concentration ratios. An advan-
energies that defines an EDS peak. The rationale behind the tage of taking the double ratio of (P/B) in Eq. (20.10) is the
development of the method is that since the characteristic suppression of matrix effects, to a first order.
and background X-rays are generated within nearly the same Alternatively, the P/B value for an element in the sample
depth distribution, they are subject to the same composi- can be compared with the P/B value for the same element in
tional related absorption and atomic number effects. It is a standard provided there is no significant difference in the
assumed that the percentage of characteristic X-rays absorbed matrix composition between sample and standard. If the
by the sample is the same as the percentage of continuum mean atomic number of a given sample is always the same,
X-rays of the same energy which are absorbed. In the ratio the Kramers’ relationship shows that the background radia-
P/B, the absorption factor (A) is no longer relevant as it has tion is proportional to atomic number. If the mean atomic
the same value in the numerator as the denominator and thus number of the sample is always the same, then
cancels. Since backscattered electrons are being lost due to
changes in atomic number (Z), there is a similar decrease in Ci = hi ( P / B )i (20.11)
the efficiency of production of both peak and background.
Because the reduced X-ray production affects both peak and
background in a similar (although not identical way), this where hi is a constant for each element. If it is possible to
factor is also cancelled out to a first order when the ratio P/B analyze all the elements and calculate the concentration of
is measured. Additionally, because nearly all biological and/ elements such as C, H, O, and N by stoichiometry, then rela-
or organic materials consist of low atomic number matrix tive concentrations can be readily converted to absolute
elements (Zmax = 10) the secondary fluorescence effect (F) is concentrations.
low and can be treated as a secondary order correction. If there is a significant change in the composition between
Strictly, these assumptions only hold true for homogeneous the unknown and standard(s), then a correction must be
samples as the characteristic and background X-rays will vary applied based upon the dependence of the continuum upon
with changes in the average atomic number of the sample. atomic number, following the original Marshall–Hall thin
However, this is not considered to have any significant effect in section method:
cases where the P/B ratio method is applied to fully hydrated
specimens which contain 85–90 % water or to dried organic ( C / C ) = h ( P / B ) / ( P / B ) ( Z
i j ij i j i
2
/ Ai ) / ( Z j2 / Aj )
material containing a small amount of light element salts. The (20.12)
ratio of peak area to the background immediately beneath the
peak is relatively insensitive to small changes in surface geom- where Z and A are the atomic number and weight.
etry. However, the sample surface should be as smooth as is The peak-to-background ratio method has been found to
practicable because uneven fracture faces give unreliable X-ray be as efficient and accurate for biological materials as the
data because of preferential masking and absorption. more commonly used ZAF algorithms, which have been
20 Spectra are processed by the following procedure. The designed primarily for analyzing non-biological bulk sam-
peaks in the spectra of the unknown and a standard of simi- ples. Echlin (1998) gives details of the accuracy and precision
lar composition are fit by an appropriate procedure, such as of the method as applied to hydrated and organic samples.
multiple linear least squares, to determine the peak area for For the analysis of a frozen hydrated tea leaf standard where
element i, Pi. The spectrum after peak-fitting and subtraction independent analysis by atomic absorption spectroscopy was
is then examined again to determine the background inten- available for comparison, peak-to-background corrections
sity remaining in the peak region of interest, giving the cor- generally gave results within ±10 % relative for trace Mg, Al,
responding Bi at the same photon energy. Once accurate P/B Si, and Ca over a range of beam energies from 5 to 20 keV.
References
339 20
References Ritchie N, Newbury D, Davis J (2012) EDS Measurements at WDS preci-
sion and accuracy using a silicon drift detector. Microsc Microanal
18:892
Echlin P (1998) Low-voltage energy-dispersive X-ray microanalysis of
Ritchie N, Newbury D (2012) Uncertainty estimates for electron probe
bulk biological specimens. Microsc Microanal 4:577
X-ray microanalysis measurements. Anal Chem 84:9956
Hall T (1968) “Some aspects of the microprobe analysis of biologi-
Roomans GM (1981) Quantitative electron probe X-ray microanalysis of
cal specimens” in Quantitative Electron Probe Microanalysis, ed.
biological bulk specimens. SEM/1981/II (AMF O’Hare, IL, SEM, Inc.,)
Heinrich K. (U.S. Government Printing Office, Washington, DC) 269
p 345
Kitazawa T, Shuman H, Somlyo A (1983) Quantitative electron probe
Roomans GM (1988) Quantitative X-ray microanalysis of biological
analysis: problems and solutions. Ultramicroscopy 11:251
specimens. J Electron Microsc Tech 9:19
Lifshin E, Doganaksoy N, Sirois J, Gauvin R (1999) Statistical consider-
Small J, Heinrich K, Newbury D, Myklebust R (1979) Progress in the
ations in microanalysis by energy-dispersive spectrometry. Microsc
development of the peak-to-background method for the quantita-
Microanal 4:598
tive analysis of single particles with the electron probe. SEM/1979/
Marshall D, Hall T (1966) “A method for the microanalysis of thin films”,
II, p 807
in X-ray Optics and Microanalysis, eds. Castaing R, Deschamps P,
Statham P, Pawley J (1978) A new method for particle X-ray microanaly-
Philibert J. (Hermann, Paris) 374
sis based on peak-to-background measurements. SEM/1978/I,
Newbury D, Ritchie N (2015a) Review: performing elemental microanal-
p 469
ysis with high accuracy and high precision by Scanning Electron
Vicenzi E, Eggins S, Logan A, Wysoczanski R (2002) Microbeam charac-
Microscopy/Silicon Drift Detector Energy Dispersive X-ray Spec-
terization of Corning archeological reference glasses: new addi-
trometry (SEM/SDD-EDS). J Mats Sci 50:493
tions to the Smithsonian microbeam standard collection. J Res
Newbury D, Ritchie N (2015b) Quantitative electron-excited X-ray
National Inst Stand Technol 107:719
microanalysis of Borides, Carbides, Nitrides, Oxides, and Fluorides
Ziebold T (1967) Precision and sensitivity in electron microprobe analy-
with Scanning Electron Microscopy/Silicon Drift Detector Energy-
sis. Anal Chem 39:858
Dispersive Spectrometry (SEM/SDD-EDS) and NIST DTSA-II. Microsc
Ziebold T, Ogilvie R (1964) An empirical method for electron microanal-
Microanal 21:1327
ysis. Anal Chem 36:322
341 21
References – 357
“Trace analysis” refers to the measurement of constituents example, Au-Cu or Au-Ag. When element “A” is present as a
presents at low fractional levels. For SEM/EDS the following major constituent, as the concentration of “A” is reduced and
arbitrary but practical definitions have been chosen to desig- replaced by “B,” the matrix effects are likely to change signifi-
nate various constituent classes according to these mass con- cantly as the composition is changed. The impact of “B” on
centration (C) ranges: electron scattering and X-ray absorption of “A” is likely to pro-
Major: C > 0.1 mass fraction (greater than 10 wt%) duce a complex and non-linear response for NA as a function
Minor: 0.01 ≤ C ≤ 0.1 (1 wt% to 10 wt%) of CA. However, when the element of interest “A” is present in
Trace: C < 0.01 (below 1 wt%) the trace concentration range, the matrix composition is very
nearly constant as the concentration of “A” is lowered and
Note that by these definitions, while “major” and “minor” replaced by “B,” resulting in a monotonic dependence for NA
constituents have defined ranges, “trace” has no minimum. as a function of CA, as shown in . Fig. 21.1, known as a “work-
Strictly, the presence of a single atom of the element of inter- ing curve,” which in the case of a dilute constituent will be
est within the electron-excited mass that is analyzed by X-ray linear. Consider that we have a known point on this linear plot
spectrometry represents the ultimate trace level that might be with the values (Ns, Cs) in the trace concentration range that
measured for that species, but such detection is far below the corresponds to the measurement of a known standard or is
practical limit for electron-excited energy dispersive X-ray the result of a quantitative analysis of an unknown. The slope
spectrometry of bulk specimens. In this section trace analysis m of this linear function can be calculated from this known
down to levels approaching a mass fraction of C = 0.0001 (100 point and the 0 concentration point, for which NA = 0 + Ncm,
ppm (ppm) will be demonstrated. While such trace measure- since there will still be counts, Ncm, in the “A” energy window
ments are possible with high count EDS spectra, achieving due to the X-ray continuum produced by the other element(s)
reliable trace measurements by electron-excited energy dis- that comprise the specimen:
persive X-ray spectrometry requires careful attention to iden-
tifying and eliminating, if possible, pathological contributions slope = ( N s − N cm ) / ( Cs − 0 ) (21.1)
to the measured spectrum from unexpected remote radiation
sources such as secondary fluorescence and backscattered The slope-intercept form (y = mx + b) for the linear expres-
electrons (Newbury and Ritchie 2016). sion for the X-ray counts in the “A” energy window, NA, as a
function of concentration, CA, can then be constructed as:
(NDL,CDL)
the BSE energy distribution; (2) the absorption effect, where NDL
the mass absorption coefficient depends on all elements pres-
?
ent; and (3) the secondary fluorescence effect, where the
absorption of characteristic and continuum X-rays with pho- Ncm
ton energies above the ionization energy of the element of
21 interest leads to additional emission for that element. Consider
the number of X-rays, NA, including characteristic plus con-
0 CDL Cs
tinuum, measured in the energy window that spans the peak
for element “A” as a function of the concentration CA in a spe- Concentration, CA
cific mixture of other elements “B,” “C,” “D,” and so on. A
simple example would be a binary alloy system where there is .. Fig. 21.1 Linear working curve for a constituent at low
complete solid solubility from pure “A” to pure “B,” for concentration in an effectively constant matrix
21.2 · E stimating the Concentration Limit of Detection, CDL
343 21
photon energies, the concentration limit of detection, CDL, 21.2.1 Estimating CDL from a Trace or Minor
must have a finite, non-zero value that will be found at some Constituent from Measuring a Known
point (NDL, CDL) along the linear response between (Ncm, 0) Standard
and (Ns, Cs). The work of (Currie 1968) can be used to
define the condition at which the counts from the charac- . Figure 21.2 shows a high count silicon drift detector
teristic X-ray emission can be distinguished with a high (SDD)-EDS spectrum of K493 (and the residual spectrum
degree of confidence above the natural statistical fluctua- after fitting for O, Si, and Pb), in NIST Research Material
tions in the background counts: the characteristic counts glass with the composition (as-synthesized) listed in
must exceed three times the standard deviation of the . Table 21.1. . Table 21.1 also lists the measured peak inten-
background: sity Ns and the background Ncm determined for this spectrum
with the EDS spectrum measurement tools in DTSA-II. For a
N A > 3 N cm
½
(21.3)
single measurement, the values for Cs, Ns, and Ncm inserted in
Eq. (21.5c) gives the estimate of CDL for each trace element, as
Thus, at the concentration limit of detection, CDL:
also listed in . Table 21.1. If n = 4 repeated measurements
N DL = continuum + characteristic = N cm + 3 N cm
½
(21.4a)
were made (or a single measurement was performed at four
times the dose), CDL would be lowered by a factor of 2.
Examination of the values for CDL in . Table 21.1 reveals
N DL − N cm = 3 N cm
½
(21.4b)
more than an order-of-magnitude variation depending on
atomic number, for example, CDL = 52 ppm for Al while
Substituting these conditions for (NDL, CDL) in Eq. (21.2): CDL = 754 ppm for Ta. This strong variation arises from dif-
N DL = ( N s − N cm ) / ( Cs − 0 ) CDL + N cm (21.5a) ferences in the relative excitation (overvoltage) and fluores-
cence yield for the various elements, differences in the
N DL − N cm = 3 N cm
½
= ( N s − N cm ) / ( Cs − 0 ) CDL (21.5b) continuum intensity, and the partitioning of the characteris-
tic X-ray intensity among widely separated peaks for the
CDL = 3 N cm
½
/ ( N s − N cm ) Cs (21.5c) L-family of the higher atomic number elements, for example,
Ce and Ta.
Equation (21.5c) enables estimation of CDL from the results of
a single analysis of an unknown or from a single measure- 21.2.2 Estimating CDL After Determination
ment of a known standard to provide a value for Cs. The cor-
of a Minor or Trace Constituent
responding measured EDS spectrum is used to determine Ns
and Ncm. If n repeated measurements are made, Ns and Ncm
with Severe Peak Interference
are then taken as averages, Ṅs and Ṅcm, over the n measure- from a Major Constituent
ments, and Eq. (21.5c) becomes
Because of the relatively poor energy resolution of EDS, peak
CDL = 3 N cm
½
/ ( N s − N cm ) n1/ 2 Cs (21.6) interference situations are frequently encountered. Multiple
linear least squares peak fitting can separate the contribu-
tions from two or more peaks within an energy window. This
CDL is an estimate of the concentration level of a constituent
effect is illustrated in . Fig. 21.3 for Corning Glass A, the
that can just be detected with a high degree of confidence.
composition of which is listed in . Table 21.2 along with the
Quantification at CDL is not reasonable because the error
DTSA II analysis. There is a significant interference for K and
budget is dominated by the variance of the continuum. To
Ca upon Sb and Sn. The initial qualitative analysis identified
achieve meaningful quantitation of trace constituents,
K and Ca, as shown in . Fig. 21.3(a). When MLLS peak fit-
(Currie 1968) further defines a minimum quantifiable con-
ting is applied for K and Ca, the Sn and Sb L-family peaks are
centration, CMQ, which requires that the characteristic inten-
revealed in the residual spectrum, . Fig. 21.3(b). The limit of
sity exceed 10 Ncm½: Inserting this criterion in Eq. 21.5c gives
detection calculated from the peak for Sn L determined from
peak and background intensities determined from this resid-
CMQ = 10 N cm
½
/ ( N s − N cm ) Cs (21.7)
ual spectrum is CDL = 0.00002 (200 ppm).
K493_20kV10nA11%DT200s
K493
10 0000
E0 = 20 keV
0.1-20 keV = 15.1 million counts
10 000
Counts
1000
100
10
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0
Photon energy (keV)
16 000 K493_20kV10nA11%DT200s
Residual[K493_20kV10nA11%DT200s]
14 000
K493
12 000 Residual after peak fitting
10 000
Counts
8 000
6 000
4 000
2 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0
Photon energy (keV)
.. Fig. 21.2 SDD-EDS spectrum (0.1–20 keV = 29 million counts) of NIST Research Material Glass K493 at E0 = 20 keV and residual after MLLS peak
fitting for O, Si, and Pb
.. Table 21.1 Limits of detection estimated from a known standard K493 (E0 = 20 keV 0.1–20 keV = 29 million counts)
Element Mass conc Ns (counts) Ncm (counts) CDL (mass conc) CDL (ppm)
O 0.2063
Si 0.1304
Pb 0.6413
21.3 · M
easurements of Trace Constituents by Electron-Excited Energy Dispersive X-ray Spectrometry
345 21
a
CorningA_20kV10nA117%DT1750s
120 000 Residual_CorningA_20kV10nA117%DT
40 000
20 000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
Ba = 0.00050
40 000
Co = 0.0013
20 000
0
2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5
Photon energy (keV)
.. Fig. 21.3 a SDD-EDS spectrum (0.1–20 keV = 7.8 million counts) of Corning Glass A at E0 = 20 keV and residual spectrum after MLLS peak fitting
for K and Ca. b Expansion of K and Ca region showing detection of Sn and Sb L-family X-rays
trace constituents that might be present in regions of the spec- correction factors for these constituents at trace levels in Si
trum that consist only of the X-ray continuum background. with DTSA-II, although this is generally a small correction.
An example is shown in . Fig. 21.4 for high purity Si. The
spectrum consists of the Si K-shell X-rays, the associated coin-
cidence peak, and the X-ray continuum background. Consider 21.3 easurements of Trace Constituents
M
that the task is to estimate CDL for several elements, for exam- by Electron-Excited Energy Dispersive
ple, Al, Cr, and Cu. In the absence of a specimen of Si with X-ray Spectrometry
known trace or minor levels of these elements, a reasonable
estimate of CDL can be made by determining the threshold 21.3.1 I s a Given Trace Level Measurement
k-ratio relative to a pure element, as illustrated for Cr and Cu
Actually Valid?
with the spectra superimposed in . Fig. 21.4(b). Using the
energy window for Cr K-L2,3 (CrKα), the continuum intensity
Trace analysis with high count EDS spectra can be performed
in the Si spectrum at Cr K-L2,3 is measured, Ncm_Si-at-Cr, and is
to concentrations levels down to 0.0001 mass fraction
divided by the Cr K-L2,3 intensity from the Cr spectrum at the
(100 ppm) in the absence of interferences and 0.0005
equivalent dose, giving the k-ratio kDL for detection:
(500 ppm) when peak interference occurs. The careful ana-
lyst will always ask the question, Is a given trace measure-
kDL = 3 N cm
½
_Si -at -Cr / N Cr (21.8) ment valid? That is, does the measured trace constituent
actually originate within the interaction volume of the speci-
Values of for kDL for Al, Cr, and Cu as measured for this Si men that is excited by primary electron beam, or is it the
spectrum are listed in . Table 21.3. These k-ratios can be result of remote excitation of another part of the specimen or
converted into CDL values by calculating the ZAF matrix from components of the SEM itself?
346 Chapter 21 · Trace Analysis by SEM/EDS
a 100 0000
Si_20kV10nA14%DT
Si
100 000
E0 = 20 keV
Counts
10 000
1000
100
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
b 100 0000
Si_20kV10nA14%DT
Cr_20kV10nA9%DT
Si Cu_20kV10nA10%DT
Cr
100 000 Cu
E0 = 20 keV
Counts
10 000
1000
100
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 21.4 a SDD-EDS spectrum of Si (20 keV; 1000 nA-s; 0.1–20 keV = 22 million counts) with energy windows defined for Al, Cr, and Cu.
b Spectrum for Si with the spectra of Cr and Cu superimposed
.. Table 21.3 Estimated limits of detection kDL for a Si Range of characteristic-induced fluorescence
spectrum with 110 million counts (0.1–20 keV)
Range of direct electron excitation
in 75Ni-25Fe at E0 = 25 keV
Element kDL ZAF CDL (mass conc) CDL (ppm)
99%
interaction volume into the K412 glass to excite secondary
fluorescence. The calculated spectrum shown in . Fig. 21.6
shows the presence of an apparent trace level of Fe (and to a Ni K-L3 in 75Ni-25Fe
Range of secondary X-ray excitation
lesser extent, Ca, Si, Al, and Mg) in the Cu, corresponding to
(characteristic fluorescence)
k = 0.0028 relative to a pure Fe standard. The Fe k-ratio as a of Fe K-L3 by Ni K-L3
function of beam position in the Cu is also shown in
. Fig. 21.6. Even with the beam placed in the Cu at a distance
of 40 μm from the K-412, there is an apparent Fe trace level .. Fig. 21.5 Range of secondary fluorescence of FeKα by NiKα in a
in the Cu of k = 0.0004, or about 400 ppm. 75Ni-25Fe alloy at E0 = 25 keV
348 Chapter 21 · Trace Analysis by SEM/EDS
180
160 Cu, 10 µm from K412
140 E0 = 25 keV Fe
120 Fe std
100
Cu
80
60 kFe = 0.0028
40
20
0
0 1 2 3 4 5 6 7 8 9 10
Photon energy (keV)
0.005
0.004
0.003
kFe
0.002
0.001
20 µm 0
0 20 40 60
Beam distance from interface ( m)
.. Fig. 21.6 DTSA-II Monte Carlo calculation of fluorescence across a iron, giving kFe = 0.0028. The inset map of the distribution of secondary
planar boundary between copper and SRM 470 (K-412 glass). The FeKα X-ray production shows the extent of penetration of
beam is placed in the copper at various distances from the interface. characteristic Cu Kα and Cu Kβ and continuum X-rays into the K-412
The spectrum calculated for a beam at 10 μm from the interface shows glass to fluoresce FeKα. Simulations at other distances give the
a small Fe peak, which is ratioed to the intensity calculated for pure response plotted in the graph
NIST DTSA II Simulation: Cubic Particle shows no Ti characteristic X-rays. When secondary fluores-
Embedded in a Bulk Matrix cence is included in the simulation, a small Ti peak is
. Figure 21.7(a) shows the results of a simulation of a 1-μm observed, corresponding to an artifact trace level k = 0.0007
cube of K-411 glass embedded in a titanium matrix and (700 ppm), demonstrating the long range of the primary
excited with a beam energy of 20 keV. For this size and beam X-rays and the creation of a trace level artifact.
energy, the primary electron trajectories penetrate through When variable pressure SEM operation is considered, the
the sides and bottom of the cube leading to direct electron large fraction of gas-scattered electrons creates X-rays from
excitation of the titanium matrix, which is seen as a major regions up to many millimeters from the beam impact point.
peak in the calculated spectrum. When the cube dimension Depending on the surroundings, this gas scattering can
21 is increased to 20 μm, the beam trajectories at E0 = 20 keV are greatly modify the EDS spectrum from what would originate
contained entirely within the K-411 cube. DTSA-II allows from the region actually excited by the focused beam.
calculations with and without implementing the secondary . Figure 21.7(c) shows this effect as simulated with DTSA II,
fluorescence calculation. When secondary fluorescence is resulting in a large peak for Ti, which is not present in the
not implemented, the calculated spectrum . Fig. 21.7(b) specimen but which is located in the surrounding region.
21.3 · M
easurements of Trace Constituents by Electron-Excited Energy Dispersive X-ray Spectrometry
349 21
2 mm SiK TiKa
1.21E-3 Emission 8.95E-3 Emission
Si K-M3 Ti K-L3
Si Ti
15 000 Fe
O
10 000 Ca
Counts
Mg
Ti Fe
Ca Fe
5 000
0
0 1 2 3 4 5 6 7
Photon energy (keV)
10 000 Fe
Ca
8 000
6 000 Ti Fe
Ca Ti Fe
4 000
2 000
0
0 1 2 3 4 5 6 7 8 9 10
Photon energy (keV)
100
10
1
0 1 2 3 4 5 6 7 8 9 10
Photon energy (keV)
.. Fig. 21.7 a DTSA-II Monte Carlo calculation of a 1-μm cubical Ti with and without calculation of secondary fluorescence. c 20-μm
particle of K411 glass embedded in a Ti matrix with a beam energy of cubical particle of K411 glass embedded in a Ti with calculation of
20 keV, including maps of the distribution of SiK (particle) and TiKα (sur- secondary fluorescence and with calculation of gas scattering in VPSEM
rounding matrix). b 20-μm cubical particle of K411 glass embedded in a operation—water vapor; 133 Pa (1 Torr); 10-mm gas path length
350 Chapter 21 · Trace Analysis by SEM/EDS
Si Ti
c
20 000 Noisy[MC simulatoin of a 20.000 mcubic inclusion of K411 in Ti] #1
Noisy[MC simulatoin of a 20.000 m cubic inclusion of K411 in Ti] #1
18 000
20 µm cubic inclusion of K411 in Ti E0 = 20 keV:
16 000 O High vacuum mode (with fluorescence)
Variable pressure mode (water vapor;
14 000
10 mm GPL; 133 Pa = 1 torr)
12 000 Fe Mg
Counts
10 000 Ca
8 000
Ti Fe
6 000 Ca
Fe
4 000
2 000
0
0 1 2 3 4 5 6 7 8 9 10
Photon energy (keV)
Noisy[MC simulatoin of a 20.000 m cubic inclusion of K411 in Ti] #1
10 000 Noisy[MC simulatoin of a 20.000 m cubic inclusion of K411 in Ti] #1
1 000
Counts
100
10
1
0 1 2 3 4 5 6 7 8 9 10
Photon energy (keV)
window
BSE
BSE
Collimator &
Chamber electron trap
wall
Specimen X-rays
entrance window or other hardware. However, the reality of generated in the specimen plane within a circular area with
the EDS measurement is likely to be quite different from a diameter of several millimeters, a feature that is important
this ideal case, at least at the trace constituent level, as a for X-ray mapping applications, where the beam is scanned
consequence of electron backscattering, shown schemati- over large lateral areas and X-rays must be accepted from
cally in . Fig. 21.9. For targets of intermediate and high any beam position within the scanned area. Moreover, the
atomic number, a significant fraction of the incident beam acceptance region is three dimensional with a vertical
is emitted as backscattered electrons, and the majority of dimension of several millimeters along the beam axis. To
these BSEs retain more than half of the incident beam determine the true acceptance volume of the EDS collima-
energy. After leaving the specimen, these BSEs are likely to tor, low magnification (maximum scanning area) X-ray
strike the objective lens and the walls of the specimen mapping of a target such as a blank aluminum sample stub
chamber as well as other hardware, where they generate the provides a direct view of the transmission of the EDS colli-
characteristic (and continuum) X-rays of those materials. mator as a function of x-y position and as a function of the
The EDS detector collects X-rays from any source with a z-position, as shown in . Fig. 21.11. For this example, any
line-of-sight to the detector, so to minimize remote BSE- X-ray generated in a large volume (at least 2.5 × 3 × 10 mm)
induced contributions to the measured spectrum, the EDS can potentially be collected by this EDS system despite the
is equipped with a collimator whose function is to restrict otherwise effective collimator. Three important sources of
the view of the EDS, as illustrated schematically in uncontrolled remote excitation within this collimator
. Fig. 21.9. The solid angle of acceptance of the EDS is sub- acceptance volume are shown in . Fig. 21.12: (1) beam elec-
stantially reduced by the collimator, minimizing remote trons scattering off the edge of the final aperture (magenta
contributions from the lens and chamber walls. While the trajectory); (2) beam electrons being stopped by the final
collimator provides a critical improvement to the measured aperture and generating the characteristic and continuum
spectrum, it is important for the analyst to understand its X-rays of the aperture material (e.g., Pt; blue dashed trajec-
inevitable limitations. The actual acceptance solid angle tory); and (3) re-scattering of BSEs that have struck the final
must be constructed by looking out from the detector lens and return to the specimen (red trajectory). Both of
through the aperture of the collimator, as shown in these sources can create X-rays several millimeters or more
. Fig. 21.10. The typical collimator accepts X-rays from the beam impact location.
352 Chapter 21 · Trace Analysis by SEM/EDS
window
BSE
BSE
Collimator &
Chamber electron trap
wall
Specimen X-rays
Green =
Extent of
Several mm specimen
X-ray sources
NOT excluded
by collimator
16 mm 18 mm 20 mm
21
1 mm
Al machined surface
0 10 20 30 40 50 60 70 80 90 100
21.4 · Pathological Electron Scattering Can Produce “Trace” Contributions to EDS Spectra
353 21
.. Fig. 21.12 Possible sources
Aperture Conventional SEM,
of remote excitation: beam elec- Final lens scattering pathological scattering
trons scattering off edge of final
aperture, beams stopped by
aperture generating characteris- Final aperture
tic and continuum X-rays, and
re-scattering of backscattered Remote X-rays EDS
electrons from the lens detector
window
BSE
BSE
Collimator &
Chamber
electron trap
wall
Specimen X-rays
Green =
Several mm Extent of
specimen
X-ray sources
NOT excluded
by collimator
EDS
detector
window
Collimator &
Chamber electron trap
wall
Green =
Extent of
specimen
X-ray sources
NOT excluded
“In-hole Spectrum” by collimator
21
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
.. Fig. 21.14 Sequence of measurements for determining pathological at the center of the 200 μm diameter hole of the platinum aperture. A
scattering effects by the “in-hole” method. For the same dose, EDS spectra logarithmic display reveals the very low counts in the “in-hole” spectrum
are measured on the brass block, aluminum stub, platinum aperture, and compared to the spectra measured on the various materials
21.4 · Pathological Electron Scattering Can Produce “Trace” Contributions to EDS Spectra
355 21
.. Fig. 21.15 Schematic
diagram of the “in-hole” Aperture Conventional SEM,
configuration with a pure scattering pathological scattering
element target placed at the Final
center of a multi-material target lens
EDS
detector
BSE
window
Green =
Extent of
Spectrum from high purity flat
specimen
scattering target, e.g., C, Ta, X-ray sources
surrounded by different materials, NOT excluded
e.g., Ag-epoxy, Ti, Al by collimator
Carbon
E0 = 20 keV
12% deadtime, 112 kHz
500s (12,500 nA-s)
0.1 – 20 keV integral
= 56 million counts
Counts
Ag
Ti
0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
356 Chapter 21 · Trace Analysis by SEM/EDS
Ag = 0.000003
2TaMa,b
Ti = 0.000005 Ta
Counts
E0 = 20 keV
Ag Ti
8% Deadtime
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20
Photon energy (keV)
window
Green =
Spectrum from strong
Extent of
topographic scatterer, e.g., SrF2,
specimen
surrounded by different
X-ray sources
materials, e.g., Pt, Ti
NOT excluded by
collimator
21 Multi-material “topographic scatter” test specimen
References
357 21
.. Fig. 21.19 Example with a
C
pyramid of high purity SrF2 as the
scattering target surrounded by a
SrL+SrL
SrL α,β
carbon tab (1 cm diameter) on a
2.5 cm diameter brass (Cu and Zn) SrF2
CK
disk. The Ni signal that is
observed likely arises from the SrF2_EDGE_20kV10n
SrKα
SrLg
AMED96kHz90T100s2k
Ni-coatings of the specimen
stage components SrF2
Remote scattering peaks
E0 = 20 keV
NiKα
Counts
CuKα
ZnKα
SrKβ
0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Photon energy (keV)
High count EDS spectra can be used to achieve limits of Currie LA (1968) Limits for qualitative detection and quantitative deter-
mination. Anal Chem 40:586–593
detection approaching a mass concentration of CDL = 0.0001
Newbury DE, Ritchie NWM (2016) Measurement of trace constituents by
(100 ppm) when there are no peak interferences from higher electron-excited X-ray microanalysis with energy dispersive spec-
concentration constituents, and CDL = 0.0005 (500 ppm) trometry. Micros Microanal 22(3):520–535
when significant peak interference does occur. However, the Williams DB, Goldstein JI (1981) Artifacts encountered in energy disper-
analyst must carefully test the SEM/EDS measurement envi- sive X-ray spectrometry in the analytical electron microscope. In:
Heinrich KFJ, Newbury DE, Myklebust RL, Fiori CE (eds) Energy dis-
ronment to ensure that the trace measurement is meaningful
persive X-ray spectrometry, National Bureau of Standards Special
and not a consequence of pathological remote scattering Publication 604 (U.S. Department of Commerce, Washington, DC),
effects. pp 341–349
359 22
References – 380
1e-3
1e-4
Si K X-rays/e–
1e-5
1e-6
1e-7
0 2 4 6 8 10 12
22 Overvoltage
22.1 · What Constitutes “Low” Beam Energy X-Ray Microanalysis?
361 22
Same scale
Si
Si_5keV
Si_4keV
Si_3keV
100000 Si_2.8keV
Si_2.6keV
Si_2.4keV
Si_2.2keV
O
Si_2.0keV
10000 Si_1.9keV
Si
C
Counts
1000
100
Si
10
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
Same scale
Si
Si_5keV
Si Si_4keV
40000 Si_3keV
5 keV Si_2.8keV
Si_2.6keV
4 keV Si_2.4keV
Si_2.2keV
30000 3 keV Si_2.0keV
Si_1.9keV
Counts
2.8 keV
2.6 keV
20000
2.4 keV
2.2 keV
O
1.9 keV
Si
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
.. Fig. 22.2 Silicon at various incident beam energies from 5 keV to 1.9 keV showing the decrease in the peak-to-background with decreasing overvoltage
.. Fig. 22.3 Periodic table Elemental measurement strategy for conventional beam energy analysis
illustrating X-ray shell choices
for developing analysis strategy Principal shell used for identification
within the conventional beam E0 = 20 keV
energy range, E0 = 20 keV K-shell K-L
U0 > 1.25 (Ec < 16 keV)
(Newbury and Ritchie, 2016)
EDS resolution: 129 eV (FWHM, MnKα)
L-shell L-M
Li Be
Marginally detectable B C N O F Ne
Na Mg Al Si P S Cl Ar
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
362 Chapter 22 · Low Beam Energy X-Ray Microanalysis
.. Fig. 22.4 Periodic table Elemental measurement strategy for conventional beam energy analysis
illustrating X-ray shell choices for
developing analysis strategy at Principal shell used for identification
E0 = 10 keV
the lower end of the conventional
K-shell K-L
beam energy range, E0 = 10 keV U0 > 1.25 (Ec < 8 keV)
(Newbury and Ritchie, 2016)
EDS resolution: 129 eV (FWHM, MnKα)
L-shell L-M
Li Be Marginally detectable B C N O F Ne
Na Mg Al Si P S Cl Ar
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
.. Fig. 22.5 Periodic table Elemental measurement strategy for low beam energy analysis
illustrating X-ray shell choices for
developing analysis strategy for Principal shell used for identification
E0 = 5 keV
the upper end of the low beam K-L
U0 > 1.25 (Ec < 4.0 keV) K-shell
energy range, E0 = 5 keV (Newbury
and Ritchie, 2016)
EDS resolution: 129 eV (FWHM, MnKα)
L-shell L-M
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
22
22.1 · What Constitutes “Low” Beam Energy X-Ray Microanalysis?
363 22
.. Fig. 22.6 Periodic table Elemental measurement strategy for low beam energy analysis
illustrating X-ray shell choices
for developing analysis strategy Principal shell used for identification
E0 = 2.5 keV
within the low beam energy
K-shell K-L
range, E0 = 2.5 keV (Newbury and U0 > 1.25 (Ec < 2.0 keV)
Ritchie, 2016)
EDS resolution: 129 eV (FWHM, MnKα)
L-shell L-M
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
.. Fig. 22.7 Periodic table Elemental measurement strategy for low beam energy analysis
illustrating X-ray shell choices
for developing analysis strategy Principal shell used for identification
E0 = 2.0 keV
within the low beam energy
K-shell K-L
range, E0 = 2.0 keV (Newbury and U0 > 1.25 (Ec < 1.6 keV)
Ritchie, 2016)
EDS resolution: 129 eV (FWHM, MnKα)
L-shell L-M
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
364 Chapter 22 · Low Beam Energy X-Ray Microanalysis
.. Fig. 22.8 Periodic table Elemental measurement strategy for low beam energy analysis
illustrating X-ray shell choices
for developing analysis strategy Principal shell used for identification
E0 = 1.0 keV
within the low beam energy
K-shell K-L
range, E0 = 1.0 keV (Newbury and U0 > 1.25 (Ec < 0.8 keV)
Ritchie, 2016)
EDS resolution: 129 eV (FWHM, MnKα)
L-shell L-M
K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr
Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn Sb Te I Xe
Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi Po At Rn
Fr Ra Ac
Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr
0.3
E0 = 2.5 keV
0.2
.. Fig. 22.9 Range of production of Na K-shell X-rays in various matrices, as calculated with the Kanaya–Okayama range equation
366 Chapter 22 · Low Beam Energy X-Ray Microanalysis
1.0
Absorption correction, A
0.9
0.8
0.7 O K-L2,3
Cu L3-M4,5
0.6
0.5
0 2 4 6 8 10 12 14 16 18 20
Beam energy (keV)
Au, while at E0 = 2.5 keV, the range for Na collapses to 0.12 μm typically the case for low energy photons, the matrix correc-
(120 nm) in a C matrix to 0.02 μm (20 nm) in Au. tion for absorption diminishes rapidly (i.e., approaches
unity) as the beam energy is reduced, as shown in . Fig. 22.10
for O K-L2,3 and Cu L3-M4,5 in Cu2O (measured relative to
22.2.2 educed Matrix Absorption
R MgO and Cu).
Correction
When the range of X-ray production is reduced by lowering 22.2.3 ccurate Analysis of Low Atomic
A
the beam energy, the generated X-rays undergo lower absorp- Number Elements at Low Beam
tion because of the reduced path length to the surface. This Energy
can be a strong effect, because X-ray absorption follows an
exponential relationship: Low atomic number elements with Z ≤ 10 have characteristic
X-ray energies below 1 keV, and these low energy photons
I / I 0 = exp − ( µ / ρ ) ρ s (22.6) suffer especially high absorption. By minimizing absorption
through operation at low beam energy, Li and Be can be
where I0 is the original intensity and I is the intensity detected; and B, C, N, and F, can be quantitatively analyzed
remaining after passing through a distance s (cm) of a mate- with accuracy such that the analyzed value is generally
rial of density ρ (g/cm3) and of mass absorption coefficient, within ± 5 % relative to the true value, as presented in
μ/ρ (cm2/g). For strongly absorbed photons, which is . Table 22.1 (borides), . Table 22.2 (carbides), . Table 22.3
(nitrides), and . Table 22.4 (oxides) (Newbury and Ritchie stoichiometric compounds (MgO, GaN) as peak-fitting ref-
2015). Examples of EDS spectra and the residual spectrum erences and standards. Note that with the exception of the Si
after fitting are shown in . Fig. 22.11 (Cr borides), . Fig. 22.12 K-family, the L-shell and M-shell characteristic X-rays of the
(Cr3C2), . Fig. 22.13 (Fe3N), and . Fig. 22.14 (Cu oxides). metallic elements were used as the analytical peaks because
These examples of analyses for low atomic number elements the low beam energy was not adequate to ionize the K-shells
in compounds with NIST DTSA II used pure elements and of these elements (Ti, Cr, Fe, Ni, Cu, Zr).
368 Chapter 22 · Low Beam Energy X-Ray Microanalysis
Cr
CrB2_5kV20nA10%DT
50000 Cr Cr2B_5kV20nA10%DT
45000 B E0 = 5 keV CrB_5kV20nA10%DT
40000 CrB2
35000 CrB
30000 Cr2B
Counts
25000
20000
15000
10000
5000
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
Photon Energy (keV)
CrB2_5kV20nA10%DT
35000 Residual_CrB2_5kV20nA10%DT
30000 E0 = 5 keV
CrB2
25000 Fitting residual
Counts
20000
15000
10000
5000
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
Photon energy (keV)
.. Fig. 22.11 EDS spectra of chromium borides: CrB2, CrB and Cr2B (upper) and residual after peak fitting for B and Cr in CrB2 (lower); E0 = 5 keV
Simulation of Cr3C2_1_GLR_5kV25nASLOW38kH12DT_60s
30000 Residual_Cr3C2_5kV25nA12%DT
25000
E0 = 5 keV
Cr3C2
20000
Counts
Fitting residual
15000
10000
5000
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Photon energy (keV)
.. Fig. 22.12 EDS spectrum of chromium carbide, Cr3C2 and residual after peak fitting for C and Cr; E0 = 5 keV
22
22.3 · Challenges and Limitations of Low Beam Energy X-Ray Microanalysis
369 22
Fe3N_5kV25nA13%DT
80000 Residual_Fe3N_5kV25nA13%DT
E0 = 5 keV
Fe3N
60000 Fitting residual
Counts
40000
20000
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Photon energy (keV)
.. Fig. 22.13 EDS spectrum of iron nitride, Fe3N and residual after peak fitting for N and Fe; E0 = 5 keV
Cu2O_5kV25nA5%DT
CuO_5kV25nA5%DT
250000
E0 = 5 keV
Cu2O
200000 CuO
Counts
150000
100000
50000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
.. Fig. 22.14 EDS spectra of copper oxides, Cu2O and CuO; E0 = 5 keV
low beam energy X-ray microanalysis is much more con- spectrum of BaCl2 with E0 = 5 keV is shown in . Fig. 22.16.
strained. As discussed above, as the beam energy is reduced, Due to the low fluorescence yield of ionizations in the Ba
the atomic shells that can be ionized become more M-shell, the Ba M-family peaks are seen to have a relatively
restricted. A beam energy of 5 keV is the lowest energy that low peak-to-background, despite Ba being present in this
provides access to measureable X-rays for elements of the case as a major constituent (mass concentration C = 0.696),
periodic table from Z = 3 (Li) to Z = 94 (Pu), as shown in making the measurement of Ba when present as a minor to
. Fig. 22.5. If the beam energy is reduced to E0 = 2.5 keV, trace constituent even more problematic. A practical prob-
EDS X-ray microanalysis of large portions of the periodic lem that arises when analyzing with the Ba M-family peaks is
table is no longer possible because no atomic shell with use- the difficulty in obtaining suitable Ba M-family peak refer-
ful X-ray yield can be excited or effectively measured for ences that are free of interferences from other elements.
these elements, creating the situation shown in . Fig. 22.6. While BaCl2 is interference-free in the Ba M-family region,
Further decreases in the beam energy results in losing BaF2 and BaCO3 are not, as shown in . Fig. 22.16. However,
access to even more elements, with only about half of the BaCl2 shows evidence of degradation under the electron
elements measureable at E0 = 1 keV, and many of those only beam, possibly changing the local compositions and thus dis-
marginally so. qualifying it as a standard. Despite degradation under the
Even to achieve the elemental coverage depicted for beam, BaCl2 can serve as a peak reference, while BaF2 or
E0 = 5 keV in . Fig. 22.5, low beam energy EDS X-ray micro- another Ba-containing compound or glass that is stable
analysis requires measurement of characteristic X-rays that under electron bombardment can serve as a standard.
are not normally utilized in conventional beam energy analy- Despite these challenges, successful analysis of the high tran-
sis for certain elements. Thus Ti must be measured with the sition temperature superconducting material YBa2Cu3O7-X at
Ti L-family when E0 ≤ 5 keV, as shown in . Fig. 22.15. E0 = 2.5 keV with CuO, Y2O3, and BaF2 as the standard and
Similarly, for Ba, the Ba L-family around 4.5 keV is the usual BaCl2 as the peak reference is demonstrated in . Fig. 22.17
choice for microanalysis, but the Ba L3 excitation energy is and . Table 22.5, where analyses with oxygen done directly
5.25 keV, and thus the Ba L-family not excited with E0 = 5 keV, against a standard (ZnO) and by the method of assumed oxy-
forcing the analyst to utilize the Ba M-family. The EDS gen stoichiometry of the cations are presented.
Ti_5kV30nA3%DT
50000
Ti
40000 E0 = 5 keV
Counts
30000
20000
10000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
22
22.3 · Challenges and Limitations of Low Beam Energy X-Ray Microanalysis
371 22
35000 BaCl2_5kV25nA4%DT
BaCl2
30000 E0 = 5 keV
25000
Counts
20000
15000
10000
5000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
140000 BaCl2_5kV25nA4%DT
BaCO3_5kV30nA4%DT
BaF2_5kV30nA4%DT
120000
E0 = 5 keV
100000 BaCl2
BaF2
Counts
80000 BaCO3
60000
40000
20000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Photon energy (keV)
.. Fig. 22.16 EDS spectrum of barium chloride, showing the Ba M-family (upper); EDS spectra of BaCl2, BaF2, and BaCO3 (lower); E0 = 5 keV
123TC_2.5kV50nA3%DT
Residual_123TC_2.5kV50nA3%DT
60000
E0 =2.5 keV
50000 YBa2Cu3O7-X
Fitting residual
40000
Counts
30000
20000
10000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
.. Fig. 22.17 EDS spectrum of YBa2Cu3O7-X, and residual after peak fitting for O K-L2, the Ba M-family and Cu L-family; E0 = 2.5 keV
372 Chapter 22 · Low Beam Energy X-Ray Microanalysis
Element Cav mass conc RDEV % σrel, % Cav mass conc RDEV % σrel, %
ZnS_5kV50nAMED5eV51kHz10DT_100s
400000
350000
ZnS
300000 E0 = 5 keV
250000
Counts
200000
150000
100000
50000
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Photon energy (eV)
.. Fig. 22.18 EDS spectrum of ZnS illustrating concept of the energy axis of the spectrum and the corresponding depth of X-ray generation;
E0 = 5 keV
22.3 · Challenges and Limitations of Low Beam Energy X-Ray Microanalysis
373 22
22.3.3 t Low Beam Energy, Almost
A 4 eV. Note the large increase in intensity in this region as the
Everything Is Found To Be Layered beam energy is lowered from 10 keV to 2.5 keV due to the
increased contribution from O K-L2,3 as the fraction of the
Most “pure” elements have surface layers such as native oxide, interaction volume represented by the surface oxide increases.
hydration layers, and others that compromise the require- Obtaining an adequate standard and peak reference for Ti for
ment for uniform composition throughout the electron- low beam energy analysis is thus problematic. Even when a
excited volume of both the unknown and the standard(s). For compound expected to be oxygen-free such as TiSi2 is selected,
example, when “pure” silicon is used as a standard, the inten- there still appears to be excess intensity due to O K-L2,3, as
sity of the O K-L2,3 peak, which arises from the SiO2 layer on shown in . Fig. 22.20. Thus, it may be necessary to use advanced
Si, increases relative to the Si K-L2,3 peak as the beam energy preparation, such as in situ ion milling to clean the surface of Ti
is lowered, as seen in . Fig. 22.19. In conventional analysis to reduce the oxygen contribution to the spectrum.
with E0 ≥ 10 keV, the deviation from “pure” silicon that this The conductive coating that is applied to eliminate surface
surface oxide represents does not constitute a significant charging in insulating specimens becomes more significant
source of error since the range is so much greater than the as the beam energy is decreased. This effect is illustrated in
native oxide thickness. However, for low beam energy analy- . Fig. 22.21 for spectra of the mineral benitoite (BaTiSi3O9)
sis, the surface oxide constitutes an increasingly significant recorded over a wide range of incident beam energies, where
fraction of the beam excitation volume as the beam energy is the peak for C K-L 2,3 is barely detectable at E0 = 20 keV but
reduced, introducing an increasingly larger error because of becomes one of the most prominent peaks in the spectrum at
the uncertainty in the standard composition. E 0 = 2.5 keV. The analyst should try to minimize the carbon
The presence of the O K-L2,3 peak from a surface oxide is contribution to the spectrum by using the thinnest acceptable
especially problematic when it interferes with the characteristic carbon layer, less than 10 nm thick, and it may be necessary
peak of interest, such as the Ti L-family, as shown in . Fig. 22.20. to explore the use of ultrathin (~1 nm) heavy metal coatings
O K-L2,3 (0.525 keV) is separated from Ti L1-M2 (0.529 keV) by as an alternative if it is desired to analyze for carbon.
Si_3keV
Si_2.8keV
Si_2.6keV
Si_2.4keV
Si_2.2keV
Si_2.0keV
Si_1.9keV
Counts
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
Si_3keV
Si Si_2.8keV
Si_2.6keV
1.9 keV Si_2.4keV
Si_2.2keV
2.0 keV Si_2.0keV
2.2 keV Si_1.9keV
2.4 keV
Counts
2.6 keV
2.8 keV
3.0 keV
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
.. Fig. 22.19 EDS spectra of Si over a range of beam energies, showing increase in the O K-L2 peak relative to Si K-L2; all spectra scaled to Si K-L2
374 Chapter 22 · Low Beam Energy X-Ray Microanalysis
Ti_2.5kV50nA%DT
Ti_5kV30nA3%DT
Ti_10kV20nA6%DT
Ti
E0 = 2.5 keV
E0 = 5 keV
E0 = 10 keV
Counts
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Photon energy (keV)
Ti_2.5kV50nA2%DT
TiSi2_2.5kV50nA1%DT
E0 = 2.5 keV
Ti
TiSi2
Counts
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
.. Fig. 22.20 (Upper) EDS spectra of Ti at various beam energies showing increase in the O K-L2 peak relative to Ti L-family peaks; (lower) EDS spec-
tra of Ti and TiSi2 at E0 = 2.5 keV
Finding an unexpected composition due to surface modi- obtained, showing no further undiscovered constituents. The
fication is a common experience when performing low beam analysis results presented in . Table 22.6 reveal significant
energy analysis of materials that must be analyzed in the as- concentrations of C and O in the TiB2. Note the greater vari-
received condition. Without special surface preparation to ance in the C and O contaminants compared to the B and Ti
expose the interior of the material, such as grinding and pol- host elements.
ishing or ion beam milling, the modified surface region dom-
inates the analysis. . Figure 22.22 (upper spectrum) shows an Analysis of Surface Contamination
example of TiB2, where inspection of the fitting residual after Low beam energy analysis samples such a shallow near-
analyzing for B and Ti shows significant peaks for C and surface region that unexpected contamination layers can
O. When these elements are included in the analysis, the fit- dominate an analysis. This can lead to the confounding situa-
ting residual shown in . Fig. 22.22 (lower spectrum) is tion where the analysis can be correct, but what is being
22
22.3 · Challenges and Limitations of Low Beam Energy X-Ray Microanalysis
375 22
Benitoite_10kV20nA7%DT
Benitoite_5kV30nA4%DT
Benitoite Benitoite_2.5kV50nA2%DT
Benitoite_20kV10nA7%DT
BaTiSi3O9
2.5 keV
5 keV
Counts
10 keV
20 keV
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
Benitoite_10kV20nA7%DT
Benitoite_5kV30nA4%DT
Benitoite_2.5kV50nA2%DT
Benitoite_20kV10nA7%DT
Counts
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Photon energy (keV)
.. Fig. 22.21 EDS Spectra of benitoite (BaTiSi3O9) over arrange of beam energies showing relative increase in the C K-L2 peak as the beam energy
decreases
measured is unanticipated. An example is shown in The results of the quantitative analysis after this first repolish-
. Fig. 22.23, which shows a low beam energy SDD-EDS spec- ing, which are presented in . Table 22.7b, show analytical
trum of NIST SRM 481 (alloy 20Au-80Ag) where the surface totals near unity but large RDEV values for Ag and Au, espe-
was prepared metallographically more than 30 years earlier. cially for the 20Au-80Ag alloy. This large deviation from the
The spectrum shows distinct peaks due to S and Cl from the SRM values is likely to be a consequence of the tarnish forma-
formation of a surface tarnish layer. Quantitative X-ray tion process selectively removing Ag from the alloy. After two
microanalysis with DTSA-II confirms the high concentra- additional repolishing steps with 1 μm and 0.25 μm diamond
tions of S and Cl and very large RDEV values for Ag and Au, abrasives (. Tables 22.7c and 22.7d), this perturbed surface
as shown in . Table 22.7a. The specimen mount was re-pol- layer was finally removed, exposing the SRM alloy, with the
ished with 0.25-μm diamond abrasive, which eliminated the DTSA-II analysis values closely matching the SRM certificate
S- and Cl- rich layer, as seen in the spectrum in . Fig. 22.23. values.
376 Chapter 22 · Low Beam Energy X-Ray Microanalysis
TiB2_2p5kV50nA2%DT
Residual_TiB2_2p5kV50nA2%DT
40000
E0 = 2.5 keV
30000 TiB2
Residual: B and Ti L fitting
Counts
20000
10000
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Photon energy (keV)
TiB2_2p5kV50nA2%DT
Residual_TiB2_2p5kV50nA2%DT
40000 ResidualTiB2_fit_B_C_O_Ti
E0 = 2.5 keV
30000
TiB2
Counts
10000
0
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Photon energy (keV)
.. Fig. 22.22 (Upper) EDS spectrum of TiB2 and residual spectrum after fitting for B K-L2 and Ti L-family revealing peaks for C K-L2and O K-L2; (lower)
after fitting for C K-L2and O K-L2; E0 = 2.5 keV
22
22.3 · Challenges and Limitations of Low Beam Energy X-Ray Microanalysis
377 22
20000
20Au-80Ag alloy, original surface
20Au-80Ag alloy, 1st-repolishing (0.25 µm)
E0 = 5 keV
15000
Counts
10000
5 000
0
0 1 2 3 4 5
Photon energy (keV)
15000
NIST DTSA-II 20Au80Ag_repolished
Quantitative analysis 20Au80Ag_residual
20Au80Ag, 1st-repolishing
E0 = 5 keV
10000
Counts
5000
0
1 2 3 4
Photon energy (keV)
.. Fig. 22.23 NIST SRM 481 (Au-Ag alloys). Analysis of an old (>30 years) metallographic preparation at E0 = 5 keV, and the spectrum after repolish-
ing with 0.25 μm diamond abrasive
22
378
.. Table 22.7a Analysis of SRM 481 (Au-Ag alloys); 1970s metallographic preparation, original surface; E0 = 5 keV; standards: S (FeS2); Cl (KCl); Ag, Au
Alloy Raw analytical S (norm mass σ(%)5 Cl (norm mass σ(%) 5 Au (norm mass σ(%)5 loc RDEV(%) Ag (norm mass σ(%) 5 loc RDEV (%)
total conc) loc conc) loc conc) conc)
40Au–60Ag 0.9959 0.0311 5.7 0.0965 10 0.3440 6.1 −14 0.5284 2.2 −12
Chapter 22 · Low Beam Energy X-Ray Microanalysis
60Au–40Ag 0.9951 0.0094 5 0.045 22 0.5754 4.7 −4.2 0.3706 4.6 −7.2
80Au–20Ag 1.005 0.0051 8.3 0.0365 4.5 0.7340 0.67 −8.3 0.2244 1.6 +12
.. Table 22.7b Analysis of SRM 481 (Au-Ag alloys); surface after first repolishing with 0.25-μm diamond; E0 = 5 keV; standards: Ag, Au
Alloy Raw analytical Au (norm) σ(%) 5 loc RDEV (%) DTSA-II error Ag (norm) σ(%) 5 loc RDEV (%) DTSA-II error
total budget (%) budget (%)
20Au–80Ag 0.9731 0.3050 7.1 +36 0.34 0.6950 3.1 −10.4 0.75
40Au–60Ag 0.9920 0.4460 0.96 +11.4 0.25 0.5540 0.77 −7.6 1.1
60Au–40Ag 0.9907 0.6322 1.0 +5.3 0.21 0.3678 1.8 −7.9 1.5
80Au–20Ag 0.9930 0.8264 0.24 3.2 0.17 0.1736 1.1 −13 2.1
.. Table 22.7c Analysis of SRM 481 (Au-Ag alloys); surface after second repolishing with 1- and 0.25-μm diamond; E0 = 5 keV; standards: Ag, Au
Alloy Raw analytical Au (norm) σ(%) 5 loc RDEV (%) DTSA-II error Ag (norm) σ(%) 5 loc RDEV (%) DTSA-II error
total budget (%) budget (%)
20Au–80Ag 1.005 0.2398 0.87 % +6.9 % 0.38 % 0.7602 0.28 % −2.0 % 0.64 %
40Au–60Ag 0.9983 0.4045 0.23 +1.1 0.27 0.5955 0.16 −0.64 0.98
60Au–40Ag 0.9897 0.6084 0.13 +1.3 0.21 0.3916 0.21 −1.9 1.4
80Au–20Ag 0.9998 0.8055 0.26 +0.62 0.19 0.1945 1.1 −2.5 1.9
20Au–80Ag 1.005 0.2398 0.87 +6.9 0.38 0.7602 0.28 −2.0 0.64
22.3 · Challenges and Limitations of Low Beam Energy X-Ray Microanalysis
.. Table 22.7d Analysis of SRM 481 (Au-Ag alloys); surface after third repolishing with 1- and 0.25-μm diamond; E0 = 5 keV; standards: Ag, Au
Alloy Raw analytical Au (norm) σ(%) 5 loc RDEV (%) DTSA-II error Ag (norm) σ(%) 5 loc RDEV (%) DTSA-II error
total budget (%) budget (%)
20Au–80Ag 1.017 0.2251 0.13 −0.11 0.39 0.7749 0.13 +0.35 0.61
40Au–60Ag 0.9988 0.3909 0.25 −2.3 0.28 0.6091 0.16 +1.6 0.93
60Au–40Ag 0.9931 0.5931 0.12 −1.2 0.22 0.4069 0.18 +1.9 1.4
80Au–20Ag 0.9957 0.7979 0.17 −0.32 0.18 0.2021 0.68 +1.2 1.9
22 379
380 Chapter 22 · Low Beam Energy X-Ray Microanalysis
22
381 23
Analysis of Specimens
with Special Geometry:
Irregular Bulk Objects
and Particles
23.1 The Origins of “Geometric Effects”: Bulk Specimens – 382
References – 411
Backscatter coefficient
specimen shape, orientation, or size influence the
0.6
measured X-ray spectrum.
2. The specimen is homogeneous in composition over the 0.5
volume excited by the electron beam from which the
characteristic and continuum X-rays are emitted, 0.4
including the secondary radiation induced by absorption
of the primary electron-excited radiation. 0.3
0.6
shape of the specimen (1) modify the interaction of the elec-
trons with the specimen so as to affect the generated X-ray 0.4
intensity and (2) alter the length of the absorption path along
which the generated X-rays travel to escape the specimen and 0.2
reach the detector so as to affect the measured X-ray
intensity. 0.0
Because electron backscattering depends on the local sur- 0 20 40 60 80
face inclination to the incident beam, tilted samples generate Tilt angle (degrees)
fewer X-rays compared to a sample at normal beam inci-
dence (0° tilt). An illustration of this effect for bulk copper, as
23 calculated with the Monte Carlo simulation embedded in
.. Fig. 23.2 Emitted Cu K-L2,3 X-ray intensity calculated as the k-ratio
relative to the intensity at a tilt of 0°, vs. surface tilt θ (inclination), for
NIST DTSA-II, is shown in . Fig. 23.1. Even at normal beam Cu at E0 = 20 keV as calculated with the Monte Carlo electron trajectory
incidence where the backscattered electron (BSE) coefficient simulation embedded in NIST DTSA-II
23.1 · The Origins of “Geometric Effects”: Bulk Specimens
383 23
.. Fig. 23.3 Schematic illus- Geometric Effects: surface roughness affects local
tration of the effects of surface absorption path to reach detector
topography on the X-ray absorp-
tion path length within the
specimen
I/I0 = exp[-(µ/ρ)ρs]
Depth
of surface
scratch
Normal flat Reduced
Extended
bulk target absorption Extended
absorption
absorption path due to absorption
path due to
path scratch path due to
ridge
scratch
Standard
I / I 0 = exp − ( µ / ρ ) ρ s (23.1) .. Fig. 23.4 Absorption as a function of path length for Al K-L2,3
(1.487 keV) passing through Si (Kcrit = 1.838 keV), and Si K-L2,3
where I0 is the initial intensity, I is the intensity that remains (1.740 keV) passing through Al (Kcrit = 1.559 keV)
after passing through a path length s (cm), (μ/ρ) is the mass
absorption coefficient (cm2/g) for the photon energy of interest for absorption as a function of path length for two contrasting
that depends on the absorption contributions of all elements cases: Al K-L2,3 (1.487 keV) passing through Si (Kcrit = 1.838 keV),
present, and ρ is the density (g/cm3). Considering the entire and Si K-L2,3 (1.740 keV) passing through Al (Kcrit = 1.559 keV).
energy range of the generated X-ray spectrum, which extends Because Si K-L2,3 is 0.181 keV above the critical ionization
from a practical minimum threshold of 100 eV to the incident energy for Al, Si is very strongly absorbed by Al (μ/ρ = 3282 cm2/g)
beam energy, E0 (the Duane–Hunt limit), as the X-ray photon such that there is no penetration beyond approximately 6 μm.
energy decreases, absorption generally increases. Absorption is By comparison, Al K-L2,3 is below the critical ionization energy
especially strong if the photon energy is less than 1 keV above for Si, so it much less strongly absorbed (μ/ρ = 535 cm2/g), with
the critical ionization energy for any elemental constituent in approximately 50% of Al K-L2,3 intensity still remaining after
the specimen. An example of this effect is shown in . Fig. 23.4 6-μm penetration through Si.
384 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
.. Table 23.1 Characteristic X-ray intensity measured on gold after various stages of grinding and polishing (Yakowitz and Heinrich 1968)
Surface condition AuMα (2.123 keV) AuMα (2.123 keV) AuLα (9.711 keV) AuLα (9.711 keV)
Coeff. variation,% 3 n1/2/n,% Coeff. variation,% n1/2/n,%
ED
S
k-ratio
0.6
O K-L2,3 (0.523 keV)
Fe L (0.704 keV)
0.4 FeK-L2,3 (6.400 keV)
0.2
23 0.0
0.01 0.1 1 10
Scratch depth and period (micrometers)
23.3 · Consequences of Attempting Analysis of Bulk Materials With Rough Surfaces
385 23
etching is usually required to produce contrast from grains at approximately 1% relative along the Mg and Fe concentra-
different crystallographic orientations and from composi- tion axes, while the results for one location fall significantly
tionally distinct phases by creating surface relief through dif- outside this cluster. This anomalous value was found to be
ferential chemical attack and dissolution or by staining associated with a shallow scratch that remained on the pol-
through chemical reactions. Such microscopic physical relief ished surface (location noted on the inset SEM image).
creates unwanted topography similar to mechanically pro- When this highly polished surface was degraded by direc-
duced scratches that can affect SEM/EDS analysis. tional grinding with 1-μm diamond grit, 20 analyses at ran-
Additionally, in some cases chemical etching can actually domly selected locations produce a much wider scatter in the
modify the chemical composition of the surface region, so normalized Mg and Fe concentrations, as shown in . Fig. 23.7,
that it is no longer representative of the bulk of the material. a direct consequence of the effect of surface geometry.
Creating an even more severe topographic feature by
gouging the polished surface of K411 with a diamond scribe
23.3 onsequences of Attempting Analysis
C created the crater seen in the SEM(ET+) SE + BSE image
of Bulk Materials With Rough Surfaces shown in . Fig. 23.8a. Many locations in this gouge crater
were analyzed, and the results are plotted in . Fig. 23.8b,
To illustrate the impact of surface topography on microanal- showing a very wide range of Mg-Fe results. For comparison,
ysis results, a microscopically homogenous glass (NIST SRM note that the 20 Mg-Fe results from the highly polished sur-
470 K411) containing several elements (O, Mg, Si, Ca, and face, including the outlier seen In . Fig. 23.6, are contained
Fe) that provide a wide range of range of photon energies, as within the small red box noted on the plot in . Fig. 23.8b.
listed in . Table 23.2, was analyzed with a range of surface
topography (Newbury and Ritchie 2013b). Analysis was per-
formed with NIST DTSA-II at E0 = 20 keV using elemental .. Table 23.2 NIST SRM 470 (Glass K411)
(Mg, Si, Fe) and multi-element (SRM 470 glass K412 for Ca)
standards, with oxygen calculated by assumed stoichiometry, Element Mass concentration Characteristic X-ray
followed by normalization of the raw result. When analyzed energy (keV)
in the ideal, highly polished (100-nm alumina final polish) O 0.4236 0.523
flat form, the analyzed concentrations for the Mg and Fe con-
stituents, selected because of their wide difference in photon Mg 0.0885 1.254
energies, measured at 20 randomly selected locations show Si 0.2538 1.740
the distribution of results plotted in . Fig. 23.6. The mean of
Ca 0.1106 3.690
the 20 analyses falls within +1.8 % relative for Fe and −1.0 %
relative for Mg (SRM certificate values). Of the 20 analyzed Fe 0.1121 6.400
locations, 19 fall within a symmetric cluster that spans
Outlier possibly a
surface finish artifact 1% relative
11.3
8.6 8.65 8.7 8.75 8.8 8.85 8.9
Mg (normalized weight percent)
386 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
11.45
11.4
11.35
Fe (weight percent)
Mean compared to
SRM values
11.3 Fe 11.41 (11.21%) +1.8% rel
Mg 8.76 (8.85 %) -1.0% rel
11.25
K411_Highly polished
11.2
K411_1-mu_Diamond
SRM
11.15 Certificate
values
11.1
11.05
11
7.8 7.9 8 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9
Mg (weight percent)
b
Analysis of K411: Bulk polished and tool gouge crater
40
35 BULK-K411
BULK-600grit
30
Fe (normalized weight percent)
1
25
Extent of ideal
20
flat specimen
measurements
15
10
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Mg (normalized weight percent)
388 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
b
Analysis of K411: Bulk polished, 600 Grit, In-hole, Chips, Shards
80
70 BULK
Shards_overscan
percent)
11.4
60
Shards_fixed-beam
Fe (normalized weight percent)
11.3
8.6 8.8
50 Macroscopic chips Mg (normalized weight
percent)
Surface voids
40
Bulk-600grit
30
20
10
0
23 0 1 2 3 4 5 6 7 8 9 10 11 12
Mg (normalized weight percent)
.. Fig. 23.9 a Examples of deep narrow pits produced in K411 glass b Plot of the normalized Mg and Fe concentrations calculated for mea-
by impacts of a diamond scribe; microscopic particles (major dimen- surements at various locations on these objects combined with the
sions <50 μm); and macroscopic particles (major dimensions >500 μm). measurements previously plotted
23.4 · Useful Indicators of Geometric Factors Impact on Analysis
389 23
FeS2
Errors in normalized analysis
Element Ideal mass concentration (k-ratio protocol with CuS and Fe, DTSA-II)
S 0.5345
Fe 0.4655 Location 10 Norm conc relative error Location 11 Norm conc RDEV (%)
S 0.2727 -49% S 0.5015 -6%
Fe 0.7273 56% Fe 0.4985 7%
EDS Raw Total 0.0391 Raw Total 0.8773
.. Fig. 23.10 Fragment of pyrite (stoichiometric FeS2) analyzed at various locations; conditions: E0 = 20 keV; DTSA-II calculations with Fe and CuS
as standards, followed by normalization
measurement conditions, but a low total can also reveal the geometric shapes. Note that for this data set, the raw analytical
presence on an unexpected elemental constituent that is not total varies from 0.03 to 1.30 mass fraction (3–130 weight wt
in the list of elements analyzed. For example, an oxidized %). For this particular composition (K411 glass), the RDEV for
inclusion in a metallic alloy will typically have an oxygen Mg and Fe is within a range of 10% relative when the analytical
mass fraction of approximately 0.3, leading to a sharp total is in the range 0.8–1.2 mass fraction (80–120 wt %).
decrease in the analytical total to ~ 0.7 (70 %) if oxygen is not Different compositions are likely to have different sensitivities
recognized and included in a standards-based analysis of that to deviations in accuracy, but the general experience is that
location compared to the surrounding metallic region. Thus, when the raw analytical total ranges from 0.9 to 1.1 mass frac-
even for conventional analysis of ideal specimens, the raw tion (90–110 wt %), the impact of the geometric factors on the
analytical total conveys useful information and should always analysis will be minimized.
be inspected.
When a specimen deviates from the ideal flat condition and
geometric factors affect the analysis, the raw analytical total 23.4.2 The Shape of the EDS Spectrum
gives a direct indication, providing a standards-based–matrix
correction factor protocol is being used. Note that “standard- A second powerful indicator that can alert the careful analyst to
less analysis” does not provide this critical information on the the possible impact of geometric factors on an analysis is the
raw analytical total if this protocol reqwuires a forced normal- shape of the EDS spectrum. The shape of the X-ray continuum
ization to unity mass fraction (100 wt %) since the electron (bremsstrahlung) background from an ideal flat specimen has
dose information is not considered. “Standardless analysis” distinctive properties. Consider the spectrum of pure boron,
schemes that use a locally measured elemental spectrum to selected because of the absence of significant characteristic
establish the dose relationship to the vendor’s standard inten- peaks above the energy of boron (0.185 keV), as shown in
sity database or which use the peak-to-background method . Fig. 23.12. A small peak of oxygen that arises from the inevi-
(see below) can provide a meaningful analytical total. table surface oxide on the boron can be seen in this spectrum, as
. Figures 23.11a, b shows the calculated normalized concentra- well as the artifact silicon peak from the absorption and internal
tions as a function of the raw analytical total for Mg and Fe in fluorescence of the silicon window support grid and the silicon
K411 from the suite of spectra obtained from the various dead layer of the detector. Otherwise, the spectrum consists
390 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
30
20
10
0
0 20 40 60 80 100 120 140
Raw analytical total (weight percent)
10
Mg (normalized weight percent)
Bulk K411
4 All Geometries
1s
2
0
0 20 40 60 80 100 120 140
Raw analytical total (weight percent)
only of the X-ray bremsstrahlung that occurs at all photon ener- absorption coefficient increases as the photon energy
gies, Eν, up to the incident beam energy, E0 (Duane–Hunt limit). decreases. Absorption is an exponential effect, so that even-
Inspecting this spectrum starting at high photon energy, as the tually the increased absorption overwhelms the increase in
photon energy Eν decreases, the intensity of the continuum, Icm, the continuum intensity so that the intensity reaches a max-
increases, a consequence of the physics of the generation of the imum. For boron with E0 = 20 keV this maximum occurs at
bremsstrahlung, which has the following form: approximately 1.3 keV. Because of the effect on the electron
23 . range and the subsequent X-ray absorption, the exact loca-
I cm ≈ iB Z ( E0 / Ev − 1) (23.2) tion of the maximum in bremsstrahlung intensity depends
on beam energy as well as the specific element(s) acting as
where iB is the beam current, and Ż is the average atomic the absorber. Moreover, for complex compositions the
number of the target. For a given material, the mass numerous characteristic peaks are superimposed on the
23.5 · Best Practices for Analysis of Rough Bulk Samples
391 23
B_20kV10nA
100000
10000
Counts
1000
100
10
0 2 4 6 8 10 12 14 16 18 20
Photon energy (keV)
B_20kV10nA
10000
8000
6000
Counts
4000
2000
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Photon energy (keV)
background, making it more difficult to discern the position and polishing would modify or destroy the material that is
of the bremsstrahlung maximum. Nevertheless, the general actually necessary for the final result. How should the analyst
shape of the EDS bremsstrahlung continuum is a powerful proceed in such a case?
indicator of geometric effects that modify X-ray produc- The analysis of rough surfaces is inevitably going to be
tion, both bremsstrahlung and characteristic X-rays. compromised compared to analysis of the ideal flat polished
. Figure 23.13 shows this effect for ideal flat polished K411 specimen. The analyst must seek to obtain the best possible
glass and several shards. The deviation in spectral shape result under the circumstances, so the analytical strategy
from the ideal case is readily apparent, and analysis of spec- must be carefully considered. Electrons that backscatter off
tra with such severe deviations in shape results in large rough surfaces are likely to produce remote excitation of
deviations in the relative accuracy of the normalized quan- X-rays from material(s) that are likely to differ from the
titative results. location where the beam is striking, as shown schematically
in . Fig. 23.14. It may be thought that the collimator on the
EDS will restrict the view of the EDS to just the region
23.5 est Practices for Analysis of Rough
B directly excited by the incident beam. This is not the case.
Bulk Samples The collimator typically permits acceptance of X-rays with
at least 50 % efficiency from an area that is 5 mm in diameter
The optimum approach to the analysis of a rough specimen is or larger. The exact transmission response depends on the
obviously to prepare a polished flat surface, but the analyst particular EDS detector and its collimator, but the region of
may be confronted with a situation where no physical modi- transmission can be readily determined by mapping a uni-
fication of the as-received specimen is permitted. That is, the form target, for example, an aluminum SEM mounting stub,
rough surface itself is the object of interest, so that grinding at the lowest magnification setting (maximum sized scan
392 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
K411_FLAT_15kV
K411_SHARDA_15kV
K411_SHARDC_15kV
100000 K411_SHARDD_15kV
K411_SHARDE_15kV
10000
Counts
1000
100
10
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
K411 K411_FLAT_15kV
K411_SHARDA_15kV
4000 E0 = 15 keV K411_SHARDC_15kV
K411_SHARDD_15kV
Flat bulk K411_SHARDE_15kV
3000 Shard A
Shard C
Counts
Shard D
2000
Shard E
1000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 23.13 EDS spectra of K411 glass in the flat polished condition, and from four shards, showing the deviation in the spectral shape from the
ideal; E0 = 20 keV
Beam
ED
S
BSE
23
.. Fig. 23.14 SEM (ET+) SE + BSE image of an irregular surface and schematic illustration of electron backscattering from a tilted surface causing
remote X-ray excitation
23.5 · Best Practices for Analysis of Rough Bulk Samples
393 23
.. Fig. 23.15 X-ray mapping experiment to
determine extent of collimator acceptance. Large
scale low magnification map (3 × 2.5 mm) of an
aluminum stub. The graph shows the intensity
measured at the center of a series of such maps
recorded at different working distances
1 mm
0 10 20 30 40 50 60 70 80 90 100
1200
1000
Al counts/pixel
800
600
400
200
0
8 10 12 14 16 18 20 22
Working distance (mm)
field) of the SEM, as shown in . Fig. 23.15. The aluminum Optimizing the EDS spectrum measured from a rough,
stub was mapped with the specimen plane located at the irregularly shaped surface requires careful consideration of
manufacturer’s specified ideal working distance for this the selection of the location on the specimen to be measured.
SEM/EDS. The Al intensity map encoded with a pseudo- The analyst must be aware of the location of the EDS relative
color scale shows that the transmission varies from maxi- to the measured location to avoid the situation illustrated in
mum (100 %) along the top of the image to values in the . Fig. 23.16a, where the beam location leads to an X-ray path
range 60–70 % at the bottom. The top-to-bottom asymme- that must pass through the bulk of the specimen to reach the
try in this map reveals that the collimator on this EDS sys- EDS, leading to extremely high absorption. Ideally, using a
tem is actually misaligned, since with proper orientation the specimen stage with several rotation axes, a rotation about a
maximum of the collimator transmission should be at the vertical axis will bring the feature of interest to directly face
image center (coincident with the optic axis). The transmis- the EDS, thus minimizing the absorption, as shown in
sion function of the collimator as a function of vertical dis- . Fig. 23.16b. A further rotation about a horizontal axis
tance along the SEM optical axis can be determined by places the feature perpendicular to beam to minimize back-
repeating the mapping at different working distances. The scattering and remote X-ray excitation (. Fig. 23.16c). Note
graph in . Fig. 23.15 shows the intensity measured at the that although backscattering is minimized by establishing
center of each map, revealing a decrease of approximately normal beam incidence (effectively a zero tilt angle), the
40 % as the working distance was increased from 10 to backscattered electrons are broadly emitted with a cosine dis-
20 mm. This collimator thus allows high transmission from tribution so that while the majority are emitted at high angles
a large volume of space, with dimensions of at least there still remains a small but significant fraction emitted at
3 × 2.5 × 10 mm, so that any X-rays produced in this volume low angles to the surface that may strike nearby features and
with a line-of-sight to the EDS detector will contribute to excite the surrounding materials, contributing to the spec-
the measured spectrum. trum measured at the beam impact position.
394 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
ED
surface: a initial position gives
ED
S
S
high absorption due to X-ray path
through bulk of specimen; b rota-
tion about a vertical axis brings Better!
Bad!
feature to directly face EDS; c
rotation about a horizontal axis
places feature perpendicular to
beam to minimize backscattering
and remote X-ray excitation
c
ED
S
23.6 Particle Analysis tilted surface of a particle acts to modify electron backscat-
tering, which alters X-ray production, while deviations from
23.6.1 ow Do X-ray Measurements
H the ideal flat surface alter the X-ray path to the detector,
of Particles Differ From Bulk which modifies X-ray absorption compared to a flat bulk tar-
Measurements? get. The geometry of a particle leads to additional loss of
beam electrons due to penetration through the sides and
The analysis of microscopic particles whose dimensions bottom of the particle, as shown in the DTSA-II Monte Carlo
approach or are smaller than the interaction volume in a simulations in . Fig. 23.17, which depict trajectories in a
bulk target of the same composition is subject to effects simi- 1 μm-diameter aluminum particle at different beam ener-
lar to analyzing rough, bulk surfaces but with additional gies. At the highest energy simulated, E0 = 30 keV, most tra-
challenges. Like the rough bulk case, the curved or locally jectories pass through the particle with some lateral
1 µm
Al
23
.. Fig. 23.17 DTSA-II Monte Carlo simulation of electron trajectories particle are blue. Green shows trajectories that emerge from the Al par-
in a 1-μm-diameter Al sphere on a bulk C substrate at various beam ticle which change to orange when they enter the C substrate
energies; incident beam diameter = 50 nm. Trajectories inside the Al
23.6 · Particle Analysis
395 23
0.1 µm 0.2 µm
0.5 µm
1 µm
Al spheres on bulk C
E0 = 20 keV
2 µm 1 µm
.. Fig. 23.18 DTSA-II Monte Carlo simulation of electron trajectories emerge from the Al particle which change to orange when they enter
in Al spheres of various diameters at E0 = 20 keV on a bulk C substrate. the C substrate
Trajectories inside the Al particle are blue. Green shows trajectories that
scattering. As the beam energy is lowered and the electron energy, electron penetration through the particle into the
range decreases, penetration through the bottom and sides substrate occurs, and the C peak of the substrate increases
diminishes, and at E0 = 5 keV the interaction volume is com- relative to the Al-peak from the particle.
pletely contained within the 1-μm-diameter aluminum par-
ticle. The beam penetration effect also depends on the
particle size, as shown in . Fig. 23.18 for particles of various 23.6.2 ollecting Optimum Spectra
C
sizes at E0 = 20 keV, and on composition, as shown for parti- From Particles
cles with a range of atomic numbers in . Fig. 23.19. Moreover,
as opposed to the backscattered electrons in the high angle Before meaningful qualitative and quantitative analysis of
portion of the cosine distribution which are likely to leave particles can be attempted, it is important to optimize the
the vicinity of the particle without further interaction, the EDS spectrum that is collected. As illustrated in the trajec-
beam electrons that penetrate through the sides and bottom tory plots in the Monte Carlo simulations shown in
of the particle are likely to reach the supporting substrate . Figs. 23.17, 23.18, and 23.19, and the calculated spectra
where they will create X-rays of the substrate material that shown in 23.20, because of the impact of particle size and
contribute to the overall spectrum measured. This effect can shape (geometry) on electron interactions, the EDS spec-
be seen in . Fig. 23.20, which shows the Al and C peaks cal- trum of a particle will always be compromised compared to
culated by the Monte Carlo simulation for 1-μm diameter Al the spectrum of a material of identical composition in the
spherical particles. At a beam energy of E0 = 5 keV, the elec- ideal flat, bulk form. The analyst must be aware of the major
tron trajectories are contained within the particle which factors that modify particle spectra and seek to minimize
effectively acts like a bulk target. No electrons penetrate the these effects. The particle spectrum can be optimized through
sides or bottom to reach the substrate so there is no C contri- careful sample preparation and by understanding of the fac-
bution to the EDS spectrum. With increasing incident tors that affect the strategy for beam placement.
396 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
Al Ti Cu
Ag Hf Au
1 µm
.. Fig. 23.19 DTSA-II Monte Carlo simulation of electron trajectories blue. Green shows trajectories that emerge from the particle which
at E0 = 20 keV in 1 μm-diameter spheres of various elements (Al, Ti, Cu, change to orange when they enter the C substrate
Ag, Hf, Au) on a bulk C substrate. Trajectories inside the particles are
Particle Sample Preparation: Bulk Substrate must measure a blank spectrum of the tape since the polymer
Because beam electrons can penetrate through the sides and base of the tape frequently contains elements such as oxygen
bottom of a particle and reach the underlying substrate, the in addition to carbon. If carbon is of analytical interest, other
measured spectrum will always be a composite of contribu- low atomic number substrates are available, including high
tions from the particle and the substrate, as shown in purity boron and beryllium (but beware of the health hazard
. Fig. 23.21 for a K411 glass particle (~1 μm in diameter) on of handling beryllium, especially in the form of beryllium
a bulk carbon substrate. To obtain a spectrum that is repre- oxide which may be released by surface abrasion). Other pure
sentative of the particle material alone, it is important when- element substrates such as aluminum and silicon can be used
ever possible to choose a substrate consisting of an element(s) if that element(s) is not of interest, but because of the high
not contained in the particle of interest. As part of a quality degree of excitation of the Al K-L2,3 and Si K-L2,3 characteris-
measurement system, the EDS spectrum of the blank sub- tic X-rays under typical operating conditions, a significant
strate material should be measured to determine what ele- fraction of the EDS deadtime will be taken up by the substrate
ments are present at the major, minor, and trace level. X-rays, diminishing the analytical information collected per
Additionally, it is desirable that the characteristic peak(s) of unit time on the particle of interest, as well as contributing
the substrate element(s) should not interfere with character- coincidence peaks that might be misinterpreted as minor or
istic peaks from the constituents of the particle. trace constituents, for example, 2 Al K-L2,3 (2.974 keV) is close
Carbon is an excellent choice for a bulk substrate since it is to the energy of Ag L3-M4,5 (2.981 keV) and 2 Si K-L2,3
available in high purity, is mechanically robust, and supplied (3.480 keV) is close to the energy of Sn L3-M4,5 (3.440 keV)
by various vendors as planchets with different surface finishes, When depositing particles on a substrate, the area density
including a highly polished, glassy surface that is nearly fea- should be minimized to avoid situations where electrons
tureless as a background for SEM imaging of small particles. scattered off the particle being analyzed can strike nearby
The low energy of the C K X-ray (0.285 keV) is unlikely to particles and excite X-rays, which will then contribute an
cause interference with most elements of interest. In addition artifact (“cross talk”). While this remote excitation is likely to
to bulk substrates, carbon is often used in the form of carbon- be at the equivalent of a minor or trace level constituent, it is
23 critical to understand such contributions when low level con-
infused tape with a sticky surface to which particles will read-
ily adhere. If a carbon tape preparation is used, the analyst stituents are of interest.
23.6 · Particle Analysis
397 23
70000
60000
50000
40000
30000
20000
10000
0
0 200 400 600 800 1000 1200 1400 1600 1700 2000 2200 2400
Photon energy (eV)
200000
150000
100000
50000
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Photon energy (eV)
c Noisy[MC simulation of a 1.000 mm
280000 E0 = 20 keV diameter sphere of Al on C] 20keV
260000
240000
220000
220000
180000
Intensity
counts
160000
140000
120000
100000
80000
60000
40000
20000
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Photon energy (eV)
d 220000 Noisy[MC simulation of a 1.000 mm
diameter sphere of Al on C] 30keV
200000 E0 = 30 keV
180000
160000
140000
Intensity
counts
120000
100000
80000
60000
40000
20000
0
0 200 400 600 700 1000 1200 1400 1600 1800 2000 2200 2400
Photon energy (eV)
.. Fig. 23.20 EDS spectra calculated with the DTSA-II Monte Carlo absence of the C peak from the substrate at E0 = 5 keV, its appearance
simulation for a 1-μm-diameter Al particle on a C substrate at vari- at E0 = 10 keV and the increase relative to the Al peak as the beam
ous beam energies: a 5 keV, b 10 keV, c 20 keV, and d 30 keV. Note the energy increases
398 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
K411_sphere-on-tape_20kV10nA
K411_sphere-on-film_20kV10nA
Counts
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
K411_sphere-on-tape_20kV10nA
K411_sphere-on-film_20kV10nA
K411
E0 = 20 keV
C – tape substrate
Counts
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 23.21 Comparison of EDS spectra similar sized K411 spheres mounted on a bulk carbon substrate (red) and on a 20-nm-thick carbon film
on a copper support grid blue). Note artifact Cu peaks arising from lateral electron scattering from particle to excite the support grid
ED
S
b
beam on the side of the particle away from the EDS results in
an extended absorption path. The generated X-rays must
pass through a large mass of the particle to reach the detector.
c This effect is illustrated schematically in . Fig. 23.23, where
absorption paths for X-rays generated near the maximum
penetration of beam electrons in the particle are compared
for three beam positions: top center, and at positions directly
facing and away from the EDS. Because absorption follows
an exponential dependence on path length and becomes
increasingly significant for low energy photons below
approximately 4 keV, the low energy portion of the spectrum
will show the strongest absorption effects. Spectra recorded
at the top of a 5-μm-diameter particle and on the side away
from the EDS are compared in . Fig. 23.24 demonstrating a
factor of two difference in the intensity measured at the
energy of Mg K-L2,3 (1.254 keV). For a beam placed on the
curved side of the particle facing the EDS, the absorption
path length is actually reduced relative to the top center, lead-
.. Fig. 23.22 Particles deposited on a thin (~20-nm) carbon film sup- ing to extra emission from the low energy photons, creating
ported on a copper grid shown at various magnifications. a Nominal 200 about 30 % excess for Mg K-L2,3, as shown in the comparison
X, field width 351 by 263 micrometers; b nominal 1 kX, field width 70 by of spectra in . Fig. 23.25. . Figure 23.26 shows the impact of
53 micrometers; c nominal 5 kX, field width 14 by 11 micrometers beam placement on the accuracy of quantitative analysis, as
discussed below. It is thus critical that the analyst is always
The Importance of Beam Placement aware of the relative position of the EDS in the SEM image
The placement of the beam on the particle relative to the EDS when choosing locations to analyze. A reliable way to locate
detector position can have a strong effect on the measured the EDS is to record an X-ray spectrum image and examine
spectrum. For particles of intermediate size where the inter- selected X-ray intensity maps, as shown in . Fig. 23.27. The
action volume is contained within the particle, placing the general rule that the source of the apparent illumination
400 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
K411_sphere_beam-top-center_20kV10nA
100000 K411_sphere_r/2_away_20kV10nA
10000
Counts
1000
100
10
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
3000 K411_sphere_beam-top-center_20kV10nA
K411_sphere_r/2_away_20kV10nA
2500 K411
15-µm sphere (20 nm C film on Cu grid)
2000 E0 = 20 keV
Beam position: Top Center
Counts
1000
500
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 23.24 Comparison of EDS spectra recorded at the top center and on the side away from the EDS
appears to come from the position of the detector immedi- structure within the particle. As described below, overscan-
ately reveals the true EDS position. For this particular parti- ning is NOT a means by which an “average” composition can
cle, the eclipsing effect of the extended absorption path on be found by quantitative microanalysis.
the backside of the particle is readily apparent, even in the
relatively energetic Co K-L2,3 (6.930 keV) intensity map.
23.6.3 -ray Spectrum Imaging:
X
Overscanning Understanding Heterogeneous
“Overscanning” is a strategy by which the beam is continu- Materials
ously scanned over the particle (or rough surface, heteroge-
neous material, etc.) while the EDS spectrum is being X-ray spectrum imaging (XSI) involves collecting a com-
collected. If the magnification control of the SEM is continu- plete EDS spectrum at every pixel location visited by the
ously variable, then the size of the scan raster can be adjusted scanned beam. When applied to particle analysis, the XSI
to bracket the particle, minimizing the fraction of the time provides the analyst with an abundance of information
that the beam spends on the surrounding substrate, as shown which can be recovered by post-collection processing of the
in . Fig. 23.28 (yellow box). Alternatively, the scan raster can XSI datacube. Composition heterogeneity down to the sin-
be adjusted to fill as much of the particle image as possible gle pixel level can be detected and interpreted. The particle
while remaining within the bounds of the particle, thus X-ray intensity images shown in . Fig. 23.27a, b were
avoiding direct beam placement on the substrate, as shown in extracted from an XSI, and the localization of Ti and Mo in
. Fig. 23.28 (green box). inclusions is immediately obvious. Solidification dendrites
23 While useful for gaining qualitative information on the are outlined in the Ti K-L2,3 map, a feature that is not obvi-
constituents of a particle, overscanning only has utility if the ous in any of the other elemental intensity maps. Software
particle (or other target object) is homogeneous. Overscanning tools enable the analyst to extract spectra that are represen-
discards valuable information on any possible inhomogeneous tative of individual components of the microstructure, such
23.6 · Particle Analysis
401 23
100000
K411_sphere_beam-top-center_20kV10nA
K411_sphere_r/2_toward_20kV10nA
10000
Counts
1000
100
10
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
K411_sphere_beam-top-center_20kV10nA
K411_sphere_r/2_toward_20kV10nA
6000
5000
K411
4000 15-µm sphere (20 nm C film on Cu grid)
Counts
E0 = 20 keV
Beam position: Top Center
3000
Beam position: R/2, toward EDS
2000
100
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (keV)
.. Fig. 23.25 Comparison of EDS spectra recorded at the top center and on the side facing the EDS
BSE Al
50 µm
EDS
Cr Co
Mo Ti
Ni EDS Mo Ti Ni
23 b 50 µm
.. Fig. 23.27 a SEM BSE image and X-ray intensity maps for Al K-L2,3, Mo L3-M4,5, and Ni K-L2,3, and a color overlay with Mo L3-M4,5 = red, Ti
Cr K-L2,3, and Co K-L2,3. Note the shadowing effect of the particle thick- K-L2,3 = green, and Ni K-L2,3 = blue. Note the localization of Ti and Mo in
ness causing attenuation of X-ray intensity. b X-ray intensity maps for inclusions, and the solidification dendrites outlined by Ti
23.6 · Particle Analysis
403 23
detector (SDD)-EDS inevitably shows the shadowing cre-
ated by specimen geometry, such shadowing can be greatly
diminished by collecting the XSI with an array of SDD-EDS
detectors placed symmetrically around the specimen, as
shown in . Fig. 23.29.
Ni Ti Total
Ni 50 µm Ti Total
.. Fig. 23.29 Comparison of X-ray spectrum imaging with a single EDS and with an array of four EDS detectors; note reduction in shadowing
404 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
Monte Carlo Simulation Fe in K411 Spheres, E0 = 20 keV .. Table 23.3 Composition of K411 glass and analysis of flat,
1.2 bulk target at E0 = 20 keV (Standards: Mg, Si, CaF2, Fe; oxygen by
stoichiometry)
1.0
NIST SRM 470 certificate values
Fe k-ratio (bulk K411)
.. Fig. 23.30 Monte Carlo calculation of the emission of Fe K-L2,3 from observed for spherical particles with diameters of 10 μm and
K411 spheres of various diameters at E0 = 20 keV. The intensity is nor- greater.
malized by the Fe K-L2,3 intensity calculated for a flat bulk target of K411
“Particle Absorption Effect”
The surface curvature of a particle can reduce the absorption
for an incident beam energy of 20 keV. For this plot, the Fe path to the detector compared to the interaction volume in a
K-L2,3 intensity emitted from the particle has been divided by flat bulk target, as shown schematically in . Fig. 23.31. Since
the intensity calculated for a flat bulk K411 target to deter- X-ray absorption depends exponentially on the absorption
mine the k-ratio. Note that as the particle diameter increases path length, surface curvature can significantly modify the
from zero, the Fe K-L2,3 k-ratio increases and asymptotically measured intensity, creating the “particle absorption effect.”
approaches unity (i.e., equivalent to the flat bulk target) for a The magnitude of the particle absorption effect depends
diameter of approximately 10 μm. Thus, for the case K411 at strongly on the energy of the characteristic photons involved.
E0 = 20 keV, bulk behavior for Fe K-L2,3 (6.400 keV) is For the example shown in . Fig. 23.32, the Fe K-L2,3
beam
de
tec
to
r
Additional
absorption path Absorption
in flat bulk target path
in particle Electron
trajectory
23
23.6 · Particle Analysis
405 23
.. Fig. 23.32 Emission of Fe
K-L2,3, O K-L2 and Mg K-L2,3 (nor- Monte Carlo calculations of K411 Spheres (E0 = 20 keV)
malized to O K-L2 and Mg K-L2,3
from bulk K411); note excursion
1.6
in the O K-L2,3 and Mg K-L2,3
above unity (i.e., exceeding the 1.8
emission from a flat, bulk target)
1.2
0.8
O
0.6 Mg
FeK
0.4
0.2
0.0
0 2 4 6 8 10
Diameter (micrometers)
radiation at 6.40 keV is sufficiently energetic that it under- modification of the absorption path. Particle geometry and
goes relatively little absorption in the bulk case so that the its impact on X-ray absorption must be considered when
modification of the absorption path length by particle sur- selecting beam locations on a particle for analysis. It is criti-
face curvature produces a negligible effect. However, the cal that the analyst always be aware of the position of the EDS
lower energy characteristic photons such as O K-L2 detector relative to the X-ray source, as demonstrated in
(0.525 keV) and Mg K-L2,3 (1.254 keV) in K411 suffer signifi- . Fig. 23.24, to minimize the effects of particle geometry.
cant absorption in a flat bulk target, so that when these pho-
tons are generated in a spherical particle, the reduced
absorption path in the direction toward the EDS leads to an 23.6.5 ncertainty in Quantitative Analysis
U
increase in X-ray emission compared to a flat bulk target. of Particles
. Figure 23.32 shows that as the particle diameter increases
from zero, the k-ratios for O K-L2 and Mg K-L2,3 initially Quantitative analysis of particles is performed by following
increase similarly to FeKα as a result of the particle mass the same k-ratio/matrix correction protocol used for flat, bulk
effect. However, for larger particles the reduced absorption specimens. However, it must be recognized that particle
path of the curved particle surface causes the emitted O geometry modification of the interaction of beam electrons
K-L2 k-ratio to actually exceed unity (i.e., higher emission and the subsequent propagation of X-rays introduce factors
than bulk behavior) for a particle diameter of 1.6 μm, reach- which violate the fundamental assumption of the bulk quan-
ing a maximum of 1.35 relative to bulk at a particle diameter tification method, namely that the only reason the X-ray
of 2.8 μm. For Mg K-L2,3, the emission exceeds unity for a intensities measured in the target being analyzed are different
particle diameter of 2 μm and reaches a maximum of 1.17 at from the standards is that the composition(s) is different.
a diameter of 3.0 μm. For particle diameters beyond the Thus, with the impact of the geometric factors, the analytical
intensity maxima, the O K-L2 and Mg K-L2,3 k-ratios gradu- accuracy of the conventional k-ratio/matrix correction proto-
ally decrease with increasing particle diameter, asymptoti- col is inevitably compromised. The critical question to con-
cally approaching the equivalent of bulk behavior at 25 μm sider is the degree to which the uncertainty budget is increased
diameter for O K-L2, and 18 μm diameter for Mg K-L2,3. by the systematic error contribution of the particle effects.
Thus, for spherical particles of the K-411 composition mea-
sured with E0 = 20 keV, effectively bulk behavior is observed he Analytical Total Reveals the Impact
T
for all characteristic X-ray energies for particles with diame- of Particle Effects
ters greater than 25 μm for a beam position at the top center The analytical total is the sum of the calculated concentra-
of the particle (detector take-off angle 40°). As demonstrated tions, including oxygen by stoichiometry if calculated.
in . Fig. 23.23, deviations in the beam placement either . Table 23.3 shows the results of the analysis of K411 glass in
toward the EDS or away have significant effects due to the the form of a flat, bulk target. The beam energy was
406 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
.. Table 23.4 Analysis of a 1.3-μm-diameter spherical particle of K411 glass with fixed beam located at particle center (standards: Mg, Si,
CaF2, Fe; oxygen by stoichiometry)
Element SRM value Analysis Rel error (%) Normalized Rel error (%)
.. Table 23.5 Analysis of a 6.1-μm-diameter spherical particle of K411 glass with fixed beam located at particle center (standards: Mg, Si,
Ca [K412 glass], Fe; oxygen by stoichiometry)
Element SRM value Analysis Rel error (%) Normalized Rel error (%)
E0 = 20 keV, and the standards used pure elements Mg, Si, and have such large systematic errors as to offer no realistic
Fe, while SRM 470 K412 glass was used as the Ca standard. meaning. To compensate for the mass effect and thus place
The analytical total was 0.9789 (the sum of the SRM certifi- the concentrations on a meaningful basis, internal normal-
cate values is 0.9886), and the relative errors were all well ization can be applied:
within the ±5 % error envelope, with the largest error at
−2.2 % relative for Ca. Cn (i ) = Ci / ΣCi (23.3)
. Table 23.4 gives the results of applying the k-ratio/
matrix correction factor protocol to the analysis of a Note that normalization is only useful if all constituents pres-
1.3-μm-diameter spherical particle of K411 glass, with the ent in the analyzed volume are included in the total, includ-
EDS spectrum collected with the beam placed at the center of ing any such as oxygen that are calculated by assumed
the particle image. The intensity of each elemental constitu- stoichiometry rather than measured directly. After normal-
ent has been measured relative to the same suite of standards: ization, the relative errors are reduced in magnitude, as given
pure elements Mg, Si, and Fe, and CaF2 as the Ca standard. in . Table 23.4, but the values for Mg (−14 %) and Fe (+26 %)
Oxygen has been calculated by the method of assumed stoi- remain well outside the bulk analysis error histogram.
chiometry. The analytical total is 0.3480, and the relative . Table 23.5 presents similar measurements and calcula-
errors for the raw calculated concentrations are large and tions for a 6.1-μm-diameter K411 particle for which the ana-
negative, for example, −70 % relative for Mg and −56 % rela- lytical total is 1.091. This particle diameter is sufficiently large
tive for Fe. These large negative relative errors (a negative so that the X-ray production for the higher energy photons,
relative error indicates that the calculated concentration Ca K-L2,3 and Fe K-L2,3, has nearly reached equivalence to the
underestimates the true concentration) for both the low pho- flat, bulk target, resulting in relative errors of −5 % or less.
ton energy peaks (e.g., Mg K-L2,3 and Si K-L2,3) and the high For this particle size, the lower energy photon peaks, Mg
photon energy peaks (e.g., Ca Kα and Fe Kα) are a result of K-L2,3 and Si K-L2,3, are still strongly influenced by the parti-
the particle mass effect reducing all X-ray intensities com- cle absorption effect, causing relative errors that are large and
23 pared to bulk behavior. Clearly, these raw concentrations positive, since more of these low energy photons escape than
23.6 · Particle Analysis
407 23
.. Fig. 23.33 Relative errors
observed for various sizes of Analysis of K411 Spheres (E0 = 20 keV)
spherical particles of K411 glass
measured with a fixed beam 40
placed at the center of the par-
ticle image
Mg
30
Si
Ca
20 Fe
-10
-20
0 2 4 6 8 10 12 14 16 18
Diameter (micrometers)
is the case for a flat, bulk target. When these raw concentra- error to the other higher atomic number constituents. In the
tions are normalized, the relative errors for Mg and Si are case of oxygen, the method of assumed stoichiometry is gen-
reduced, but the relative errors for Ca and Fe are increased erally much more effective.
after normalization.
These examples demonstrate the complex interplay of Does Overscanning Help?
X-ray generation and propagation as influenced by particle Because of the difficulty in analyzing particles, overscan-
geometry. While normalization of the raw calculated concen- ning the particle during EDS collection is thought to obtain
trations is necessary to put particle analyses on a realistic a spectrum that averages particle effects. In reality, even for
concentration basis, the uncertainty budget for particle anal- homogeneous particles, overscanning does not decrease
ysis is substantially increased compared to that for the ideal the relative uncertainties but can actually cause an increase.
flat target. . Figure 23.33 plots the relative error envelope for . Figure 23.34a plots the relative errors in the normalized
normalized concentrations as a function of particle diameter concentrations for the analysis of Mg and Fe in K411 spheres
for spherical particles of K411 glass. For particles whose of various sizes, comparing point beam analyses centered on
dimensions are substantially smaller than the bulk interac- the particle image with continuous overscanning during EDS
tion volume, the relative errors in the normalized concentra- collection. Mg and Fe are chosen because the large separa-
tions are large and increase as the particle size decreases. The tion in characteristic X-ray energy provides sensitivity to
relative errors decrease as the particle diameter increases, the action of the particle mass effect, which is the only sig-
eventually converging with the flat bulk case for particles nificant influence on energetic Fe K-L2,3, while both the mass
above approximately 25-μm diameter. effect and the absorption effect influence Mg K-L2,3. While
Normalization is most successful when applied to compo- the error range for point beam analysis is substantially larger
sitions where the measured characteristic X-rays have similar than the ideal error histogram for flat bulk target analysis, the
energies, for example, Mg K-L2,3, Al K-L2,3 and Si K-L2,3. effect of overscanning is actually to shift the distribution of
Although low atomic number elements such as oxygen can be results to even more severe relative errors. This is a result of
measured directly, the high absorption of the low energy pho- the non-linear nature of X-ray absorption, which can be seen
tons of the characteristic X-ray significantly increases the in the beam placement measurements shown in . Figs. 23.24
effect of the particle absorption effect, so that normalization and 23.25. A similar effect is seen for irregular shards in
introduces large errors and effectively transfers increased . Fig. 23.34b.
408 Chapter 23 · Analysis of Specimens with Special Geometry: Irregular Bulk Objects and Particles
30
10
0 Point beam
Overscan
-10
-20
-30
-40 -30 -20 -10 0 10 20
Mg relative error (%)
80 Point beam
Overscan
60
Fe relative error (%)
40
20
-20
-40
-60 -40 -20 0 20 40
Mg relative error (%)
23.6.6 Peak-to-Background (P/B) Method Wernisch, 1991a, b, c) is based on the observation that
although the characteristic X-ray intensity emitted from an
Specimen Geometry Severely Affects the irregularly shaped object is highly dependent on local geo-
k-ratio, but Not the P/B metric effects, the P/B ratio measured between the character-
Another approach to establishing quantitative X-ray micro- istic X-rays and the continuum X-rays of the same energy is
analysis for objects of an irregular shape, such as particles much less sensitive to specimen geometry. . Table 23.6 con-
and rough surfaces, is the peak-to-background (P/B) method, tains measurements of the k-ratio (measured relative to bulk
which is an extension of the Marshall–Hall method for the K411) and the P/B from the spectra of the SRM-470 (K-411
23 correction of mass loss in beam-sensitive materials (see glass) shards in Fig. 23.13. The shard spectra show significant
7 Chap. 20) (Marshall and Hall, 1966; Hall, 1968). The P/B
deviations from the spectrum of the polished bulk K411,
method (Small et al., 1978, 1979a,b; Statham and Pawley, especially at low photon energies below 4 keV. Although the
1978; Statham, 1979; Wendt and Schmidt, 1978; August and k-ratio for Mg measured for these shards varies by a factor of
23.6 · Particle Analysis
409 23
and continuum generation processes, at least to a first order for
.. Table 23.6 K411 shards
photons of the same energy. Both types of radiation have a
Sample Element P/B k-ratio similar, although not identical, depth distribution; thus, the
(relative absorption paths to the detector are alike. As the same photon
to bulk energy is chosen for characteristic and continuum X-rays, the
K411)
geometric absorption effect is thus comparable for both. When
Shard A Mg 4.52 0.545 making corrections for an irregularly shaped object, the exact
absorption path is very difficult to determine. Because the con-
Shard C Mg 4.49 0.339
tinuum radiation of the same photon energy is following the
Shard D Mg 4.52 1.132 same path to the detector that the characteristic radiation fol-
Shard E Mg 4.73 0.389
lows, regardless of local object shape, this continuum intensity
IB can be used as an automatic internal normalization factor to
Bulk Mg 4.57 1.00 compensate for the major geometric effects. Furthermore, the
Shard A Si 16.35 0.617 P/B ratio is independent of probe current (and thus need not
be measured); yet, the quantification results need not be nor-
Shard C Si 17.32 0.548
malized. Because of this, both standards-based and standard-
Shard D Si 14.95 1.06 less P/B algorithms have been implemented that provide an
Shard E Si 17.33 0.447 estimate of the analytical total. More sophisticated models
have been developed that account for the second-order (i.e.,
Bulk Si 15.80 1.00
subtler) differences between the distributions of characteristic
Shard A Ca 6.78 0.835 and continuum radiation generation (August and Wernisch,
Shard C Ca 6.58 0.866
1991a, b, c).
.. Table 23.7 Relative Deviation from Expected Value (RDEV) observed with peak-to-background corrections compared to the raw
concentrations and normalized concentrations after conventional k-ratio/matrix corrections
Spherical particle: IN-100 alloy, 88 μm diameter, with beam placed at 22 μm from the top center on the backside of particle
standard identical in composition to the unknown object is collimator has a large acceptance area at the specimen. To
generally not available. However, an estimate of the concen- minimize the extraneous background contributions, the
trations of elements in the unknown object is always avail- small particles should be mounted on a thin (approximately
able in the ZAF procedure, including the first step, where 10–20 nm) carbon film supported on a metal grid (typically
Ci = ki /Σk. The value of IB,bulk can therefore be estimated copper, as used in the transmission electron microscope) and
from the background measured on pure element standards: mounted over a blind hole drilled into a carbon block. The
continuum contribution from such a thin film is negligible
I B, bulk = ∑C j I j, B,std (23.6) relative to particles as small as approximately 250 nm in
j
diameter.
where Ij,B,std is the pure element bremsstrahlung at the energy
of interest and Cj is the concentration of element j. An exam-
ple of an analysis of a complex IN-100 particle with conven- 23.7 Summary
tional k-ratio/matrix corrections (including normalization of
the raw values) and using the k-ratio/matrix corrections aug- 1. Particle analysis is inevitably compromised compared to
mented with the P/B method is given in . Table 23.7. The analysis of ideal flat samples, leading to an increased
relative deviation from the expected value (RDEV) for each error budget.
element is reduced compared with the simple normalization 2. Careful attention must be paid to optimizing particle
procedure, especially for Al, which is highly absorbed when sample preparation to minimize substrate contributions
measured on the backside of the particle. to the spectrum and to reduce contributions from
The special advantage of the P/B method is that it can be nearby particles.
applied to spectra obtained with a focused probe directed at 3. Quantitative analysis of particles follows the k-ratio/
a specific location on a particle. Thus, particles that have a matrix correction protocol. The analytical total that
chemically heterogeneous sub-structure can be directly stud- results from this procedure is an indication of the
ied. To be effective, the P/B method requires spectra with magnitude of particle geometry effects (mass effect and
high counts. Because the ratio of background intensities is absorption effect).
used to scale the particle peak intensities, the statistical 4. Normalization of the raw concentrations (including
uncertainty in the background ratio propagates into the error oxygen by stoichiometry, if appropriate) is necessary
in each concentration value in addition to the statistics of the to place the calculated composition on a realistic
characteristic peak. Even more importantly, the P/B method basis.
depends on the background radiation originating in the 5. Large relative errors, exceeding 10%, are encountered after
excited volume of the specimen only, and not in the sur- normalization. The analytical errors are exacerbated when
rounding substrate. When an irregularly shaped object such low and high photon peaks must be used for analysis.
as a particle becomes small relative to the bulk interaction 6. The analytical errors generally increase as the particle
volume, the penetration of the beam into the substrate means size decreases.
that the continuum continues to be produced, even if the 7. Overscanning does not decrease the analytical errors,
substrate is a low atomic number element such as carbon. As and may well increase the errors depending on the
23 noted above, the energy-dispersive X-ray spectroscopy particular combination of elements being analyzed.
References
411 23
References spheres for microanalysis. In: Johari O (ed) Scanning electron
microscopy/1978/I. IITRI, Chicago, p 445
Small JA, Heinrich KFJ, Newbury DE, Myklebust RL (1979a) Progress in
August H-J, Wernisch J (1991a) Calculation of the depth distribution
the development of the peak-to-background method for the
function for continuous radiation, Scanning 13:207–215
quantitative analysis of single particles with the electron probe in
August H-J, Wernisch J (1991b) Calculation of depth distribution func-
Scanning Electron Microscopy/1979/II, ed. Johari, O. (IITRI,
tions for characteristic x-radiation using an electron scattering
Chicago), p 807
model. I—theory, X-Ray Spectrom, 20:131–140
Small JA, Newbury DE, Myklebust RL (1979b) Analysis of particles and
August H-J, Wernisch J (1991c) Calculation of depth distribution func-
rough samples by FRAME P, a ZAF method incorporating peak-to-
tions for characteristic x-radiation using an electron scattering
background measurements, in Microbeam analysis, ed. Newbury, D.
model. II—results, X-Ray Spectrom, 20:141–148
(San Francisco Press, San Francisco), p 243
Hall T (1968) Some aspects of the microprobe analysis of biological
Small JA, Heinrich KFJ, Newbury DE, Myklebust RL, Fiori CE (1980)
specimens. In: Heinrich K (ed) Quantitative electron probe micro-
Procedure for the Quantitative Analysis of Single Particles with the
analysis. U.S. Government Printing Office, Washington, DC, p 269
Electron Probe. In: Heinrich KFJ (ed) Characterization of Particles.
Marshall D, Hall T (1966) A method for the microanalysis of thin films. In:
National Bureau of Standards Special Publication 533, Washington,
Castaing R, Deschamps P, Philibert J (eds) X-ray optics and micro-
pp 29–38
analysis. Hermann, Paris, p 374
Statham PJ, Pawley JB (1978) A new method for particle X-ray micro-
Newbury D, Ritchie W (2013a) Quantitative SEM/EDS, Step 1: What
analysis based on peak to background measurement. In: Johari O
Constitutes a Sufficiently Flat Specimen? Microsc Microanal
(ed) Scanning electron microscopy/1978/I. IITRI, Chicago, p 469
19(Suppl 2):1244
Statham P (1979) A ZAF procedure for microprobe analysis based on
Newbury D, Ritchie W (2013b) Is Scanning Electron Microscopy/Energy
measurement of peak-to-background ratios. In: Newbury D (ed)
Dispersive Spectrometry (SEM/EDS) Quantitative? Scanning 35:141
Microbeam Analysis – 1979. San Francisco Press, San Francisco, pp
Small JA, Heinrich KFJ, Fiori CE, Myklebust RL, Newbury DE, Dilmore MF
247–253
(1978) The production and characterization of glass fibers and
Wendt M, Schmidt A (1978) Improved reproducibility of energy-disper-
spheres for microanalysis, In: Johari, O (ed) Scanning electron
sive X-ray microanalysis by normalization to the background, Phys
microscopy/1978/I, IITRI, Chicago, p 445
Status Solidi (a) 46:179
Small JA, Heinrich KFJ, Newbury DE, Myklebust RL (1979) SEM/1979/II,
Yakowitz H, Heinrich KFJ (1968) Quantitative electron probe microanaly-
SEM, Inc., AMF O’Hare, Illinois, p 807
sis: fluorescence correction uncertainty. Mikrochim Acta 5:182
Small JA, Heinrich KFJ, Fiori CE, Myklebust RL, Newbury DE, Dillmore MF
(1978) The production and characterization of glass fibers and
413 24
Compositional Mapping
24.1 Total Intensity Region-of-Interest Mapping – 414
24.1.1 Limitations of Total Intensity Mapping – 415
References – 439
SEM images that show the spatial distribution of the elemen- per elemental window can vary widely from a few counts
tal constituents of a specimen (“elemental maps”) can be cre- per pixel to several thousand or more. A typical strategy to
ated by using the characteristic X-ray intensity measured for avoid saturation is to collect 2-byte deep X-ray intensity
each element with the energy dispersive X-ray spectrometer data that permits up to 65,536 counts per energy region per
(EDS) to define the gray level (or color value) at each picture pixel. In common with the practice for BSE and SE images,
element (pixel) of the scan. Elemental maps based on X-ray the final elemental map will be displayed with a 1-byte
intensity provide qualitative information on spatial distribu- intensity range (0–255 gray levels). To maximize the con-
tions of elements. Compositional mapping, in which a full trast within each map it is necessary to nearly fill this dis-
EDS spectrum is recorded at each pixel (“X-ray Spectrum play range so that it is common practice to automatically
Imaging” or XSI) and processed with peak fitting, k-ratio scale (“autoscale”) the measured intensity in a linear fashion
standardization, and matrix corrections, provides a quantita- to span slightly less than 1-byte. The displayed gray levels
tive basis for comparing maps of different elements in the are scaled to range from near black, but avoiding black (gray
same region, or for the same element from different regions. level zero) to avoid clipping, to near white, but avoiding full
white (gray level 255), to prevent saturation. The counts are
expanded for elements that span less than 1-byte in the
24.1 otal Intensity Region-of-Interest
T original data collection, while the counts are compressed for
Mapping elements that extend into the 2-byte range. An example of
such total intensity mapping is shown in . Fig. 24.2, which
In the simplest implementation of elemental mapping, presents a set of maps for Si, Fe, and Mn in a cross section of
energy regions are defined in the spectrum that span the a deep-sea manganese nodule with a complex microstruc-
characteristic X-ray peak(s) of interest, as shown in ture. The EDS spectrum shown in . Fig. 24.1 also reveals a
. Fig. 24.1. The total X-ray intensity (counts) within each typical problem encountered in simple intensity window
energy region, IXi, consisting of both the characteristic peak mapping. Manganese is one of the most abundant elements
intensity, including any overlapping peaks, and the contin- in this specimen, and the Mn K-M2,3 (6.490 keV) interferes
uum background intensity, is digitally recorded for each with Fe K-L2,3 (6.400 keV), which is especially significant
pixel, creating a set of x-y-IXi image arrays for the defined since iron is a minor/trace constituent. To avoid the poten-
suite of elements. Depending on the local concentration of tial artifacts in this situation, the analyst can instead choose
an element, the overvoltage U0 = E0/Ec for the measured the Fe K-M2,3 (7.057 keV) which does not suffer the inter-
characteristic peak, the beam current, the solid angle of the ference but which is approximately a factor of ten lower
detector, and the dwell time per pixel, the number of counts intensity than Fe K-L2,3. While sacrificing sensitivity, the
Mn-nodule_20kV25nA
Mn
1000000
800000
Counts
600000
Si
400000
Fe
200000
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Photon energy (KeV)
.. Fig. 24.1 EDS spectrum measured on a cross section of a deep-sea manganese nodule showing peak selection (Si K-L2, Mn K-L2,3, and Fe
K-M2,3) for total intensity elemental mapping
24
24.1 · Total Intensity Region-of-Interest Mapping
415 24
Si Fe
20 µm
Si Fe Mn
Mn
.. Fig. 24.2 Total intensity elemental maps for Si K-L2, Mn K-L2,3, and Fe K-M2,3 measured on a cross section of a deep-sea manganese nodule,
and the color overlay of the gray-scale maps
success of this strategy is revealed in . Fig. 24.2, where the 24.1.1 imitations of Total Intensity
L
major Mn and minor Fe are seen to be anti-correlated. Mapping
. Figure 24.2 also contains a primary color overlay of
these three elements, encoded with Si in red, Fe in green, and While total intensity elemental maps such as those in
Mn in blue, a commonly used image display tool which pro- . Fig. 24.2 are useful for developing a basic understanding of
vides an immediate visual comparison of the relative spatial the spatial distributions of the elements that make up the
relationships of the three constituents. The appearance of specimen, total intensity mapping is subject to several sig-
secondary colors shows areas of coincidence of any two ele- nificant limitations:
ments, for example, the cyan colored region is a combination 1. By selecting only the spectral regions-of-interest, the
of green and blue and thus shows the coincidence of Fe and amount of mass storage is minimized. However, while all
Mn. The other possible binary combinations are yellow (red spectral regions-of-interest are collected simultaneously,
plus green, Si + Fe) and magenta (red plus blue, Si + Mn), if the analyst needs to evaluate another element not ori
which are not present in this example. If all three elements ginally selected when the data was collected, the entire
were present at the same location, white would result. image scan must be repeated to recollect the data with that
new element included.
416 Chapter 24 · Compositional Mapping
2. Total intensity maps convey qualitative information only. 4. When peak interference occurs, the raw intensity in
The elemental spatial distributions are meaningful only a given energy window may contain contribu-
in qualitative terms of interpreting which elements are tions from another element, as shown in . Fig. 24.1
present at a particular pixel location(s) by comparing dif- where the region that includes Fe K-L2,3 also con-
ferent elemental maps of the same area, for example, using tains intensity from Mn K-M2,3. While choosing the
the color overlay method. Since the images are recorded non-interfered peak Fe K-M2,3 gives a useful result
simultaneously, the pixel registration is without error even in the case of the manganese nodule, if the speci-
if overall drift or other distortion occurs. However, the men also contained cobalt at a significant level, Co
intensity information is not quantitative and can only con- K-L2,3 (6.930 keV) would interfere with Fe K-M2,3
vey relative abundance information within an individual (7.057 keV) and invalidate this strategy. The peak
elemental map (e.g., “this location has more of element A interference artifact can be corrected by peak fitting,
than this location because the intensity of A is higher”). or by methods in which the measured Mn K-L2,3
The gray levels in maps for different elements cannot be intensity, which does not suffer interference in this
readily compared because the X-ray intensity for each particular case, is used to correct the intensity of the
element that defines the gray level range of the map is Fe K-L2,3 + Mn K-M2,3 window using the known Mn
determined by the local concentration and the complex K-M2,3/K-L2,3 ratio.
physics of X-ray generation, propagation, and detection 5. The total intensity window contains both the charac-
efficiency, all of which vary with the elemental species. teristic peak intensity that is specific to an element
The element-to-element differences in the efficiency of and the continuum (background) intensity, which
X-ray production, propagation, and detection are embed- scales with the average atomic number of all of the
ded in the raw measured X-ray intensities, which are elements within the excited interaction volume but is
then subjected to the autoscaling operation. Unless the not exclusively related to the element that is generat-
autoscaling factor is recorded (typically not), it is not pos- ing the peak. For the map of an element that consti-
sible after the fact to recover the information that would tutes a major constituent (mass concentration C >0.1
enable the analyst to standardize and establish a proper or 10 wt %), the non-specific background intensity
basis for inter-comparison of maps of different elements, contribution usually does not constitute a serious
or even of maps of the same element from different areas. mapping artifact. However, for a minor constituent
Thus, the sequence of gray levels only has interpretable (0.01 ≤ C ≤ 0.1, 1 wt % to 10 wt %), the average atomic
meaning within an individual elemental map. Gray levels number dependence of the continuum background
cannot be sensibly compared between total intensity maps can lead to serious artifacts. . Figure 24.3 shows an
of different elements, for example, the near-white level in example of this phenomenon for a Raney nickel alloy
the autoscaled maps of . Fig. 24.2 for Si, Fe, and Mn does containing major Al and Ni with minor Fe. The
not correspond to the same X-ray intensity or concentra- complex microstructure has four distinct phases, the
tion for three elements. Because of autoscaling, it is not compositions of which are listed in . Table 24.1, one
possible to compare maps for the same element “A” from of which contains Fe as a minor constituent at a
two different regions, even if recorded with the same dose concentration of approximately C = 0.04 (4 wt %).
conditions, since the autoscaling factor will be controlled This Fe-rich phase can be readily discerned in the Fe
by the maximum concentration of “A,” which may not be gray-scale map, where the intensity of this phase,
the same in two arbitrarily chosen regions of the sample. being the highest iron-containing region in the
3. This lack of quantitative information in elemental total image, has been autoscaled to near white. In addition
intensity maps extends to the color overlay presentation to this Fe-containing phase, there appears to be
of elemental maps seen in . Fig. 24.2. The color over- segregation of lower concentration levels of Fe to the
lay is useful to compare the spatial relationships among Ni-rich phase relative to the Al-rich phase. However,
the three elements, but the specific color observed this effect is at least partially due to the increase in
at any pixel only depicts elemental coincidence not the continuum background in the Ni-rich region
absolute or relative concentration. The particular color relative to the Al-rich region because of the sharp
that occurs at a given pixel depends on the complex difference in the average atomic number. For trace
physics of X-ray generation, propagation, and detec- constituents (C < 0.01, 1 wt %), the atomic number
tion as well as concentration, and the autoscaling of the dependence of the continuum background can
separate maps that precedes the color overlay, which dominate the observed contrast, creating artifacts in
distorts the apparent relationships among the elemental the images that render most trace constituent maps
constituents, also influences the observed colors. nearly useless.
24
24.2 · X-Ray Spectrum Imaging
417 24
Al Fe
Ni Al Fe Ni
20 µm
.. Fig. 24.3 Total intensity elemental maps for Al K-L2 (major), Fe K-L2,3 (minor), and Ni K-L2,3 (major) measured on a cross section of a Raney
nickel alloy, and the color overlay of the gray-scale maps
A complete spectrum is
recorded at each pixel.
y
rg
ne
yE
ra
AlK
X-
NiL
40
30 X NiKα
Counts
20
NiKβ
10
0
100 200 300 400 500 600 700 800 900 1000
AlK
150
100
Counts
50
0
100 200 300 400 500 600 700 800 900 1000
Channel (X-ray Energy)
Regardless of how the measured photons are cataloged, while the lower spectrum from a different pixel in an Al-rich
the XSI captures all possible elemental information about the region has a maximum of 150 counts. An alternative view of
region of the specimen being mapped within the limitations the datacube is shown in . Fig. 24.5, where the datacube can
set by the excitation energy (E0), the dose (beam current be thought of as a stack or deck of x-y image cards, where
multiplied by pixel dwell time, iB * τ), and the EDS spectrom- each x-y image corresponds to a different energy with an
eter performance (solid angle and efficiency). . Figure 24.4 energy range (“card thickness”) equal to the energy channel
shows the conceptual nature of the data. The [x, y, I(E)] data- width, for example, typically 5 eV or 10 eV. If a card is
cube can be thought of as an x-y array of spectra I(E). The selected from the stack with an energy that corresponds to a
individual pixel spectra can be selected by the analyst for characteristic peak, then the card shows the elemental map
inspection by specifying the x-y location. Depending on the for that peak, as shown in . Fig. 24.5. A card with an energy
electron beam current, the pixel dwell time, and the solid that corresponds to spectral background will have fewer
angle of the EDS detector, the counts in an individual pixel counts than a peak card, but such a background card may
spectrum may be low. For the example XSI of Raney nickel still have discernible microstructural information because of
shown in . Fig. 24.4, the upper spectrum taken from an Ni- the atomic number dependence of the X-ray continuum, as
rich area has a maximum of 40 counts per energy channel, shown in . Fig. 24.5.
24
24.2 · X-Ray Spectrum Imaging
419 24
50 µm
Y
y
rg
ne
yE
ra
X-
1 keV
.. Fig. 24.5 X-ray spectrum image considered as a stack of x, y images, each corresponding to a specific photon energy Ep
In the simplest case, by capturing all possible X-ray informa- SUM Spectrum
tion about an imaged region, the XSI permits the analyst to Simple but highly effective software tools that are present in
define regions-of-interest for total intensity mapping at any nearly all vendor XSI platforms as well as the open source soft-
time after collecting the map datacube, creating elemental ware (NIH ImageJ-Fiji and NIST Lispix) include those that
maps such as those in . Fig. 24.2. If additional information is calculate “derived spectra.” Derived spectra are constructed by
required about other elements not in the initial set-up, there systematically applying an algorithm to the datacube to
is no need to relocate the specimen area and repeat the map extract carefully defined information. The most basic derived
since the full spectrum has been captured in the XSI. spectrum is the “SUM” spectrum, illustrated schematically in
The challenge to the analyst is to make efficient use of the . Fig. 24.6. Conceptually, each energy “card” in the XSI data-
XSI datacube by discovering (“mining”) the useful informa- cube is selected and the counts in all pixels on that card are
tion that it contains. Software for aiding the analyst in the added together, as indicated by the systematic route through
interpretation of XSI datacubes ranges from simple tools in all pixels shown in . Fig. 24.6. This summed count value is
open source software to highly sophisticated, proprietary then placed in the corresponding energy bin of the SUM spec-
vendor software that utilizes statistical comparisons to auto- trum under construction, and the process is then repeated for
matically recognize spatial correlations among elements the next energy card until all energies have been considered.
present in the specimen region that was mapped (Kotula The resulting SUM spectrum has the familiar features of a
et al. 2003). conventional spectrum: characteristic X-ray peaks and the
420 Chapter 24 · Compositional Mapping
Sum Spectrum:
Add the counts from
every pixel in each
energy plane; the
SUM from a plane
becomes the intensity
in the corresponding
energy channel of the
derived spectrum.
y
rg
ne
yE
ra
X-
X
Al K-L2,3
3000000
2000000
Counts
Ni L-M
1000000 Ni K-L2,3
Ni K-M2,3
0
100 200 300 400 500 600 700 800 900 1000
Channel (X-ray Energy)
.. Fig. 24.6 Concept of the SUM spectrum derived from an X-ray spectrum image by adding the counts from all pixels on an individual image at
a specific photon energy, Ep
X-ray continuum background, as well as artifacts such as channel and view the corresponding energy card to reveal the
coincidence peaks and Si-escape peaks. The SUM spectrum in elemental map for that peak, as shown in . Fig. 24.7. By select-
. Fig. 24.6 was calculated from the same Raney nickel XSI ing a band of adjacent channels that spans the peak and aver-
datacube used for . Fig. 24.4. Note that the count axis now aging the counts for each x-y pixel in the set of energy cards,
extends to more than 3 million counts, much higher than the an elemental image with reduced noise is obtained, as also
count axes of the individual pixel spectra in . Fig. 24.4 because shown in . Fig. 24.7. Systematically selecting each of the
of the large number of pixels that have been added together to prominent characteristic X-ray peaks, total intensity maps for
construct the SUM spectrum. The characteristic peaks that all of the major elemental constituents can be obtained. The
can be recognized in the SUM spectrum (scaled to the highest SUM spectrum can be treated just like a normally recorded
intensity) represent the dominant, most abundant elemental spectrum. By expanding the vertical scale or changing from a
features contained in the XSI. The analyst can select a peak linear display to a logarithmic display, lower relative intensity
24
24.2 · X-Ray Spectrum Imaging
421 24
3000000
Counts
0 NiKh CuKh
100 200 300 400 500 600 700 800 900 1000
730 766
.. Fig. 24.7 Using the SUM spectrum to interrogate an X-ray spectrum image to find the dominant elemental peaks for the region being
mapped
peaks corresponding to minor and trace constituents can be is a needle you are looking for!) can be recognized with the
recognized, and the corresponding total-intensity region-of- “MAXIMUM PIXEL” derived spectrum (Bright and Newbury
interest images can be constructed, subject to the increas- 2004). As shown in . Fig. 24.8, the MAXIMUM PIXEL derived
ing influence of the continuum background incorporated in spectrum is calculated by making the same tour through all of
the window. For these lower abundance constituents, it is the pixels on each energy card as is done for the SUM spec-
necessary to use a band of energy channels to reduce the noise trum, but rather than adding the pixel contents, the algorithm
in the resulting elemental image, and such images are sub- now locates the maximum intensity within a card regardless of
ject to the artifacts that arise from the atomic number depen- what pixel it comes from to represent that energy value in the
dence of the X-ray continuum noted above. constructed MAXIMUM PIXEL spectrum. An example of the
application of the MAXIMUM PIXEL spectrum to an XSI is
MAXIMUM PIXEL Spectrum shown in . Fig. 24.9, where an unanticipated Cr peak is recog-
The SUM spectrum reveals dominant elemental features in the nized. When the Cr image is selected from the XSI, the Cr is
mapped region. Rare, unexpected elemental features, which in seen to be localized in a small cluster of pixels. Note that in the
the extreme case may occur at only a single pixel (i.e., looking plot of the SUM, LOG10SUM, and MAXIMUM PIXEL spectra
for a “needle-in-a-haystack” when you don’t even know that it in . Fig. 24.9, the Cr peak is only visible in the MAXIMUM
422 Chapter 24 · Compositional Mapping
gy
er
y En
ra
X-
X
AlK
200
150
Counts
100 NiL
SiK NiKα
50 ClKα CaKα CrKα CuKα
NiKβ
0
100 200 300 400 500 600 700 800 900 1000
PIXEL spectrum. Because Cr is represented at only a few pix- form are typical choices). Another approach is the creation
els, in the SUM spectrum the Cr intensity is overwhelmed by of a pixel mask by selecting all pixels within a particular
the continuum intensity at all of the other pixels, so that the Cr elemental map (or BSE or SE image) whose intensities fall
peak is not visible even in the logarithmic expansion shown in within a defined range above a specified gray level thresh-
the LOG10SUM plot. old. Pixels that satisfy this intensity criterion can occur any-
After the elemental maps have been created from the where in the x–y plane of the XSI and do not have to be
XSI, additional image processing tools can be brought to contiguous. Selections can range down to an individual
bear. Spatial regions-of-interest can be defined in the ele- pixel. An example of this process is shown in . Figure 24.10,
mental maps or in corresponding BSE or SE images where the Fe-rich phase has been selected from the Fe total
recorded simultaneously with the XSI. These spatial regions intensity map (shown in color overlay with Al and Ni) to
can be selected by simple image processing tools that define create a mask (shown as a binary image in the inset) from
a group of contiguous pixels that fall within a particular which the SUM spectrum representing the Fe-rich phase is
shape (square, rectangle, circle, ellipsoid, or closed free constructed.
24
24.2 · X-Ray Spectrum Imaging
423 24
Cr
10 m Al Cr Ni
AlK
1.0
0.9 NiL
NiKα
0.8
0.6 FeKα
Relative intensity
0.5
Log10 SUM
0.4
0.2
0.1 SUM
0.0
500 1000 1500
Sum Log sum Max running sum log running sum running max
.. Fig. 24.9 Use of the MAXIMUM PIXEL spectrum to identify and locate an unexpected Cr-rich inclusion in Raney nickel alloy
424 Chapter 24 · Compositional Mapping
Al
1000000 FINAL_-1900u-Mask_
NiL highFe_phase
Ni
Al + Al
Al + NiL
100000 Fe
Log10 Counts
10000
Al Fe Ni
1000
100
0 1 2 3 4 5 6 7 8 9 10
20 µm
50000 FINAL_-1900u-Mask_
highFe_phase
45000
Al + Al
40000
35000
Al + NiL
30000
Counts
25000 Fe mask
20000
15000
10000
5000
0
0 1 2 3 4 5 6 7 8 9 10
Photon Energy (keV)
.. Fig. 24.10 Creation of a pixel mask (inset) using the Fe elemental map for Raney nickel to select pixels from which a SUM spectrum is constructed
for the Fe-rich phase
24.3 Quantitative Compositional Mapping elemental intensities are then quantified by a “standardless
analysis” procedure to calculate the individual pixel concen-
The XSI contains the complete EDS spectrum at each pixel (or tration values. Alternatively, rigorous standards-based quan-
the position-tagged photon database that can be used to titative analysis can be performed on the pixel-level intensity
reconstruct the individual pixel spectra). Quantitative com- data with NIST DTSA II by utilizing the scripting language to
positional mapping implements the normal fixed measure- sequentially calculate the pixel spectra in a datacube. MLLS
ment location quantitative analysis procedure at every pixel in peak fitting to extract the characteristic intensities, k-ratio
a map. Qualitative analysis is first performed to identify the calculation relative to a library of measured standards, and
peaks in the SUM and MAXIMUM PIXEL spectra of the XSI matrix corrections to yield the local concentrations of these
to determine the suite of elements present within the mapped elements for each pixel. An example of the NIST DTSA-II
area, noting that all elements are unlikely to be present at procedure applied to an XSI measured on the manganese
every pixel. The pixel-level EDS spectra can then be individu- nodule example of . Fig. 24.2 is presented in . Fig. 24.11,
ally processed following the same quantitative analysis proto- where the Fe K-L2,3 total intensity map, which suffers signifi-
col used for individually measured spectra, ultimately cant interference from Mn K-M2,3, shows a change in the
replacing the gray or color scale in total intensity elemental apparent level of Fe in the center of the image after quantita-
maps, which are based on raw X-ray intensities and which are tive correction, as well as changes in several finer-scale details.
nearly impossible to compare, with a gray or color scale based Quantitative compositional maps for major constituents
on the calculated concentrations, which can be sensibly com- displayed as gray-scale images are virtually identical to raw
pared. Most vendor compositional mapping software extracts intensity elemental maps, as shown for the major constitu-
characteristic peak intensities by applying multiple linear ents Al and Ni in . Figs. 24.12 and 24.13. Significant differ-
24 least squares (MLLS) peak fitting or alternatively fits a back- ences between raw intensity maps and compositional maps
ground model under the peaks. These extracted pixel-level are found for minor and trace constituents, where correction
24.3 · Quantitative Compositional Mapping
425 24
.. Fig. 24.11 Quantitative
compositional mapping with
DTSA-II on the XSI of a deep-
sea manganese nodule from
. Fig. 24.2: direct comparison of
the total intensity map for Fe and
the quantitative compositional
map of Fe; note the decrease in
the apparent Fe in the high Mn
portion in the center of the image
after quantification and local
changes indicated by arrows (yel- Raw Fe K-L2,3 + Mn K-M2,3
low shows extension of high Fe
region; magenta shows elimina-
tion of dark features)
20 µm
for the continuum background significantly changes the of autoscaling noted above still apply. Various pseudo-color
gray-scale image. As shown for the minor Fe constituent in scales, in which bands of contrasting colors are applied to
. Figs. 24.12 and 24.13, the false Fe contrast between the Ni- the underlying data, are typically available in image pro-
rich and Al-rich phases that arises due to the atomic number cessing software. Such pseudo-color scale can partially
dependence of the X-ray continuum is eliminated in the Fe overcome the display limitations of gray-scale presentation,
compositional map. but the resulting images are often difficult to interpret. An
While rendering the underlying pixel data of quantita- example of a five band pseudo-color scale applied to the
tive compositional maps in gray scale is a useful starting compositional maps using NIST Lispix is shown in
point, the problem of achieving a quantitatively meaningful . Fig. 24.15. An effective display scheme for quantitative
display remains. Because of autoscaling, the gray scales of elemental maps that enables a viewer to readily compare
the Al, Ni, and Fe maps in . Fig. 24.12 do not have the same concentrations of different elements spanning major, minor
numerical meaning. The Al and Ni concentrations locally and trace ranges can be achieved with the Logarithmic
reach high enough levels to correspond to brighter gray Three-Band Encoding (Newbury and Bright 1999). A band
levels and have sufficient range to create strong contrast of colors is assigned to each decade of the concentration
with direct gray-scale encoding, as seen in . Figs. 24.12 and range with the following characteristics:
24.13. The Fe constituent which is present at a concentra- Major: C > 0.1 to 1 (mass fraction) deep red to red pastel
tion with a maximum of approximately 0.04 mass fraction Minor: 0.01 ≤ C ≤ 0.1 deep green to green pastel
(4 wt %) never exceeds the minor constituent range, so the Trace: 0.001 ≤ C < 0.01 deep blue to blue pastel
Fe map appears dark with little contrast in the quantita-
tive map of . Fig. 24.13. Autoscaling of the Fe-map in The quantitative elemental maps are displayed with
. Fig. 24.14 improves the contrast, but the same limitations Logarithmic Three-Band Encoding in . Fig. 24.16, and the
426 Chapter 24 · Compositional Mapping
Al Fe
20 µm
Ni Al Fe Ni
.. Fig. 24.12 Raney nickel alloy XSI: total intensity maps for Al, Fe, and Ni with color overlay
Al Fe
20 µm
Ni Al Fe Ni
24 .. Fig. 24.13 Quantitative compositional maps of the Raney nickel alloy XSI in 24.13; note low contrast in the Fe image presented without
autoscaling
24.3 · Quantitative Compositional Mapping
427 24
Al Fe
20 µm
Ni Al Fe Ni
.. Fig. 24.14 Quantitative compositional maps Raney nickel alloy XSI with contrast enhancement of the Fe map
Al Fe
20 µm
Ni
.. Fig. 24.15 Five band pseudo-color presentation of the quantitative compositional maps for Al, Fe, and Ni derived from the Raney nickel
alloy XSI
428 Chapter 24 · Compositional Mapping
major-minor-trace spatial relationships among the elemental from the atomic-number dependence of the X-ray continuum
constituents are readily discernible. seen in the raw Fe intensity image, as shown in . Fig. 24.17.
The Logarithmic Three-Band Encoding of the Fe compo- The Logarithmic Three-Band Encoding shows that this appar-
sitional map in . Fig. 24.16 reveals apparent Fe contrast in the ent contrast occurs near the lower limit of the trace range. Is
Al-rich and Fe-rich phases, and this contrast differs markedly this trace Fe contrast meaningful or merely an uncorrected
BSE Al
Ni Fe
20 µm
0.001 0.01 0.1 1.0
0.1 1.0 10 100 wt%
.. Fig. 24.16 Logarithmic Three-Band Color encoding of the quantitative compositional maps for Al, Fe, and Ni derived from the Raney nickel alloy XSI
Fe Fe
20 µm
Fe raw intensity
elemental map
0.001 0.01 0.1 1.0
0.1 1.0 10 100 wt%
Fe compositional map
24 .. Fig. 24.17 Direct comparison of the Fe total intensity map and the Logarithmic Three-Band Encoding of the Fe quantitative compositional
map. Note the distinct changes in the contrast within the trace concentration (C < 0.01) regions of the image
24.3 · Quantitative Compositional Mapping
429 24
Raney1-Mask_
Al
Fe-rich phase highFe_phase
1000000
Ni
100000
Ni
Counts (Log)
Fe
O
10000
Mn
Cr
Cu
Zn
1000
Ni
Ni
Fe
Cu
Zn
Mn
Cr
Al
100
Ni
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
Al
Raney1-Mask_
Ni
Ni
Fe
18000
Al+Al
16000
Counts
14000
Al+NiL
12000
O
10000 Cr Mn
8000
6000
Cr
Ni
Mn
Cu
4000
Zn
Ni
Fe
2000
Mn
Cr
Cu
Zn
Al
Ni
0
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
.. Fig. 24.18 Raney nickel alloy XSI: mask of pixels corresponding to the Fe-rich phase and the corresponding SUM spectrum; the Fe peak corresponds
to C = 0.044 mass fraction
artifact? To examine this question, the analyst can use pixel constituent, with its highest level in the intermediate-Ni phase
masks of each phase selected from the Ni compositional map where CFe = 0.0027 (2700 ppm), falling to CFe = 0.00038
(or the BSE image) to obtain the SUM spectrum, as shown in (380 ppm) in the high-Ni phase and to CFe = 0.00030
. Figs. 24.18 (Fe-rich phase), 24.19 (Al-rich phase), 24.20 (300 ppm) in the high-Al phase. Additionally, trace Cr at a
(Ni-intermediate phase), and 24.21 (Ni-rich phase). These similar concentration is found in the Al-rich phase along with
SUM spectra confirm that Fe is indeed present as a trace the trace Fe.
430 Chapter 24 · Compositional Mapping
Raney1-Mask_ High
Al
Al-LowNi_phase
1000000
Counts (Log)
Ni
O
Ni
100000
Cr
Fe
Cu
Zn
10000
Ni
Ni
Cu
Zn
Cr
Fe
Al
1000
Ni
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
160000 Ni Raney1-Mask_High
Ni
O
Al
Al-LowNi_phase
Ni
140000
Al
120000
100000 Cr
Al+Al
Counts
Fe = 0.00030
80000 Cr
Al+NiL
Fe
60000
Cu
Zn
40000
Ni
Ni
20000
Cr
Fe
Cu
Zn
Al
Ni
0
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
.. Fig. 24.19 Raney nickel alloy XSI: mask of pixels corresponding to the Al-rich phase and the corresponding SUM spectrum; note the low level
peaks for Fe and Cr; the Fe-peak corresponds to C = 0.00030 = 300 parts per million
24
24.4 · Strategy for XSI Elemental Mapping Data Collection
431 24
Raney1-Mask_
Al
Intermediate Ni-rich phase Intermediate_Ni_phase
Ni
10000000
Ni
1000000
Counts (Log)
Fe
100000
Al+Al
Al+NiL
10000
Ni
Ni
Fe
Al
Ni
1000
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
Ni
240000 Raney1-Mask_
Al
O
Ni
Ni
220000
Cr Fe, 0.0027 Intermediate_Ni_phase
Al
200000
Fe
Cr
180000
Mn
160000
Cu
140000
Zn
Ni si-escape
Counts
120000
100000
80000
60000
Ni
40000
Ni
Fe
Cr
Cu
Mn
Zn
20000
Al
Ni
0
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
.. Fig. 24.20 Raney nickel alloy XSI: mask of pixels corresponding to the intermediate Ni-rich phase and the corresponding SUM spectrum; note the
low level peak for Fe; the Fe-peak corresponds to C = 0.0027 = 2700 parts per million
a cceptable dead-time will depend on the level of spectral arti- coincidence artifacts, which scale with dead-time, must be
facts that the analyst is willing to accept: (1) If the mapping considered. Coincidence peaks for Al + NiL and Al + Al, as
software collects the XSI at constant pixel dwell time without well as the coincidence continuum between these peaks, are
dead-time correction, the resulting map can be subject to illustrated in . Figs. 24.18, 24.19, 24.20, and 24.21. This por-
severe artifacts if the dead-time is so high that the peak of the tion of the EDS spectrum contains K-shell peaks for S and Cl,
OCR versus ICR response is exceeded at some pixels during L-shell peaks for Mo, Tc, Ru, Rh, Pd, and Ag, and M-shell
mapping. As shown in . Fig. 24.23, operating at very high peaks for Hg, Tl, Pb, and Bi. If none of these elements is of
dead-time can result in the same OCR being produced by two interest at the minor or trace level, then the Al + NiL and
widely different ICR values, which may effectively correspond Al + Al coincidence can be ignored and the advantages of
to two different concentrations. Artifacts produced by this high throughput realized for the other elements of interest.
effect are shown in the Al elemental maps recorded at high However, if this spectral region is required for one or more of
dead-time shown in . Fig. 24.24. Dead-time-corrected data these elements, a lower dead-time should be selected by
collection in elemental mapping avoids this artifact. (2) If the reducing the beam current to reduce coincidence, which
analyst is interested in minor and/or trace level constituents, depends strongly on the input count rate.
432 Chapter 24 · Compositional Mapping
Raney1-Mask_
10000000
Al
high_Ni_phase
Ni
High Ni-rich phase
Ni
1000000 O
Counts (Log)
Cr
Cu
100000
Fe
Zn
Al+Al
Al+NiL
10000
Ni
Ni
Cu
Zn
Cr
Fe
Al
Ni
1000
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
Raney1-Mask_
Ni
Ni
Al
O
130000
high_Ni_phase
120000
Al Cr Ni
110000
100000 Fe = 0.00038
Cr
Mn
90000
Cu
Fe
80000
Counts
70000
Ni si-escape
Zn
60000
50000
40000
Ni
30000
Ni
20000
Cu
Cr
Fe
Mn
Zn
10000
Al
Ni
0
0 1 2 3 4 5 6 7 8 9 10
Energy (keV)
.. Fig. 24.21 Raney nickel alloy XSI: mask of pixels corresponding to the high Ni-rich phase and the corresponding SUM spectrum; note the low
level peak for Fe; the Fe-peak corresponds to C = 0.00038 = 380 parts per million
24.4.2 Choosing the Pixel Density the empty area of the pixel could be “filled in” by increasing
the number of pixels to reduce the pixel size, but this would
The size of the scanned area, the number of image pixels (nX, lead to extremely large XSI data structures that would require
nY) and the pixel dwell time, τ, are critical parameters that very long accumulation times. A practical upper limit for
the analyst must choose when defining an XSI data collec- XSI mapping is typically 1024 × 1024 pixels, which for a
tion. An estimate of the size of the lateral extent of the X-ray spectrum of 4096 channels of 2 bytes intensity depth would
interaction volume, obtained either from the Kanaya– produce an XSI of 8 Gbytes in the uncompressed RAW for-
Okayama X-ray range or from Monte Carlo simulation, is mat. To reduce the mass storage as well as the subsequent
also useful, especially for small-area, high-magnification processing time for quantitative compositional mapping cal-
mapping. Choosing the size of the scanned area (magnifica- culations, the analyst may choose 512 × 512 or 256 × 256
tion) depends on the lateral extent of the specimen features pixel scan fields, especially when a small area is scanned
that are the objective of the mapping measurement. As is the (high magnification operation) leading to overlapping pix-
case with SEM imaging using BSEs and/or SEs, when large els. A pixel overlap of approximately 25 % serves to fill all
areas are being scanned (low magnification operation), the space in the square pixels, but further oversampling provides
pixel size may be greater than the lateral extent of the X-ray no additional information, so that the analyst would be bet-
source size, which is a convolution of the incident beam size ter served by lowering the magnification to cover more
and the interaction volume for X-ray production, so that specimen area with the chosen pixel density, or alternatively,
much of the pixel area is effectively unsampled. In principle, choose a lower pixel density.
24
24.4 · Strategy for XSI Elemental Mapping Data Collection
433 24
.. Fig. 24.22 Throughput of an
SDD-EDS system consisting of SDD-EDS Throughput
four 10-mm2 detectors with the Combined output
outputs summed at three differ- 1.4 106
of four 10-mm2 detectors
ent operating time constants Ideal no deadtime
response
1.2 106
145 eV
> 1.2 MHz
0.0 100
0 100 1 106 2 106 3 106 4 106 5 106
Input Count Rate (X-rays/second)
3e+5
2e+5
1e+5
0
0.0 5.0e+5 1.0e+6 1.5e+6 2.0e+6 2.5e+6
BSE 5% deadtime
Al
Al Al
.. Fig. 24.24 Al map in Raney nickel recorded at 5 %, 60 %, and 80 % dead-time; note changes in the high Al region at 80 % dead-time compared
to 5 % dead-time
24.4.3 Choosing the Pixel Dwell Time An XSI was recorded with 640 by 480 pixels with a 64-μs pixel
dwell at on OCR of 750 kHz for a total mapping time of 20 s.
Once the analyst has chosen the pixel density, the product of The raw intensity maps for Cu, Zn, and Pb and their color
the number of pixels in a map and the pixel dwell time gives overlay are shown in . Fig. 24.25b. While very noisy on the
the total mapping time. The OCR of the EDS and the pixel single pixel level, these maps nevertheless reveal the localiza-
dwell time determine the number of counts in the individual tion of the lead corresponding as expected to the bright region
pixel spectra, which sets the ultimate limit on the composi- in the SEM-BSE image. The color overlay, however, shows
tional information that can be subsequently recovered from numerous dark areas within the particle which do not corre-
the XSI. The high throughput of SDD-EDS, especially when spond to Cu, Zn, or Pb. With this short pixel dwell, the pixel
clusters of detectors are used, provides OCR of 105/s to 106/s, level EDS spectra contain only about 50 counts total, the effect
enabling various XSI imaging strategies determined by the of which can be seen in the noisy derived MAXIMUM PIXEL
number of counts in the individual pixel spectra. spectrum in . Fig. 24.26. (An additional derived spectrum,
the RUNNING MAXIMUM PIXEL, which is averaged over
“Flash Mapping” three adjacent energy “cards” to reduce the effect of the low
By operating with a high OCR, major constituents can be count, is also shown.) The SUM spectrum shown in
mapped in less than 60 s, a mode of operation that can be . Fig. 24.26, consisting of all counts recorded in 20 s, approxi-
termed “flash” mapping, which is useful for surveying mately 15 million, contains abundant information. In addition
unknowns. An example of a flash mapping survey of a leaded- to the peaks for Cu, Zn,and Pb, a major peak for Ni is observed.
brass particle is shown in . Fig. 24.25, where the SEM-BSE When the raw elemental map for Ni is constructed from the
image in . Fig. 24.25a reveals a high-atomic-number inclusion. XSI, the missing regions in . Fig. 24.25b are filled in, as shown
24
24.4 · Strategy for XSI Elemental Mapping Data Collection
435 24
Leaded brass particle
a BSE (solid state, summed)
10 µm
b Cu
Pb
10 µm
Zn Cu Pb Zn
What is missing?!
.. Fig. 24.25 Leaded brass particle XSI: a SEM-BSE image; note bright inclusion; b XSI recorded with 640 × 480 pixels; 64 μs/pixel = ~ 20 s total
scan and total intensity images for Pb, Cu and Zn with color overlay; note apparently missing regions of particle (dark)
436 Chapter 24 · Compositional Mapping
Ni,Cu,ZnL
PIXEL spectrum, and RUNNING 1.0
MAXIMUM PIXEL spectrum (aver-
0.9
CuKa
aged over three consecutive
NiKa
energy “cards”); note unexpected 0.8
ZnKa
PbM
Ni peak
CuKb
0.7
ZnKb
PbLa
Relative Intensity
0.6
PbLb
0.5 LOG10 SUM
MAXIMUM PIXEL
0.4
0.3
RUNNING MAX (3 card average)
0.2
0.1
SUM
0.0
500 1000 1500 2000
Sum Log sum Max log running sum running sum running max
Photon Energy (10-eV Channel number)
in . Fig. 24.27, which also shows the color overlap of Cu, Zn, the examples in . Figs. 24.25, 24.26, and 24.27. Of course, by
and Ni. This example demonstrates the value of XSI imaging accumulating more counts above this threshold, progres-
to augment the information obtained from the atomic number sively lower concentration contrast details can be revealed
contrast of the SEM-BSE image. In . Fig. 24.27b, the SEM- for the major constituents. For problems involving minor
BSE image easily distinguishes the Pb-rich inclusion from the and/or trace constituents, longer pixel dwell times are nec-
brass matrix but shows no distinct contrast from the Ni-rich essary to accumulate at least 500–5000 counts per pixel
regions relative to the Cu-Zn brass matrix. Ni, Cu, and Zn are spectrum, and the beam current should be reduced to keep
only separated by one unit of atomic number, so that the BSE the dead-time generally below 10 % to minimize
atomic number contrast between these phases is very weak coincidence peaks. This dead-time condition can be relaxed
and dominated by the contrast produced by the Pb-rich region if the coincidence peaks, which are only produced by high
relative to the brass matrix. The SEM-BSE contrast situation is count rate parent peaks associated with major constituents,
further complicated by the topographic contrast of the com- do not interfere with the minor/trace constituent peaks of
plex surface of the particle. Element-specific compositional interest.
imaging reveals the details of the complex microstructure of An example of the compositional details that can be
this particle. observed at the level of approximately 5000 counts per pixel
spectrum is shown for a complex Zr-Ni-V alloy with minor
High Count Mapping Ti, Cr, Mn and Co in . Fig. 24.28. Excellent gray-scale (after
The strategy for elemental mapping data collection depends autoscaling) contrast is obtained between the phases which
on the nature of the problem to be solved: the most critical have relatively small changes in composition for the individ-
question is typically, “What concentration levels are of ual elements. Although the single pixel spectra do not have
interest?” If minor and trace level constituents are not adequate counts for robust quantification, the analyst can use
important, then short duration (60 s or less), low pixel den- the images to form pixel masks that contain much higher
sity (256 × 256 or fewer) XSI maps with a high OCR will total counts. The compositional values noted in . Fig. 24.28
usually contain adequate counts, a minimum of approxi- are based on quantifying SUM spectra taken from the two
mately 50–500 counts per pixel spectrum, depending on phases that are specified by the arrows in the elemental maps,
the particular elements and overvoltage, to discern concen- and the very small statistical error reported reflects the very
tration-based contrast for major constituents, as shown in high count SUM spectra.
24
24.4 · Strategy for XSI Elemental Mapping Data Collection
437 24
a
Cu Ni
10 µm
Zn Cu Ni Zn
b BSE Pb
10 µm
Cu Ni
.. Fig. 24.27 Leaded brass particle XSI: a total intensity images for and Zn total intensity maps; the SEM-BSE image is insensitive to the
Cu, Zn, and Ni with color overlay showing Ni filling in empty areas seen compositional contrast between the Cu-Zn regions and the Ni-rich
in . Fig. 24.25b; b direct comparison of SEM-BSE image and Cu, Ni, regions
438 Chapter 24 · Compositional Mapping
b Co Ti
50 µm
Mn Cr
24
References
439 24
References Kotula P, Keenan MR, Joseph R, Michael JR (2003) Automated analysis of SEM
X-ray spectral images: a powerful new microanalysis tool. Microsc Micro-
anal 9:1
Bright D (2017) NIST Lispix, a software engine for image processing,
Newbury D, Bright D (1999) Logarithmic 3-band color encoding: Robust
available free at: 7 http://www.nist.gov/lispix/doc/contents.htm
method for display and comparison of compositional maps in elec-
Bright D, Newbury D (2004) Maximum pixel spectrum: a new tool for
tron probe X-ray microanalysis. Microsc Microanal 5:333
detecting and recovering rare, unanticipated features from spec-
Newbury D, Ritchie N (2013) Elemental mapping of microstructures by
trum image data cubes. J Microsc 216:186
scanning electron microscopy –energy dispersive X-ray spectrom-
Gorlen KE, Barden LK, Del Priore JS, Fiori CE, Gibson CC, Leapman RD
etry (SEM-EDS): extraordinary advances with the Silicon Drift
(1984) Computerized analytical electron microscope for elemental
Detector (SDD). J Anal At Spectrom 28:973
mapping. Rev Sci Instrum 55:912
441 25
Attempting Electron-Excited
X-Ray Microanalysis
in the Variable Pressure
Scanning Electron Microscope
(VPSEM)
25.1 Gas Scattering Effects in the VPSEM – 442
25.1.1 Why Doesn’t the EDS Collimator Exclude the
Remote Skirt X-Rays? – 446
25.1.2 Other Artifacts Observed in VPSEM X-Ray Spectrometry – 448
References – 459
While X-ray analysis can be performed in the Variable Pres- ressures of 2500 Pa are possible, permitting liquid water
p
25 sure Scanning Electron Microscope (VPSEM), it is not pos- to be maintained in equilibrium when the specimen is
sible to perform uncompromised electron-excited X-ray cooled to ~ 3 °C. Such specimen chamber pressures are sev-
microanalysis. The measured EDS spectrum is inevitably eral orders of magnitude higher than that of a conventional
degraded by the effects of electron scattering with the atoms high-vacuum SEM, which typically operates at 10−2 Pa to
of the environmental gas in the specimen chamber before the 10−4 Pa or lower. As the beam emerges from the high vac-
beam reaches the specimen. The spectrum is always a com- uum of the electron column through the final aperture into
posite of X-rays generated by the unscattered electrons that the elevated pressure of the specimen chamber, elastic scat-
remain in the focused beam and strike the intended target tering events with the gas atoms begin to occur. Although
mixed with X-rays generated by the gas-scattered electrons the volume density of the gas atoms in the chamber is very
that land elsewhere, micrometers to millimeters from the low compared to the density of a solid material, the path
microscopic target of interest. length that the beam electrons must travel typically ranges
It is critical to understand how severely the measured from 1 mm to 10 mm or more before reaching the speci-
spectrum is compromised, what strategies can be followed to men. As illustrated in . Fig. 25.1, when elastic scattering
minimize these effects, and what “workarounds” can be occurs along this path, the angular deviation causes beam
applied in special circumstances to solve practical problems. electrons to substantially deviate out of the focused beam
The impact of gas-scattered electrons on the legitimacy of the creating a “skirt.” The unscattered beam electrons follow
analysis depends on the exact circumstances of the VPSEM the expected path defined by the objective lens field and
conditions (beam energy, gas species, path length through land in the focused beam footprint identical to the situa-
the gas) and the characteristics of the specimen and its sur- tion at high vacuum but with reduced intensity due to the
roundings. Gas scattering effects increase in significance as gas scattering events that rob the beam of electrons. The
the constituent(s) of interest range in concentration from electrons that remain in the beam behave exactly as they
major (concentration C > 0.1 mass fraction) to minor would in a high vacuum SEM, creating the same interac-
(0.01 ≤ C ≤ 0.1) to trace (C < 0.01). tion volume and generating X-rays with exactly the same
spatial distribution to produce identically the same spec-
trum. This “ideal” high vacuum equivalent spectrum rep-
25.1 Gas Scattering Effects in the VPSEM resents the true microanalysis condition. However, this
ideal spectrum is degraded by the remotely scattered elec-
The VPSEM allows operation with elevated gas pressure in trons in the skirt which generate characteristic and con-
the specimen chamber, typically 10 Pa to 1000 Pa but even tinuum (bremsstrahlung) X-rays from whatever material(s)
higher in the “environmental SEM” (ESEM), where they strike.
Unscattered
beam electron
Elastic
scattering
EDS
event with
gas atom
Limits of
EDS collimator
Skirt
1 Pa
10 Pa
80 100 Pa
1000 Pa
2500 Pa
Skirt radius (micrometers)
60
40
30
20
0
0 2 4 6 8 10
Gas path length (mm)
444 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
.. Table 25.1 Portion of the electron scattering table for VPSEM simulation. Note the unscattered fraction of the 20 keV beam, 0.701
(100 Pa, 5-mm gas path length, oxygen). The full table extends to 1000 μm
Ring Inner radius, Outer radius, Ring area, Electron count Electron Cumulative, %
μm μm μm2 fraction
0.7
0.6
3 mm GPL
5 mm GPL
0.5
10 mm GPL
0.4
0 10 20 30 40 50
Radial distance from beam center (micrometers)
0.8
0.7
0.6
3 mm GPL
0.5 5 mm GPL
10 mm GPL
0.4
0 200 400 600 800 1000
Radial distance from beam center (micrometers)
100 Pa of oxygen requires a radial distance of approximately The extent of the degradation of the measured EDS spec-
190 μm, as shown in . Fig. 25.5, which plots the skirt distribu- trum by gas scattering is illustrated in the experiment shown
tion out to 1000 μm (1 mm). The strong effect of the gas path in . Fig. 25.6. The incident beam is placed at the center of a
length on the skirt radius, which follows a 3/2 exponent in the polished cross section of a 40 wt % Cu – 60 wt % Au alloy
scattering Eq. 25.1, can be seen in . Fig. 25.5 in the plots for wire 500 μm in diameter surrounded by a 2.5-cm-diameter
3 mm, 5 mm, and 10 mm gas path lengths. Al disk. For a beam energy of 20 keV and a gas path length of
446 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
25 80768 Ch#: 170 ChkV: 1.7000 Work: 6120 Results: 3172 Marker: Si 14
AlKα AuMα 500 µm 40Cu-60Au
Spec# Counts
1 6602
2 8435
3 4632
4 4936 25 mm AI
5 4330
6 4685
CuKβ1
0 Kα FeKα
AuLι Α
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
.. Fig. 25.6 EDS spectra measured with the beam placed in the center of a 500 μm diameter wire of 40 wt % Cu–60 wt %Au surrounded by a
2.5-cm-diameter Al disk; E0 = 20 keV; gas path length = 4 mm; oxygen at various pressures
4 mm through water vapor, the EDS spectra measured over a 25.1.1 hy Doesn’t the EDS Collimator
W
pressure range from 53 Pa to 1600 Pa are superimposed, Exclude the Remote Skirt X-Rays?
showing the in-growth of the Al peak with increasing pres-
sure. Even at 53 Pa, a detectable Al peak is observed, despite Gas scattering in the VPSEM mode always degrades the inci-
the beam center being 250 μm away from the Al. As the pres- dent beam, transferring a significant fraction of the beam elec-
sure is increased, the Al peak ranges from an apparent trace trons into the skirt. The radius of the skirt can reach a
to minor and finally major constituent peak. millimeter or more from the focused beam impact. It might be
DTSA-II also simulates the composite spectrum created thought that the EDS collimator would restrict the acceptance
by these two classes of electrons as they strike the speci- area of the EDS to exclude most of the skirt. As shown in the
men. Various configurations of two different materials can schematic diagram in . Fig. 25.8, while a simple collimator
be specified, one that the unscattered beam strikes, for acts to successfully shield the EDS from accepting X-rays pro-
example, a particle, and the other by the skirt electrons, for duced by backscattered electrons striking the lens and cham-
example, the surrounding matrix. . Figure 25.7 shows ber walls, the acceptance volume near the column optic axis is
spectra simulated for the example of . Fig. 25.6, the quite large. The EDS acceptance is not defined by looking back
500-μm-diameter 40 wt %Cu–60 wt %Au wire in the Al at the detector from the specimen space as the cone of rays
disk with a 4-mm-gas path length through water vapor. The whose apex is at the beam impact on the specimen and whose
simulation of the lowest VPSEM gas pressure of 53 Pa pro- base is the detector active area (the dashed red lines in
duces a low level Al peak similar to the experimental mea- . Fig. 25.8). While the red lines define the solid angle of the
surement. Thus, even at this low pressure and short gas detector for emission from the beam impact point, the accep-
path length for which 89 % of the electrons remain in the tance region is actually defined by looking from the detector
focused beam, there are still gas-scattered electrons falling through the collimator at the specimen space (the dashed
at least 250 μm from the beam impact. As the pressure is green lines in . Fig. 25.8). The true area of acceptance can be
progressively increased, the in-growth of the Al peak due to readily determined by conducting X-ray mapping measure-
the skirt electrons is well modeled by the Monte Carlo ments. . Fig. 25.9 shows a series of measurements of X-ray
simulation. maps of a machined Al disk taken at the lowest available
25.1 · Gas Scattering Effects in the VPSEM
447 25
E0 = 20 keV
12000
10000 4 mm gas path length
Counts
8000 H2O
6000
53 Pa H2O
4000
High vacuum
2000
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Photon energy (eV)
High_vacum_40Cu-60Au_20keV
1600 Pa H2O 1200Pa_H2O_40Cu-60Au_20keV
1600Pa_H2O_40Cu-60Au_20keV
133Pa_H2O_40Cu-60Au_20keV
800Pa_H2O_40Cu-60Au_20keV
800 Pa H2O
400 Pa H2O
133 Pa H2O
Scaled to AuM
53 Pa H2O
High vacuum
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Photon energy (eV)
.. Fig. 25.7 DTSA-II Monte Carlo simulations of the specimen and gas scattering conditions of . Fig. 25.6. Upper plot: high vacuum and 53 Pa
(4 mm GPL, H2O). Lower plot: high vacuum and various pressures from 53 Pa to 1600 Pa
Window
BSE
BSE
Collimator &
electron trap
Specimen X-rays
Several mm
Green =
Extent of
Several mm specimen
X-ray sources
NOT excluded
by collimator
448 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
25 1200
Al counts/pixel
1000
800
600
400
200
0
8 10 12 14 16 18 20 22
Working distance (mm)
10 mm 12 mm 14 mm
16 mm 18 mm 20 mm
1 mm
Al machined surface
0 10 20 30 40 50 60 70 80 90 100
.. Fig. 25.9 Collimator acceptance volume as determined by intensity measured at the center of the scan field as a function of
mapping an Al disk at various working distances (10–20 mm) at the working distance
widest scanned field (i.e., lowest magnification); the plot shows the
C Ka
Gas = H2O
H2O
Space# Counts
6 mm gas path 1 788
2 924
C disk, 25 mm diam 3 1523
4 2129
5 2753 EDS
2000 Pa (15 torr)
6 3451
1600 Pa (12 torr)
1200 Pa (9 torr)
O Ka
800 Pa (6 torr)
400 Pa (3 torr)
266 Pa (2 torr)
Carbon
133 Pa (1 torr)
53 Pa (0.4 torr)
.. Fig. 25.10 Generation of O K X-rays from the environmental gas as a function of chamber pressure
remain within the 50 μm diameter glass shard, as evidenced If the environmental gas can contribute to the spec-
by the negligible C intensity in the spectrum. An O peak is trum, can the gas also absorb X-rays from the specimen?
also observed, at least some of which is actually from the Because of the low gas density, this effect might be expected
specimen. When the pressure is increased by a factor of five to be negligible, and as listed in . Table 25.2, which is cal-
to 1330 Pa (10 torr), the C intensity rises significantly culated for a 40 –mm-path through the gas from the X-ray
because the skirt now extends beyond the boundary of the source at the beam impact on the specimen to the EDS, for
particle, and the O peak intensity increases by a factor of the lowest pressure considered, 10 Pa, over 99 % of the
four, all of which is due to the environmental gas. It is also X-rays of all energies leaving the specimen in the direction
worth noting that in addition to characteristic X-rays from of the detector arrive there, even for F K, the energy of
the environmental gas, there is increased bremsstrahlung which, 0.677 keV, is just above the O K-shell absorption
generation as well from the inelastic scattering of beam and energy, 0.535 keV, which results in a large mass absorption
backscattered electrons with the gas atoms. In . Fig. 25.11, coefficient. When the pressure is increased to 100 Pa, the
the background is substantially higher for the spectrum loss of F K due to absorption increases to ~ 6 %, and at a
measured at the elevated pressure, leading to a reduced pressure of 2500 Pa (18.8 torr), ~ 80 % of the F K radiation
peak-to-background, which is easily seen for the Zn L-family is lost to gas absorption, and even Al K-L2 suffers a 20 %
and Al K-L2 peaks, an effect which makes for poorer limits of loss in intensity compared to the conventional high vac-
detection. uum SEM situation.
450 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
25
Extra C from
skirt on substrate
Extra O from environmental H2O
5741 #
Ch : 147 Ch kV: 1.4700 Work: 969 Results: 1506 Marker: Pb B2
E0 = 20 keV O 0.245
3 mm GPL Al 0.0265
H 2O
Si 0.140
266 Pa (2 torr)
Zn 0.0402
1330 Pa (10 torr) 50 mm chip
Ba 0.0896
Ta 0.0409
Pb 0.418
Al Ba L
Extra continuum
due to gas scattering
Zn L
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
Photon energy (keV)
.. Fig. 25.11 Modification of the measured X-ray spectrum by gas scattering, including in-growth of the O K peak from contributions of the
environmental gas as well as increased background due to increased bremsstrahlung created by the gas scattering
0.9
0.7
0.6
0.5
0.4
0 10 20 30 40 50
Radial distance from beam center (micrometers)
4. T, the sample chamber temperature (K): The scattering Major constituent: concentration C > 0.1 mass fraction
skirt is reduced by operating at the lowest possible tem- Minor constituent: 0.01 ≤ C ≤ 0.1
perature. Trace constituent: C < 0.01
5. L, the gas path length (m). The shorter the gas path
length, the smaller the gas scattering skirt, as shown in Depending on the exact nature of the specimen, gas scat-
. Fig. 25.4, where the skirt is compared for gas path tering will almost always introduce spectral artifacts at the
lengths of 3 mm, 5 mm, and 10 mm. Note that the gas trace and minor constituent levels, and in severe cases arti-
path length appears in Eq. (25.1) with a 3/2 power, so facts will appear at the level of apparent major constitu-
that the skirt radius is more sensitive to this parameter ents.
than the other parameters in Eq. (25.1). A modification
to the vacuum system of the VPSEM that minimizes the
gas path length consists of using a small diameter tube to 25.2.2 Favorable Sample Characteristics
extend the high vacuum of the electron column into the
sample chamber. Given that the LVSEM operating conditions have been
selected to minimize the gas scattering skirt, what specimen
types are most likely to yield useful microanalysis results? If
25.2.1 Workarounds To Solve Practical most of the gas scattering skirt falls on background material
Problems that contains an element or elements that are different from
the elements of the target and of no interest, then by follow-
Gas scattering effects can be minimized but not avoided, ing a measurement protocol to identify the extraneous ele-
and for many VPSEM measurements and in situ experi- ments, the measured spectrum can still have value for
ments, the microscopist/microanalyst may be significantly identifying the elements within the target area, always recog-
constrained in the extent to which any of the parameters in nizing that the target is being excited by the focused beam
Eq. (25.1) can actually be changed to reduce the gas scatter- and the skirt.
ing skirt without losing the advantages of VPSEM opera-
tion. The measured EDS spectrum is always compromised, Particle Analysis
but by carefully choosing the problems to study, successful Particle samples comprise a broad class of problems related to
X-ray analysis can still be performed. A useful way to con- the environment, technology, forensics, failure analysis, and
sider the impact of gas scattering is the general concentra- other areas. Particles are very often insulating in nature so that
tion level at which the remote scattering corrupts the a conductive coating is required for examination in the con-
measured spectrum: ventional high vacuum SEM, and the complex morphologies
452 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
of particles often make it difficult to apply a suitable coating. germanium, or gold (often as a thick film evaporated on
25 The VPSEM with its charge dissipation through gas ionization silicon). Again, whatever the choice of substrate, the X-ray
is an attractive alternative to achieve successful particle imag- spectrum of the bare substrate should be measured to
ing. When VPSEM X-ray analysis measurements are needed establish the analytical blank prior to analyzing particles
to characterize particles, specimen preparation is critical to on that substrate.
achieve a useful result. There are many methods available for
particle preparation, but the general goal for successful . Figure 25.13a shows a VPSEM image of a particle cluster
VPSEM X-ray analysis is to broadly disperse the particles on a prepared on carbon tape (which has a blank spectrum con-
suitable substrate so that the unscattered beam and the inner- sisting of major C and minor O from the polymer base) and
most intense portion of the skirt immediately surrounding the EDS X-ray spectrum obtained with the beam placed on
the beam can be placed on individual particles without excit- particle “A”. The spectrum is seen to contain a high intensity
ing nearby particles. The more distant portions of the beam peak for Si, lower intensity peaks for O, Al, and K, and peaks
skirt may still excite other particles in the dispersion, but the just at the threshold of detection for Ca, Ti, and Fe. With the
relative fraction of the electrons that falls on these particles is other particles nearby, how much of this spectrum can be
likely to be sufficiently small that the artifacts introduced in reliably assigned to particle “A”? A local “operational blank”
the measured spectrum will be equivalent to trace (C < 0.01 can be measured by placing the beam on several nearby sub-
mass fraction) constituents. strate locations so that focused beam only excites the sub-
1. When particles are collected on a smooth (i.e., not tortu- strate while the skirt continues to excite the specimen over its
ous path) medium such as a porous polycarbonate filter, extended reach. Examples of the “working blanks” for this
the loaded filter can be studied directly in the VPSEM particular specimen are shown in . Fig. 25.13b, c and are
with no preparation other than to attach a portion of the revealed to be surprisingly similar, considering the separa-
filter to a support stub. Prior to attempting X-ray mea- tion of location “BL3” from “BL1” and the particle array.
surements of individual particles, the X-ray spectrum of Inspection of these working blanks show Al and Si at the
the filter material should be measured under the VPSEM minor level, and Ca and Fe at the trace level, as estimated
operating conditions as the first stage of determining the from the peak-to-background ratio. Thus, the interpretation
analytical blank (that is, the spectral contributions of all of the spectrum of particle “A” can make use of the local ana-
the materials involved in the preparation except the lytical blank, as shown in . Fig. 25.13d. The major Al and Si
specimen itself). In addition to revealing the elemental peaks are not significantly perturbed by the low blank contri-
constituents of the filter, this blank spectrum will also butions for these elements, and the minor K is not present in
reveal the contribution of the environmental gas to the the working blank and therefore it can be considered valid.
spectrum. However, the low levels of Ca and Fe observed in the blank
2. When particles are to be transferred from the collection are similar to those in the spectrum on particle “A”, and thus
medium, such as a tortuous path filter, or simply obtained they should be removed these from consideration as legiti-
from a loose mass in a container, the choice of the sample mate trace constituents. Note that despite the size of particle
substrate is the first question to resolve. Conceding that “A,” which is approximately 50 μm in its longest dimension, it
VPSEM operation will lead to significant remote scat- must be remembered that measurement sensitivity to any
tering that will excite the substrate, the sample substrate possible heterogeneity within particle “A” is likely to be lost in
should be chosen to consist of an element that is not of the VPSEM mode because of the large fraction of the inci-
interest in the analysis of individual particles. Carbon dent current that is transferred within the 25-μm radius, as
is a typical choice for the substrate material, but if the demonstrated in . Table 25.1. Another example of the use of
analysis of carbon in the particles is important, then an the working blank is shown in . Fig. 25.13e for particle “D”.
alternative material such as beryllium (but beware of the Here the Ca is much higher than in the working blank, and so
health hazards of beryllium and its oxide) or boron can it can be considered a legitimate particle constituent, as can
be selected. If certain higher atomic number elements can the Mg, which is not in the working blank. The low Fe peak is
be safely ignored in the analysis, then additional materials similar in both the particle spectrum and the working blank,
may be suitable for substrates, such as aluminum, silicon, so it must not be considered legitimate.
.. Fig. 25.13 a VPSEM image of a cluster of particles and an EDS X-ray location BL3 so that only the beam skirt excites the particles. d Use of
spectrum measured with the beam placed on one of them, particle “A.” analytical blank to aid interpretation of the EDS spectrum of Particle “A.”
b EDS X-ray spectrum measured with the beam placed on the substrate e Use of analytical blank to aid interpretation of the EDS spectrum of
at location BL1 so that only the beam skirt excites the particles. c EDS Particle “A”
X-ray spectrum measured with the beam placed on the substrate at
25.2 · What Can Be Done To Minimize gas Scattering in VPSEM?
453 25
a
3453 Ch#: 261 Ch kV; 2.6100 Work: 51 Results: 0 Marker: Fe 26
SiKa
BL2 C
B
D BL1
O Ka
Al Ka
MgKa K Ka
C Ka Ti Ka FeKa
S Ka
CaKa Ti Kb1 FeKb1
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
b
1396 Ch#: 274 ChkV:: 2.7400 Work: 101 Results: 0 Marker: Fe 26
C ka
BL3
BL2
B
C
D BL1
Operational blank at “BL1”
A
Ka
O
Si Ka
Al Ka
CaKa
FeKa
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
454 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
25 c 3718
718 Ch#: 266 Ch kV: 2.6600 Work: 269 Results: 0 Marke:r:Fe 26
Marke:r
C Ka
BL3
BL2 C
B
D BL1
O Ka
SiKa
Al Ka
CaKa
FeKa
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
d
SiKa
AlKa Operational blank adjacent to A
CaKa
FeKa
30.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
SiKa
BL2 C
B
D BL1
0 Ka
ALKa
K Ka
MgKa S Ka TiKa FeKa
C Ka CaKa TiKb1 FeKb1
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
Particle “D”
BL3
CaKa
E
BL2 C
B
O Ka D BL1
MgKa
C Ka S Ka K Ka TiKa FeKa
TiKb1 FeKb1
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
25.2.3 Unfavorable Sample Characteristics SEM-BSE image, there are three different phases present
25 denoted D, I, and B, with different Al/Ni ratios. The spectral
EDS analysis in VPSEM is most problematic for specimens intensities for Al and Ni show differences in the low pressure
consisting of densely packed microscopic features, for exam- EDS spectra shown in . Fig. 25.15a that are sufficient to read-
ple, most solid materials, natural and synthetic, with a micro- ily distinguish the phases despite the gas scattering
structure. The measured EDS X-ray spectrum in such a case (E0 = 20 keV; 50 Pa; water vapor; 6-mm gas path length).
depends very strongly on the VPSEM conditions, the dimen- When the pressure is increased to 665 Pa, two of the phases
sions of the target area of interest, and the exact nature of the can no longer be distinguished in the EDS spectrum shown
surrounding microstructure. An example is presented in in . Fig. 25.15b. This loss of phase recognition is also seen as
. Fig. 25.14, which shows the microstructure of Raney nickel, a loss of contrast in X-ray mapping, as seen in the elemental
an aluminum-nickel catalyst. Based on the contrast in this intensity maps shown in . Fig. 25.16a, b.
I
Raney nickel B
D
E0 = 20 keV
BSED
100 µm
25.2 · What Can Be Done To Minimize gas Scattering in VPSEM?
457 25
ALKa
Soec# Counts
1 510
NiLa1 NiKa
NiKb1
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
ALKa
Soec# Counts
1 481
NiLa1 NiKa
NiKb1
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Photon energy (keV)
.. Fig. 25.15 a EDS spectra obtained at locations “D”, “I,” and “B.” E0 = 20 keV; water vapor; 50 Pa, 6-mm gas path length. b EDS spectra obtained at
locations “D,” “I,” and “B.” E0 = 20 keV; water vapor; 665 Pa, 6-mm gas path length
458 Chapter 25 · Attempting Electron-Excited X-Ray Microanalysis in the Variable Pressure Scanning Electron Microscope (VPSEM)
25 a
AI K-L2
E0 = 20 keV
133 Pa
665 Pa
1330 Pa 2000 Pa
50 µm
Ni K-L2
E0 = 20 keV
133 Pa
665 Pa
1330 Pa 50 µm
2000 Pa
.. Fig. 25.16 a Elemental intensity map for Al K-L2 at various pressures (E0 = 20 keV, water vapor and 6-mm gas path length). b Elemental intensity
map for NiKα at various pressures (E0 = 20 keV, water vapor and 6-mm gas path length)
References
459 25
References Danilatos GD (1993) Introduction to the ESEM instrument. Micros Res
Tech 25:354
Newbury D (2002) X-Ray microanalysis in the variable pressure (envi-
Danilatos GD (1988) Foundations of environmental scanning electron.
ronmental) scanning electron microscope. J Res Natl Inst Stand
Microscopy Adv Electronics Electron Phys 71:109
Technol 107:567
461 26
References – 470
Typically, a pure metal such as copper, nickel, aluminum, 26.2 Sample Preparation
silicon, among others, is used and the number of counts in
the range of energies around the K characteristic lines is Regardless of whether you are performing standards-based
used as the proxy for the probe current, or the entire spec- or standardless quantitative analysis, you will need to ensure
trum can be integrated from a threshold, for example, that your unknowns and standards, if used, are prepared in a
0.1 keV to E0. suitable fashion for analysis. How you prepare your samples
Precise calibration of the probe current using a spectrum will depend very much on the type of samples that you ana-
takes patience. For 0.1 % precision, you will need to integrate lyze. For the most accurate analysis of bulk materials, includ-
at least 1,000,000 counts in the range of energies used for the ing that utilizing the standardless protocol, the specimen
calibration. must be prepared with a flat surface, with the degree of flat-
464 Chapter 26 · Energy Dispersive X-Ray Microanalysis Checklist
ness becoming increasingly critical for measurements with (iii) Complex standards:
X-rays having energies below 1 keV—e.g., Be, B, C, N, O, F— In many cases, complex standards similar to the
where the surface roughness should be reduced below 50 nm unknown sample will produce the most accurate
26 root mean square. Some samples require very little prepara- results. However, complex standards often are more
tion (e.g., a silicon wafer) and others (anything that isn’t flat) difficult to work with and it is often hard to find
require a lot. You may need to embed the samples and stan- reliable compositional data. Use of complex stan-
dards in epoxy mounting medium (preferably conductive) dards is an advanced topic and rarely justified
and use suitable equipment to grind and polish the samples. unless suitable pure element and simple compound
Appropriate preparation protocols are so specialized that standards are not available.
rather than provide an exhaustive list of possible procedures,
the analyst is referred to the rich literature on sample prepa-
ration: 26.2.2 Peak Reference Materials
55 Echlin, P., Handbook of Sample Preparation for Scanning
Electron Microscopy and X-ray Microanalysis (Springer, Peak reference spectra serve as examples of the shape of an
New York, 2009) element’s characteristic peaks. A peak reference can serve as
55 Geels, K., Metallographic and Materialographic Speci- a reference for one or more families. Standards materials can
men Preparation, Light Microscopy, Image Analysis, and be used as references if there are no interferences for that ele-
Hardness Testing (ASTM, West Conshohocken, PA, ment. Paradoxically, some materials suitable to serve as peak
2006) references are not suitable for use as standards, generally due
55 McCall, J., Metallographic Specimen Preparation: Optical to instability under the electron beam; for example, BaCl2
and Electron Microscopy (Springer, New York, 2012). provides an excellent peak reference for the Ba M-family, but
55 Vander Voort, G., Metallography, Principles and Practice it is unstable under electron bombardment and thus cannot
(ASM International, Materials Park, OH, 1999) serve as a standard.
Standards are specially prepared materials that serve to pro- 26.3.1 Calibrating the EDS Detector
vide X-ray data for the elements present in the unknown. You
will need a standard for each element in the unknown mate- Most EDS detectors allow you to configure two different char-
rial. Note that some standards can serve for two or more ele- acteristics of the detector—the pulse process time and the
ments, for example, FeS2. energy calibration. Some detectors provide other more
To be useful, a standard must be— advanced options. You will need to consult the manufacturer’s
1. Suitably sized—Practically speaking, most standards range documentation to determine the optimal setting for these
in size from tens of micrometers to a tens of millimeters parameters. In general, the goal should be to ensure that the
2. Polished to a smooth, flat surface with surface texture of detector is configured the same way day-in/day-out even if it
less than 100 nm (50 nm for low energy X-rays like oxy- means making small compromises to the ultimate performance.
gen)
3. Mounted in a manner that facilitates placement on the Selecting a Pulse Process Time Constant
stage perpendicular to the beam 1. Most EDS detectors will allow you to make a trade-off
4. Conductive or coated with a conductive material between X-ray throughput and detector resolution.
5. There are several different classes of standards, in order Higher throughput comes at the price of poorer detector
of the ease of use: resolution. This setting is called different things by vari-
(i) Pure element standards: ous vendors—throughput, pulse processor setting, shap-
The easiest standards to use are pure elements and ing time, or time-constant.
this is where a novice should start whenever 2. Typically, a pulse processor time constant selected in the
possible. middle of the allowed range represents the best trade-off.
(ii) Simple compound standards: Despite common belief, the highest resolution setting is
Some elements are not stable as pure elements—N, rarely the optimal choice as this setting is usually accom-
O, F, Cl, Br, S—and must be provided as com- panied by severe throughput limitations. Usually, it is
pounds. The easiest to use are typically stable, better to select a moderate resolution and obtain a mod-
stoichiometric compounds like alumina (Al2O3) or erate throughput.
calcium fluoride (CaF2). With stable stoichiometric 3. On a modern silicon drift detector (SDD), throughput is
compounds, the true composition of the standard typically not limited by the pulse-process time but rather
is unambiguous. Ideally, the compound is chosen through pulse pile-up (coincidence) events. Selecting a
so that there are no interferences between the very fast process time won’t actually improve usable
element’s characteristic peaks. throughput.
26.3 · Initial Set-Up
465 26
4. Be sure to turn off “adaptive shaping” or other mecha- just used to calibrate your detector as a record of the detec-
nisms that adapt the process time dynamically depend- tors performance. The program extracts efficiency, calibra-
ing upon X-ray flux. Adaptive shaping changes the shape tion and other pieces of instrumental data and records
of the characteristic peaks as the count rate varies due to them in a database. This database can then be exported as
variations in local composition and makes linear peak control charts or tabulations. It is a nice way to make the
fitting less accurate. most of the calibration data you have already spent the time
to collect.
Energy Calibration
1. Select one material that you will always use to calibrate aintaining the Working Distance/
M
your detector. Mount a piece of it near your Faraday cup. Specimen-to-EDS Distance
Copper is a good choice but there are other materials Maintaining consistent experimental conditions is critical
that have both low energy and higher energy X-ray for achieving rigorous quantitative microanalysis. A criti-
peaks that can be used for calibration. cal parameter is the specimen-to-EDS distance, SEDS, since
2. Select a channel width and channel count that fully cov- the solid angle of the EDS varies as 1/S2EDS. The SEM and/or
ers the range of beam energies that you may use. A EDS manufacturer(s) will have specified the ideal SEM
width of 10 eV/channel and 2,048 channels is a good working distance (WD) at which the EDS central axis
choice for a 20-keV beam energy. A width of 10 eV/ intersects the SEM optic axis. An electron probe microana-
channel and 4,096 channels is suitable for higher-energy lyzer uses a fixed-focus optical microscope with a very
work. Since EDS spectra comprise relatively small size shallow depth-of-focus to bring the specimen to this ideal
files and computer storage is inexpensive, consider 5 eV WD position (to which the wavelength dispersive spec-
per channel, especially for low photon energy work to trometers are also focused) on a consistent basis. While
provide adequate channels to describe the peak. very useful, such an optical microscope is rarely available
Whatever choice is made, it is important to consistently on an analytical SEM, so that another method must be used
use that choice. to select the working distance properly and consistently.
3. Each day before you start collecting data, collect and The following procedure assumes the use of a flat polished
store an initial spectrum from your selected calibration specimen.
material, for example, Cu. Use this spectrum and your 1. Load the specimen so that its Z-height is approximately
vendor’s software to calibrate the energy axis. Usually correct for the ideal working distance specified by the
this involves adjusting the electronics to ensure that manufacturer. Most SEMs provide a mechanical mount-
the measured peaks are centered around the correct ing reference frame to approximately set this initial
channels. Most modern detectors automatically and specimen altitude.
dynamically adjust the zero offset and the calibration is 2. Using the manufacturer’s specified value of the ideal
just a matter of the software automatically selecting a SEM working distance for EDS (e.g., 10 mm), set the
gain that produces the desired number of eV/channel. SEM objective lens strength to focus at this optimal
Older detectors may require a manual gain and zero working distance, making use of whatever objective
offset adjustment using external potentiometers. lens meter reading is available to monitor this adjust-
Regardless of the mechanism, at 10 eV/channel a 0.1 % ment.
mis-calibration will correspond to a single channel 3. Select a low magnification to begin (e.g., 100×). Despite
error in the position of a peak at 10 keV, so the detec- care in polishing, there are almost always a few fine
tor calibration should ideally be maintained to better scale scratches or other irregularities to be found.
than 0.1 %. Fortunately, modern detectors are able to Locate one, and use the stage Z-motion (i.e., motion
hold this precision of calibration for days or weeks. along the optic axis of the SEM) to bring this scratch
4. Once the detector is calibrated, set the beam energy and into approximate focus, without adjusting the objective
probe current to established nominal values and collect a lens strength. Increase the magnification in stages to
spectrum for a established live time. Save this spectrum 5000× and refine the focus with the Z-axis motion, not
as a demonstration that your detector is performing cor- by changing the objective lens strength. If the SEM
rectly. DTSA-II provides tools to extract and track per- stage automation system permits, save this
formance metrics and then plot the results over time as a Z-parameter.
control chart. 4. This procedure is likely to be more reproducible at lower
probe currents where the convergence angle is larger
Quality Control and depth-of-focus is smaller.
Sooner or later, you will be asked by a client to demonstrate 5. Consistency is critical. Always collect your spectra at
that your instrument was performing correctly when their this ideal working distance. Before collecting each spec-
data was collected. The easiest way to satisfy this require- trum ensure that the sample surface is at the optimal
ment and to impress the client is to keep a long term record working distance by setting the objective lens/working
and present the data as a control chart. DTSA-II provides distance and bringing the sample into focus using the
functionality which permits you to archive the spectra you vertical stage axis.
466 Chapter 26 · Energy Dispersive X-Ray Microanalysis Checklist
The analytic total is the sum of the mass fractions measured • Analytical Procedure
for each element in the unknown. The analytic total should The analytical procedure should provide sufficient detail that
be close to unity if all elements have been recognized and you or someone else with the correct instrumentation could
included in quantitative calculations. When a typical mate- replicate the measurement.
rial is analyzed under typical conditions, the analytical total 1. Scanning electron microscope
may reasonably vary 1 % or 2 % from unity due to measure- 5 Manufacturer and model
ment uncertainties. Simple measurements based on energetic 5 Beam energy
K peak transitions can be reliably measured to better than a 5 Nominal probe current
percent. More complex measurements involving low energy 2. X-ray detector
X-rays (like carbides and oxides) are likely to have larger 5 Manufacturer and model
deviations. A deviation of more than a percent or two should 5 Configuration settings
inspire you to start asking questions. 3. Other aspects
1. Have I missed an element? 5 Picoammeter
(i) Check the residual. Is there a peak that hasn’t been 5 Software
quantified? 4. Standards
(ii) Is it possible that there is an unmeasurable element 5 Identity, composition, source, live time, probe current,
like H, He, or Li in the unknown? sample preparation, coating
2. Is there a problem with the measurement process? 5. References
(i) Is the sample preparation adequate? 5 Identity, composition, characteristic line assignment,
(ii) Is the sample tilted? sample preparation, coating
(iii) Is the sample at the correct working distance? 6. Unknown
(iv) Is the probe current being measured accurately? 5 Identity, sample preparation, coating
(v) Did the probe current drift? 7. How the locations for analysis were selected. Was it based
(vi) Did the specimen charge? on the customer’s directions or if the analyst selected the
locations, what criteria were used?
from mass concentration (weight fraction) to atomic frac- v ariation between locations for each element. Compare
tion or oxide fraction includes normalization. this value with the uncertainty estimate for a single mea-
3. DTSA-II provides an estimate of the measurement uncer- surement to detect heterogeneity.
26 tainty for each element. The estimate should make clear
which factors were considered and whether the estimate • Conclusions
should be considered an estimate of accuracy or just an esti- Conclusions should be pithy. You should be very careful only
mate of precision. A measurement without an uncertainty to report that which is directly supported by the measure-
estimate is open to misuse. The client may assume that the ment results. In other words, stick to the facts and avoid con-
result is more accurate than it really is and draw conclusions jecture. Don’t answer questions that go beyond the data and
that cannot be justified by the data. Alternatively, the client your personal expertise.
may not trust the data or may assume that it is less accurate
than it is and fail to draw conclusions that are justified. Either
way, data presented without uncertainties is of limited utility.
Reference
4. If the spectra represent a nominally homogeneous
International Organization for Standardization (2012) Standard file for-
region (or one you suspect to be), add descriptive statis- mat for spectral data exchange ISO 22029:2012. 7 https://www.iso.
tics (mean, standard deviation) summarizing the org/standard/56211.html
471 27
X-Ray Microanalysis
Case Studies
27.1 Case Study: Characterization of a Hard-Facing
Alloy Bearing Surface – 472
References – 479
27.1 ase Study: Characterization of a Hard- . Fig. 27.4 and the SEM-BSE image in . Fig. 27.5, which
C
Facing Alloy Bearing Surface shows strong atomic number contrast. SEM-EDS analysis
with DTSA-II gave the results presented in . Fig. 27.5, which
Background: As part of a study into the in-service failure of the confirm the gray scale sequence as a function of composi-
bearing surface of a large water pump, characterization was tion, with the sharp rise in W (see table in . Fig. 27.5) domi-
requested of the hard-facing alloy, which was observed to have nating backscattering.
27 separated from the stainless steel substrate, causing the failure. The information provided by SEM-EDS enabled metal-
This problem illustrates the critical importance of careful lurgists to modify the hard-facing alloy composition and
specimen preparation of a macroscopic object with centime- deposition conditions to eliminate the void formation during
ter dimensions to locate regions of microscopic interest with deposition, producing satisfactory service behavior.
micrometer dimensions. Metallographic preparation pro-
duced a polished cross section of an intact portion of the Stainless steel substrate
hard-facing alloy layer as deposited onto the stainless steel Voids
base, as shown in . Fig. 27.1. SEM-EDS analysis with
DTSA-II gave the results shown in . Fig. 27.2 for the stain-
less steel base and at one location in the hard-facing alloy.
EDS elemental mapping produced the images shown in
. Fig. 27.3. The SEM-BSE image (. Fig. 27.1) revealed the
presence of numerous voids that were predominantly located
(white arrows) at the interface between the hard-facing alloy
and the stainless steel substrate, with a smaller population of
voids located within the hard-facing alloy (yellow arrows).
The elemental maps (. Fig. 27.3) revealed that the hard-fac-
ing alloy layer had a very complex microstructure with dis-
tinct heterogeneity. The voids were found to be closely
associated with regions with that had elevated Cr, both near
the hard-facing alloy-stainless steel interface and within the
.. Fig. 27.1 SEM-BSE image of the cross-section of a hard-facing alloy
hard-facing alloy. These voids severely compromised the per-
deposited on a stainless steel substrate. Note the voids at the interface
formance of the bearing surface. between the hard-facing alloy and the stainless substrate (white arrows),
Complex fine-grained regions were also observed within as well as a smaller population of voids entirely within the hard-facing
the hard-facing alloy, as shown in the elemental maps in alloy (green arrows)
200 mm
BSE MAG: 100 ´ HV: 20.0 kV WD: 11.0 mm
Si Cr Mn Fe Co Ni
0.0377 0.0915 0.0006 0.0228 0.3859 0.4616
±0.0002 ±0.0002 ±0.0001 ±0.0001 ±0.0005 ±0.0007
27.1 · Case Study: Characterization of a Hard-Facing Alloy Bearing Surface
473 27
Ni Co
Cr Ni_Cr_Co
200 µm
.. Fig. 27.3 Elemental mapping, with color overlay: Ni = red; Cr = green; Co = blue
Ni W
Cr Ni_W_Cr
5 µm
Cr Fe Co Ni W
Cr Fe Co Ni W
.. Fig. 27.5 SEM-BSE image and DTSA-II analyses of selected grains in the fine-scale region
Steel screw
27.2 ase Study: Aluminum Wire Failures
C
in Residential Wiring
Thermal damage to
Background: In the early 1970s, aluminum wire was used plastic case
extensively as a substitute for more expensive copper wire in
residential and commercial wiring, specifically for 110 V Thermal damage to
electrical outlets that used steel screw compressive clamping Brass
wire insulation
of the wire against a brass or steel plate. The aluminum wire– plate Al wire
steel screw junctions were observed to fail catastrophically 1 cm
through a process of overheating, leading in extreme cases to
.. Fig. 27.6 Residential electrical outlet wired with aluminum. The
initiation of structural fires (Meese and Beausoliel 1977; laboratory test was interrupted after the thermal event initiated and
Rabinow 1978). . Figure 27.6 shows an example of the dam- was automatically detected, but significant thermal damage to the
age to the wire-screw junction and the surrounding plastic plastic casing and wire insulation still occurred
housing and wire insulation caused during an overheating
event observed in a laboratory test. This failure was a puz-
zling occurrence, since aluminum is an excellent electrical
conductor and was long used successfully in high voltage
electrical transmission lines. Moreover, the vast majority of
Al wire–screw connections provided proper service without
overheating. However, those connections that did fail in ser-
vice often produced such catastrophic effects that the critical
evidence of the initiation of the failure was destroyed.
Capturing an event like that shown in . Fig. 27.6 required
intensive laboratory studies in which thousands of junction
boxes were tested and continuously monitored with thermal
sensors until a failure initiated, which was then automatically
interrupted to prevent complete destruction of the evidence.
This problem illustrates the “macro to micro” sampling
problem. The failure mechanism was eventually discovered
by SEM/EDS characterization to have a microscopic point of
origin, but this microscopic failure origin with micrometer
dimensions was hidden within a complex macroscopic struc-
ture with centimeter dimensions. Solving the problem
required a careful sample preparation strategy to locate the
unknown feature(s) of interest. The metallographer mounted
the entire Al-wire/steel screw/brass plate assembly in epoxy, .. Fig. 27.7 Metallographic mount (2.5-cm diameter) showing the
as shown in . Fig. 27.7, and sequentially ground and p olished cross section of the steel screw, aluminum wire, and brass plate
27.2 · Case Study: Aluminum Wire Failures in Residential Wiring
475 27
Al
BSE 500 µm
1
2
Fe Zn
.. Fig. 27.9 SEM-BSE image and elemental maps for Al, Fe, and Zn of the anomalous contact zone
476 Chapter 27 · X-Ray Microanalysis Case Studies
Al 0.759 a/o
Fe 0.241 a/o
FeAl3
Al 0.719 a/o
Fe 0.281 a/o
Fe2Al5
Mounting
epoxy
the damage seen in . Fig. 27.6. (Note that the practical solu-
tion to this problem was to modify the wire connections to
provide much greater springiness to eliminate the opening of
gaps that allowed arcing to occur.)
Note cracks
BSE
Note cracks
Mn O
20 µm
Ni Mn O Ni
.. Fig. 27.13 SEM/EDS X-ray spectrum imaging elemental maps for Mn, O, and Ni and color overlay (Mn = red; O = green; Ni = blue). Note cracks are
observed in the O map but are much less visible in Mn and Ni
478 Chapter 27 · X-Ray Microanalysis Case Studies
27
Mn Fe
20 µm
Ni Mn Fe Ni
.. Fig. 27.14 SEM/EDS X-ray spectrum imaging elemental maps for Mn, Fe, and Ni and color overlay (Mn = red; Fe = green; Ni = blue)
Mn Si
Mn Si Al 20 µm
Al
.. Fig. 27.15 SEM/EDS X-ray spectrum imaging elemental maps for Mn, Si, and Al and color overlay (Mn = red; Si = green; Al = blue)
References
479 27
.. Table 27.1 Fixed-beam analysis at the center of . Fig. 27.12; NIST DTSA II analysis with pure element and microanalysis glass standards
O Na Mg Al Si K Ca Ti Mn Fe Ni Cu
0.2742 ± 0.0420 ± 0.0196 ± 0.0210 ± 0.0458 ± 0.0100 ± 0.0267 ± 0.0031 ± 0.4412 ± 0.0834 ± 0.0196 ± 0.0134 ±
0.0003 0.0005 0.0002 0.0001 0.0002 0.0000 0.0001 0.0000 0.0003 0.0002 0.0001 0.0002
Al Fe
Mn 20 µm
.. Fig. 27.16 SEM/EDS X-ray spectrum imaging maps after quantitative analysis with DTSA-II presented with logarithmic three-band encoding for
Al, Fe, and Mn. Note Fe-enrichment band
Cathodoluminescence
28.1 Origin – 482
References – 489
28.1 Origin
750 380
Cathodoluminescence (CL) is the emission of low energy
photons in the range from approximately 1 eV to 5 eV
(infrared, visible, and ultraviolet light) as a result of inelas- Infrared Ultraviolet
tic scattering of the high energy beam electrons (. Fig. 28.1).
1240 620 410 310 250 nm
Materials that can emit such photons are insulators or semi-
conductors which have an electronic structure with a filled
28 valence band of allowed energy states that is separated by a
gap of disallowed energy states from the empty conduction 0 1 2 3 4 5 eV
band, as shown schematically in . Fig. 28.2a. Inelastic scat-
tering of the beam electron can transfer energy to a weakly .. Fig. 28.1 Range of photon energies and wavelengths for cathodo-
bound valence electron promoting it to the empty conduc- luminescence
tion band, leaving a positively charged “hole” in the con-
duction band. When a free electron and a positive hole are
attracted and recombine, the energy difference is expressed a “Intrinsic cathodoluminescence”
as a photon, as illustrated in . Fig. 28.2b. Because the pos- Electronic structure has a filled valence band and an empty
sible energy transitions and the resulting photon emission conduction band separated by a band gap of disallowed energy
states’ Initial excitation by inelastic scattering of beam electron
are defined by the intrinsic properties of a high purity mate- to promote valence band electron into conduction band
rial, such as the band-gap energy but also including energy
levels that result from physical defects such as lattice vacan- Conduction band: unoccupied states
cies, rather than by the influence of impurity atoms, this
type of CL is referred to as “intrinsic CL emission.” Since
the valence electron promoted to the conduction band can
receive a range of possible kinetic energies depending on Band gap (Egap ≈ few eV) of disallowed states
the details of the initial scattering, the photons emitted dur-
ing free electron–hole recombination can have a range of
energies, resulting in broad band CL photon emission.
Valence band: occupied states
Because of the great mismatch in the velocity of the high “Hole”
CL ZnS
28.2 Measuring Cathodoluminescence
28.2.1 Collection of CL
28.2.2 Detection of CL b
The photomultiplier (PM) is a useful detector for CL because .. Fig. 28.4 a CL emission from ZnS observed with a defocused
220-μm-diameter beam with 500 nA of beam current at E0 = 20 keV. b
of its sensitivity to photons in the energy range of interest, its
Corresponding white light illumination image with the electron beam
fast response (hundreds of picoseconds to nanoseconds), and blanked into a Faraday cup
484 Chapter 28 · Cathodoluminescence
28
Light pipe
Photomultiplier
50 µm
.. Fig. 28.8 CL imaging (RGB) of a carbonado, a rare variety of dia- halos leaving the defects in the diamond structure (Magee et al. 2016)
mond. The polycrystalline material interacted with uranium during (Images courtesy of E. Vicenzi (Smithsonian Institution))
the Precambrian and MeV radiation damage is evident as metamict
486 Chapter 28 · Cathodoluminescence
Silicate
glass ZrO2
Resorbed
core
28 Zircon
Growth
zoning
.. Fig. 28.9 CL imaging (RGB) of an impact zircon collected using using a low energy beam (Zanetti et al. 2015) (Images courtesy of
E0 = 20 keV (left) and 5 keV (right). The details of zircon disproportionation E. Vicenzi (Smithsonian Institution))
to ZrO2 and silicate glass, as well as the zircon interior, are more evident
emission were investigated by spectral peak fitting. The During the charging and discharging of lead acid battery
derived central wavelength map reveals the strain level of plates, a variety of lead containing phases can form on the
the material with noticeable shifts appearing along the surface of the lead plate. Two important phases are lead sul-
pyramid ridges. The RGB composite image is created from: fide (PbS) and lead sulfate (PbSO4). Note that these com-
high strain state GaN (yellow), low strain state (purple), pounds are similar in backscattering making it difficult to
defect (red). determine PbS from PbSO4 in images. . Figure 28.11a shows
a secondary electron image of the surface of a lead plate after
Lead-Acid Battery Plate Reactions it has been exposed to conditions that may occur in a lead-
It is possible to detect relatively strong CL signals with modi- acid battery. Numerous euhedral crystals were observed on
fied BSE detectors, as most of these solid state detectors are the surface of the lead plate. The EDS spectra indicated Pb
very sensitive to light. The challenge is to exclude BSE from and S as the major constituents and possibly O. Oxygen can
the detected signal. The easiest way to do this is to coat a glass be detected in the EDS spectrum of PbSO4 from an ideal flat
coverslip with a conductive transparent material like indium specimen, as shown in the DTSA-II simulation in
tin oxide (ITO). The glass cover slip prevents the electrons . Fig. 28.11c, but because of the high absorption from Pb, the
from interacting with the solid state detector and the con- time requirement for mapping O to locate PbSO4 becomes
ductive transparent coating prevents charging while allowing prohibitive. Moreover, given the complex topography of the
the CL signal to reach the detector. sample shown in . Fig. 28.11a, mapping oxygen is likely to be
It is interesting to note that some of the low vacuum or badly compromised by strong X-ray absorption artifacts
variable pressure SEMs that are commercially available use from the topography. CL can be of great use in this system as
the CL light generated by the interaction of the secondary PbS does not exhibit CL while PbSO4 strongly exhibits CL,
electrons with the gas in the chamber to produce a signal enabling these compounds to be rapidly distinguished. In
during variable pressure operation. One can use the same order to determine the likely compound, a simple CL system
detector system, which consists of a glass light guide located consisting of a light guide attached to a photomultiplier tube
near the sample that is coupled to a photomultiplier, for was used. . Figure 28.11b is an image obtained using this
direct detection of CL emission. This type of detector is sen- simple CL detector. Note that the euhedral crystals strongly
sitive to low levels of light and thus when used in high vac- exhibit CL, and this clearly indicates that these crystals are
uum mode can be a very simple but effective CL detector. most likely PbSO4 and not PbS.
28.3 · Applications of CL
487 28
a b
c
d
.. Fig. 28.10 CL study of GaN structures grown on Si: a SEM SE image (E0 = 2 keV); b panchromatic CL image; c CL spectrum image data analyzed
to derive the central wavelength value; d RGB composite (Example courtesy of D. Stowe, Gatan, Inc.)
488 Chapter 28 · Cathodoluminescence
a b
28
c
300 000
Pbs_15keV
PbSO4_15keV
250 000
PbS
PbSO4
200 000 E0 = 15 keV
DTSA-II simulation
Counts
150 000
100 000
50 000
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0
Photon energy (keV)
.. Fig. 28.11 a SEM SE image of deposits on lead-acid battery. b CL image of the same area; c DTSA-II simulation of the EDS spectra of PbS and
PbSO4 from an ideal flat surface at E0 = 15 keV (courtesy of Sandia National Laboratory)
References
489 28
a b
4000
c
3500
3000
2500
Counts
2000
1500
1000
500
0
300 350 400 450 500 550 600 650 700 750 800
nm
.. Fig. 28.12 Acetaminophen, E0 = 20 keV: a SEM SE image; b panchromatic CL image; c CL spectrum (Example courtesy of S. Wight, NIST)
Despite the general vulnerability of organic compounds to Magee C, Teles G, Vicenzi E, Taylor W, Heaney P (2016) Uranium irradia-
tion history of carbonado diamond; implications for Paleoarchean
radiation damage under electron bombardment, some organic
oxidation in the São Francisco craton (South America). Geology 44.
compounds can be examined with CL spectrometry. doi: 10.1130/G37749.1
. Figure 28.12 shows an SEM SE image and a panchromatic CL Zanetti M, Wittman A, Carpenter PK, Joliff BL, Vicenzi E, Nemchin A,
image of acetaminophen (paracetamol, N-(4-hydroxyphenyl) Timms N (2015) Using EPMA, Raman LS, Hyperspetcral CL, SIMS,
ethanamide N-(4-hydroxyphenyl)acetamide) along with a CL and EBSD to study impact melt- induced decomposition of zircon.
Microsc Microanal 21:S3 1447–1448
spectrum showing broad CL bands.
491 29
Characterizing Crystalline
Materials in the SEM
29.1 Imaging Crystalline Materials with Electron
Channeling Contrast – 492
29.1.1 Single Crystals – 492
29.1.2 Polycrystalline Materials – 494
29.1.3 Conditions for Detecting Electron Channeling Contrast – 496
References – 514
λe =h/p (29.2a)
a b
–
(220)
–
(110) [001]
29
)
00
(4
)
00
(1
5° 5 mm 5° 5 mm
.. Fig. 29.3 a Wide field scanning (BSE image) of a Si single crystal planes are marked, as well as the parallel channeling bands defined
wafer whose surface is parallel to the (001) plane, thus looking along when the scan angle to the planes equals ± θB. Images processed with
the [001] pole; E0 = 15 keV. b The traces of two different sets of crystal ImageJ-Fiji CLAHE function
a b
1° 500 µm
2° 1 mm
.. Fig. 29.4 a, b Progressively smaller scanning areas (higher magnifications) restrict the angular range and thus the portion of the electron
channeling pattern that is observed. Images processed with ImageJ-Fiji CLAHE function
29.1.2 Polycrystalline Materials nucleate randomly in the liquid and grow in size until they
encounter each other, producing a three-dimensional
Most crystalline materials are not single crystals. During microstructure of randomly oriented crystalline units
solidification from a molten state, numerous crystals called “grains.” Materials fabrication processes such as
29.1 · Imaging Crystalline Materials with Electron Channeling Contrast
495 29
1 mm
100 µm
a b
29
.. Fig. 29.13 These series of EBSD patterns show the effect of sample tilt on the EBSD pattern quality. a 60°, b 50°, and c 40°
For EBSD, we tilt the sample at a high angle with respect those that lost only 0.2 kV. The pattern is formed from these
to the electron beam. This tilt is important to obtaining the low loss electrons and thus the resolution is quite good. Thus,
high quality patterns that are needed for EBSD. As shown in although EBSD involves backscattered electrons, it is a tech-
. Fig. 29.13, the tilt angle has a large effect on the pattern nique with much higher spatial resolution than standard
quality. The best pattern qualities are obtained at higher tilt backscattered electron imaging. Typically, resolutions of bet-
angles with some optimal condition between the pattern ter than 0.1 μm can be achieved for a beam voltage of 20 kV
quality and the ease of imaging of the highly tilted sample. in the transition metals. The electrons that have lost larger
. Figure 29.14a shows the backscattered electron energy dis- amounts of energy will contribute the background intensity
tribution that is developed from a highly titled sample (70° in the EBSD pattern. These electrons are the main reason
tilt) compared to 0° tilt. Note that the energy is highly peaked why there is limited contrast in EBSD patterns without some
toward the operating voltage of the microscope at 70° tilt form of background removal either through division or sub-
when compared to the same distribution from an electron traction or some other computational method of removal
beam that is normal to the sample surface. Further, since the (Michael and Goehner 1996).
EBSD pattern is clearly visible, they must form from only the The high tilt angle limits the spatial resolution attainable
electrons that have lost either no or a small amount of their due to the elongation of the electron beam foot print on the
energy. The electrons that have lost only a small amount of sample surface. The resolution parallel to the tilt axis is much
their energy are emitted from the sample in regions that are better than the resolution perpendicular to the tilt axis due to
very close to the original beam foot print on the surface of the high sample tilt angles used to acquire EBSD patterns.
the sample. This is shown in . Fig. 29.14b, which is a Monte The resolution perpendicular to the tilt axis is related to the
Carlo simulation of the distance from the initial beam impact resolution parallel to the tilt axis by
point that the electrons emerge from the sample surface for
Lperp = Lpara (1 / cos θ ) (29.3)
all the backscattered electrons, those that lost up to 2 kV and
29.2 · Electron Backscatter Diffraction in the Scanning Electron Microscope
501 29
a b
>19.8 kV
# Backscattered electrons
# Backscattered electrons
>18 kV
All electrons
.. Fig. 29.14 a Monte Carlo electron trajectory simulations of the back- electron distributions from Ni at 20 kV and a sample tilt of 70° for different
scattered electron distributions for Ni at 20 kV at a tilt of 70° (red) and for a levels of energy loss as a function of distance from the beam impact point
sample at 0° tilt (blue) that is normal to the electron beam. b Backscattered
where θ is the sample tilt with respect to the horizontal, Lperp significant amounts of energy contributing to the overall
is the resolution perpendicular to the tilt axis, and Lpara is the intensity of the pattern background. To compensate for this,
resolution parallel to the tilt axis. The resolution perpendicu- methods of removing the background are utilized. Some sys-
lar to the tilt axis is roughly three times the resolution parallel tems use a software-generated background to compensate for
to the tilt axis for a tilt angle of 70° and increases to 5.75 times the background and increase the pattern contrast. A more
for a tilt angle of 80°. Thus, it is best to work at the lowest established approach uses a background image obtained by
sample tilt angles possible consistent with obtaining good scanning over a large number of grains that produce the
EBSD patterns. background signal. This background is then used to normal-
ize the raw EBSD pattern to produce a pattern with high con-
trast (Michael and Goehner 1996).
29.2.4 ow Does a Modern EBSD System
H EBSD cameras are generally able to collect patterns at
Index Patterns higher angular resolution per pixel than is needed for most
orientation mapping. However, if high accuracy is needed, it
Modern EBSD systems all use the same basic steps to go from is possible to use the full resolution of the camera to produce
a collected EBSD pattern to indexing and to the determina- EBSD patterns. The disadvantage of using the full resolution
tion of the crystal orientation with respect to a reference of the EBSD camera is that longer exposure times are needed
frame. Once the pattern is collected, the necessary informa- to produce usable EBSD patterns. This longer exposure can
tion for indexing of the pattern (indexing refers to assigning really slow down map acquisition. It is possible to bin the
a consistent set of crystallographic directions to the pattern camera resolution, usually by factors of 2, to produce lower
with respect to a given or a set of given candidate structures) resolution patterns but with the ability to collect the patterns
must be derived from the diffraction pattern. Once the pat- at a much higher collection rate. Thus, if the high angular
tern has been indexed correctly (various vendors use differ- resolution is not required, the camera should be used at a
ent measures of what constitutes correct indexing), it is a reduced resolution, (2 × 2, 4 × 4, or 8 × 8 binning) to speed up
simple matter to determine the crystallographic orientation the acquisition. The tradeoff is between orientation accuracy
represented by the EBSD pattern. and speed of the measurement.
The most important factor in obtaining a quality orienta- Once the camera is set correctly for the given experiment,
tion is that the system have good quality patterns for index- the patterns should be of high quality with good contrast.
ing. The quality of the detected patterns needed for a given The next step is for the software to extract the line positions
experimental measurement can be influenced by choices from the collected pattern. In all modern systems this is
made over how the camera is operated. EBSD patterns are of accomplished with an algorithm called the Hough Transform,
low inherent contrast mainly due to electrons that have lost which takes straight lines and transfers them into points
502 Chapter 29 · Characterizing Crystalline Materials in the SEM
with the beam’s size not being larger than the features of deal with this issue is the use of dynamic focus as discussed
interest within the sample. Modern EBSD cameras have been elsewhere in this book. In brief, the dynamic focus adjusts
developed to be more sensitive and capable of dealing with the focus so that the electron beam focus is changed with
lower beam currents. This in practice means that one should respect to the position of the electron beam on the sample.
try settings that generate 1–2 nA as a starting point. Higher This is of most importance for low magnification detailed
beam currents will be desirable for faster EBSD map acquisi- scans but may still provide some advantages at higher mag-
tion. Once these conditions are set, a few patterns should be nifications as well. Once the dynamic focus has been prop-
collected. If patterns are of poor quality it may be necessary erly set up, one can collect the desired SEM images of the
to revisit the sample preparation steps used and modify them sample.
in order to reduce the surface damage. Although some fine EBSD is useful for a range of samples that include elec-
29 tuning of the microscope operating conditions may improve trical conductors, like metals through electrically insulating
the pattern quality, microscope operating conditions cannot oxide. For insulating samples or metal samples mounted for
make up for poor quality patterns due to inadequate sample metallography, methods must be employed to reduce charg-
preparation. ing of the sample. Sample charging is an important issue as
If good quality patterns are obtained, it now an excellent charging in the extreme can result in pattern distortions to
time to determine that appropriate match units have been the point that the patterns are no longer useful and if only a
selected from a database of crystallographic structures. For minor effect sample drift may result causing the resulting
the highest-quality EBSD work, one should try and calibrate images and maps to be distorted. It is sometimes adequate
the EBSD camera geometry for each experiment. Modern to sputter coat samples to ensure conductivity. However,
systems will use one of the known phases from the sample (if this coating should be kept as thin as possible and it is best
you have a known phase) and the crystallographic data file to use metal films rather than carbon coating as the conduc-
selected to try and index one or multiple diffraction patterns. tivity of metal coating are usually much higher. If the sam-
Once a match has been found, the software will attempt to ple cannot be coated there is another approach that has been
vary the calibration parameters to select the parameters that found to be quite useful and that is the use of variable pres-
give the best fit to the observed EBSD pattern. Once this is sure in the sample chamber. The introduction of a low pres-
done one can assume the system is adequately calibrated. It is sure, generally only a few 10–20 Pa, is required to reduce
recommended though that multiple points be tried to make charging effects for EBSD. If a pressure that is too high is
sure calibration is optimal. utilized there will be a noticeable degradation of the pattern
As discussed, EBSD is not a method that is routinely used quality due to scattering of the electrons between the sample
to determine the crystal structure of the sample, although and the phosphor screen. Thus, some trial and error may be
there has been work in this area, but it is a method that required to set the conditions correctly (El-Dasher and
requires suitable match units to successfully index the EBSD Torres 2009).
patterns. Excellent quality EBSD patterns will not be indexed There is no absolutely correct way to set up EBSD map
if an incorrect candidate unit cell has been selected. One parameters due to the variety of sample types and the
should attempt to index a few of the patterns obtained from information that may be required. There are a few simple
the sample with the selected candidate unit cells. If indexing things that can be done to ensure that quality orientation
is not possible, then it may be necessary to change the candi- data are collected. One, the most important is to determine
date unit cell and attempt indexing. In the case of a multi- the step size that will be used for the EBSD map. If too
phase sample it is important to collect patterns from all of the large of a step size is selected, important details of the
phases and insure that they can be indexed from the list of microstructure may be missed; and conversely if too fine of
candidate unit cells. a step size is used, the resulting maps may be of high qual-
ity but will have taken much longer to acquire than if a
Adjust SEM and Select EBSD Map Parameters larger step size had been selected. One strategy for an
The analysis is now at the point where the quality of the EBSD unknown sample is to start with a coarse step size so that a
patterns and therefore the sample preparation has been rather quick map can be obtained and allow better sam-
assessed and found to be adequate. The sample has been cor- pling strategies to be developed. The large step size map
rectly positioned within the microscope with respect to col- will allow the user to determine if the candidate match
lecting data that is referenced to the sample and the phase or phases are appropriate, get an estimate of the
microscope operating conditions have been optimized. The grain size and to begin to visualize the microstructure of
EBSD collection is nearly ready to proceed with a few other the sample. The large-step-size image will then allow a bet-
details. ter selection of step size for more detailed studies. One rule
First, due to the high tilt of the sample it will be apparent of thumb is that at least 10 points across the diameter of a
at lower magnifications that the areas of the sample away grain are required for a reasonable assessment of the
from the focus point will be out of focus. The best way to grain size.
29.2 · Electron Backscatter Diffraction in the Scanning Electron Microscope
505 29
Run the Automated Map physical direction. In order to understand inverse pole figure
Once all of the previous steps have been carried out it is now maps it is important to understand what are pole figures and
time to collect the crystallographic orientations. Once the inverse pole figures and how to interpret them.
run is started the software will collect an EBSD pattern at Pole figures are used to answer the question, Where does
each pixel, detect the bands in the EBSD pattern, calculate a particular crystallographic direction or plane fall in space
the best fit to the band positions using the candidate crystal relative to some arbitrary physical sample direction or plane?
structures, calculate the unit cell orientation, save the infor- Pole figures have been used for many years and are very com-
mation and move to the next pixel and repeat this series. mon in the preferred orientation or texture literature. In
Even though modern EBSD systems are capable of running many cases a single pole figure does not provide sufficient
100–1000 patterns each second, larger maps may consist of information and additional pole figures of other crystallo-
more than a million pixels and thus require hours or days to graphic directions are required. A pole figure is simply a ste-
collect. A map of 2000 × 1000 pixels taken at a setting that reogram with the axes defined by the external reference
allows 100 patterns to be collected and analyzed each second frame. It is common for evaporated or deposited thin films to
will require nearly 6 h to complete. Faster acquisition rates have a specific crystallographic direction parallel to the film
are available but at the expense of orientation accuracy or growth direction. . Figure 29.17 is a series of pole figures
noise. Quite often it is most efficient to run these longer from an evaporated Au thin film. In this example we show the
acquisitions overnight when the SEM is not being utilized [111] and the [110] pole figures. For the <111 > pole figure we
anyway. These long acquisition times put extra emphasis in note that there is a large number of poles plotted in the center
the microscope’s environment in order that sample drift due of the pole figure. This shows that many of the pixels in the
to temperature changes or other disturbances are minimized. data set have the <111 > direction aligned with the sample
It is also quite useful to post a note on the operating panel of normal. There are additional rings also in the <111 > pole fig-
the SEM so that the next user does not disturb an ongoing ure. These are a result of there being more than one <111 > type
acquisition or assume that the microscope has been left in a direction in a cubic crystal (in fact there are actually four of
safe condition with respect to the inserted EBSD cameras. these present). In the <110 > pole figure we see that there is a
number of poles in the center of the pole figure indicating
that there are a number of measured pixels with <110 > direc-
29.2.6 Display of the Acquired Data tions parallel to the sample normal direction.
Inverse pole figures are used to answer the question,
EBSD is somewhat unique in analytical techniques as there What crystallographic poles or planes are preferentially par-
are so many ways to present the collected data in meaningful allel or perpendicular to a specific sample direction? We usu-
ways. Of course, everyone likes to produce beautiful color ally again display these with respect to the physical axes of
maps of microstructures, but in reality it is not just the images the microscope or the sample as described for the pole fig-
that are important but the crystallographic data that is con- ures. For inverse pole figures we plot all of the directions that
tained in every point in an image that is important. In order are pointing in a specific direction of the sample. Inverse pole
to get that data, one must begin to use and understand crys- figures are extremely useful for samples where there are spe-
tallographic representations of the sample that are not simple cific axes of the sample that have a preferred crystallographic
images. Although it is beyond the scope of this chapter to direction. . Figure 29.18 is an example of inverse pole figures
describe every possible way that EBSD data can be displayed plotted for the same Au-thin film shown in . Fig. 29.17. It is
it is important to at least introduce these and how they might often helpful to show all three orthogonal directions so that
be utilized (Randle and Engler 2000). the preferred texture can be visualized. Note that the inverse
Quite often, the first map that EBSD practitioners display pole figures of . Fig. 29.18 show exactly the same informa-
is called a band contrast or an image quality map or other tion that is shown in . Fig. 29.17. It is sometimes helpful to
names depending on the particular vendor involved. These first plot the inverse pole figures as they will give an indica-
images are most commonly shown as gray-scale images tion of the samples texture without plotting pole figures of
where the gray level is scaled by some measure of the quality many different directions. The Z or surface normal direction
of the pixel by pixel EBSD pattern. The sharper or more of the inverse pole figures of . Fig. 29.18 is the one that car-
defined the pattern is the higher the gray level in the image. ries the most information about the sample. Here we see the
These images can be striking representations of the micro- high density of pixel orientations that are clustered around
structure of the sample and can reveal microstructural details the <111 > and less so around the <110 > directions.
not clearly visible in either light-optical microscope images Once we are familiar with the concepts of pole figures
or secondary or backscattered electron images in the SEM. and inverse pole figures, we can then go ahead and under-
One of the most common ways to display EBSD orienta- stand inverse pole figure maps which are one of the more
tion data is with orientation maps. More accurately these are common ways that EBSD orientation data is shown. An
called inverse pole figure maps with respect to a specific inverse pole figure map extends the idea of an inverse pole
506 Chapter 29 · Characterizing Crystalline Materials in the SEM
Y0 Pole figure
<111>
[Au_20kv_5kx_280pa_r:
Iron fcc (m3m)
Complete data set
39083 data points
Equal Area projection
Upper hemisphere
X0
29.2 · Electron Backscatter Diffraction in the Scanning Electron Microscope
507 29
111
111
101 101
111
101
.. Fig. 29.18 Inverse pole figures that show the same data as the pole direction. The X and Y pole figures show where the other <111 > poles
figures in . Fig. 29.9. The three directions X, Y, and Z (parallel to the plot. The smaller density of poles near the <110 > apex of the triangle
samples surface normal) are shown. The high density of poles plotted indicates that there are a few pixels with a <110 > direction parallel to
near the <111 > apex of the stereographic triangle indicates that many the sample surface normal
of the measured pixels had <111 > parallel to the sample surface normal
figure to an image. In order to do this we must first assign a figures (. Fig. 29.17) or inverse pole figures (. Fig. 29.18).
color to each direction in the inverse pole figure stereo- When inverse pole figure maps are displayed it is important
graphic triangle. A common way this is done is shown in to always include the reference direction, otherwise it is dif-
. Fig. 29.19. Inverse pole figure maps are then plotted by ficult or impossible for the viewer to make sense of the infor-
mapping the orientation of each pixel onto the color key mation shown. . Figure 29.20 is an inverse pole figure map
shown in . Fig. 29.19. . Figure 29.20 is an inverse pole figure with respect to the surface normal direction or the Z direc-
map produced from the same data that is shown as pole tion. Thus, the map is mostly blue, indicating that most of the
508 Chapter 29 · Characterizing Crystalline Materials in the SEM
grains have an <111 > direction normal to the sample sur- 29.2.7 Other Map Components
face. There are also some areas that are green in the map and
these areas have an <110 > direction parallel to the surface Once the orientations of the pixels in an array are known, it is
normal. The inverse pole figure map with respect to the X now possible to add additional information to the maps.
direction looks totally different to the Z inverse pole figure There are many possible components the can be plotted based
map. Both of these are shown in . Fig. 29.20. on EBSD data. One of the easiest components to add is that of
grain boundaries. Grain boundaries are the planes (which
intersect the planar surface as lines) that separate two regions
of different crystallographic orientations. It is fairly trivial to
IPF colouring Iron fcc
calculate the change in orientations between two pixels. If we
Y0
29 111 do this for an entire map we can then plot lines where the dif-
ference in orientations between two adjacent pixels exceeds a
predefined limit. It is typically assumed that a grain boundary
is represented by a change in orientation that exceeds 10°.
. Figure 29.20 has black lines plotted where the change in ori-
entation exceeds this limit and thus the map shows the size of
individual grains even though they are of consistent color due
to the grain orientations with respect to the sample normal.
.. Fig. 29.20 Inverse pole figure maps from the X-direction (left) and the Z-direction (right). This is the same data that is shown in . Figs. 29.9
and 29.10. Inverse pole figure maps show the pixel-by-pixel or spatial arrangements of the crystal orientations (Bar = 5 µm)
29.2 · Electron Backscatter Diffraction in the Scanning Electron Microscope
509 29
show up as single pixels surrounded by correctly indexed pix- the acronym for transmission electron backscattered diffraction,
els. The software searches through the data and where there is which is a rather non-physical description due to the inclusion
a single pixel of a different orientation surrounded by pixels of of both transmission and backscattered in the name) is the
a different but similar orientation, the single mis-indexed pixel transmission analog to EBSD is a way to achieve extremely high
is replaced by the average of the surrounding orientations. The spatial resolution for crystallographic orientation or phase map-
next step is to deal with the pixels that are not indexed. The ping (Keller and Geiss 2012; Trimby 2012). TKD is realized by
same procedure is applied as with the single mis-indexed pixels using an electron-transparent sample, as in the transmission
in that kernel math is used. Each individual pixel has six near- electron microscope, that is held at normal or near- normal inci-
est neighbors. If a cleaning procedure that requires six nearest dence with respect to the electron beam while a standard EBSD
neighbors to agree then the single un-indexed pixel is assigned camera is placed at the exit surface of the sample, as shown in
the average orientation of it’s nearest neighbors. These proce- . Fig. 29.21. Here the electron beam accelerating voltage must
dures allow the user to select the number of nearest neighbors be high and the sample must be thin in order for transmission of
that must agree and using less than six nearest neighbors allows the electrons to occur. The maximum beam energy typically
larger regions of pixels with no correct indexing to be replaced. available in modern SEMs is 30 kV, which requires that the sam-
Use of these procedures can be very dangerous and must be ple must be quite thin. The use of a thin sample limits the size of
disclosed when inverse pole figure maps are published. the beam interaction volume within the sample and immedi-
ately improves the spatial resolution of TKD. In addition, the
fact that the sample may be oriented normal to the incident elec-
29.2.9 ransmission Kikuchi Diffraction
T tron beam further improves the resolution to the point that
in the SEM 2-nm spatial resolution has been achieved in TKD. It is incredi-
bly fortuitous that the TKD patterns are extremely similar in
One of the limitations of EBSD is that the resolution is com- appearance to EBSD patterns, as shown in . Fig. 29.22, and can
promised by the fact that the patterns are formed by back- be collected with the same cameras and the same analysis soft-
scattered electrons and that the sample is highly tilted leading ware as is used for EBSD. There are two main disadvantages to
to the previously mentioned fact that the resolution perpen- TKD. First, it may be difficult to produce appropriate thin sam-
dicular to the tilt axis is much worse than that parallel to the ples. However, most laboratories will have access to a dual-plat-
tilt axis. The resolution can be greatly improved if the back- form FIB/SEM and thin samples prepared with FIB are perfectly
scattered volume is reduced and the geometrical factors are adequate for TKD. The second disadvantage is that when a small
reduced. One could image EBSD at reduced voltages to pixel size is needed, it is difficult to map larger regions. For
reduce the interaction volume, but this process has practical example, if a map step size of 4 nm is selected, a 1000 × 1000 pixel
limitations related to the need for increased sample prepara- map will only cover 4 × 4 μm. However, if orientation mapping
tion quality. Also, improved EBSD cameras would be needed of extremely fine grained material is needed, TKD may be the
to take advantage of lower voltage operation. Lower voltage only way to achieve the resolution needed.
operation does nothing to reduce the geometrical effects. A typical TKD map acquired at 30 kV of a thinned sam-
Transmission Kikuchi diffraction (TKD, although some in ple of polycrystalline Si layers in a semiconductor device is
the literature have referred to this method as t-EBSD, which is shown in . Fig. 29.23. This map was acquired with a step size
a b
.. Fig. 29.21 Detector and sample positioning for transmission detector. Note row of solid state detectors located on the detector for
Kikuchi diffraction. a The detector shown is an on-axis detector. collecting transmission images of the sample. b Sample and detector
Sample and detector positioned and inserted for TKD with an on-axis arrangement for TKD with a conventional style EBSD detector
510 Chapter 29 · Characterizing Crystalline Materials in the SEM
29
.. Fig. 29.22 TKD patterns collected at 30 kV from a thin sample of austenite. The imaging conditions and sample thickness result in either
typical-appearing Kikuchi patterns, as shown in the left image; or if the sample is very thin, spot patterns can be collected (right)
a b
.. Fig. 29.23 TKD of polycrystalline Si layers in a semiconductor device. These maps were acquired at 30 kV with a 6-nm step size. a Band quality
image of the Si layers. b Orientation map with respect to the growth direction of the polycrystalline Si layers
of 6 nm and demonstrates the superior resolution that can be their thermal history. One example of this will be illustrated
attained with think samples and the TKD method. with the Gibeon meteorite. There has been interest in
understanding the beautiful Widmanstatten pattern that is
seen in these meteorites and how this two phase structure
29.2.10 Application Example of ferrite (body-centered cubic crystal structure) and aus-
tenite (face-centered cubic crystal structure). Previous
pplication of EBSD To Understand Meteorite
A work had studied the formation of this structure and most
Formation of those studies had assumed that at high temperatures in
EBSD has found application in many materials studies from the parent asteroid the meteorite consisted of very large
ceramics, to semiconductors to metals and alloys. It has also grains of austenite. During the cooling of this meteorite in
has been applied to understanding metallic meteorites and space over many millions of years the austenite was assumed
29.2 · Electron Backscatter Diffraction in the Scanning Electron Microscope
511 29
.. Fig. 29.24 a EBSD inverse
a
pole figure map with respect
to the sample normal direction
that is constructed by tiling 220
separate 1 × 1-mm maps. The
interesting Widmanstatten pat-
tern can be seen in the large
ferrite plates that are running
nearly the length of the image.
This map contains both indexed
austenite and f errite; although
at this scale only the larger fer-
rite is visible. b This is the same
area as shown in . Fig. 29.23a,
but now we present the ferrite
as a band c ontrast or a measure
of the p attern sharpness and
the austenite in the colors of b
the inverse pole figure map
with respect to the sample
normal direction. The amazing
observation from this EBSD data
is that the a ustenite has the same
orientation throughout the large
22 × 10-mm area, which leads
to the interpretation that the
austenite is retained from the
original parent body
to have fully transformed to ferrite. On further cooling, the The ferrite pole figures are much more complicated and are a
ferrite, super-saturated with Ni, then was thought to decom- result of the many variants of ferrite that form from a single
pose to the two phase ferrite plus austenite structure. orientation of austenite.
However, EBSD has shown that this is not be the correct There are also regions in . Fig. 29.24a that are very fine
path for the observed microstructural evolution (Goldstein grained and difficult to resolve with the 3-μm step size used.
and Michael 2006). Further examination of the microstructure showed that these
. Figure 29.24a is a large-area EBSD map acquired by regions were extremely fine grained and required higher res-
mapping smaller areas of about 1 × 1 mm and tiling together olution than can be achieved with using bulk EBSD. Due to
220 of these tiles into a large area map that covers 22 × 10 mm the small feature size in these areas, TKD is an excellent
with a 3-μm step size. The sample was mechanically polished method to utilize. . Figure 29.26 is a secondary electron
using standard metallographic practice followed by a few image of a focused ion beam produced thin sample. Also
hours of vibratory polishing on colloidal silica. The entire shown is a scanning transmission electron image acquired at
map shown consists of a more than 25 million individual pix- 30 kV which demonstrates that the sample is sufficiently thin
els. The general microstructure at a low magnification is for the transmission of 30 kV electrons. . Figure 29.27 is the
clearly visible in . Fig. 29.24a. . Figure 29.24b shows a band resulting TKD map and phase information obtained from the
contrast image of the ferrite and the austenite as an inverse thin sample using an on-axis TKD detector. The step size for
pole figure map with respect to the sample surface normal. this image was 4 nm. It is now clear from these images that
Note that all of the austenite has the same or very close to the the fine-grained regions in . Fig. 29.24a consist of regions of
same orientation, as shown by the austenite all of the same single crystal austenite that can be seen in . Fig. 29.24b but
color in the inverse pole figure map. This is an important also regions of ferrite that have begun to decompose during
observation as the austenite could not have formed from pre- cooling to the equilibrium austenite plus ferrite that would be
cipitation from the ferrite but must be remnants of the origi- expected. The presence of twinned austenite precipitates is
nal large austenite grains found in the parent meteorite body somewhat surprising, but may be explained by some of the
at elevated temperatures early in the meteorites life. This is stress in the sample during transformation.
further demonstrated by the pole figures shown in This example shows how EBSD and TKD may be applied
. Fig. 29.25. The austenite pole figures show that there is only to complex microstructures and how the use of TKD is
a single orientation of austenite in the 22 mm × 10 mm area. extremely complementary to EBSD. The visualization of the
512 Chapter 29 · Characterizing Crystalline Materials in the SEM
Y
{100} {110} {111}
Z
X
29
Y
{100} {110} {111}
Z X
.. Fig. 29.25 Pole figures from the austenite (top) and the ferrite (bot- also observed from . Fig. 29.23b. The complexity of the ferrite pole
tom). In the austenite pole figures the single orientation of the austen- figures is due to the many crystallographic variants of ferrite that form
ite is shown by the arrangements of the poles that are shown. This was from the single orientation of austenite
a b
.. Fig. 29.26 A thin sample made for TKD of the fine two phase regions image of the thinned sample. b Scanning transmission electron image of
in the Gibeon meteorite. The sample was prepared with conventional FIB the sample at 30 kV
followed by low voltage ion FIB milling at 5 kV. a Secondary electron SEM
29.2 · Electron Backscatter Diffraction in the Scanning Electron Microscope
513 29
a b
.. Fig. 29.27 TKD phase and orientation maps acquired from the thin colored red and ferrite green. b Orientation map of both the austenite
sample shown in . Fig. 29.18. The data was acquired at a beam volt- and ferrite phases. Note the fine-scale twinning that occurs in the aus-
age of 30 kV with a step size of 4 nm. This data was collected using an tenite precipitates
on-axis TKD detector, as shown in . Fig. 29.13a. a Map with austenite
texture of the samples and an understanding of the fine by charging of the polymer material. It is satisfactory to
details present in the microstructure were extremely useful lightly conductively coat the specimen for EBSD analysis
in determining the cooling history and the microstructural provided the coating is kept as thin as possible while remain-
development in this meteorite. It should also not be lost on ing adequate for charge removal.
the reader that EBSD and the related TKD span a huge range Non-conductive samples may also be lightly coated for
of length scales from collecting large mm sized regions to analysis as discussed above. It is also possible to utilize the
using extremely small step sizes to elucidate the nm scalede- variable pressure mode of operation to reduce charging
tails of the microstructure. related artifacts. One must carefully choose the correct pres-
sure to be used as too high of a pressure will result in blurred
patterns due to scattering in the gas and too low of a pressure
29.2.11 Summary may not control the charging.
Proper positioning of the specimen within the SEM is criti-
EBSD and the related technique of TKD are an important cal. One must remember that the sample will be tilted to a high
part of crystalline materials characterization in the SEM. With angle for analysis. The high tilt required for EBSD will limit
EBSD the SEM can now be considered a complete materials how tall the sample is and how it must be mounted on the SEM
characterization tool that can not only take excellent quality sample stage. A sample that is too short may also present dif-
images of samples but can determine the elemental constitu- ficulties in positioning the sample in the proper location.
ents of the sample as well as the detailed crystallography.
EBSD Detector
Due to the limited space in most SEM sample chambers, it
is best to position the sample so that the sample is tilted
29.2.12 lectron Backscatter Diffraction
E appropriately and the area to be analyzed is in the field of
Checklist view. Once the sample position has been established the
EBSD detector should be introduced. The exact location of
Specimen Considerations the detector is not extremely important, but the sample to
EBSD requires a properly prepared sample that is securely detector distance should be sufficiently short so that a large
attached to an appropriate support as the sample will be tilted solid angle can be obtained and the brightest part of the pat-
to high angles to facilitate EBSD pattern acquisition. The tern located in the upper half of the detector image. It is
sample surface should be free of artifacts due to sample prep- very important to keep in mind the position of the sample
aration. Due to the high sample tilt and the generally long and the detector as there are two kinds of people who use
acquisition times required for EBSD, carbon tape is not a EBSD: Those who have hit the EBSD detector and those
good choice as it tends to creep allowing the sample to move who will. So be careful when moving the sample and the
causing drift related image issues. Samples mounted in epoxy detector. A chamberscope is absolutely required for safe
metallographic mounting materials are subject to drift caused EBSD operation.
514 Chapter 29 · Characterizing Crystalline Materials in the SEM
References – 528
30.1 Introduction the sample in terms of ion implantation and loss of crystal-
line structure. Sputtering is the process that removes atoms
The use of focused ion beams (FIB) in the field of electron from the target. Secondary electron production is important
microscopy for the preparation of site specific samples and as images formed with secondary electrons induced by ions
for imaging has become very common. Site specific sample have some important advantages over electron-induced sec-
preparation of cross-section samples is probably the most ondary electron imaging. Finally, it is important to realize
common use of the focused ion beam tools, although there that it is impossible to have an ion beam interact with a sam-
are uses for imaging with secondary electrons produced by ple without some form of damage occurring that leaves the
the ion beam. These tools are generally referred to as FIB sample different than before the ion irradiation.
tools, but this name covers a large range of actual tools. There A schematic diagram of the interactions is shown in
are single beam FIB tools which consist of the FIB column on . Fig. 30.1. Here an energetic ion is injected into a crystalline
a chamber and also the FIB/SEM platforms that include both sample. The ion enters the sample at position 1. The ion is
a FIB column for sample preparation and an SEM column for then deflected by interactions with the atomic nuclei and the
30 observing the sample during preparation and for analyzing electron charges. As the ion moves through the sample it has
the sample post-preparation using all of the imaging modali- sufficient energy to knock other atoms off their respective lat-
ties and analytical tools available on a standard SEM column. tice positions as shown at position 2. The target atoms that
A vast majority of the FIB tools presently in use are equipped are knocked off their atomic positions can have enough
with liquid metal ion sources (LMIS) and the most common energy to knock other target atoms off their atomic positions
ion species used is Ga. Recent developments have produced as shown at position 3. Some of the atoms that have been
plasma sources for high current ion beams. The gas field ion knocked from their atomic positions may reoccupy a lattice
source (GFIS) is discussed in module 31 on helium ion position or may end up in interstitial sites. There can also be
microscopy in this book. lattice sites that are not reoccupied by target atoms and are
This chapter will first review ion/solid interactions that left as vacancies. Both interstitials and vacancies are consid-
are important to our use of FIB tools to produce samples that ered damage to the crystalline structure of the sample as
are representative of the original material. This discussion shown in position 4. Most of the time, the original beam ion
will then be followed by how FIB tools are used for special- will end up coming to rest within the sample. This is termed
ized imaging of samples and how they are used to prepare ion implantation and is shown at position 5. Ion implantation
samples for a variety of SEM techniques. results in the detection of the ion beam species in the sample.
Many of the collision cascades will eventually reach the sur-
face of the sample. Sufficient energy may be imparted to
30.2 Ion–Solid Interactions knock an atom from the surface into the vacuum. This pro-
cess is called sputtering and results in a net loss of material
It is important to understand some of the ion-solid interac- from the sample as shown in position 6. At the same time
tions that occur so that the user can appreciate why certain when the ion is either entering or leaving the sample, second-
methods and procedures are followed during sample prepa- ary electrons are generated that are useful for producing
ration. There are many events that occur when an energetic images of the sample surface scanned by the ion beam. It is
ion interacts with the atoms in a solid, but for the case of SEM important to remember that scanning an energetic ion beam
sample preparation and ion imaging we are mainly interested over the surface of the sample will always result in some
in sputtering, secondary electron production and damage to damage to the sample. Understanding the interaction of ions
Collision
cascade
4
5
Interstitial
Implanted Ga atom
30.3 · Focused Ion Beam Systems
519 30
source or a plasma source to produce ion beams with variable
100 current to allow both large volume removal and fine scale
Ga polishing of the sample. Both the LMIS and the plasma
Xe
source have advantages and disadvantages. The most com-
80 mon LMIS produces Ga ions, while it is common for the
plasma sources to utilize inert gases like Ne, Ar, or Xe. Ga ion
sources have higher brightness as compared to the plasma
Ion range (nm)
60
sources and thus have higher current densities in the focused
ion spots. Ga is a fairly reactive element, however, and may
result in artifacts when implanted into some materials.
40
Plasma sources have lower brightness but higher overall cur-
rent than the LMIS sources resulting lower current densities
in the focused spot but much higher total current that allows
20
much faster material removal rates when compared to LMIS
ion sources. Both the LMIS and the plasma ion source utilize
similar optics to produce focused scanning beams of ions.
0
0 10 20 30 40 50 60 70 80 . Figure 30.3 shows a schematic of a typical ion column. This
Target atomic number column resembles a simple SEM column with the exception
that the magnetic lenses that are used to focus electrons are
.. Fig. 30.2 Calculated range for 30-kV Xe and Ga ions as a function
of atomic number
Ion source
with the target material is helpful in reducing the amount of
damage to acceptable or tolerable levels through appropriate
sample preparation techniques. For a complete review of Beam acceptance aperture
ion–solid interactions see Nastasi et al. (1996).
In module 1 on electron–beam specimen interactions, the
range that an electron travels in a sample is discussed. Generally
for medium-energy electrons (15–20 keV) the electron range
in the transition elements is on the order of 1 μm. The ranges Lens 1
of heavy ions like Ga and Xe are extremely short when com-
pared to electron ranges for similar energies. . Figure 30.2 is a
plot of the ion ranges for Xe and Ga as a function of the atomic
number of the target material. Note that the range for either Ga
or Xe ions with an energy of 30 keV is generally much less than
50 nm. Thus, ions travel very short distances in solid targets
and as a result the near surface region of the sample is where
the ion interactions take place and where we expect to see the
crystalline sample damage discussed in . Fig. 30.1 (Nastasi Beam blanking plates
et al. 1996; Ziegler and Biersack 1985).
not capable of focusing the heavier ions. For ions, the lenses
are electrostatic and require high voltages to focus energetic
ions due to their relatively large mass to charge ratio.
Sample stages must be accurate and reproducible for easy
and efficient FIB processing of samples. There can be multi-
ple stage moves during the preparation of samples in a FIB
tool and each move should be reproducible so that the opera-
tor can easily return to the region of interest. Stage accuracy
and reproducibility has become more important as auto-
mated routines have become common place during the pro-
duction of samples or during sequential milling and imaging
operations (Giannuzzi 2006; Orloff et al. 2003).
30
30.4 Imaging with Ions
.. Fig. 30.4 Spot burns on a tungsten coated silicon wafer. The beam
Although FIB is usually used to remove material via sputter- was held stationary at each point for 10 s. Top row from left to right: 24
ing, ions produce a large yield of secondary electrons. The pA, 80 pA, 0.23 nA, 0.43 nA, 0.79 nA, and 2.5 nA. Bottom row from left to
right: 9 nA, 23 nA, 47 nA and 65 nA (Bar = 200 µm)
secondary electrons signal can be collected and imaged just
like secondary electrons produced with an electron beam,
which means we can use all of the detectors that we are very where the ion beam is put in spot mode and left stationary so
familiar with from electron beam imaging. This leads to the that the substrate is milled away. The beam size is then some
use of the FIB as an imaging tool sometimes referred to as measure of the size of the spot including the halos around the
scanning ion microscopy. Secondary electrons produced milled area as ion beams suffer significant aberrations as occur
with ions are referred to as ion induced secondary electrons with lenses in the SEM. In both the LMIS and the plasma
(iSE). FIB columns are all equipped with one or more iSE cases, the ion columns are capable of producing symmetric
detectors with the most common one being the ET detector ion beams with the exception that at the larger currents vari-
as used in the SEM. iSE imaging has some advantages over SE ous optical aberrations become dominating and result in a less
imaging in the SEM. First, the ion beam produces many well defined ion beams.
more SEs per incident particle than does a similar current The collection of secondary ions requires an additional
electron beam resulting in a high signal, low noise image. detector that is sensitive to secondary ions and can reject the
Also, it is interesting to note that the iSE signal collected in a signal produced by the secondary electrons. These secondary
FIB is free of the backscattered electron component that ion detectors are often now optional on modern instruments.
reduces contrast in electron beam induced SE images. iSE . Figure 30.5 is a comparison between secondary electron
imaging of surfaces shows topographical contrast that is imaging induced by electrons and secondary ion imaging
familiar to anyone who has operated an SEM. iSE imaging of induced by ion scanning. The sample consists of a tungsten
crystalline samples can produce very striking high contrast wire and a human hair. The electron image shows good surface
grain images due to the higher propensity for ions to channel detail, but the uncoated hair is charging and the surface infor-
along specific crystal planes resulting in a varying iSE yield as mation is somewhat obscured. The secondary ion image shows
a function of grain orientation. no charging artifacts on the human hair but also shows an
During ion imaging of a sample there are other signals increased level of surface detail due to the small escape length
produced that are of limited use. The interaction with the of the secondary ions. One must always remember that imag-
sample causes secondary ions to be ejected from the sample. ing with ions will always cause some level of damage within the
These can be used to form images. However there are few sample (Giannuzzi 2006; Mayer 2007; Michael 2011).
secondary ions produced as compared to the iSE and there- . Figure 30.6 is an electron-induced secondary electron
fore the signals tend to be noisy requiring either longer scan (SE) image of free-machining brass coated with layers of Cu,
times or higher beam currents to be used, both of which will Ni, and Au. Note that free-machining brass contains particles
result in increased levels of sample damage during imaging. of Pb. . Figure 30.6 shows the contrast that we have come to
Resolution of scanning ion microscope images is not just a expect from electron beam induced SE imaging. The higher
function of the beam size that is generated by the ion column. atomic number regions appear brighter as a result of higher
The ultimate resolution of a scanning ion image is a convolu- secondary electron yield that results from the SE2 contribu-
tion of the beam size and some measure of the rate at which tions from the unavoidable backscattered electrons (BSEs).
the sample is milled. Resolution is worse for materials that mill . Figure 30.7 is an iSE image (30 kV Ga) of the identical area
quickly and better for those materials that have a slower sput- of the sample. The various layers are immediately obvious
ter rate (Orloff et al. 1996, 2003). Just like the SEM, as the cur- because the ion channeling contrast is quite strong and the
rent in the probe is increased, the beam size increases. grain structure of each layer is revealed. It can also be
. Figure 30.4 shows spot burns from an LMIS (Ga) column observed that the Pb region and the Au layer are no longer
30.5 · Preparation of Samples for SEM
521 30
.. Fig. 30.5 Images of a tungsten wire (left) and a human hair (right).
a Secondary electron image induced by 5-kV electrons. b Ion induced
secondary ion images with 30-kV Ga ions. Note the reduced charging
in the ISE image and the increased surface detail visible in the second-
ary ion image
bright relative to the brass as in . Fig. 30.6. . Figure 30.7 .. Fig. 30.7 Ion beam (30-kV) induced secondary electron imaging
demonstrates that the iSE yield is not a simple function of the of a free-machining brass (Cu-Zn alloy) that has been coated with lay-
ers of Cu, Ni and Au. Note that the various layers are much more easily
target atomic number (Joy and Michael 2014). Also,
visualized due to the high contrast crystallographic contrast. Also, note
. Fig. 30.7 demonstrates the very strong grain contrast that that the brightness of the phases can no longer be interpreted strictly
can be observed in many crystalline materials. This grain by atomic number. Here the Pb and the Au regions appear with lower
contrast is due to the way that the crystallography of the sam- signal levels than does the brass or the Cu and Ni layers
ple impacts the penetration depth of the primary beam ions
and therefore the iSE yield. Small changes in the ion beam
incident angle can change the grain contrast that is observed. 30.5 Preparation of Samples for SEM
Thus, if you are trying to determine the nature of the contrast
that is observed in a crystalline sample it is a simple matter to FIB–based sample preparation for SEM is a large field due to
tilt the sample 2–4° and observe how the contrast changes. If the versatility of modern SEMs and the many techniques that
the contrast is due to ion channeling then the contrast are utilized. One of the most common uses of FIB is for sub-
between grains should change (Giannuzzi and Michael 2013). tractive processing of the sample in that the FIB is used to
522 Chapter 30 · Focused Ion Beam Applications in the SEM Laboratory
sputter site specific regions of the sample to produce cross equipped with gas injectors that can be used during sample
sections. It is also common to produce samples that are preparation. Common precursor gasses used can deposit
manipulated from the bulk for study. One example is the tungsten, platinum, or carbon. Each of these materials works
extraction and production of thin samples for transmission quite well to protect the region of interest from the ion beam.
Kikuchi diffraction as discussed in module 29 on electron The precursor gasses are delivered to the sample surface
backscatter diffraction. FIB has become an indispensable through a small needle that is placed in very close proximity
tool in the production of electron transparent samples for to the area of interest. Either the ion beam or the electron
STEM/TEM and now the FIB can produce samples that are beam can be scanned over the area of interest. Some of the
sufficiently thin for transmission imaging in the SEM at gas molecules that are delivered through the needle absorb
30 keV using the STEM-in-SEM technique. FIB sample prep- on to the sample surface where combined action of the pri-
aration is applicable to many materials classes that are imaged mary beam (electron or ion) and the secondary electrons
in the SEM including polymers, metal, semiconductors, and produced by the interaction of the beam with the sample
ceramics. decomposes the absorbed gas. This leaves behind a deposit
30 that contains the desired material but also includes the ion
beam species (if ions were used) and some residual organics
30.5.1 Cross-Section Preparation from the precursor. Due to the short range of ions in materi-
als, the protective layer need not be very thick; but typically
In materials characterization it is often of interest to image a most applications use about 1 μm to provide protection and
section of the sample that is not visible from a planar section. for ease of subsequent milling of the sample. . Figure 30.9
In this case, the FIB is used to mill away material to expose a shows a cross section produced in a sulfide copper test cou-
cross sectional view of the sample, as shown schematically in pon where the objective of the experiment was to measure
. Fig. 30.8. There are many ways to accomplish this and in the rate of sulfide growth in accelerated aging conditions. The
many modern FIB tools this process is fully automated and first step (. Fig. 30.9a) is to deposit the protective platinum
the user must simply indicate where the cross section is to be layer on the area to be sectioned. The goal is to have the com-
made. The following example will show the steps that are pleted cross section positioned under the platinum protec-
needed for a cross section to be produced. tive layer. A coarse first cut is made with a large current ion
As was discussed earlier, any time the sample is exposed beam, the result of which is shown in . Fig. 30.9b. Although
to the ion beam damage will occur. It is necessary to image the cross section is relatively clean at this point it is not ade-
the sample during preparation so the easiest way to eliminate quate for quality SEM imaging. Further polishing of the cross
the damage to the sample from the beam is to place a protec- section is completed with lower current ion beams in this
tive layer over the area to be cross sectioned. FIB tools come case . Fig. 30.9c was completed with a 1-nA ion beam and
. Fig. 30.9d is the cross section after final polishing with a
300-pA ion beam. The total time to produce this cross sec-
tion is about 20 min. The completed cross section imaged at
Ion beam higher resolution is shown in . Fig. 30.10. Note that all of the
important microstructural features are easily observed in this
very smooth polished cross section.
Electron beam FIB prepared surfaces can be very smooth and nearly fea-
tureless. The relative brightness between different materials
can often provide sufficient image contrast. Materials con-
trast by itself may be insufficient and a method to enhance
the contrast may be needed. One way to do this is to intro-
duce gasses near the sample that react with the ion beam and
the sample surface to etch the FIB polished surface. One
chemical that does this for semiconductor devices is trifluor-
acetic acid (TFA) that etches oxides and nitrides preferen-
tially to silicon and metals. . Figure 30.11 demonstrates the
use of TFA for enhancing the contrast in FIB milled surfaces.
The sample was prepared using standard cross sectioning
Stair step cut procedures in the FIB. After cross sectioning the sample was
rotated and tilted so that the milled surface was normal to the
ion beam. A low beam current is selected and then the gas is
.. Fig. 30.8 Schematic of an FIB-prepared cross section. A stair-step introduced through a fine needle while the ion beam is
is milled using the ion beam and then the exposed section that is per- scanned over the milled surface. While milling, the sample
pendicular to the original sample surface is polished with a series of
lower current ion beams. The cross section can be immediately imaged
image provides an etching progress monitor so that the etch-
with the SEM beam that is at some inclined angle with respect to the ing progress can be terminated when the correct degree of
ion beam etching has been achieved.
30.5 · Preparation of Samples for SEM
523 30
a b
5 mm
5 mm
c d
5 mm
5 mm
.. Fig. 30.9 Required steps to making a quality cross section of the 7-nA Ga ion beam. c Further polishing with a 1-nA Ga ion beam d Final
surface of a corroded copper test coupon in the FIB. a Deposition of polishing with a 300-pA Ga ion beam to produce the completed cross
platinum protective layer. b Coarse milling of the cross section with a section
Electron beam pt
Copper sulfide
Au marker layer
200 nm Copper sulfide
copper
524 Chapter 30 · Focused Ion Beam Applications in the SEM Laboratory
a b
c d
.. Fig. 30.12 Gibeon meteorite sample produced by the lift-out method placed across the large gap in the middle and epoxied in place. d Sample
to produce large area samples suitable for EBSD, TKD, EDS, or STEM imag- after final milling of the surface. e Sample arranged on a pre-tilted sample
ing. a Milled sample ready for ex situ lift-out. b Lift-out sample on the holder for EBSD
end of the micromanipulator needle. c Cu support for the sample. It was
526 Chapter 30 · Focused Ion Beam Applications in the SEM Laboratory
a b
.. Fig. 30.14 Image reconstruction of a plated stainless steel test reconstructed as it is the collected data. b Reconstructed slice along
coupon. Each section was milled perpendicular to the sample surface. the red plane shown in d. This image is reconstructed once the slice
These reconstructions were made from a series of 360 milled slices and thickness is known. The resolution in this direction is limited by the FIB
required approximately 3 h to collect. The width of the milled area is milled slice thickness. c This is a reconstructed image of a slice parallel
20 μm. a A secondary electron image of one milled cross section that to the sample surface shown in blue in d. d Schematic of milled volume
is the green orientation shown in d. This image does not need to be
528 Chapter 30 · Focused Ion Beam Applications in the SEM Laboratory
27.8 µm
30 16.6 µm
15 µm
b
.. Fig. 30.16 EBSD 3D reconstruction of a tin whisker from serial
sectioning data in the FIB. The acquisition required 75 200-nm-thick
sections and took nearly 48 h to complete sectioning and data
acquisition
References – 539
Wavelength (nm)
1/43
While the scanning electron microscope (SEM) is the most
widely used tool for high performance imaging and micro- 0.01
1/86 Protons
analysis, it is not the only option and may not even always
be the best instrument to choose to solve a particular prob- 0.001 1/350 1 pm
He+
lem. In this chapter we will discuss how, by replacing the
Ga+
beam of electrons with a beam of ions, it is possible to pro- 0.0001
duce a high performance microscope which resembles an
10-5 10 fm
SEM in many respects and shares some of its capabilities
1 10 100 1000 104 105
but which also offers additional and important modes of
31 operation. Energy (eV)
1µm 1µm
50 nm
20 nm
50 nm
31
100 µm
1.5 mm
100 nm
while gallium (Ga+) is widely used for ion beam machining of higher accelerating voltage to operate than does a lighter ion,
materials. Historically the earliest ion beam microscope (Levi- and this heavy ion bombardment may cause significantly
Setti 1974) used protons (H+), while more recently argon more damage to the specimen under examination. Heavy
(Ar+), neon (Ne+), and lithium (Li+) beams have all also been ions also usually have multiple ionization states and these
put into use for nanoscale imaging and fabrication. At present may result in emitted beams whose incident energy is effec-
negatively charged ions are much less widely employed than tively distributed across several values spanning a wide
the corresponding positive species, but it can be safely antici- range. This energy spread, in turn, may result in chromatic
pated that they will eventually be employed for special pur- aberrations in the lenses which may then further degrade the
poses because of the additional versatility that they can offer. achievable resolution.
Although in principle any ion could be employed for In addition, while all energetic ions cause sputtering of
microscopy, using a heavier ion requires a correspondingly the target atoms to some degree and also implant into the
31.2 · Generating Ion Beams
533 31
ionized due to the high electric field gradient and acquired a
positive potential which then resulted in them being acceler-
ated away from the tip and towards the viewing screen where
they formed into an image. On October 11, 1955, Muller and
his students were able to demonstrate for the first time ever
the direct observation of an atomic structure. Almost 20
years later Professor Riccardo Levi-Setti of the University of
Chicago developed a modification of Muller’s ion source
which allowed it to generate ion beams that could be focused,
scanned, and used for imaging in the same way as electrons
in a conventional SEM (Levi-Setti 1974). It was this develop-
ment that became the basis for the source for the present
helium ion microscope (HIM).
In present-day ion beam microscopes, the emitter is once
again fabricated into the form of a needle, but now the exact
size and shape of this tip is very carefully optimized. Using
patented, and proprietary, procedures developed by the Zeiss
company the emitter tip is shaped and sharpened until it con-
tains just three atoms (Notte et al. 2006). This “trimer” con-
100 nm figuration is inherently more stable than any more random
Field Of View
CARL ZEISS SMT 800.00 nm 100.00 nm
Dwell Time
0.5 us
Date: 4/7/2010
Time: 3:19 PM
arrangement and also ensures that the maximum emitted ion
Mag (4×5 Polaroid)
142,875.00 X
Blanker Current
0.2 pA
Line Averaging
128
Acceleration V
36.5 kV
current is directed parallel to the axis of the ion beam. A
carefully placed, moveable aperture can then be used to select
.. Fig. 31.5 High resolution helium ion imaging with high depth-of- any one of the three “trimer” ion emission peaks to be used as
field of a soft tissue sample, human pancreatic cells (Sample courtesy of
the beam source for the instrument. The available ion beam
Paul Walther, Univ. of Ulm)
current varies with the magnitude of the helium, or other gas,
pressure and can reach values as high of several hundred
target, which can locally alter the target composition, ions of picoamperes.
higher mass, such as Ga+ and above, have higher sputtering Low energy ions, i.e., those with less than about 1 MeV of
rates that may substantially alter the target. While acceptable energy, are not significantly affected by magnetic fields so all
for some tasks such as thinning or machining materials, the of the lenses must be electrostatic in type rather than mag-
level of damage per incident heavy ion is undesirable for netic. The ion beam then travels along the microscope col-
imaging purposes because such significant local alteration umn until it reaches the specimen where its interaction
occurs that the fine spatial details of the specimen are lost generates ion-induced secondary electrons (iSE) as well as, in
before an image representative of the original material can be some cases, other signals. The ion emitter is adequately stable
successfully captured. and provides a bright signal for periods from 5–10 h before
the emitter needs to be re-optimized. Once the available
beam current becomes too low in intensity, or too unstable to
31.2 Generating Ion Beams be useful, then the tip must be reformed. This can be done in
situ by the operator using an automated procedure which
The approach now most commonly employed for producing takes some 10–20 min to complete.
a beam of ions for use in microscopy is a development of the Changing from the familiar He+ beam to a Ne+ beam, or
method originally employed by Prof. Erwin Muller at Penn to some other source of emission, requires that any residual
State University in the 1940s and 1950s (Muller 1965). gas in the chamber must be first pumped away. The desired
Muller’s device was a sealed metal cylinder containing helium new gas of choice can then be injected into the chamber, and
gas which was cooled to a temperature of just a few degrees the system can be brought back into operation by reforming
Kelvin (K). At one end of the cylinder was a sharply pointed the “trimer” as described earlier. The usable overall lifetime
metal needle connected to a power supply capable of sup- of these emitters is typically of the order of many months
porting a positive voltage of up to a few kilovolts, while at the when they are treated with reasonable care and attention.
other end was a fluorescent imaging screen. When neutral Although in many ways operating a helium ion beam
helium atoms drifted towards the needle they became (HIM) microscope is similar to operating a conventional
534 Chapter 31 · Ion Beam Microscopy
SEM, to fully optimize the performance of the HIM it is nec- 31.3 Signal Generation in the HIM
essary to pay more attention to certain details. The ion source
offers just two adjustable controls—the “extractor” which Energetic electrons and ions travel through solid materials
determines the field at the tip and so controls the emission of while undergoing elastic and inelastic scattering events
the source, and the “accelerator” which determines the land- until they either deposit all of their initial energy and come
ing energy of the beam on to the specimen surface. The to a halt, or are re-emitted from a sample surface and sub-
extractor module “floats” on the top of the potential deter- sequently escape while generating secondary electrons and
mined by the accelerator setting, and both the brightness and backscattered electrons or ions as they leave. Every such
the stability of the ion source are affected by the extractor beam particle trajectory is unique and so it is not possible
module settings. As the extractor potential is increased, the to predict in advance how deep, or how far, any particular
emission current rises before reaching a plateau at the so- incident ion or electron might travel. Examples of Monte
called “best imaging voltage” (BIV), which for helium ions Carlo simulations for ion beam trajectories are shown in
occurs at a field strength of about 44 V/nm depending on the . Fig. 31.6.
geometry of the tip itself. Exceeding the BIV will make the The most probable depth reached by the beam, and the
beam less stable, and can result in so much damage to the horizontal spread of the beam, can both be estimated using
31 emitter that it may become necessary to reform the tip again the formula developed by Kanaya and Okayama (1972)
before reliable operation can be restored. The “accelerator” which assumes that the range R depends only on the incident
control determines the actual landing energy of the ions on energy E of the incident particle and the density ρ of the
the specimen. For a He+ beam this energy is usually in the material through which the beam is traveling. The “K–O”
range 25–45 keV while for a heavier ion such as Ne+ the cor- range is then given as
responding value is more typically in the 20–35-keV range.
In either case the yield of secondary electrons increases with RK −O = κ E p / ρ (31.2)
the accelerator setting, but continuous operation with the
system at energies of 45 keV or higher may cause problems where RK-O is the beam range (in nm), ρ is the density of the
such as insulator breakdowns and discharges. It is therefore target material (in g/cm3), k is a constant depending on the
good practice to record and save all the experimental param- choice of particle, i.e., electrons or ions, and P is a scaling
eters likely to be encountered. constant. For example, when using a helium ion beam then
H+ Ga+
250 nm 80 nm Ne+ 40 nm
2
4
250nm 50nm
He+ Ar+
250 nm 50 nm
.. Fig. 31.6 Monte Carlo ion beam trajectory simulations for various ion species. E0 = 40 keV
31.3 · Signal Generation in the HIM
535 31
where ε is a constant whose value depends on the target
“K-O” range for Ions and Electrons material, E is the instantaneous energy of the incident
106 charged particle, and s is the distance travelled along the
electrons trajectory. The ion-generated SE yield is always larger than
105 the corresponding SE yield from electrons because ions
Range (nm)* Density (gm/cc)
2
Si
Ar+
1.5
iSE yield
1
H+
Ga+
0.5 He+
electrons
31 0
10
Beam Energy (KeV)
100
.. Fig. 31.8 Yield of secondary electrons from Si for electrons and vari-
ous ions as a function of energy
Field of View Dwell Time Mag (4x5 Polaroid)
CARL ZEISS SMT 10.00 um 1.00 um 50.0 us 12,700.00 X
Working Dist Image Size Blanker Current Detector
12.1 mm 1024x1024 0.7 pA PrimaryETDetector
31.4 urrent Generation and Data
C
Collection in the HIM .. Fig. 31.9 High spatial resolution and high depth of field HIM imag-
ing of Ga-ion-beam etched Al-Cu aligned eutectic alloy. Vertical relief
approximately 5 µm. (Bar = 1 µm) (A. Vladar and D. Newbury, NIST)
A typical ion beam microscope operates at energies selected
between 10 and 35 keV, and generates an incident beam cur-
rent of the order of 0.1 pA to >100 pA at the specimen. These limited penetration of the ion beam into materials provides
ions interact with the specimen of interest generating an iSE highly detailed images of surface, and near-surface features
secondary electron signal which can be collected by an are visible in HIM images that would likely never be evident
Everhart–Thornley (ET) detector very similar to that used in in a conventional SEM image.
the conventional SEM. The signal is then passed through an The strategy for operating the HIM is different than that of
“analog to digital” (A/D) convertor for real-time display and a conventional SEM because the incident beam energy is gen-
for storage. Because the incident ion beam currents are quite erally held fixed at the highest possible energy, typically in the
low, the A/D convertor, which samples the incoming iSE sig- range 30–40 keV, because this simultaneously optimizes both
nal at a fixed repetition period of 100 ns, will likely completely the signal-to-noise ratio and the image resolution. When there
miss many of these sparsely generated signal pulses. To over- is a requirement to examine sub-surface detail this can best be
come this problem, an additional image control labeled as achieved by exploiting the inevitable removal of surface layers
“image intensity” (II) has been added to the familiar SEM by the ion beam as it “rasters” across the sample. Material can
brightness and contrast controls. This “image intensity” con- then be removed at the rate of a few tens of nanometers per
trol allows the operator to choose between (a) the true aver- minute, with images stored every few seconds to yield a full
aging mode in which N successive A/D conversions are made three-dimensional reconstruction of the sampled volume.
and the total yield is then divided by N and (b) the true inte- Charging is an inevitable problem for the HIM. The high
gration mode in which N successive A/D conversions are SE coefficient of ion beams tends to cause positive charging at
summed and that value is then reported, or any arbitrary set- the surface, which is further exacerbated by the positive
ting between these two extremes. (J. Notte 2015, personal charge injected by the positive He+ ions. The simplest and
communication). The addition of this illumination control most reliable technique to control such positive charging is to
provides considerable additional imaging flexibility while periodically flood the specimen surface with a very low energy
ensuring that the usual “brightness” and “contrast” controls electron beam so as to re-establish charge balance before
are still able to determine the overall appearance of the image. starting the next scan, although this approach requires care to
In practice, operating the HIM is generally similar to eliminate both under- and over-compensation. An alternative
operating an SEM, but the HIM achieves superior image res- approach is to inject air into the specimen chamber through a
olution and contrast. The small beam probe diameter and the small jet positioned just a short distance away from the desired
much enlarged depth of field together produce highly sample region. This is easy to implement and requires little
detailed images of even the most complicated three dimen- supervision once an initial charge balance has been achieved.
sional structures, as shown in . Fig. 31.9 for Ga-ion-beam The HIM operating mode equivalent to the familiar SEM
etched directionally-solidified Al-Cu eutectic alloy. The “backscattered electron signal” is “Rutherford backscattered
31.5 · Patterning with Ion Beams
537 31
Ni
.. Fig. 31.10 Reaction zone at Sn-Ni interface as imaged with ORION .. Fig. 31.11 Ion channeling contrast observed in deposited copper
Plus HIM from backscattered helium striking the annular microchannel showing the distinct crystal grains. The field of view is 3 μm, and the
plate. Field of view is 30 μm. Material composition is easily distinguished incident beam was helium at E0 = 30 keV (Bar = 200 nm) (Imaged with
by atomic number contrast (Bar = 2 µm) (Sample provided by F. Altmann the E–T detector using the ORION Plus, courtesy of Carl Zeiss)
of the Fraunhofer Institute for Mechanics of Materials, Halle, Germany;
Image courtesy of Carl Zeiss)
of the ion beam itself is also one to two orders of magnitude
smaller than for an electron beam making it readily possible
imaging (RBI)” (Rutherford 1911). Some fraction of the inci- to retrieve crystal orientation data from regions at the bot-
dent ion beam is deflected though an angle greater than 90° tom of trenches and holes while still having adequate signal
as a result of high angle elastic scattering events. This elastic intensity with which to work.
scattering generates a signal carrying information about the
crystalline geometry and the orientation of the specimen. An
example of the Ni-Sn reaction zone at an interface between 31.5 Patterning with Ion Beams
Ni and Sn to form intermetallic compounds as imaged with
backscattered He ions is shown in . Fig. 31.10. The RBI sig- When ions strike a target they always remove a finite amount
nal increases in magnitude as a function of the atomic num- of material from the top surface by the process called “sput-
ber of the target material, but is also affected by a periodic tering,” which involves ion-atom collisions that dislodge
variation of the RBI signal in which the signal falls to a much atoms and transfer sufficient kinetic energy so that a small
lower intensity whenever the outer shell of electrons is filled. fraction escape the sample. While this is sometimes a prob-
This behavior occurs at the atomic numbers 2, 8, 18, and son lem when imaging is the main application of a microscope,
on, and so rather restricts the quantitative utility of this mode the ability of ion beams to remove material in a controlled
as an analytical tool. fashion from a surface is of considerable and increasing
When ions travel through a crystalline material they may value. At its simplest this capability can be applied to gener-
experience “channeling contrast,” which is a variation in the ate simple structures such as arrays of holes, or to pattern
signal level that is dependent on the orientation of any cur- substrates. The typical beam choice for this process is gallium
rently placed crystalline target material to the incident beam. because of its high sputtering coefficient, but residual gallium
When using a helium ion beam, signal variations of 40 % or becomes embedded in the material being machined. Neon is
more in magnitude can be obtained in this way so producing also a useful candidate for the gas as it can remove material at
highly visible maps of sample crystallography, as shown in about 60 % of the speed of gallium but without permanently
. Fig. 31.11 for a polycrystalline copper sample. A similar depositing anything into the target material. For the very
effect also occurs for electrons but the corresponding con- softest or most fragile materials a helium ion beam can also
trast is an order-of-magnitude smaller at approximately pattern at a slow but acceptable rate.
2–5 %. The utility of this mode of operation for analytical An important current use of this technology relies on
purposes is enhanced by the fact that the convergence angle “Graphene” as the material of choice. Graphene comes in the
538 Chapter 31 · Ion Beam Microscopy
form of flexible sheets which may be only by a few atom layers Step 4
in thickness and yet can readily be cut and shaped as required. If these are the first observations of the sample which is
Graphene provides the basis to fabricate nanoscale semicon- now in the HIM, then try recording images from some
ductor devices, and varieties of other active structures. expendable region of the specimen at several different
magnifications and currents to confirm whether or not
ion-induced beam damage is going to be a problem. In
31.6 Operating the HIM any case always start imaging at the lowest possible
magnification and work up from there. Working “high to
Using a He+, or some other, ion beam for imaging materials low” is likely to be unsatisfactory because the damage
is somewhat different from, although not really much more induced at higher magnification will be obvious as the
complex than, conventional imaging with electrons. However magnification is reduced. Also check the imaged field of
some planning prior to turning on the ion beam will help by view from time to time for signs of damage, or any
indications of drift.
Step 1
31 Choose the initial ion beam to be employed. If only one Step 5
beam type is available, for example, Ga+, then this step is Maintain a constant check on the incident ion beam
not crucial because it should imply that the required current. If this is falling, or if the images are becoming
operating choices have already been optimized. noisy and unstable, then the emitter tip should be
However, if several different ion beam species are reformed before continuing.
available then a planned strategy becomes necessary. If
there is only a limited amount of the sample then the
user should start with the softest ion beam available—
helium—so as to minimize damage in the near-surface minimizing beam damage optimizing conditions, and
region of the specimen—and image as required. Only improving specimen throughput.
when this set of images is completed should a more
aggressive beam choice be tried.
31.7 hemical Microanalysis with Ion
C
Beams
Step 2
Select the desired beam energy. The best choice is always The ability of a microscope to image a material while simul-
the highest voltage available because this will enhance taneously performing a chemical microanalysis of the mate-
image resolution, and the ion source will be running at its rial is of great value in all areas of science and technology. As
optimum efficiency. Even at modestly high energies, for a result a high percentage of all SEMs are equipped with
example 45 keV, the depth of penetration of ions into X-ray detectors for the identification and analysis of speci-
most specimens will generally be less than that of a men composition. However, X-ray generation is only possi-
conventional SEM operating at just a few hundred ble when the velocity of the incoming charged particle
electron volts. If the signal-to-noise of the image appears (electrons or ions) equals or exceeds the velocity of the orbit-
to be good enough to record useful information then it is ing electron within the sample. For an incident beam of elec-
worth trying to reduce the incident beam current by trons these two energies will only be numerically equal in
using a smaller beam current in the column to see if the value when their velocities are the same. When the kinetic
damage rate of the sample can be reduced while energy of a beam electron matches the ionization (binding)
simultaneously increasing the imaging depth of field. energy of an atomic shell electron, inelastic scattering of the
beam electron becomes possible, causing ejection of that
atomic electron. But when the incident beam is made of
Step 3 helium ions, which have a mass 7300 times heavier than an
Next choose a working distance (WD) from the source to the electron, then the energy of the ion also has to be increased
sample. Obtaining the shortest possible WD is not as by a factor of 7300× by accelerating it before atomic ioniza-
important in an ion microscope as it is in a conventional SEM tion becomes possible. X-ray microanalysis with an ion
because the aberrations of the ion beam-optical system are beam is therefore only possible with ion beam energies in
much less severe than those in conventional electron optics. the MeV range, which is far above the operating energy of
Using a longer working distance will allow safer specimen the HIM.
tilting and manipulation. Before proceeding also ensure that Possibly the most promising approach for chemical anal-
the specimen stage and its specimens are not touching ysis using ion beams at present is Time of Flight- Secondary
anything else, nor obscuring access to the various detectors, Ion Mass Spectrometry (TOF-SIMS), which directly analyzes
probes, and accessories in the column and chamber. the specimen atoms sputtered atoms from the surface by the
primary ion beam. In this method the specimen of interest is
References
539 31
contained in the sample chamber in the usual way, but on one the area selected, identifying the chemistry of the various
side of the chamber is a shutter which can be opened or compounds, clusters, and elements that are present whereas
closed at very high speed as required. The ion beam is then conventional X-ray microanalysis can only identify which
switched on for a short period of time, striking the specimen elements are present. This technology is presently only avail-
and sputtering the top two or three atomic layers of the speci- able on a few dual-beam Ga+ “focused ion beam source”
men, including neutral, negative, and positive fragments and (SEM–FIB) instruments, but many groups are now working
at the same time the shutter is opened. Which fragment of to extend this capability to the HIM.
the sample, positive or negative, is to be analyzed is deter-
mined by the polarity of the electric field applied in the
chamber. The chosen charged particles from the sample drift References
toward the shutter at which point they are abruptly acceler-
ated by a strong electric field and travel on until they reach Bethe H (1930) Theory of the transmission of corpuscular radiation
the detector which records the time taken from the original through matter. Ann Phys Leipzig 5:325
electrical pulse. This time interval will depend on the mass as Bethe HA (1933) Charged particle energy loss in solids. In: Handbook of
physics, vol 24. Springer, Berlin, p 273
well as the charge of the particle and so by measuring the Dapor M (2011) Secondary electron emission yield calculation per-
transit time required to reach the detector, the mass-to- formed using two different Monte Carlo strategies. Nucl Instr
charge ratio of each individual atom cluster can be deduced. Methods Res Phys 269:B269–1671
Given accurate “time of transit” measurement capability on Everhart T, Thornley R (1960) Wide-band detector for micro-microam-
the system, this mass-to-charge ratio data can provide rapid, pere low-energy electron currents. J Sci Instrum 37:246
Fresnel AJ (1817) Memoire sur les modifications que la reflexion imprime
reliable, and unique identification of all the atomic and clus- ala lumiere polarisse. FO 1:441
ter ions ejected during ion bombardment. After a suitable Fresnel A (1826) Diffraction of light. Mem Acad Sci Paris 5:339–487
time interval the shutter is again closed, the system is briefly Kanaya K, Okayama S (1972) Penetration and energy-loss theory of elec-
allowed to stabilize, and the operation is then repeated as trons in solid targets. J Appl Phys 5:43
often as required to achieve adequate statistics. Levi-Setti RL (1974) Proton scanning microscopy: feasibility and prom-
ise. Scanning Electron Microscopy Part 1:125
TOF-SIMS offers some significant advantages over other Muller EW (1965) Field Ion Microscopy. Science 149:591
possible approaches. The incident ion beam can either be Notte J, Hill R, McVey S, Farkas L, Percival R, Ward B (2006) An introduc-
held stationary at a point, or scanned across a chosen area as tion to helium ion microscopy. Micros. Microanal. 12:126
required to produce a chemical map. Each cycle of this pro- Ramachandra R, Griffin B, Joy DC (2009) A model of secondary electron imag-
cess will then physically remove a few atomic layers from the ing in the helium ion scanning microscope. Ultramicroscopy 109:748
Rutherford ER (1911) Radioactive Transformations. Yale University Press,
analyzed area, so repeated measurements on the same area New Haven
can eventually provide a three-dimensional view of the sam- Zhu Y, Inada H, Nakamura M, Wall J (2009) Imaging single atoms using
pled region. Most important of all, this approach to analysis secondary electrons with an aberration-corrected electron micro-
provides detailed information about the actual chemistry of scope. Nat Mater 8:808
541
Supplementary
Information
Appendix – 542
Index – 547
Appendix
to the suitability, or otherwise, of any given piece of yields for about 40 or elements, and a collection of inorganic
information. compounds and polymers.
The database currently contains several thousand individual The clear discrepancies that often exist between the compa-
values collected from more than 100 published papers and rable sets of original yield figures for the same material may be
reports spanning the period from 1898 to the present day. Since the result of surface contamination, or the result of a different
this is a “work in progress” the compilation is constantly being assumption about the appropriate emitted energy range for sec-
extended as additional values become available. As far as possi- ondary electrons (usually now taken to be 0–50 eV, although in
ble a consistent style of presentation is used so that data for dif- some early work 0–70 or even 0–100 eV was used). In addition,
ferent elements and compounds may readily be compared. All since many of the materials documented are poorly conducting
of the available data sets are grouped by element or compound the effects of charging must also be considered. For example, in
name for each of the major information groups (SE yields, BSE studies of the oxides (e.g., Whetten and Laponsky 1957) maxi-
yields, stopping powers, X-ray ionization cross sections). The mum SE yields of δ > 10 were measured using pulsed electron-
data is presented in a simple two-column format with the origin beam techniques. Clearly no non-conducting material can
of each of the data sets (numbered #1 to #n) indicated by a num- sustain this level of emission for any significant period of time,
ber in parenthesis referring back to the bibliography. since it will become positively charged and recollect its own sec-
ondaries. Similarly at higher energies, where the SE yield δ < 1
and negative charging occurs, the incident beam energy must be
Backscattered Electrons corrected for any negative surface potential acquired by the sam-
ple to give a correct result (although there is no little evidence in
The data on the backscattered electron (BSE) yield as a func- the original papers that this has been done). Consequently, all SE
tion of the atomic number of the target and of the incident yield results for insulators must be treated with caution unless
beam energy is of particular importance in Monte Carlo the provenance of the original data is well documented.
computations because it provides the best test of the scatter- Since there is no sum rule for secondary yields, data must
ing models that are used in the simulation. This data is there- be acquired for every compound of interest over the energy
fore both the starting point for the construction of a Monte range required, a task which will be a lengthy one unless suit-
Carlo model, and the source of values against which the sim- ably automated procedures can be developed and applied. In
ulation can be tested. The backscattered electron section con- addition it will be necessary to repeat many of the measure-
tains data for 40 or more elements spread across the periodic ments reported here using better techniques before any level
table, as well as for a selection of compounds. If Castaing’s of precision and accuracy can be obtained. In summary the
rule can be assumed to be correct, then the backscattering SE data is in an even less satisfactory state than that for the BS
yield of a compound can be found if the backscattering coef- electrons, even though a wider range of materials is covered,
ficients and the atomic fraction of the elements that form it because the quality of much of the data is poor.
are known. Hence a desirable long-term goal is to obtain a
complete set of BSE yield curves for elements. At present the
BSE section contains information on more than 45 elements, Stopping Powers
which is barely half of the solid elements in the periodic
table, and of this number perhaps only 25 % of the data sets The stopping power of an electron in a solid, that is, the rate
are of the highest quality, so much experimental work at which the electron transfers its energy to the material
remains to be done especially at the lower energies. through which it is passing, is a quantity of the highest
importance for all studies of electron-solid interactions since
it determines, among other parameters the electron range
Secondary Yields (Bethe 1930), the rate of secondary electron production
(Bethe 1941), the lateral distribution and the distribution in
With the increasing interest in the simulation of secondary depth of X-ray production, and the generation and distribu-
electron (SE) line profiles and images, there is a need to have tion of electron–hole pairs. Despite its importance there is no
detailed information on secondary electron yields as a func- body of experimental measurements of stopping power at
tion of atomic number and incident beam energy. Secondary those energies of interest to electron microscopy and micro-
electron emission was the subject of intense experimental analysis. Instead stopping powers, and the quantities which
study for a period of 20 years or more from the early 1930s, depend on them, have been deduced by analyzing measure-
resulting in the publication of no less than six full-length ments of the transmission energy spectrum of MeV-energy
books on the topic. This effort did not, however, produce as β-particles to yield a value for the mean ionization potential I
much experimental data as would have been expected, of the specimen (ICRU 1983), and then Bethe’s (1930) ana-
because the aim of much of the work that was done was to lytical expression for the stopping power has been invoked to
demonstrate that the SE yield versus energy curve followed a compute the stopping power at the energy of interest. While
“universal law” (Seiler 1984), and to find the parameters this procedure is of acceptable accuracy at high energies
describing this curve. As a result the data actually published (>10 keV) it is not reliable at lower energies because some of
was usually given in a normalized format that makes it diffi- the interactions included in the value of I (e.g., inner shell
cult to derive absolute values. The database currently contains ionizations) no longer contribute.
544 Appendix
The database contains experimentally determined stop- Acknowledgments The original version of the database was
ping power curves for a collection of elements and com- supported by the Semiconductor Research Corporation
pounds. The method for obtaining this information from (SRC) under contract 96-LJ-413.001, contract monitor is Dr.
electron energy loss spectra has been described elsewhere D Herr, and by a grant from International SEMATECH.
(Luo et al. 1991). The data is plotted in units of eV/Å as a func-
tion of the incident energy in keV. At the high energy end of
the profiles the data corresponds closely to values deduced References
from Bethe’s (1930) law and using the I-values from the ICRU
tables. At lower energies, however, significant deviations Ashley JC, Tung CJ, and Ritchie RH (1979) Surf Sci 81:409
Bethe HA (1930) Ann Phys 5:325
occur as the Bethe model becomes physically unrealistic Bethe HA (1941) Phys Rev 59:940
although good agreement has been found with values com- Bishop H (1966) Ph.D. Thesis University of Cambridge
puted from a dielectric model of the solids (Ashley et al. 1979). Campbell-Swinton AA (1899) Proc Roy Soc 64:377
The stopping power of a compound is the weighted sum I.C.R.U (1983) “Stopping powers of electrons and positrons”, Report #37
of the stopping power of its constituents; thus a key priority to International Committee on Radiation Units (ICRU:Bethesda, MD)
Luo S, Zhang X, Joy DC (1991) Rad. Effects and Defects in Solids 117:235
for future work should be to complete the set of stopping Seiler H (1984) J Appl Phys 54:R1
power profiles for elements rather than to acquire more data Starke H (1898) Ann Phys 66:49
on compounds. Whetten NR, Laponsky AB (1957) J Appl Phys 28:515
12. Bruining H (1942) Die Sekundt-Elektronen Emission Fester Korper. 67. Smick AE, Kirkpatrick P (1945) Phys Rev 67:153
Springer-Verlag, Berlin 68. Wittry DB (1966) Proc. 4th Conf. on X-ray optics and microanalysis.
13. Neubert G, Rogaschewski S (1980) Phys Stat Sol (a) 59:35 In: Castaing R et al. (ed). Hermann, Paris, p 168
14. Kanter M (1961) Phys Rev 121:1677 69. Rothwell TE, Russell PE (1988) In: Newbury DE (ed) Microbeam
15. Drescher H, Reimer L, Seidel M (1970) Z Angew Physik 29:331 analysis—1988. San Francisco, San Francisco Press, p 149
16. Sommerkamp P (1970) Z Angew Physik 4:202 70. Whetten NR (1962) In: Methods in experimental physics. NY,
17. Czaja W (1966) J Appl Phys 37:4236 Academic Press, p IV
18. Whetten NR, Laponsky AB (1957) J Appl Phys 28:515 71. Green M, Cosslett VE (1968) Brit J Appl Phys (J Phys D) 1:425
19. Whetten NR (1964) J Appl Phys 35:3279 72. Sudarshan TS (1976) IEEE Trans. on Electrical Insulation EI-11:32
20. Dawson PH (1966) J Appl Phys 37:3644. (Note that this SE data was 73. LaVerne JA, Mozumder A (1985) J Phys Chem 89:4219
taken for energy cutoffs varying from 30eV to 100eV) 74. Al-Ahmad KO, Watt DE (1983) J Phys D Appl Phys 16:2257
21. Bruining H, De Boer JM (1938) Physica V:17 75. Ishigure N, Mori C, Watanabe T (1978) J Phys Soc Japan 44:973
22. Cosslett VE, Thomas RN (1965) Brit J Appl Phys J Phys D 16:774 76. Garber FW, Nakai MY, Harter JA, Birkhoff RD (1971) J Appl Phys
23. Koshikawa T, Shimizu R (1973) J Phy D Appl Phys 6:1369 42:1149
24. Johnson JB, McKay KG (1954) Phys Rev 93:668 77. Luo S (1994), Ph.D Thesis University of Tennessee
25. Dionne GF (1975) J Appl Phys 46:3347 78. Kalil F, Stone WG, Hubell HH, Birkhoff RD (1959) ORNL Report 2731
26. Kanaya K, Kawakatsu M (1972) J Phys D Appl Phys 5:1727 79. Arifov UA, Khadzhimukhamedov Kh. Kh. Makhmudov (1971) In:
27. Hazelton RC, Yadlowsky EJ, Churchill RJ, Parker LW (1981) IEEE Trans Secondary emission and structural properties of solids. In: Arifov
on Nucl Sci NS-28:4541 UA (ed) Consultants Bureau, New York, p 30
28. Gross B, von Seggern H, Berraissoui A (1983) IEEE Trans on Electrical 80. Arifov UA, Kasymov A Kh (1971) In: Arifov UA (ed) Secondary emis-
Insulation EI-22:23 sion and structural properties of solids. Consultants Bureau,
29. Kishimoto Y, Ohshima T, Hashimoto H, Hayashi T (1990) J Appl New York, p 78
Polymer Sci 39:2055 81. Septier A, Belgarovi M (1985) IEEE Trans. on Electrical Insulators
30. Matskevich TL (1959) Fiz Tvend Tela Acad Nauk SSSR 1:227 EI-20:725
31. Kazantsev AP, Matskevich TL (1965) Sov Phys Sol State 6:1898 82. Ahearn AJ (1931) Phys Rev 38:1858
32. Willis RF, Skinner RF (1973) Solid State Comm 13:685 83. Flinn EA, Salehi M (1980) J Appl Phys 51:3441
33. Gair S, quoted in Burke EA (1980) IEEE Trans. on Nucl Sci NS-27:1760 84. Flinn EA, Salehi M (1979) J Appl Phys 50:3674
34. Sternglass EJ (1954) Phys Rev 95:345 85. Dunn B, Ooka K, Mackenzie JD (1973) J Am Cer Soc 56:9
35. Palluel P (1947) Compt Rendu 224:1492 86. Joy DC, Joy CS, SEMATECH Report TT# 96063130A-TR, August
36. Clark E (1935) Phys Rev 48:30 1996
37. Tawara H, Harison KG, De Heer FJ (1973) Physica 63:351 87. Kaneff W (1960) Annalen der Physik 7:84
38. Hink W, Pashke H (1971) Phys Rev A4:507; also Hink W, Pashke H 88. Petzel B (1960) Annalen der Physik 6:55
(1971) Z Phys 244:140 89. Hovington P, Joy DC, Gauvin R, Evans N (1995) unpublished data.
39. Isaacson M (1972) J Chem Phys 56:1813 For a summary see Joy DC, Luo S, Gauvin R, Hovington P, Evans N
40. Colliex PC, Jouffrey B (1972) Phil Mag 25:491 (1996) Scanning Microscopy 10:653–666
41. Egerton RF (1975) Phil Mag 31:199 90. Buhl R (1959) Zeit F Phys 155; see also Keller M (1961) Zeit F Phys
42. Glupe G, Melhorne W (1971) J Physique Suppl 32 C4 – 40; also Glupe 164:274
G, Melhorne W (1967) Phys Lett 25A:274 91. Miyake S (1938) Proc Phys Math Soc Japan 20:280
43. Shima K, Nakagawa T, Umetani K, Fikumo T (1981) Phys Rev A24:72 92. Lashkarev VE, Chaban AS (1935) Phys Z Soviet 8:240
44. Davis DV, Mistry VD, Quarles CA (1972) Phys Lett 38A:169 93. Westbrook GL, Quarles CA (1987) Nucl Instrum and Methods Phys
45. Platton H, Schiwietz G, Nolte G (1985) Phys Lett 107A:83 Res B24-25:196
46. Fitting HJ (1974) Phys Stat Sol (a)26:525 94. Motz JW, Placious RC (1964) Phys Rev 136:662
47. Luo S, Zhang X, Joy DC (1991) Rad Eff and Defects in Solids 117:235 95. Broyles C, Thomas D, Haynes S (1953) Phys Rev 89:715
48. Sorenson H (1977) J Appl Phys 48:2244 96. Lay H (1934) Z Physik 91:533
49. Kurrelmeyer B, Hayner LJ (1937) Phys Rev 52:952 97. Frey WF, Johnston RE, Hopkins JI (1958) Phys Rev 113:1057
50. Kamiya M, Kuwako A, Ishii K, Morita S, Oyamada M (1980) Phys Rev 98. Konstantinov AA, Sazonova TE, Perepelkin VV (1960) Izv Akad
A22:413 Nauk SSSR, Ser Fiz 24:1480; also from Konstantinov et al. (1961) Izv
51. McDonald SC, Spicer BM (1988) Phys Rev A37:985 Akad Nauk SSSR Ser Fiz 25:219; Konstantinov et al. (1964) Izv Akad
52. Hoffman DHH, Brendel C, Genz H, Low W, Muller S, Richter A (1979) Nauk SSSR Ser Fiz 28:107; Konstantinov et al. (1965) Izv Akad Nauk
Z Phys A293:187; also -Hoffman DHH, Genz H, Low W, Richter A SSSR Ser Fiz 29:302
(1978) Phys Lett 65A:304 99. Gray PR (1956) Phys Rev 101:1306
53. Hink W, Ziegler A (1969) Z Phys 226:222 100. Data taken from miscellaneous references in Fink RW, Jopson RC,
54. Luo S “Study of Electron-Solid Interactions,” PhD Thesis, University Mark H, Swift CD (1966) Rev Mod Phys 38:513
of Tennessee May 1994 101. Data taken from miscellaneous references in Bambynek W,
55. Meyer F, Vrakking JJ (1973) Phys Lett 44A:511; see also Vrakking JJ, Crasemann B, Fink RW, Freund HU, Mark H, Swift CD, Price RE, Rao
Meyer F (1974) Phys Rev A 9:1932 PV (1972) Rev Mod Phys 44:716
56. Hippler R, Sneed K, McGregor I, Kleinpoppen H (1982) Z Phys A307:83 102. Jopson RC, Mark H, Swift CD (1962) Phys Rev 128:2671
57. Quarles C, Semaan M (1982) Phys Rev A26:3147 103. Sidhu NPS, Grewal BS, Sahota HS (1988) Xray spectrometry
58. Jessenberger J, Hink W (1975) Z Phys A275:331 17:29
59. Shima K (1980) Phys Lett 237 104. Fischer B, Hoffmann KW (1967) Z fur Physik 204:122
60. Pockman LT, Webster DL, Kirkpatrick P, Harworth K (1947) Phys Rev 105. Henrich VE, Fan JCC (1974) J App Phys 45:5484
71:330 106. Bronstein IM, Fraiman BS (1969) In: Vtorichnaya Elektronnaya
61. Myers HP (1952) Proc Roy Soc 215:329 Emissiya. Nauka, Moskva, p 340
62. Hubner H, Ilgen K, Hoffman KW (1972) Z Phys255:269 107. El Gomati MM, Assad AMD (1997) Proc. 5th European Workshop
63. Kiss K, Kalman G, Palinkas J, Schlank B (1981) Acta Phys Hung 50:97 on Microbeam Analysis (in press)
64. Salem SI, Moreland LD (1971) Phys Lett 37A:161 108. Rossouw CJ, Whelan MJ (1979) J Phys D Appl Phys 12:797
65. Green M (1964) Proc Roy Soc 83:435 109. Allawadhi KL, Sood BS, Raj Mittal, Singh N, Sharma JK (1996) X-ray
66. Hansen H, Flammersfeld A (1966) Nucl Phys 79:134 Spectr 25:233
546 Appendix
110. Horakeri LD, Hanumaiah B, Thontadarya SR (1997) X-ray Spectr 125. An Z, Tang CH, Zhou CG, Luo ZM (2000) J Phys B At Mol Opt Phys
26:69 33:3677–3684
111. Mearini GT, Krainsky IL, Dayton JA, Wang X, Zorman CA, Angus JC, 126. Llovet X, Merlet C, Salvat F (2001) J Phys B At Mol Opt Phys
Hoffman DF, Anderson DF (1995) Appl Phys Lett 66:242 35:973–982
112. Thomas S, Pattinson EB (1969) Brit J Appl Phys II-2:1539 127. Hilleret N, Bojko J, Grobner O, Henrist B, Scheuerlein C, Taborelli M
113. He FQ, Long X, Peng X, Luo Z, An Z (1996) Nucl Inst Meth Phys v (2000) “The secondary electron yield of technical materials and its
114: 213 Variation with surface treatments” in Proc.7th European Particle
114. He FQ, Long X, Peng X, Luo Z, An Z (1997) Nucl Inst Meth Phys v Accelerator Conference, Vienna
129: 445, v 6: 697 128. Campos CS, Vasconcellos MAZ, Llovet X, Salvat F (2002)
115. Suszcynsky DM, Borovsky JE, Goertz CK (1992) J Geophysical Res Measurements of L-shell X-ray production cross sections of W, Pt,
E2 97:2611–2619 and Au by 10–30 keV electrons. Phys Rev A, 66
116. Horakeri LD, Hanumaiah, Thontadarya SR (1998) X-ray 129. Prasad MS (2002) personal communication
Spectrometry 27:344 130. Ushio Y, Banno T, Matuda N, Baba S, Kinbara A (1988) Thin Solid
117. An Z, Li TH, Wang LM, Xia XY, Luo ZM (1996) Phys Rev 54:3067 Films 167:299
118. Luo Z, An Z, Li T, Wang LM, Zhu Q, Xia X (1997) J Phys B At Mol Opt 131. Lin Y (2003) personal communication
Phys 30:2681 132. Sim K-S, Kamel N (2003) Phys Lett A 317: 378
119. Luo Z, An Z, He F, Li T, Long X, Peng X (1996) J Phys B At Mol Opt 133. Sorenson H, Schou J (1982) J Appl Phys 53:5230
Phys 29:4001 134. Pinato CA, Hessel R (2000) J Appl Phys 88:478
120. He FQ, Peng XF, Long XG, Luo ZM, An Z (1997) Nucl Instrum and 135. Martsinovskaya EG (1965) Soviet Phys Solid State 7:662
Methods in Physical Research B129:445 136. Yang KY, Hoffman RW (1987) Surf Interf Anal 10:121
121. Peng X, He F, Long XG, Luo ZM, An Z (1998) Phys Rev A58:2034 137. Dawson PH (1966) J Appl Phys 36:3644
122. Thiel BL, Stokes DJ, Phifer D (1999) Secondary electron yield curve 138. Merlet C, Llovet X, Salvat F (2008) Phys Rev A 78:022704
for liquid water, microscopy and microanalysis 5(Suppl 2)(Proc. 139. Merlet C, Llovet X, Fernandez-Varea JM (2006) Phys Rev A
Microscopy & Microanalysis 99):282–283 73:062719
123. Llovet X, Merlet C, Salvat F (2000) J Phys B At Mol Opt Phys 140. Merlet C, Llovet X, Salvat F (2004) Phys Rev A69:032708
33:3761–3772
124. Schneider H, Tobehn I, Ebel F, Hippler R (1993) Phys Rev Lett
71:2707–22709
547 AA–EP
Index
A –– collection of 491
–– detection of 491–492
–– sample preparation for 510
–– spatial resolution 507–508
Aluminum wire failures 482–483 –– measurement 491 Electron beam, interaction volume change
Anaglyph stereo pair presentation 218 –– origin 490–491 –– elastic scattering 3–4
Ancient impact zircons 493 Compositional mapping –– elastic scattering cross section 4
Angular distribution, backscattered –– limitations of 427–429 –– inelastic scattering 2–3
electrons 24–25 –– MAXIMUM PIXEL spectrum 433–435 –– Monte Carlo calculations
Auger electron emission 48 –– quantitative compositional mapping –– beam electron interaction volume 7–8
Auger electron spectroscopy 37 436–441 –– composition 8–9
Avogadro’s number 4 –– SUM spectrum 431–433 –– electron interaction volume 7
–– total intensity region-of-interest –– electron trajectory simulation 4–5
mapping 426–427 –– incident beam energy 9–10
B –– X-ray spectrum imaging (XSI) 429–431
–– EDS dead-time 441–443
–– Monte Carlo simulation (CASINO
simulation) 5–7
Backscattered electrons (BSEs) 2, 549, vii –– elemental mapping data collection –– size of the interaction volume 11–12
–– angular distribution 24–25 441–450 –– specimen tilt 10–11
–– atomic number contrast 19–20 –– flash mapping 444–445 –– specimen atoms 2
–– energy distribution 29–30 –– high count mapping 445–450 Electron-excited X-ray microanalysis, geometric
–– η vs. atomic number, Z 17–19 –– pixel density 443–444 effects 398
–– η vs. surface tilt, θ 21–23 –– pixel dwell time 444–450 Electron interaction volume 363–364
–– numerical measure 16 Concentration limit of detection (CDL) Electron optical brightness, β 83
–– origin 16 –– reference value 362–363 Electron optical parameters
–– spatial distribution 25–29 –– trace or minor constituent 359–362 –– astigmatism 85–87
–– topographic contrast—number effects 23–24 Continuous energy loss approximation 2 –– beam convergence angle, α 80–81
Backscattered electron yield data 584–563 –– beam current 79–80
Beam convergence angle, α 80–81
D
–– beam current density 80
Beam current 79–80, 214 –– beam diameter 79
Beam current density 80 –– beam energy 78–79
Beam diameter 79 A Database of Electron-Solid Interactions (Joy) 37
–– beam solid angle 81–83
Beam energy 78–79, 297, 330. See also Low Detective quantum efficiency (DQE) 101–103
–– electron optical brightness, β 83
beam energy SEM; Low beam energy X-ray DTSA-II EDS software
–– focus 83–87
microanalysis –– design 255
Electron probe microanalyzers (EPMAs) 91
–– compositional contrast with backscattered –– fundamental concepts 256–264
Electron–solid interactions, database of 548
electrons 213 –– GUI 267–279
Energy dispersive X-ray microanalysis checklist
–– high resolution SEM imaging 214 –– Monte Carlo simulation 267
–– analytical procedure 477
–– vs. secondary electrons (SE) yield 40 –– motivation 254
–– calibrating the EDS detector
–– topographic contrast –– optional tables 279–283
–– detector position 473–474
–– with backscattered electrons 213 –– overview 254–255
–– energy calibration 472–473
–– with secondary electrons 213 –– platform 254
–– probe current 474
Beam placement 414 –– simulation 264–267
–– pulse process time constant 472
Beam solid angle 81–83 –– three-leg stool
–– quality control 473
Bethe expression 2, 3 –– experiment design 256
–– sample orientation 473
Brightness equation 83 –– quantification 255
–– working distance/pecimen-to-EDS
BSEs. See Backscattered electrons (BSEs) –– simulation 255
distance 473
Bulk biological and organic specimens 355–356 –– collecting data
Bulk specimens
–– origins of geometric effects 396–397 E –– collecting peak-fitting references 475
–– collecting spectra from unknown
–– particle analysis EDS. See Energy dispersive X-ray spectrometry (EDS) 475–476
–– optimum spectra 410–414 Electron backscattering diffraction (EBSD) xiii –– collecting standards 475
–– quantitative analysis of particles 415–420 –– align sample 510–511 –– experiment optimization 474
–– uncertainty in 420–422 –– check for EBSD patterns 511–512 –– exploratory spectrum 474
–– X-ray measurements 408–410 –– checklist –– reference spectra 475
–– X-ray spectrum imaging (XSI) 414–415 –– acquisition parameters 522 –– selecting standards 474
–– with rough surfaces 399–401 –– candidate crystallographic phases 522 –– data analysis 476
–– EBSD detector 521–522 –– instrumentation
Energy dispersive X-ray spectrometry (EDS) xi –– imaging with ions 528–529 –– scale bars 107
–– adequate counts 297–298 –– ion–solid interactions 526–527 –– by scanning action 106–107
–– beam current 330–332 –– SEM, sample preparation 530 –– stereomicroscopy
–– beam energy 297, 330 –– 3D techniques and imaging 532–536 –– qualitative stereomicroscopy 118–121
–– coincidence peaks 294–295 –– quantitative stereomicroscopy 121–124
G
–– detection process –– surface measurement 117–118
–– coincidence peaks 231–233 ImageJ-Fiji
–– peak broadening 228–231 –– calibrated structure in 109–110
Geometric factors
–– Si absorption edge 233–234 –– routine linear measurements with 110
–– bulk specimens 396–397
–– Si-escape peak 231 ImageJ universe 204
–– electron-excited X-ray microanalysis 398
–– Si internal fluorescence peak 233–234 Imaging crystalline materials
–– useful indicators of
–– detector dead-time 297 –– acquired data 512–513
–– EDS spectrum 404–406
–– detector time constant 297 –– application 518–521
–– raw analytical total 401–404
–– electron-excited EDS operation –– cleaning EBSD data 516–517
–– rough bulk samples 406–408
–– beam current 235–237 –– electron channeling contrast
Graphical user interface (GUI) 90, 204
–– channel width and number 235 –– electron backscattering diffraction
–– EDS time constant 234 (EBSD) 504–512
–– solid angle 235
–– exciting characteristic X-rays 287–288
H –– instrument conditions 504
–– specimen preparation 504
–– fluorescence yield 288 Hard-facing alloy bearing surface 480–482 –– map components 513–516
–– lower photon energy region 300–301 Helium ion microscope (HIM) xiv –– polycrystalline materials 502–503
–– manual peak identification 301–305 High resolution imaging –– single crystals 500–502
–– manual qualitative analysis 296–297 –– achieving visibility 178 –– transmission Kikuchi diffraction (TKD)
–– minor and trace constituents 301 –– beam footprint 162–164 517–518
–– parameters –– delocalized signals 162–164 Incident beam energy 9–10
–– calibration 330 –– instrumentation considerations 162 Ink-Jet printer deposits 224–226
–– solid angle 330 –– pathological specimen and instrumentation International Union of Pure and Applied
–– spectrum channel energy width behavior Chemistry (IUPAC) 51
328–329 –– contamination 179 Ion beam microscopy
–– spectrum energy span 328–329 –– instabilities 179 –– chemical microanalysis 546
–– time constant 329–330 –– pathological specimen behavior 178–179 –– generating ion beams 541–542
–– pathological electron scattering –– pixel size 162–164 –– helium ion microscope (HIM)
–– trace analysis artifacts 364–369 –– secondary electron contrast 164–165 –– current generation and data
–– peaks, identifying 299 –– beam range 167 collection 543–545
–– principles, qualitative EDS analysis 287 –– bright edge effect 167 –– operating the 545–546
–– QC project 246–249 –– critical dimension metrology 167–169 –– patterning with 545
–– QC tools within DTSA-II 246 –– isolated edges 165–167 –– signal generation in 542–543
–– quality assurance issues 286 –– with secondary electrons 169–171 –– patterning with ion 545
–– quality measurement environment –– beam energy strategies 171–173 –– usefulness 538–541
–– detector geometry 237–240 –– low loss BSEs 176–178
–– SE1 signal 173–176
J
–– detector orientation 240–245
–– energy calibration linearity 245 –– SEM 162
–– optimal working distance 240 HIM. See Helium ion microscope (HIM)
–– process time 240 Java Runtime Environment (JRE) 205
–– Si escape peak 293
I
–– silicon drift detector (SDD)
–– low X-ray flux 250–251 Image defects
K
–– moderate resolution 251 –– charging Kanaya–Okayama range 11–12, 60, 182
–– output count rate with live-time dose –– control charging artifacts 153–155 k-ratio
249–250 –– define 148–149 –– analytical total 310
–– resolution and peak position stability 250 –– in SEM images 149–152 –– converting sets of 310
–– software tools 298–299 –– contamination 157–158 –– define 308–309
–– trace level measurement 363–364 –– image defocusing (blurring) 113–117 –– element by difference 311
–– X-ray absorption 288–289 –– Moiré effects 158–159 –– estimate CZ 310–311
–– X-ray energy database 289–292 –– projection distortion (foreshortening) –– matrix corrections 316–317
Energy distribution, backscattered 111–113 –– matrix effects, physical origin of 317
electrons 29–30 –– radiation damage 155–157 –– normalization 310
EPMAs. See Electron probe microanalyzers Image formation –– oxygen by assumed stoichiometry 311
(EPMAs) –– calibrating the image 107–109 –– quantitative electron-excited X-ray
Everhart–Thornley detector 99–100, 128, –– image defects microanalysis
129–130, 213, 215 –– image defocusing (blurring) 113–117 –– standardless analysis 314–316
–– projection distortion (foreshortening) –– standards-based k-ratio protocol
F 111–113
–– image dimensions 107
313–314
–– reporting composition
Fiji 204–206 –– ImageJ-Fiji, calibrated structure in 109–110 –– atomic fraction 312
Focused ion beams (FIB), xiii –– ImageJ-Fiji, routine linear measurements –– mass fraction 311–312
–– cross-section preparation 530–532 with 110 –– oxide fractions 312–313
–– focused ion beam systems 527–528 –– magnification 107 –– stoichiometry 312
549 AE–SP
Index
L
Scanning electron microscope (SEM) imaging
N
checklist
–– beam current
Landing energy 79
–– high resolution imaging 214
Lead-acid battery plate reactions 494–495 National Institutes of Health (NIH) 204
–– low contrast features 214
Light-optical analogy, Everhart–Thornley NIST DTSA II simulation 364
–– beam energy
(positive bias) detector 131–133
–– compositional contrast with backscattered
O
Long-range secondary X-ray fluorescence 364
electrons 213
Low beam energy SEM
–– high resolution SEM imaging 214
–– backscattered electron signal
Open Microscopy Environment –– topographic contrast with backscattered
characteristics 182–185
(OME) 204 electrons 213
–– constituent 182
Overscanning 414 –– topographic contrast with secondary
–– extremely low beam energy imaging 187
electrons 213
–– high depth resolution SEM 185–186
P
–– electron detector
–– for high lateral resolution SEM 185
–– backscattered electron detectors 213
–– Kanaya–Okayama range 182
–– Everhart–Thornley detector 213
–– secondary electron 182–185 Particle absorption effect 415–420
–– electron signals available
Low beam energy X-ray microanalysis Particle analysis
–– backscattered electrons 212
–– advantage of –– optimum spectra 410–414
–– beam electron range 212
–– improved spatial resolution 381–382 –– quantitative analysis of particles 415–420
–– secondary electrons (SEs) 212–213
–– low atomic number elements 382–385 –– uncertainty in 420–422
–– image interpretation
–– reduced matrix absorption correction 382 –– X-ray measurements 408–410
–– annular BSE detector 215
–– challenges and limitations –– X-ray spectrum imaging (XSI) 414–415
–– contrast encoding 215
–– reduced access to elements 385–387 Particle mass effect 415
–– direction of illumination 214
–– surface layers 388–393 Particle sample preparation 413–414
–– Everhart–Thornley detector 215
–– vertical heterogeneity 387–388 Plugins 206–209
–– observer’s point of view 214
–– constitutes 376–380
–– semiconductor BSE detector 215
Q
–– low beam energy analysis range 381
–– image presentation
–– peak selection strategy 380–381
–– live display adjustments 214
Quantitative compositional mapping, –– post-collection processing 214
R
Magnification vs. pixel dimension 163 Scanning electron microscope (SEM)
Manganese nodule 483–488 instrumentation
Modeled detectors –– detective quantum efficiency (DQE)
Robust light-optical analogy 131
–– aluminum layer 261–263 101–103
–– azimuthal angle 261 –– detector characteristics
–– crystal thickness 261
–– dead layer 263
S –– bandwidth 97
–– electron detectors, angular
–– detector area 261 Scanning electron microscope (SEM) measures for 96–97
–– elevation angle 261 –– compositional microstructure ix –– energy response 97
–– energy scale 261 –– crystal structure xii–xiii –– electron beam parameters 78
–– gold layer 261–263 –– dual-beam platforms, combined electron –– electron detectors
–– material editor dialog 264–266 and ion beams xiii–xiv –– abundance 95
–– Mn Kα, resolution at 261 –– electron-optical parameters vii–ix –– angular distribution 95
–– nickel layer 261–263 –– elemental composition x–xii –– backscattered electrons 97–99
–– number of channels 261 –– specimen property information ix–x –– kinetic energy response 95–96
550 Index
Scanning electron microscope (SEM) Shallow surface relief 222–224 –– scanning electron microscopy at elevated
instrumentation (cont.) Silicon drift detector (SDD) pressures
–– electron optical parameters –– low X-ray flux 250–251 –– focused electron beam 193–196
–– astigmatism 85–87 –– moderate resolution 251 –– image resolution 196–198
–– beam convergence angle, α 80–81 –– output count rate with live-time dose –– specimen charging 191–192
–– beam current 79–80 249–250 –– water environment of specimen 192–193
–– beam current density 80 –– resolution and peak position stability 250 –– solve practical problems 461
–– beam diameter 79 Single pixel measurement 222 –– unfavorable sample characteristics 462–468
–– beam energy 78–79 Spatial distribution, backscattered electrons –– X-ray spectrometry 458–460
–– beam solid angle 81–83 –– depth distribution 26–28
–– electron optical brightness, β 83 –– Monte Carlo simulation 25
–– focus 83–87
–– imaging modes 87–88
–– radial distribution 28–29
Spectrum imaging (SI) xi
X
–– high-current mode 90–92 Stopping powers 549–550 X-ray ionization cross sections 550
–– high mode 88–90 X-rays
T
–– low-voltage mode 94–95 –– characteristic of
–– resolution mode 93–94 –– families 50–51
–– secondary electron detectors –– fluorescence yield 48–50
Thin section analysis 353–355
–– Everhart–Thornley detector 99–100 –– intensity 53–56
3D viewer plugin tool 218, 219
–– through-the-lens (TTL) electron –– isolated atoms 53–54
Transmission Kikuchi diffraction (TKD)
detectors 100 –– nomenclature 51
517–518
–– specimen current 100–101 –– origin 46–47
Scanning electron microscopist and X-ray –– thick, solid specimens 55–56
microanalyst (SEMXM) 204
Scanning transmission electron microscope
V –– thin foils 54–55
–– weights of lines 51–53
(STEM) image 175–176 Variable pressure scanning electron microscopy –– continuum (bremsstrahlung) 56–57
Secondary electron energy (VPSEM) vii –– absorption 65–75
spectrum 34–35 –– contrast in 200–201 –– continuum intensity 57–58
Secondary electrons (SE) 2, 549, vii –– conventional SEM high vacuum environment –– depth distribution function,Φ(ρz) 64–65
–– angular distribution of 39–40 –– beam integrity 190 –– electron-excited X-ray spectrum 58–59
–– energy distribution 34–35 –– difference from 190–191 –– fluorescence 75–76
–– escape depth 35–37 –– Everhart–Thornley secondary electron –– Monte Carlo simulation 62–63
–– origin 34 detector 190 –– range of 60–62
–– spatial characteristics of 40–43 –– minimizing contamination 190 X-ray spectrum imaging (XSI)
–– yield vs. atomic number 37–38 –– stable electron source operation 190 –– compositional mapping 429–431
–– yield vs. beam energy 40 –– EDS collimator 456–458 –– EDS dead-time 441–443
–– yield vs. specimen tilt 38–39 –– elevated pressure microscopy, detectors for –– elemental mapping data collection
Secondary electron yield data 553–584 –– backscattered electrons, passive scintillator 441–450
Secondary yields 549 detector 198–199 –– flash mapping 444–445
SEM/EDS –– secondary electrons, gas amplification –– high count mapping 445–450
–– limits of detection for 358–359 detector 199–200 –– pixel density 443–444
–– remote excitation sources 370–373 –– favorable sample characteristics 461–462 –– pixel dwell time 444–450
Semiconductors 493–494 –– gas scattering effects in 452–456, 460–461 –– particle analysis 414–415