Springer Series in Materials Science 2

Edited by Ulrich Gonser

Springer-Verlag Berlin Heidelberg GmbH

Springer Series in Materials Science
Advisors: M.S. Dresselhaus . H. Kamimura· K.A. Muller
Editors: U. Gonser· R. M. Osgood, Jr.. M. B. Panish . H. Sakaki
Managing Editor: H. K. V. Lotsch

27 Physics of New Materials

Editor: F. E. Fujita

28 Laser Ablation
Principles and Applications
Editor: J. C. Miller
29 Elements of Rapid Solidification
Fundaments and Applications
Editor: M. A. Otooni
30 Process Technology for Semiconductor Lasers
Crystal Growth and Microprocesses
By K. Iga and S. Kinoshita
31 Nanostructures and Quantum Effects
By H. Sakaki and H. Noge

Volumes 1-26 are listed at the end of the book.

Martin von Allmen Andreas Blatter

Laser-Beam Interactions
with Materials
Physical Principles and Applications

Second, Updated Edition

With 78 Figures and 7 Tables

Springer
Dr. Martin von AlImen Dr. Andreas Blatter
Or. GRAFAG PX Holding
CH-4563 Gerlafingen Bd. des Eplatures 42
Switzerland CH-2304 La Chaux-de-Fonds
Switzerland

Series Editors:
Prof. Dr. U. Gonser M. B. Panish, Ph. D.
Fachbereich 12.1, Gebăude 22/6 AT&T BeII Laboratories
Werkstoffwissenschaften 600 Mountain Avenue
Universităt des Saarlandes Murray HiI!, NJ 07974-2070, USA
0-66041 Saarbriicken, Germany

Prof. R. M. Osgood, Jr. Prof. H. Sakaki

Microelectronics Science Laboratory Institute of Industrial Science
Oepartment of Electrical Engineering University of Tokyo
Columbia University 7-22-1 Roppongi, Minato-ku
Seeley W. Mudd Building Tokyo 106, Japan
New York, NY 10027, USA

Managing Editor:
Dr.-Ing. Helmut K. V. Lotsch
Springer-Verlag, Tiergartenstrasse 17
0-69121 Heidelberg, Germany

Second Edition 1995

Updated Printing 1998

ISSN 0933-033X
ISBN 978-3-540-59401-7 ISBN 978-3-642-57813-7 (eBook)
DOI 10.1007/978-3-642-57813-7

Library of Congress Cataloging-in-Publication Data.

Von AlImen. M. Laser -beam interactions with materials: physical principles and applications / Martin von Allmen.
Andreas Blatter. - 2nd updated ed. p. cm. - (Springer series in materials science; v.2) Includes bibliographical
references and index. ISBN 978-3-540-59401-7(softcover: alk. paper) 1. Materials-Effect of radiation ono 2. Laser
heams-Industrial applications. 1. Blatter. Andreas. 1960 -. II. TiUe. III. Series. TA418.6.V63 1995 620.1"1228-
dc20 95-32586

This work is subject to copyright. AII rights are reserved. whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting. reproduction on
microfilm ar in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted
only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission
for lIse must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German
Copyright Law.
Springer-Verlag Berlin Heidelberg 1987, 1995
<[)
Originally Pllblished by Springer-Verlag Berlin Heidelberg New York in 1995

The lIse of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in
the absence of a specific statement.that such names are exempt from the relevant protective laws and regulations and
therefare free for general use.
Typesetting: PSTM Technical Word Processor
SPIN: 10663729 54/3144-543210 - Printed on acid-free paper
Preface

Eight years after the appearance of the first edition of Laser-Beam Interac-
tions with Materials, the topic seems far from losing any of its vigour and
fascination - if the rate of papers published is any indication. A number of
interesting new applications have appeared in the meantime, and a tremen-
dous amount of work on process characterisation has been done. Neverthe-
less, the main ideas of eight years ago are still there, and so are some of the
old puzzles.
This second edition, which owes its existence to the friendly reception
of the first one, comes with a number of corrections, updates and timely
additions (notably a new section on laser deposition) but preserves both the
original layout of the monograph and its emphasis on the physics behind the
phenomena.

Bern, September 1994 M. von Allmen

A. Blatter

v
Preface to the First Edition

Lasers, having proven useful in such diverse areas as high-resolution spec-

troscopy and the guiding of ferryboats, are currently enjoying great popu-
larity among materials scientists and engineers. As versatile sources of
"pure" energy in a highly concentrated form, lasers have become attractive
tools and research instruments in metallurgy, semiconductor technology and
engineering. This text treats, from a physicist's point of view, some of the
processes that lasers can induce in materials.
The field of laser-material interactions is inherently multidisciplinary.
Upon impact of a laser beam on a material, electromagnetic energy is con-
verted first into electronic excitation and then into thermal, chemical and
mechanical energy. In the whole process the molecular structure as well as
the shape of the material are changed in various ways. Understanding this
sequence of events requires knowledge from several branches of physics. A
unified presentation of the subject, for the benefit of the materials research-
er as well as the advanced student, is attempted here. In order to keep the
book reasonably trim, I have focused on laser effects in solids such as thin
films and technological materials. Related topics not covered are laser-in-
duced chemical reactions in gases and liquids and laser effects in organic or
biological materials.
This monograph grew out of a series of lectures which I gave for gra-
duate-level students in applied physics at the University of Bern. Its layout
reflects the diversity of the subject - Chapter 2 draws essentially from physi-
cal optics, Chapters 3 and 4 from materials science, and Chapter 5 from
fluid dynamics and plasma physics. While experts in certain fields covered
here may find the treatment of their speciality rather less than exhaustive, I
hope that the integrated treatment attempted will serve at least two pur-
poses: It should provide access and orientation for students and newcomers
in the large and diverse field of laser-material interactions, and it may, per-
haps, uncover certain relations and connections not always obvious to the
specialist. Readers' comments pointing out errors or inconsistencies will, in
any case, be most welcome.
It is a pleasant duty to thank all those who have contributed to this
book: colleagues who have given me access to and education on their work.
as well as critical readers of preliminary versions of the manuscript who
helped in eliminating some of its shortcomings. Finally, this book would not
have appeared without the dedicated work of R. Fluck and E. Krlihenbuhl

VII
who made most of the drawings and kept track of the paperwork, H.P.
Weber who gave support, and - most important of all - my wife who smo-
othed the atmosphere and brewed many, many cups of coffee.

Bern, August 1986 M. von Allmen

VIII
Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Experimental Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2. Absorption of Laser Light .... . . . . . . . . . . . . . . . . . . . .. 5

2.1 Fundamental Optical Properties .... . . . . . . . . . . . . . . .. 5
2.1.1 Plane-Wave Propagation. . . . . . . . . . . . . . . . . . . .. 5
2.1.2 Macroscopic Material Properties . . . . . . . . . . . . . . . 8
2.1.3 Nonmetals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.4 Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 12
2.2 Modified Optical Properties . . . . . . . . . . . . . . . . . . . . . . 15
2.2.1 Self-Focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.2 Phenomena Due to Free-Carrier Generation ........ 20
2.2.3 Semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.4 Insulators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.5 Reflection by Hot Metals . . . . . . . . . . . . . . . . . . . . 31
2.3 Phase Transitions and Shape Effects . . . . . . . . . . . . . . . . . 32
2.3.1 Surface Corrugation . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.2 Shaping Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3.3 Hole Drilling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3.4 Evaporation and Plasma Effects . . . . . . . . . . . . . . . . 39

3. Heating by Laser Light . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.1 Temperature Distributions . . . . . . . . . . . . . . . . . . . . . . . 41
3.1.1 Thermalization and Heat Transport . . . . . . . . . . . . . . 41
3.1.2 Solution of the Heat-Flow Equation . . . . . . . . . . . . . . 44
3.1.3 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.4 Moving Sources . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1.5 Variable Parameters . . . . . . . . . . . . . . . . . . . . . . . 50
3.1.6 Impact of Absorption Phenomena on Temperature .... 52
3.2 Heat-Treatment Processes . . . . . . . . . . . . . . . . . . . . . . . 55
3.2.1 Annealing with Laser Beams . . . . . . . . . . . . . . . . . . 55
3.2.2 Thermal Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2.3 Crystallization of Amorphous Semiconductor Layers ... 58
3.2.4 Compound Synthesis . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.5 Transformation Hardening .... . . . . . . . . . . . . . . . 65

IX
4. Melting and Solidification . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.1.1 Regimes of Laser Remelting . . . . . . . . . . . . . . . . . . 69
4.1.2 Heat Flow and Latent Heat . . . . . . . . . . . . . . . . . . . 72
4.1.3 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.1.4 Interface Kinetics. . . . . . . . . . . . . . . . . . . . . . . . . 81
4.1.5 Nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.2 Regrowth of Ion-Implanted Substrates . . . . . . . . . . . . . . . . 85
4.2.1 Semiconductor Substrates . . . . . . . . . . . . . . . . . . . . 85
4.2.2 Segregation and Trapping . . . . . . . . . . . . . . . . . . . . 89
4.2.3 Metallic Substrates . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3 Surface Alloying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3.1 Semiconductor Substrates . . . . . . . . . . . . . . . . . . . . 97
4.3.2 Constitutional Supercooling . . . . . . . . . . . . . . . . . . 98
4.3.3 Metallic Substrates ... . . . . . . . . . . . . . . . . . . .. 104
4.4 Melt Quenching . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 107
4.4.1 Glass Formation . . . . . . . . . . . . . . . . . . . . . . . .. 107
4.4.2 Silicon-Based Systems . . . . . . . . . . . . . . . . . . . . . 113
4.4.3 Metal-Based Systems . . . . . . . . . . . . . . . . . . . . .. 114

5. Evaporation and Plasma Formation . . . . . . . . . . . . . . . . .. 115

5.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.1.1 The Thermodynamics and Kinetics of Evaporation ... 116
5.1.2 Hydrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.1.3 Ionization of the Vapor . . . . . . . . . . . . . . . . . . . . 122
5.1.4 Gas Breakdown . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.2 Evaporation at Moderate Irradiance Levels . . . . . . . . . . . . 125
5.2.1 Beam Heating and Evaporation . . . . . . . . . . . . . . . 125
5.2.2 Vapor Expansion and Recoil ... . . . . . . . . . . . . .. 128
5.2.3 Drilling, Welding, Cutting . . . . . . . . . . . . . . . . . . 131
Hole Drilling . . . . . . . . . . . . . . . . . . . . . . . . . .. 131
Penetration Welding . . . . . . . . . . . . . . . . . . . . .. 134
Cutting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.3 Absorption Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.3.1 Laser-Supported Combustion Waves . . . . . . . . . . . . 138
5.3.2 Plasma-Enhanced Coupling . . . . . . . . . . . . . . . . .. 142
5.3.3 Laser-Supported Detonation Waves . . . . . . . . . . . .. 146
5.3.4 Effects of LSDWs
on the Beam-Material Interaction . . . . . . . . . . . . . . 148
5.4 Phenomena at Very High Irradiance . . . . . . . . . . . . . . . . 150
5.4.1 The Self-Regulating Plasma . . . . . . . . . . . . . . . . . 152
5.4.2 Laser-Driven Deflagration Wave . . . . . . . . . . . . . . 156
5.4.3 Inertial Confinement . . . . . . . . . . . . . . . . . . . . .. 157

x
 5.5 Pulsed Laser Deposition. . . . . . . . . . . . . . . . . . . . . . .. 159
5.5.1 Solid Film Growth from a Vapor . . . . . . . . . . . . . . 160
5.5.2 The PLD Process . . . . . . . . . . . . . . . . . . . . . . .. 162

A. Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
A.l Selected Material Data . . . . . . . . . . . . . . . . . . . . . . . . , 166
A.2 Green's Functions for Solving the Heat-Flow Equation .... , 168
A.3 Numerical Solution of the Heat-Flow Equation . . . . . . . . . , 173
A.4 Units and Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
A.4.1 Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . , 175
A.4.2 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
A.4.3 Subscripts. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 177

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

Subject Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 193

XI
1. Introduction

Laser beams, in a way, are a material scientist's dream: They can deliver
concentrated "pure" energy to almost any material, and they do so exactly
where, when, and in the quantity desired. Once available, lasers simply had
to be used in materials science, if only because they were there. Things
have, of course, turned out to be slightly more complicated, as we discuss in
the following. Nevertheless, the fascination of transforming materials by
laser beams will certainly continue to catch the fantasy of scientists and
engineers for many years to come.
The historical development of the field has strongly been influenced by
that of the laser technology itself. Experimental investigations of laser ef-
fects on materials began to appear soon after the first demonstration of the
ruby laser in 1960. Due to the poor reproducibility and beam quality of the
early lasers, this work was largely qualitative. It was mainly devoted to mat-
erial evaporation, in which careful dosage of the laser energy is not so cru-
cial. The popularity of the ruby laser was soon challenged by the newly dev-
eloped Nd lasers (Nd-doped YAG or glass), which are inherently more
stable and deliver optically superior beams, although they happen to emit in
the near-infrared rather than the visible spectrum, which makes absorption
in many materials more difficult. Later, with lasers and other equipment
becoming more sophisticated, experiments on controllable material melting
(without evaporation) became feasible, and laser processing of semiconduc-
tor structures and thin films began to attract enormous interest. Related
activities have largely dominated the literature since the late seventies [1.1].
Nd lasers, unlike the ruby laser, are also capable of Continuous-Wave
(CW) operation and offer the possibility of continuous, rather than pulsed,
processing [1.2]. Steady progress is being made in the development of
diode- rather than flashlamp-pumped Nd lasers which have dramatically
improved energy efficiencies. For most of the rugged industrial applications
like welding or cutting, however, the laser of choice is the CO2 laser,
because it is technologically simple; it compensates for its unfavorable in-
frared wavelength with versatility, high continuous power and energy-effi-
ciency. Another CW laser often used is the Ar laser which emits in the
green spectral regime, but it has a low efficiency and is not available for
high power levels. A most promising new type is the excimer laser, a pulsed
device which emits in the ultraviolet and combines high power with accept-
able efficiency [1.4]. As they are becoming cheaper and more versatile,
excimer lasers are increasingly replacing some of the earlier laser types.
With photon energies of 4 -:- 7 eV, 1 excimer beams are not only absorbed
efficiently in most materials, but are also able to break chemical bonds
directly upon absorption, i.e. before even heat is created. This opens up
regimes of interaction not accessible with other laser sources.
An even more recent development is the advent of lasers with sub-pico-
second pulse durations. As their available power increases steadily, femto-
second lasers start being explored also in the field of laser materials pro-
cessing. Most studies today have dealt with microscopic material structuring
by ablation. Femtosecond pulse durations are comparable or shorter than
the time constants for electronic relaxation in the material. This is a fasci-
nating new aspect, but the observed effects do not yet seem to be well un-
derstood and will not be treated in the present edition.
An alternative to laser beams as energy sources in materials processing
are particle beams, in particular medium-energy electron beams. They have
certain advantages with respect to absorption [1.5], and machines capable
of delivering pulse energies or CW powers adequate for efficient processing
are becoming available. If the existing differences in absorption characteris-
tics are taken into account, one usually finds that electron-beam-induced
material effects are practically identical to effects due to laser irradiation.
An obvious disadvantage of particle beams, apart from their more expensive
hardware, is the necessity to perform the processing in vacuum.

1.1 Experimental Aspects

Performing experiments on laser-induced material effects requires (in addi-

tion to diagnostic elements) three basic pieces of equipment: A laser, a
beam delivery system, and a fixture for the sample or workpiece. Laser
types suitable for materials processing, and more specific equipment data
are found in specialized periodicals [1.6]. A typical laboratory-type beam
delivery system may consist of nothing more than a simple lens and a set of
attenuating filters to adjust the beam power, while industrial systems often
contain elaborate setups of adjustable mirrors and lenses capable of guiding
the beam to any desired location within the working area. Optical fibres
provide a convenient and flexible alternative for beam guiding, but low-loss
fibres for CO 2 and excimer wavelengths are still being sought. In high-
power applications the delivery system is usually fitted with a device pro-
tecting the beam impingement area, such as a screen against debris or an
inert-gas nozzle. Adequate safety precautions should, in any case, be taken
to protect operators from exposure to the laser beam and from debris [1.7].
1 The symbol -:- is used throughout the text as a shorthand for "from -to" or "between ".

2
 The sample fixture in laboratory experiments typically consists of an
adjustable table, sometimes fitted with automated scanning equipment,
sometimes with a microscope. Scanning is usually done linearly by means
of an x,y table, or alternatively in a circular way by some sort of turntable.
Certain experiments (e.g., those involving reactive materials) must be per-
formed in vacuum or in an inert atmosphere, and therefore require the
sample fixture to be surrounded by a suitable chamber into which the beam
is directed through optical windows.
Another important consideration in designing experiments on laser-
solid interaction is diagnostics. For quantitative work, the beam power or
pulse energy should be monitored by a suitable calibrated detector. Much of
what is known today about the mechanism of absorption of intense light has
been learned by monitoring (in real-time) the laser light reflected or trans-
mitted by the sample. Other diagnostic techniques often employed in-situ
during laser processing include spectroscopy of emitted light (including
pyrometry), transient sample temperature or conductivity measurements,
measurements of thermionic emission as well as acoustic measurements.
Somewhat more costly techniques are real-time mass spectrometry of sam-
ple disintegration products, in-situ Auger or Rutherford backscattering
spectroscopy and time-resolved X-ray diffraction. The real-time informa-
tion gathered by this instrumentation is usually supplemented, of course, by
subsequent examination of the irradiated material, involving all the analyti-
cal techniques currently available to the materials scientist.

1.2 Outline

This monograph treats laser effects in metals, semiconductors and insula-

tors, with an emphasis on technological materials and structures. It is not,
however, intended to be an instruction manual (although some technical in-
formation on specific processes is given), but aims rather at providing the
physical insights from which instruction manuals are derived. Related topics
such as laser-induced chemical reactions in gases and liquids [1.8], and laser
effects in organic or biological materials [1.9], are beyond the scope of this
text.
The organization of the book is as follows. After this introduction,
Chap.2 deals with the absorption of light in solids. Intense laser beams tend
to modify the optical properties of the irradiated matter by a number of phe-
nomena, and the bulk of the chapter is devoted to such effects - from simple
heating through the production of free carriers to nonlinear-index phe-
nomena. Also treated are coupling effects due to macroscopic phenomena
like melting and surface rippling.

3
 The remammg chapters describe phenomena due to absorbed laser
light, loosely following a sequence of increasing sample temperature or
energy density. Chapter 3 is concerned with solid-state thennal processing,
usually done with continuous beams of moderate power. Here a section is
devoted to analytical calculations of the sample temperature under various
conditions of irradiation and material response, followed by discussions of
various processes of current interest. Chapter 4 treats laser remelting in its
various fonns. To emphasize interconnections between the various phe-
nomena observed in this regime, a section on modelling (including numeri-
cal simulation) and fundamental aspects of melting, soldification and phase
formation precede the sections that describe individual applications.
Finally, Chap.5 deals with the most "energetic" laser effects: evaporation
and ionization. It starts out again with a section on fundamentals, mainly in-
tended to facilitate understanding of the following sections, which then
describe phenomena pertinent to a wide range of laser intensities - from
drilling and welding to the production of hot vapors and plasmas. We end
our survey by returning to materials and consider deposition of thin films
from the laser-generated vapors.

4
2. Absorption of Laser Light

Laser light, in order to cause any lasting effect on a material, must first be
absorbed. As trivial as this may sound, absorption very often turns out to be
the most critical and cumbersome step in laser processing. An enormous
amount of work has been dedicated to investigating laser absorption mech-
anisms under various circumstances, and a great deal can be learned from
this work for the benefit of laser materials processing.
The absorption process can be thought of as a secondary "source" of
energy inside the material. Whilst driven by the incident beam, it tends to
develop its own dynamics and can behave in ways deviating from the laws
of ordinary optics. It is this "secondary" source, rather than the beam emit-
ted by the laser device, which determines what happens to the irradiated
material.
Section 2.1 is meant as a refresher and as a basis for subsequent discus-
sions. It summarizes the familiar optical properties of condensed matter, as
far as they are relevant to absorption. The following two sections then treat
modes of optical behavior influenced by intense laser irradiation, arising
from atomistic and from macroscopic material responses, respectively.

2.1 Fundamental Optical Properties

2.1.1 Plane-Wave Propagation

The simplest form of lightl is a monochromatic, linearly polarized plane
wave. This will, for most of our purposes, be a sufficient approximation of
a real laser beam. The electric field of a wave propagating in a homogene-
ous and nonabsorbing medium can be represented as

(2.1)

where z is the coordinate along the direction of propagation, w is the angu-

lar frequency, and A is the wavelength. The last two quantities are related

1 Here and in what follows, we shall interpret the words "optical" and "light" generously such
as to cover the whole range of wavelengths (between roughly 0.1 and 10 p.m) currently of in-
terest in laser materials processing.

5
through the phase velocity cln l , c being the speed of light, and n l the re-
fractive index of the medium [nl = 1 in vacuum; it is hardly different in air
at standard temperature and pressure (stp)] by

21r C
A= - - (2.2)
w nl

An expression analogous to (2.1) also holds for the magnetic field H. The
magnetic and electric field amplitudes are related by

(2.3)

with EO being the dielectric constant in vacuum. On average, the electric and
magnetic fields each carry the same amount of energy. However, in the
force f exerted by the electromagnetic wave on an electron

(2.4)

the contribution due to the magnetic field is smaller than that due to the
electric field by a factor of the order of vi c (v being the electron velocity),
and hence it is usually negligible. It is the term -eE in (2.4) that ultimately
produces just about every phenomenon discussed in this book.
The energy flux per unit area of the wave is termed irradiance2 and
given by

(2.5)

In the language of quantum mechanics, a wave of angular frequency wand

irradiance I corresponds to the flux II liw of photons of energy liw. Light in-
teracts with matter only in portions of whole quanta. However, while quan-
tum mechanics is required to understand the microscopic aspects of this pro-
cess, the photon fluxes in intense laser beams are enormous and classical
concepts are generally adequate to describe beam-solid interaction phe-
nomena.
The concept of a beam implies that the irradiance is maximum near the
optical axis and falls off laterally. The most common lateral distribution is a
cylindrically symmetric Gaussian

2 2
I(r) = 10 e- r /w (2.6)

2 The term "intensity", often used instead, denotes the energy flux per unit solid angle.

6
where 10 is the irradiance on axis (r = 0), and w is referred to as the beam
radius. The total power of the beam is then

P = 7rw210 . (2.7)

The distribution (2.6) strictly applies only to a laser operated in its funda-
mental (TEMoo) resonator mode, but we shall use it as an approximation
throughout. The wave front of a freely propagating Gaussian beam can usu-
ally be assumed to be approximately planar. In a first approximation, where
diffraction effects are ignored, this holds also for a focused beam in the vic-
inity of the focal point [2.1].
In absorbing media the real refractive index n t must be replaced by a
complex index n == n t +in2 . The meaning of n2 (also called the extinction
coefficient) becomes apparent when (2.2) is modified accordingly and in-
serted into (2.1): The electric field, upon propagation over a distance z,
decreases by the factor exp(wn2 z/c), indicating that some of the light
energy is absorbed. The absorption coefficient for the irradiance (2.5) is

(2.8)

The inverse of cx is referred to as the absorption length.

In inhomogeneous media the refractive index varies in space - in cer-
tain crystals it even depends on the direction of propagation. Spatial varia-
tions of the refractive index deform the wave front, bend the beam path and
cause secondary waves to split off from the primary one, as discussed in
standard textbooks on optics. We only cite the well-known formula for re-
flection of a wave perpendicularly incident from vacuum or air onto the
plane boundary of a solid with refractive index n. The ratio R of reflected-
to-incident irradiance is, in this case, given by the Fresnel expression

n -1 12
R= -I
n+l
-
. (2.9)

The reflectance and the absorption coefficient determine the amount of

beam power absorbed within the material. The power density deposited at
the depth z by a perpendicularly incident beam of irradiance I is

z
Ja(z) = 1(l-R)cx{l-exp[ - JOCX(ZI)dZ] }. (2.10)

7
This expression represents our "secondary source". The integral in the
exponential function is referred to as the optical thickness of the material
between 0 and z. In opaque materials (z » 11 a) the fraction of energy ab-
sorbed is determined by the quantity (l-R) alone, also known as the ab-
sorptance.

2.1.2 Macroscopic Material Properties

The connection between the refractive index and the properties of the medi-
um of propagation is formally provided by Maxwell's equations. For the
case of a nonmagnetic and isotropic material of dielectric constant Eand
conductivity u they may be written as

V·E =0, (2.11)

V·H =0, (2.12)
V xE = - (11 Eoc2 )aHI at , (2.13)
V xH = EEoaE/at + uE . (2.14)

Taking the curl of (2.13), recognizing that V x(V xE) == V (V· E)-V2 E and
using (2.11, 14) yields the wave equation

E a2 E u aE
V2E = - - + - - - - (2.15)
c2 at 2 E c2
o at .
Insertion of the plane wave ansatz (2.1) and use of (2.2) with the complex
index shows that the latter is related to E and u by

(2.16)

The quantity E defined by the right-hand side of (2.16) is the complex di-
electric function which can be regarded as a generalized response function
of the material. The real and imaginary parts of nand E are related by

E2 = 2n1 n2 ' (2.17)

n2 2 -
IEI-E1
...;........;-..;;..
- 2 . (2.18)

The dielectric function fully describes the response of a material to weak

electromagnetic irradiation. It depends on the light frequency in a manner
determined by the microscopic structure of the material. Let us briefly

8
review the main features of nonmetals (insulators or semiconductors) and
metals.

2.1. 3 Nonmetals

Insulators and semiconductors in the absence of excitation have only bound

electrons and are basically transparent, except in the vicinity of what can be
thought of as resonances. In the classical Lorentz model the electron is re-
presented as a harmonic oscillator driven by the oscillating force of the
wave's electric field (2.4). The oscillating electrons produce a macroscopic
polarization of the material which superposes on the electric field of the in-
cident wave. The dielectric function is simply the ratio of the total field
(wave plus polarization) to the field of the wave alone. In quantum mechan-
ics, a resonance corresponds to the transition of an electron between two
states, the energy difference ~E of which determines the resonance fre-
quency Wo = ~E/h. The classical and quantum-mechanical treatments yield
almost identical expressions for the dielectric function. For a nonmetal with
Ne bound electrons showing one single resonance we have [2.2]

(2.19)

Here fose' the oscillator strength, is a measure of the probability of the

transition, while the damping constant r describes the width of the reso-
nance that arises from the finite width of the initial and final electron
states. The optical consequences of a resonance are illustrated in Fig. 2.1. It
shows €, along with the derived quantities n, R and IX, as a function of the
light frequency, as obtained from (2.8,9,17-19). Note that a peak in the
absorption coefficient due to a resonance is invariably accompanied by a
peak in the reflectance.
Real materials have, of course, many resonances, and in their expres-
sion for € the resonance term of (2.19) is replaced by a sum over many
terms. The most important resonance arises from transitions of valence-
band electrons to the conduction band (interband transitions). To induce an
interband transition the incident photon must have an energy at least equal
to the band-gap energy Eg . The free carriers (electrons and holes) created in
pairs in interband transitions can, if present in sufficient numbers, influ-
ence the optical response of the material. Insulators have band gaps corres-
ponding to light frequencies in the vacuum ultraviolet, and carrier concen-
trations under day-light illumination remain negligible. Semiconductors, on
the other hand, have band gaps in the visible or infrared part of the spec-
trum. Free carriers (optically or thermally generated) contribute measurably
to the metal-like reflectance of many semiconductors in the visible region.

9
 Fig.2.1. Frequency dependence of the
20 dielectric function, the refractive index,
f the Fresnel reflectance and the absorp-
tion coefficient for a medium with a
w: ·
W 0
single resonance at Wo (calculated for
liwo = 4eV, lir= leV, N = 5·1Q22 cm-3)

t 2
~
cl

7
0.8
6
t 0.6
5 d
01
0.4 .2

w-

Most nonmetals show - in addition to electronic transitions - resonant

coupling to high-frequency optical phonons located in the near-infrared
region of the spectrum. Phonon coupling can be described by resonance
terms similar to the one in (2.19), but with masses and damping constants
characteristic of lattice vibrations rather than electronic vibrations.
Figures 2.2,3 show the absorption coefficient and the reflectance as a
function of wavelength for two representative nonmetals - quartz (E g =
6.geV, corresponding to a wavelength of 180nm) and crystalline Si (E g =
1.1 eV at room temperature, corresponding to A = 1.13 JLm). Both materials
exhibit the same essential features - absorption peaks around 10 JLm due to
phonon coupling, weak absorption at intermediate wavelengths and a steep
increase in absorption as the photon energy approaches the band gap.
With reference to Fig.2.3, it will be noted that the absorption coeffi-
cient of Si increases with decreasing laser wavelength only gradually near
1.13 JLm (flw = 1. leV), but rather abruptly near 0.36 JLm (flw = 3.4eV). The
explanation is that there are two different kinds of interband transitions in
materials like Si. The first kind, direct transitions, lead from valence-band
to conduction-band states with the same wave vector. Such transitions are

10
 0.8

crystalline Si02

i
0.6 E
v
'a
a: a. .... "E
III
III
u 0.4 ~
c;
g 't
u
102 0
u

.
.!!
't .2
Q.
c;

10 5
I/)
.0
a

00.1 0.2 0.5 2 5

wavelength (jJm)

Fig.2.2. Reflectance and absorption coefficient as a function of wavelength for cry-

stalline quartz

106
Silicon
~

"i
1.0 E
~
loS "E
III
0.8 'u
;:
a: 't
III 0
u u
c;
a 0.6 lei' c:
a5
0
U
.
.!!
Q; I/)
.0
a

0.2
a xlOOO
102
00.1 0.2 05 5 10
wavelength (jJm)
Fig.2.3. Reflectance and absorption coefficient as a function of wavelength for cry-
stalline silicon

11
possible in Si only at photon energies exceeding its direct gap at 3.4 eV.
The second kind of interband transitions, indirect transitions, involve val-
ence- and conduction-band states of different wave vectors. The difference
in wave vector must be provided by a phonon, as the photon itself carries
only negligible momentum. Since the fundamental gap at 1.1 eV is an indi-
rect gap, it allows only phonon-assisted transitions. The probability of such
transitions depends on the phonon occupancy and is relatively small and
temperature-dependent. Other indirect-gap materials are Ge, diamond, SiC,
GaP, AlP and AlAs, whereas materials such as Se, Te, ZnO, as well as most
II-VI and IV-VI compounds have direct fundamental gaps.
The simple picture outlined above applies only to homogeneous materi-
als. In nonmetals inhomogeneous on a scale of one wavelength or more the
optical appearance is modified by light scattering at grain boundaries or in-
clusions. This results in quite strong absorption even in materials that, in-
trinsically, would be transparent, e.g., ceramics. The effect can be rational-
ized in terms of an effective multiplication of the light path inside the mate-
rial caused by a large number of scattering events. Inclusions that are small
compared to the wavelength can, on the other hand, be treated in terms of
an "averaged" dielectric function. An expression often used to describe ran-
domly distributed inclusions (nonmetallic or metallic) with the dielectric
function £j in a matrix with the dielectric function Em was first derived by
Maxwell-Garnett in order to explain the colors shown by certain suspen-
sions or "colored glasses". The average dielectric function E follows from
[2.4]

E(£j -Em)
(2.20)
Ej +2Em

where E < 1 denotes the volume fraction occupied by the inclusions. Eq.
(2.20) predicts "resonances" leading to absorption peaks that are present
neither in the pure host nor in the inclusion material.

2.1.4 Metals

The optical response of a metal is dominated by the conduction electrons.

Since the electron gas is degenerate, only electrons in states close to the
Fermi level, referred to as free electrons, contribute to the optical proper-
ties. There is no resonance frequency for a free electron, and its only inter-
action with the lattice is by collisions. The dielectric function of a free-elec-
tron metal can be obtained from (2.19) by formally replacing the damping
constant r by the inverse collision time 1/ 7e and setting the resonance fre-
quency equal to zero, and fose equal to one. The resulting expression can be
written as

12
 (2.21)

(2.22)

is the electron plasma frequency. At w = wp (which is in the vacuum ultra-

violet for most metals) both E1 and n 1 vamsh. The variation of E and the
associated quantities with the light frequency is shown in Fig.2A. The
plasma frequency is seen to separate two regimes of rather different optical
properties: Large Rand 01. for w < wp ' and small Rand 01. for w > wp'
The optical properties of a free-electron metal for w < wp are related
to its DC conductivity ao' With the aid of the Drude formula [2.2,3]

(2.23)

4
t....
w
.: .
2
E2 E,
0

tN
2
C
i-
n,
R
7
0.8
6 t
t 0.6 5 d
Fig.2.4. Frequency dependence of
It: the dielectric function, the refrac-
0.4 ~ tive index. the Fresnel reflectance
4 and the absorption coefficient of a
0.2 free-electron metal (calculated for
3 liw(! = 8.3eV. corresponding to Ne
= ,·1022 cm·3, and IiITe = O.02eV)

-
0 Wp
W

13
useful approximations for the optical parameters can be obtained. For the
range W « liTe (far-infrared region) Eq.(2.21) gives fl =:: -(10TeIfO and f2
=:: (101 EOW, from which follows that n l =:: n2 =:: «(10/2Eow)1I2. Using (2.8,9)
yields

I- R =:: V'8Eowl (10 (2.24)

and
(2.25)

For the range I I Te < W < wp (near-infrared and visible regions for most
metals) we have n l =:: wp/2w2Te ::::: 0 and n2 =:: wplw which gives

(2.26)

and

(2.27)

In real metals the simple free-electron behavior is modified by a

number of secondary effects, in particular by interband transitions. As an
illustration, we depict in Fig.2.5 the coupling parameters for Al and Au.

a AI 106
a,Au

las .,E
~

to R,AI u
ts
O.B f
10' ;g
a: 'ii
0.6 o
CII
u u
c
.5! .§
~
u
0.4 103 e-o
~ AI and Au evaporated films 1/1
.J:l
C
0.2

0
0.1 0.2 0.5 2 5 10
wavelength (}1m)
Fig.2.S. Reflectance and absorption coefficient as a function of wavelength for alumi-
num and gold

14
The curves of both metals show oscillations caused by interband transitons,
such as the one at 0.83 ILm (l.5eV) in Al. A comparison with Fig.2.4 indi-
cates that the plasma frequency for Al is indeed in the vacuum ultraviolet,
whereas that of Au appears to be in the visible. The anomalously low appar-
ent plasma frequency of Au (which causes its yellow color) is due to transi-
tions from d-band states, which set in at nw ~ 2 eV. The shift in wp is read-
ily understood from an inspection of Figs.2.IA: The d electrons make a
bound-type contribution to the dielectric function, and since the contribu-
tion to El is positive, the point El = 0 which defines wp is shifted to a lower
frequency. Similar behavior is also found in eu and Ag.
The optical properties predicted by Fig.2.4 are, of course, bulk proper-
ties. In practice, metal surfaces usually show lower reflectance than the
bulk due to contamination (adsorbates, oxide layers, etc.) or macroscopic
defects. This explains the appreciable scatter in the reflectivity data found
in the literature. Deviations from bulk behavior may also arise from "intrin-
sic" surface phenomena such as plasmon excitation or diffuse electron
scattering, particularly in thin films. The optical behavior becomes almost
completely dominated by surface effects when the dimension of the metal is
shrunk to values of the order of the absorption length, such as in extremely
thin evaporated films or in aggregate structures consisting of small insu-
lated metallic particles [2.4].
This concludes our short survey of the "linear" optical properties of
materials. Some experimental values of reflectance and absorption length
for a number of nonmetals and metals are given in Table A.I.

2.2 Modified Optical Properties

The fundamental optical properties outlined in the previous section repre-

sent the response of materials to light that is weak enough so as not to per-
turb the states of the electrons and atoms significantly. Powerful laser irra-
diation, however, is known to alter the optical properties of many materials,
and often drastically so. Coupling is then no longer well characterized by a
static dielectric function but becomes a dynamical process. The amount of
light absorbed in the material may turn out to be smaller or larger than
expected from the fundamental optical properties.
All beam-induced changes in the optical properties of solids can be traced
back to one of three mechanisms. These are, roughly in a sequence of in-
creasing irradiance, the following:
(i) Heat production and resulting changes in the density or the electronic
characteristics of the material. Pertinent effects include thermal self-
focusing in transparent media as well as "thermal runaway" phe-
nomena in semiconductors and metals.

15
(ii) Optical generation of free carriers by interband transitions or impact
ionization in semiconductors and insulators. As a result, the absorption
coefficient increases dramatically, possibly causing explosive material
damage.
(iii) Nonlinear distortion of electron orbitals or whole molecules by the
electric field of an intense beam. A host of nonlinear optical pheno-
mena, including self-focusing and multiphoton absorption, are caused
by field effects.
Apart from these "intrinsic" phenomena, beam-solid coupling, in addi-
tion, tends to be affected by beam-induced changes in the shape of the mat-
erial, usually in connection with melting or evaporation. These will be dis-
cussed in Sect.2.3. In what follows we shall consider phenomena in solids as
related to the three mechanisms mentioned above.

2.2.1 Self-Focusing

Self-Focusing (SF) - or its opposite, self-defocusing - occurs if the real part

of the refractive index of the medium varies locally as a function of irradi-
ance. Extensive reviews on this topic were presented by Akhmanov [2.5],
Svelto [2.6], and Shen [2.7]. Since laser beams are more intense near the
beam axis than away from it, an irradiance-dependent real index of refrac-
tion has an effect somewhat similar to a lens placed in the beam path. The
"lens", as illustrated in Fig.2.6, is converging if n l increases with the irradi-
ance - the phase velocity decreases towards the beam axis, causing a plane
wave front to become concave and eventually to collapse. Similarly, the
"lens" is diverging if n l decreases as a function of irradiance. This self-de-

~ Fig.2.6. Geometrical rays (arrows) and phase

fronts of a Gaussian beam undergoing self-

.~.
focusing in a medium in which the local real
refractive index increases as a function of irra-
diance
r

16
focussing is inconsequential as far as energy deposition is concerned (al-
though it can be disturbing), but self-focusing increases the effective irradi-
ance and reinforces other irradiance-dependent phenomena. The variation
in refractive index can itself be thermal or field induced.
Let us first consider thermal self-focusing. The temperature depend-
ence of the refractive index can be thought of as consisting of two parts

dn 1 = (an 1] (an1] dp (2.28)

dT aT p + ap dT'
The first term accounts for the temperature dependence of the electronic or
molecular polarizability due to shifts in the absorption bands. Thus, in mate-
rials with band-gap energies in the ultraviolet region that decrease with tem-
perature, this term tends to be positive in the visible spectrum. The second
term, connected to thermal expansion (p being the density of the solid), is
usually negative since an 1/ap > 0 and dpldT < O. (In addition, there may
be strain-induced birefringence, which we do not consider here). Hence, the
first term may favor self-focusing while the second tends to oppose it. How-
ever, the two contributions to (2.28) have different relaxation times (ther-
mal expansion proceeds at the sound velocity and is much "slower" than
heating itself), and there may be transient self-focusing for a short pulse
even if the steady-state value of dn 1I dT is negative.
To describe thermal SF the equations for beam propagation have to be
solved simultaneously with the associated heat-conduction problem [2.5].
Consider a weakly absorbing medium in which dn 1IdT > 0, irradiated by a
parallel Gaussian beam of radius wand power P. The absorbed power cre-
ates a continuous positive thermal lens in the material along the beam path.
The minimum absorbed power required for the thermal lens to balance
beam diffusion by diffraction in steady state is found to be

AI(
P tr = dn 1 IdT (2.29)

where K is the thermal conductivity. For P exceeding P tr the beam focusses

after propagating the distance (aberrationless approximation)

(2.30)

where Zo = 271'w2n1 IX is the diffraction length of the beam. In many mater-

ials thermal SF may take place at powers as little as 1 W. Under certain con-
ditions the focused beam can produce a thermal waveguide in the medium

17
where it travels for distances » zo without divergence. In a typical experi-
ment [2.8], a self-guided filament about 50 /-tm in diameter, resulting from
thermal SF of a 3 W Ar laser beam in lead glass (nfO) = 1.75, dnl/dT =:::
1O-5K-I,K =::: 5.4·1O- 3 W/cm·Kanda =::: O.lcm- I ), was observed at a dis-
tance of 20 cm.
Electric-field- induced SF, unlike its thermal counterpart, is a high-
irradiance effect relevant only for powerful nano- or picosecond pulses.
The root of field-induced changes in the refractive index is the anharmoni-
city of all interparticle potentials. As the force (2.4) grows stronger, the
linear relation between the electric field and the polarization is eventually
lost and the harmonic oscillator model, on which (2.19) is based, fails to
reproduce the interaction accurately. The true motion of the oscillating
electrons now contains Fourier components at frequencies other than the
driving frequency: the material polarization, and hence the dielectric func-
tion, show resonances at multiples of w (including at w = 0). Related physi-
cal phenomena include the generation of frequency-shifted secondary waves
(frequency multiplication, optical mixing and stimulated scattering), as well
as a field-dependent dielectric function for the incident wave. The latter is
commonly written as

(2.31)

where E(O) is the linear dielectric function. Electronic as well as molecular

polarizability contributes to the nonlinear terms in (2.31). In isotropic
media a major nonlinear contribution is due to electrostriction. In liquids
strong fields are known to induce orientation of molecules, making the
medium optically anisotropic and increasing the average value of EI (high-
frequency Kerr effect). The nonlinear parameters E(n) in (2.31) are, in gen-
eral, complex. For the present discussion we are interested in the real parts;
the imaginary parts are related to multiphoton absorption (Sect. 2.2.4).
In dealing with field-induced SF it is convenient to introduce an inten-
sity-dependent real refractive index of the form

n I -- n(O)
I + "II (2.32)

where nfO) is the linear index and "I = E(2) 12cEO E(O).
Consider a transparent medium with "I > O. In the geometrical-optics
approximation the effect of SF on a Gaussian beam can be described in
terms of a field-induced thin lens with a focal length that depends on irradi-
ance [2.6]

(2.33)

18
The lens exactly balances diffraction if znl equals zoo This holds, independ-
ently of the incident beam radius, if the power is equal to the threshold
value

(2.34)

In the optical glass "BK-7", 'Y is 4.10- 16 cm2 /W, for which (2.34) predicts a
threshold of about 3 MW for green light (allowance for aberrations leads to
a slightly different numerical factor). The situation in which SF and diffrac-
tion just balance is known as self-trapping. For P > Ptr the beam collapses
into a focal spot after propagating along the distance (aberrationless approx-
imation)

Zo
Z = -,.=-===== (2.35)
f V'P/P tr - 1.

As the beam power varies in time so does zf and the point of self-focus
moves back and forth along the optical axis. However, the above equations
hold only for pulses long compared to the response time of the nonlinear
process. For electronic polarizability this is of the order of 10- 14 s, while
coordinated molecular motion, as involved in the Kerr effect, takes of the
order of 10- 11 s. Electrostriction has a time constant given by the transit
time of a sound wave across the beam diameter, i.e., 10-9 s or more (time
constants for thermal SF can be as long as 1s). Whenever the pulse duration
is of the order of the relevant time constant, SF is of a transient nature.
Qualitatively, the picture is that the index nonlinearity experienced by the
leading edge of a pulse is smaller than that seen by the trailing edge.
Accordingly, only the lagging portion of a pulse experiences SF. For short
pulses relaxation effects may result in effective suppression of SF. Besides,
SF is not limited to narrow Gaussian beams but can occur in beams of arbi-
trary diameter if their lateral profile is modulated. Individual peaks or crests
of sufficient power can undergo SF independently, causing broad powerful
beams to break up into many small filaments inside a nonlinear medium,
e.g., a solid-state laser amplifier. Self-focusing has also been observed in
gases [2.9].
The arguments considered so far would predict an infinite irradiance to
result, for diffraction cannot prevent the beam from collapsing into a point-
like focal spot once P > Ptr" In reality, terminal focal diameters are found
to be at least a few /Lm. The mechanism ultimately limiting SF is the genera-
tion of free carriers, either by thermal emission in absorbing media or by
optical breakdown in nonabsorbing ones [2.10]. The free-carrier contribu-
tion to the real part of the refractive index is negative (see below), and it

19
will eventually cause self-defocusing. At the same time, the free carriers
cause strong absorption. The typical result is explosive thermal damage in
the focal region.

2.2.2 Phenomena Due to Free-Carrier Generation

Free-carrier generation is the most important self-induced coupling effect

in nonmetals. Before discussing mechanisms of free-carrier generation let
us briefly consider the impact of free carriers on the optical properties of a
nonmetal.
The simplest approach is to think of the total material polarization, and
hence of the dielectric function, as a sum of lattice and carrier contribu-
tions. Remembering (2.21,22) and considering electrons only, we can write
the total refractive index as (assuming w » 11 Te)

(2.36)

where Do is the complex lattice index and wp = (Ne e 2 / me EO) 112, with Ne
being some function of the irradiance. It is assumed here that Ne is small
compared to the atomic density and, in particular, small enough that the lat-
tice and the carriers behave independently of each other (a condition often
violated in practice, as we shall see later on). A term analogous to the elec-
tron term applies for holes. The effect of free carriers on the optical proper-
ties of the nonmetal, as obvious from (2.36), is to reduce the real part and
to increase the imaginary part of n. This, according to (2.8,9), increases
the absorption coefficient and tends to decrease the reflectance for wp < w.
Only for carrier densities large enough that wp > w does the reflectance in-
crease by further increasing wp .
In semiconductors, holes tend to be mobile and contribute significantly
to absorption. Since electrons and holes come in equal numbers, it is con-
venient to treat them together and to write the total absorption coefficient
in the form

(2.37)

where ao is the lattice absorption coefficient, Neh = Ne = Nh is the density

of carrier pairs, and Eeh is the absorption cross section of a carrier pair,
defined as

e2 1 1
-E-n--c-w- [-m-e*-Te- + -m-;-r,-h-] . (2.38)
o I 2

20
Here m; and 7 x (x = e, h) are the effective masses and collision times of
electrons and holes, respectively [2.11]. Note that l;eh scales essentially with
},,2, making free-carrier absorption mainly relevant for infrared beams. For
example, in Si experimental values of l;eh at 20°C are around 5 .10- 18. cm-2
at 1.06 IAffi [2.12] and around 1.10- 16 cm-2 at 10.6 pm [2.13].
Let us now consider mechanisms of carrier generation. Free-carrier
phenomena are of a somewhat different nature in semiconductors and in in-
sulators, and we shall consider the two types of material separately.

2.2.3 Semiconductors

Even far-infrared beams with liw < Eg and of moderate irradiance can
create significant densities of free carriers in semiconductors, due to ther-
mal excitation. The equilibrium free-carrier density in an intrinsic semicon-
ductor at temperature T is given by

(2.39)

from which it follows that the absorption coefficient (2.37) increases with
temperature faster than exp(lIT) if ao is constant; in applying (2.39) it is
generally necessary to allow for a temperature dependence of the band gap
[2.14]. For illustration, Fig.2.7 shows measured and calculated absorption
coefficients at 10.6 pm in crystalline n-type Si of various doping levels as a
function of temperature [2.15]. The additional absorption by the free car-
riers causes stronger heating which leads to even more carriers in a sort of
positive feedback loop. This results in typical "thermal runaway" behavior,
observed particularly with far-infrared radiation where lattice absorption is
small and free carriers absorb efficiently.
Photons with liw > Eg create carriers by a mechanism far more effi-
cient than heating, namely interband absorption, in which every absorbed
photon leaves behind a carrier pair. Optically generated carriers and their
effects have received considerable attention in connection with laser anneal-
ing of Si by nanosecond pulses. The rate of generation of carrier pairs by
interband absorption in an intrinsic semiconductor is aoIlliw, were ao is the
coefficient for interband absorption, and I is the local irradiance which de-
pends on both interband and free-carrier absorption [2.16]. To model the sit-
uation, carrier generation as well as carrier losses, due to diffusion and re-
combination, must be allowed for. The resulting carrier density follows
from a differential equation of the form

(2.40)

21
 TEMPERATURE IKI Fig.2.7. Absorption coefficient at h =
1O.6/Lm in phosphorus-doped Si at vari-
ous doping levels, as a function of tem-
perature. (Dots and squares are meas-
ured, solid lines calculated) [2.15]

10

(bl 500 700 900

TEMPERATURE IKI

where the first term describes generation and the second diffusion of car-
rier pairs. The ambipolar diffusion constant Damb allows for the attraction
between ~ectrons and holes. The last term of (2.40) is the recombination
rate (for Neh > Neh this term would describe thermal carrier generation).
The recombination lifetime Tr must be considered as a function of both tem-
perature and carrier density. In most semiconductors the dominant recombi-
nation mode at large carrier densities is Auger recombination in which the
recombination energy Eg is given to a third carrier (electron or hole), rather
than to the lattice. Auger transitions are thus three-particle processes with
lITr ex Neh2.
Under conditions typical for pulsed laser annealing, the carrier density
can reach values where the energy absorbed by the free carriers exceeds
that absorbed by the lattice. Moreover, at densities above 10 17 or 10 18 cm- 3
collisions among the carriers begin to dominate over collisions with the lat-
tice, and carriers behave collectively, i.e., as a plasma. It is well known that
thermally excited carriers in a semiconductor have a Boltzmann distribution
- this is merely the limiting form of the Fermi distribution for carriers with
E-Ep »kT. Optically excited carriers tend to have more complicated dis-
tributions, depending on the properties of the light source as well as the
solid. Light absorption by free carriers sends electrons into states high up in
the conduction band and holes deep down in the valence band. Carriers
excited by short, intense pulses "remember" the pulse spectrum in their

22
energy distribution [2.17]. This nonthermal distribution relaxes by means of
carrier-lattice as well as carrier-carrier collisions. Whilst at low concentra-
tions carriers reach thermal equilibrium with the lattice individually, in the
plasma regime a thermal distribution is established among the carriers be-
fore equilibrium with the lattice is reached [2 .18]. One may then speak of
carrier (electron and hole) temperatures that differ from the lattice tempera-
ture. In a Boltzmann plasma the carrier temperature Tc is related to the
average carrier excess energy (Ec) by (Ec) = (3/2)kTc (c = e or h). Elec-
tron and hole plasmas exchange energy by e-h collisions and by Auger
processes, and for times long enough for the electrons to thermalize among
themselves electron and hole temperatures may be taken as equal. A sche-
matic representation of the electron plasma inside a laser-irradiated semi-
conductor, indicating typical values for the various time constants involved,
is shown in Fig. 2.8.
Let us now consider the role played by the carrier plasma in the beam-
solid interaction. The plasma may be regarded as an independent system
which absorbs light and exchanges energy with the lattice. The plasma gains
energy from interband absorption (an amount of liw-E g per carrier pair) as
well as from free-carrier absorption. Furthermore, if Auger processes domi-
nate the recombination, an amount Eg is added to the plasma energy for
every recombination event. The plasma loses energy by carrier diffusion -
every carrier pair carrying away its mean excess energy (Ec) - and by
phonon emission which heats the lattice . The latter is , apart from residual
phonon absorption, the only source of lattice heating . The plasma thus acts
as a sort of funnel through which the absorbed energy must go before it is

e-lattice
coll isions

fiw
in terband
absorption
- Eg r-~~------------------------~~~-----

Fig.2.8 . Schematic diagram of electron energies in a laser-irradiated semiconductor,

showing the free-electron plasma exchanging energy with the beam and the lattice. An
analogous diagram holds for holes

23
given to the lattice. The resulting average excess carrier energy - taking
electrons and holes to be in equilibrium with each other - follows from an
expression of the form [2.19]

(2.41)

The first term represents energy gained from interband and free-carrier
absorption, while the second is the energy liberated upon Auger recombina-
tion. The last two terms represent energy lost to diffusing carriers and
energy transferred to the lattice, respectively. Here 3kT12 is the equilibrium
carrier excess energy, and TE is an effective carrier-lattice energy relaxa-
tion time. Eq.(2.41) must be solved together with (2.40) and the heat-flow
equation for the lattice temperature. If the irradiance changes slowly in
comparison to the time constants in (2.41), the plasma density and energy
will assume steady-state values which make gain and loss rates equal. The
main difficulty in such a calculation is that the material parameters Damb
E g , ao, Eeh , Tr and TE depend on the lattice temperature and, via screening
effects, also on the density and temperature of the plasma. In Si, carrier
lifetimes and energy relaxation time are expected to saturate at carrier den-
sities of the order of 1021 cm3 [2.18]. As an illustration of this regime,
Fig.2.9 exhibits the surface temperature of crystalline Si irradiated by a Nd
laser pulse, as calculated numerically by Lietoila and Gibbons [2.19], with
and without inclusion of absorption by the carrier plasma (also shown is the
influence of the temperature dependence of Eg). At the instant when the
lattice reaches the melting point (1412°C), the carrier density turns out to
be 2· 1020 cm- 3 .
The assumption of a Boltzmann distribution of carriers is justified only
as long as the carrier density remains small compared to the density of
states in the energy range covered by the plasma; otherwise the plasma
becomes degenerate. The main effect of degeneracy is to limit the rate of
energy exchange between carriers due to the lack of available unoccupied
states. Also affected, and for the same reason, is the Auger recombination
rate which is found to increase only as the square, instead of the cube, of
the carrier density in degenerate plasmas [2.20]. The decreasing rates of
carrier scattering and recombination cause an increase in the energy diffus-
ing out of the light-absorbing zone by the flow of hot carriers. Yoffa [2.18]
predicted that the ambipolar diffusivity Damb should become larger than
usual thermal diffusivities by about 3 orders of magnitude. Under such cir-
cumstances the spatial distribution of deposited energy would clearly be det-
ermined by the carrier diffusion length, rather than by the light-absorption
length. On the other hand, diffusion may to some extent be counteracted by

24
 energy (J I cm 2 )
0 2 468 10
14:)0 I
I

" crystalline Si
1200 I
I

,.... I

u FCA included I
I

!... 1000 I
Eg :Eg (T) " Eg : constant
....C1I
:J 800
"0
....
C1I
a. 600

-
E
C1I
400

200
No FCA
..................................................................................................................................
00 10 40 2050 60 70
30
time (ns)
Fig.2.9. Calculated surface temperature in crystalline Si irradiated by a Nd-Iaser beam
of 160 MW Icm 2 as a function of time (solid line). Different curves result if free car-
rier absorption (FCA) is ignored (dotted) or included with a constant (dashed), rather
than temperature-dependent, band gap. [2.19]

a local decrease in band gap caused by lattice heating or by plasma effects,

which tends to confine the carrier plasma (see below).
Dense carrier plasmas can store energy not only as thermal energy but
also in the form of plasma oscillations. A share of the order of 10% of the
total plasma energy was estimated to consist of plasmons in laser-irradiated
Si with Nc :::::: 1020 cm- 3 [2.18]. Plasmons can be treated as quasi-particles
which dissipate energy in collisions with the lattice. The interaction is
strongest if the plasmon energy equals the relevant phonon energy (typically
below 0.1 eV, corresponding to carrier densities below 10 17 cm- 3 ). In much
denser plasmas, plasmons tend to decouple from the lattice unless the plas-
mon energy reaches or exceeds Eg . In this event plasmon decay via e-h pair
creation as well as recombination by plasmon emission become possible.
Such processes have been observed in Ge and Si irradiated by picosecond
pulses [2.21,22].
The creation of a dense carrier plasma, which involves the removal of a
significant fraction of the binding electrons, should have an impact also on
the properties of the lattice itself. First, moving electrons from valence-
band to conduction-band states reduces the size of the band gap in most
semiconductors, while in some (like HgTe and HgSe) it causes the reverse
effect [2.23]. A local narrowing of the band gap - be it caused by plasma
effects or by local lattice heating - creates an electric field which tends to
confine the plasma. Lietoila and Gibbons [2.19] included the effect in their

25
computer modeling of laser-irradiated Si (Fig.2.9), but found it to be insig-
nificant in their case.
The plasma should also affect the mechanical strength of the semicon-
ductor lattice. The removal of electrons from (bonding) valence-band to
(antibonding or plane wave) conduction-band states tends to soften the lat-
tice and to reduce the phonon frequencies. The reduced phonon energies
result in an increased number of excited phonons at any given temperature,
further destabilizing the lattice. Van Vechten and associates [2.24] proposed
that a large density of photoexcited carriers could lead to a fluid-like state of
the material at lattice temperatures far below the normal melting point,
thereby bringing about the observed crystallization of amorphous Si by na-
nosecond laser pulses without thermal melting. As main experimental sup-
port for this hypothesis the researchers cited measurements of lattice tem-
perature based on Raman scattering, which indicate insignificant lattice
heating even for irradiances above the observed annealing threshold [2.25].
The interpretation of the Raman data remains a subject of controversy,
however [2.26]. Particularly hard to explain by the plasma annealing hypo-
thesis is the observed duration (up to some 10-7 s) of a state of enhanced
reflectivity of the irradiated Si [2.27], readily explained by the metallic
nature of molten Si if a thermal melting mechanism is assumed. Deforma-
tion-potential scattering of energetic electrons (Ee :::: 1eV) in Si yields
energy relaxation times of the order of 10- 12 s, and screening seems unable
to increase this time by anything near the amount required [2.28]. Circum-
stantial experimental evidence for thermal melting in nanosecond laser
annealing of Si has been presented by a number of researchers [2.29-31].
Experiments with picosecond and sub-picosecond pulses indicate that the
relaxation of energetic carriers in Si is much faster than that of thermal
ones. Malvezzi et al. [2.32] inferred ambipolar diffusivities of 1/20 of the
near-equilibrium value, leading to a hot-carrier diffusion length of only
some 10 nm. These and other experiments show that thermal melting in Si
is delayed by plasma effects by at most a few picoseconds [2.33]. As an
illustration, Fig.2.10 shows time-resolved reflectance measured at A = 1
/-tm in crystalline Si irradiated by red 90 femtosecond pulses (A = 620nm)
of various fluences [2.34]. Here, for 0.63F th (F th denoting the threshold
fluence for surface melting) the reflectance is seen to decrease, indicating
that the free-carrier density corresponds to the condition wp < w for the
probe frequency. The reflectance then slowly returns to its static value as
the carrier plasma diffuses away from the surface. For fluences at and
above Fth , the reflectance first increases, indicating that now wp > w (cor-
responding to Ne > 5 X 10 21 cm-3 ), but then quickly decreases and increases
again, apparently due to melting. After a few picoseconds the reflectance
assumes a quasi-static value determined by the thickness of the metallic
melt layer formed. At 4F th explosive surface damage occurs.

26
 0.4
~
! .ftI*-··.
1:-~d~,a-a-.,;--;....-......~-• •- __~_
2.5FTH
a
0.3 87' ............... .
4.0F~'"H--------- •. _

0.2
-
-0
c:z::
c:z::
en
~ 0.1

o I I I
10 15 lips) 20

-0.1

Fig.2.t O. Time-resolved reflectance at >. = I 11m in crystalline Si irradiated by red 90

fs pulses of various fluences. Fth denotes the melt threshold fluence. [2.34]

2.2.4 Insulators

Let us now consider transparent media, assuming that hw < E g . Even in

such materials, free-carrier generation is observed at high irradiance due
either to multiphoton transitions or to impact ionization.
In a multiphoton interband transition, n photons, the combined
energy of which exceeds Eg , are absorbed simultaneously. The simultane-
ous absorption of n photons by an atom requires, according to the uncerta-
inty principle, that those n photons be incident upon its cross-sectional area
within a time interval ot = hi liw = 2 7r I w, or one period. The probability of
such an event is proportional to the photon flux I/liw raised to power n
[2.35]. The rate of carrier-pair generation may be written as

a
-N NE(n)(l/liw)n = a(n)-
I
at eh
=
0 - 0 liw
(2.42)

where E6 n) denotes the cross section for n-photon lattice absorption (with
dimensions m2n sn-1) and a6 n) the corresponding absorption coefficient.
Both quantities are related to the nonlinear dielectric function of order n
appearing in (2.31) (for large n the situation becomes somewhat more
complicated due to field-induced shifts in the band structure). Multiphoton
ionization as a mechanism contributing to free-carrier production is of prac-
tical relevance only for small n, since the cross sections become exceed-
ingly small for larger n. For example, E6 3) in NaCI (E g = 8.leV) is about
10-80 cm6 s2, while E6 5) is 5 '10- 141 cm 10s 4 [2.36]. The corresponding ab-

27
sorption coefficients ad 3) and adS), calculated for ruby-laser light (hw =
1.78eV) and its second harmonic, respectively, and for an irradiance of
10 10 W/cm2 (which is close to the measured threshold for optical break-
down in NaCl), are 0.3 cm- l and 8.10-4 cm- 1 , respectively. The relevance
of multiphoton ionization in the present context is limited to short wave-
lengths (small n) or sub-nanosecond pulses, for which avalanche break-
down, to be treated below, is impossible. In all other cases of practical inter-
est the dominant mechanism of carrier generation in insulators is impact
ionization.
Impact ionization is the inverse process of Auger recombination:
Upon colliding with a lattice atom an energetic carrier is slowed down while
creating an additional low-energy carrier pair (an extra electron being
knocked off the atom, leaving behind a hole). In the presence of an electric
field the new carriers are accelerated again until the process repeats, and a
carrier avalanche develops. The mechanism was originally proposed to
explain electric breakdown in insulators by an external DC field. In the opti-
cally driven process the electrons are energized by photon absorption (the
hole contribution is generally negligible due to the low mobility of holes in
large-gap materials). The number of free electrons is, except for losses to
diffusion and recombination, doubled in each step. If some 108 electrons
per cm3 are initially present in a material (this many may be expected even
in reasonably pure crystals at room temperature and daylight illumination),
then it takes 10 10 = 234 additional electrons per cm3 , or about 34 genera-
tions, to bring the electron density to 10 18 cm- 3 which makes the crystal
essentially opaque. Once this happens, formation of a microplasma [2.37]
and explosive material damage is likely to occur given the high irradiance
required for the electron avalanche to develop in the first place. The
mechanical and thermal consequences of this phenomenon, known as "opti-
cal breakdown" will be considered within a broader context in Chap.5. Let
us discuss here more specifically the conditions under which it occurs.
As long as the degree of ionization is small (Neh < N), an energetic
electron may be expected to undergo an ionizing collision as soon as it has
reached a threshold ionization excess energy (measured from the bottom of
the conduction band) close to Eg • The rate of creation of new electrons is
then proportional to the number of existing ones times their average rate of
energy gain, a(Ee)/ at (time as well as ensemble average)

(2.43)

It is evident from this that the electron density increases exponentially with
time if electron losses are negligible. The essential question is thus at what
rate do free electrons gain net energy inside the irradiated material. The

28
rate of energy change experienced by an average electron as a function of
the electrical field of the beam can be expressed classically as

oE
(2.44)

Here oE is the average amount of energy transferred to the lattice in a colli-

sion. (Electron-electron collisions are negligible during most of the time in
which the avalanche develops). Since the electrons' kinetic energies can
reach several electron volts, the collision frequencies and absorption cross
sections differ from those of low-energy electrons. Collision times gener-
ally decrease with electron energy, and they become very short - of the
order of 10- 15 s - for energies of a few electron volts in any material. One
consequence of the very short collision times experienced by energetic elec-
trons is obvious from (2.44): If Te is of the order of 10- 15 s, then WTe < 1,
and the gain term should be independent of the light frequency, up to wave-
lengths in the red. This is in accordance with experimental evidence: The
breakdown threshold fields for large-gap substances are found to be practi-
cally the same from DC up to the ruby-laser frequency. For shorter wave-
lengths the gain term in (2.44) decreases like w-2 and the threshold should
increase accordingly. Eventually multiphoton ionization will thus have a
lower threshold and dominate. The crossover wavelength depends on the
band gap and the collision frequency, and is expected to be in the visible
region in the case of the alkali halides [2.38].
Equations (2.43,44), while giving the correct trends. turn out to be of
little use for calculating realistic breakdown thresholds for avalanche ioniza-
tion. The reason is that Te and oE depend on energy in different ways,
which means that electrons in certain energy ranges may gain net energy
while in others they lose energy. A more satisfactory approach was taken by
Holway [2.39] who used a Fokker-Planck equation to calculate the time his-
tory of the electron distribution function under the influence of the oscillat-
ing electrical field. Holway found that fluctuations in electron energy can
produce energetic electrons even if the average electron gains no energy at
all. For illustration, Fig.2.11 shows electron distribution functions at vari-
ous times in sapphire (E g = 6eV) during irradiation by an intense ruby-laser
pulse. It clearly demonstrates that avalanche breakdown is due to the
"lucky" rather than to the average electrons. In view of this, what has been
called the experimental "irreproducibility of the damage threshold" by Bass
and Barrett [2.40] is no surprise. Indeed, the concept of "lucky" electrons
implies that avalanche ionization should be of a statistical nature, at least
for short pulses. Bass and Barrett found that the probability of a pulse pro-
ducing breakdown varies with the field strength E inside the material like
exp(-b/E) , b being a material constant related to the band gap. This was

29
-
~
I II
C
~
0.2

U
<II
Qj
0.1

0.05

energy (eV)

Fig.2.11. Spreading of free electron distribution in sapphire due to irradiation by a

ruby laser pulse that creates an electric field of 5· 107 V/cm. The steady-state distribu-
tion c, reached after about 0.1 ps, is independent of the chosen initial distribution a; b is
an intermediate distribution. (2.39]

true under the condition that the radiation was tightly focused in order to
reduce the probability of absorbing defects within the focal volume, and for
beam powers low enough to exclude self-focusing [2.41].
It is clear that all predictions of breakdown thresholds are useless if the
electrical field inside the material is not connected to the known irradiance
in a simple manner. Bloembergen [2.42] showed that the presence of
cracks, pores or inclusions with dimensions exceeding about 10 nm in a die-
lectric leads to local enhancements of the electrical field (and therefore to a
reduction of apperant threshold irradiance), which increases with the refrac-
tive index of the material. Reduction factors of 2..;.-5 were predicted in low-
index materials such as alkali halides or optical glasses. Similarly, the
apparent breakdown threshold is lower for multimode than for single-mode
beams, due to transient irradiance maxima present in the former [2.43].
Finally, the threshold for optical breakdown increases substantially
towards very short pulse durations, due to the finite multiplication time of
the electron avalanche. Under conditions typical for optical breakdown in

30
insulators, the avalanche build-up time is of the order of 0.1 -;- 1 ns. The
breakdown threshold for shorter pulses is determined by multiphoton ab-
sorption which is an "instantaneous" process. Carrier diffusion out of the
focal region also leads to an increase in the avalanche threshold for rela-
tively long pulses of sharply focused radiation. Table A.2lists typical values
of the breakdown threshold for a choice of wavelengths, pulse durations and
materials [2.44].

2.2.5 Reflection by Hot Metals

Even metals are not immune to beam-induced changes in their optical pro-
perties: their reflectance is usually observed to decrease under irradiation.
The effect is thermal in nature and makes metals susceptible to thermal
runaway, particularly in the infrared where reflectances tend to be high.
The bulk reflectance of metals decreases, as a rule, with temperature
because the apparent electron-lattice collision time shortens. In addition,
hot metals are reactive and irreversible changes in reflectance due to chemi-
cal reactions at the surface (oxidation, etc.) tend to occur, except under
high vacuum conditions. Reliable experimental data on the reflectance of
hot metals are scarce, but at least for the infrared we can get some informa-
tion as follows. The total absorptance of a metal can be thought of as com-
posed of three contributions, due to free electrons (fe) , interband transi-
tions (ib) and surface (surf) effects

(l-R) ::::: (I-R)(fe) + (l-R)(ib) + (l-R)(surt) (2.45)

No general statements can be made concerning the temperature de-

pendence of the last two terms. They depend, respectively, on details of the
band structure, and on the state and the reactivity of the metal surface.
However, we can relate the temperature dependence of the free-electron
term to that of the DC conductivity 0"0' which is usually known. Taking the
free-electron density and mass to be independent of temperature we can use
(2.24,26) to write

1 - R(T) ] (fe) y'0"0 (T0)/0"0 (T) for w« liTe

[ { (2.46)
l-R(To)
- 0"0 (TO)/O"O(T) for w» liTe·

Above room temperature the conductivities of typical metals decrease

roughly linearly with temperature [2.45], and the same holds for liquid
metals at not too high temperatures [2.46]. Upon melting most metals show
an additional drop in conductivity. Table A.3 lists free-electron reflectances
at various temperatures and compares them to experimental values at A = 1

31
",m for extremely clean and for ordinary samples [2.45-47]. The relevance
of the phenomenon discussed here can be judged by comparing the (1-
R)(fe) and (1-R)(exp) values at 300 K. Obviously the effect is relevant only
for the highly reflective metals.
In addition to the rather slight decrease in metal reflectance predicted
by free-electron theory, more drastic reductions in apparent reflectance are
observed in the melting and evaporation regimes. While early explanations
involved effects like a metal-dielectric transition in the liquid surface layer
caused by thermal expansion [2.48], or the smearing-out of the density jump
at the surface by evaporation beyond the critical point [2.49], there is now
little doubt that reduced reflection under laser irradiation (which mayor
may not indicate enhanced absorption in the condensed material) arises for
one of two reasons: Mechanical deformation of the surface, possible at
moderate irradiance, or plasma effects, which require irradiances sufficient
for strong evaporation. Such effects are discussed in the next section.

2.3 Phase Transitions and Shape Effects

Powerful laser beams not only tend to affect the intrinsic optical properties
but also the shape of an irradiated material. This, in tum, influences beam-
solid coupling in various ways. Shape effects are almost always related to
melting or evaporation.

2.3.1 Surface Corrugation

Melting a surface by laser radiation typically leaves its trace in the form of
permanent ripples or corrugations. The patterns are often unrelated to the
beam profile and appear even if the incident beam is smooth. A large vari-
ety of patterns have been described in the literature (a few are shown in
Fig.2.12), not all of which appear to be fully understood as yet. Explaining
a particular pattern requires answers to at least two questions, which mayor
may not be related: (i) what is the source of the pattern, and (ii) by what
physical mechanism is it imprinted in the material?
In practice, by far the most frequent answer to question (i) is light
scattering at material imperfections or dust particles. The scattered radia-
tion, being coherent with the incident beam, interferes with the latter to
form a pattern of modulated irradiance which now acts on the material. The
basic features of many patterns can be understood from simple geometrical
optics. A familiar pattern is that produced by a point-like scattering center
(such as a dust particle) located in the ambient air, a distance z away from

32
Fig.2 . 12. Laser-produced surface structures in (a) Pd, (b) (lll)Si and (c) (lOO)Si ,
irridated by (a) 50 ns and (b,c) 200 Ils Nd-Iaser pulses just above the melting threshold:
Pattern (a) is due to light scattering as described by (2.47) with () = 0, z ==0 . Patterns
(b and c) are apparently unrelated to scattering

the sample surface. It consists of a system of confocal ellipses (circles for

normal incidence) described by

(2.47)

where BJ = cos¢sinO and B2 = 2z/(nAcosO). Here 0 is the beam incidence

angle, measured from the surface normal, ¢ is the polar angle in the sample
surface measured from the plane of incidence, and rn is the distance from
the vertical through the scatterer to the nth fringe. An example of such a
pattern is that of Fig.2.12a. In the case of linear rather than point-like
scatterers, such as scratches in the surface, fringes are parallel to the
scatterer and have a spacing of AI (1 ±sinO) [2.50].
Often, however, patterns are more complex than simple geometrical
optics would indicate. The efficiency of interference in a particlar direction
depends on the state of polarisation of the incident beam. The formation of
fringes with spacings of AI cosO running parallel to the polarization direction
of a p-polarized incident beam were observed on nominally smooth surfaces
[2.51]. If the scattered waves propagate inside the material rather than in
air, one would expect the fringe spacing to scale with Aln!, rather than with
A. Fringes with such spacing have, indeed, been observed in transparent die-
lectrics [2.52]. A generalized treatment of interference of scattered waves
in terms of nonradiative "radiation remnants" propagating along the surface
of a material was formulated by Sipe et al. [2 .53].
Surface deformations are the most common and most efficient scatter-
ers, but "latent" patterns, such as lateral variations in lattice temperature or
in free-carrier density [2 .54] seem able to result in some scattering even on

33
geometrically smooth surfaces. Stimulated scattering at surface polariton
waves has been observed in molten Ge [2.55] and quartz [2.56]. Whatever
the origin, scattered fields from different scattering features interact, and
diffraction at ripples gives rise to secondary ripples. The resulting patterns
contain Fourier components at several spatial frequencies. Thus, once a
fringe pattern has been physically imprinted on a surface, it can perpetuate
itself coherently in subsequent, fully or partially overlapping laser shots
[2.57]. This may explain the gradual build-up of periodic damage patterns
that is frequently observed during repeated irradiation by pulses too weak to
cause any conspicuous damage individually [2.58]. Once heavy damage has
been done, surface patterns tend to deflect or scatter a large part of the inci-
dent beam, so a reduction in specular reflection is detected [2.59].
If light scattering of any kind is the answer to question (i), then ques-
tion (ii) is how the modulated irradiance is physically transformed into a
persisting variation of surface geometry. The basic sequence is that the ma-
terial melts, undergoes deformation, and finally - after irradiation - resolidi-
fies, making the deformation permanent. The mechanism of deformation
depends on the absorbed fluence as well as on the material [2.60]. In Si or
Ge irradiated near the melting threshold it appears to be related to the phase
transition itself. It has been shown [2.61] that the strong decrease in absorp-
tance upon melting of Si (Table A.l) leads to an instability at irradiances
just above the melt threshold: The power absorbed by the solid is sufficient
for melting, but the power absorbed by the melt is insufficient to prevent
resolidifaction! Hence neither a homogeneous solid nor a homogeneous melt
is a stable configuration, and a pattern of molten patches must form spon-
taneously. The ratio of molten to solid surface is a function of the irradi-
ance, while the feature size is determined by the absorption length in the
solid [2.62]. The melt pattern actually formed would be random for a mathe-
matically homogeneous beam, but even the slightest modulation of irradi-
ance, by interference or other mechanisms, will produce regularities in the
pattern.
At irradiances well above the melt threshold uniform melting occurs,
and lateral variations of the melt temperature become the relevant force for
imprinting patterns in the material. The surface tension of liquids decreases
with temperature, and liquid tends to be pulled away from hotter towards
cooler regions (Marangoni effect). This results in meniscus-shaped defor-
mations in the case of stationary beams and in ripple formation in the case
of scanned beams [2.63]. For short pulses the modulation of surface temper-
ature results in a modulation of both the depth and the lifetime of the melt.
In layered samples the elemental distribution is found to be accordingly
modulated after solidification [2.64]. Also often related to surface rippling
are acoustic phenomena. Pulsed melting of Si causes an abrupt local
increase in density, which acts as a strong source for acoustic waves [2.31].
Surface-acoustic or capillary waves frozen-in after irradiation have been rel-

34
ated to surface ripples with wavelengths in the micrometer range and are
unrelated to light scattering [2 .65,66]. Yet another potential source of sur-
face ripples in the melt regime are surface oxides, which can stay solid on
top of a melt layer and tend to become wrinkled like the skin on milk [2.67].
At even higher irradiance the molten surface is shaped by evaporation.
Stimulated light scattering at surface corrugations driven by evaporation
recoil in liquid metals was demonstrated [2.68]. On polished surfaces strong
evaporation results in extensive damage , manifested optically by diffuse
rather than specular reflection [2.69]. Enhanced absorption in the material
may result if the depth of the surface modulation is sufficient for multiple
reflections to occur .

2.3 .2 Shaping Surfaces

The phenomena resulting in surface corrugation upon melting are usually
seen as nuisances in laser materials processing, but they can also be put to
good use. Examples include optical recording as well as various techniques
that could be referred to as "laser polishing" .
In optical recording, sub-micrometer sized holes - later to be read out
as information bits - are imprinted into sensitive films by tightly focussed
nanosecond diode-laser pulses. The irradiation induces a bell-shaped tem-
perature distribution and hence a radial variation of surface tension. The
molten film pulls away from the hot center to pile up at the cool rim, but the
resulting depression (say, of radius r) opposes the flow by a capillary pres-
sure 2a/r. If the irradiation persists, the two forces reach a balance and the
resulting depression assumes an equilibrium shape [2.70]. However, in nan-
osecond recording a steady state is normally not reached and the hole shapes
are influenced by viscous and inertial forcess [2.71]. Gases trapped inside
the film or adsorbed at the surface tend to complicate the process [2.72], but
are often used intentionally to assist in the hole-formation process. The
laser imprints shown in Fig . 2.13 illustrate the action of radial material flow,
bulging due to gas-assisted film detachment and hole opening.

Fig.2.13 . Scanning electron micrographs of

laser-recorded imprints in an amorphous
11137 Sb so Gel3 film produced with 20 JLS pulses
from a Ar+ laser [2.72]

35
Fig.2.14. Scanning electron micrographs of a CVD-produced TiC coating on graphite
before (top) and after laser remelting with a 10 ns Nd:YAG laser pulse [2 .77]

36
 Capillary forces are also a key ingredient in laser polishing of ragged
surfaces . Tuckerman and Schmitt [2.73] demonstrated in 1985 a simple and
effective method to planarize thin metal films in integrated-circuit intercon-
nect structures. It consists of simply melting the contact by a short laser
pulse, causing it to resolidify in a near-planar configuration enforced by its
high surface tension. The concept has since been extended to faceted ref-
ractory films such as titanium carbide [2.74] and sapphire [2.75] and to
films on soft or heat-sensitive substrates like plastics [2.76] or graphite
[2 .77]. The technique has a beneficial self-regulating feature as Fig .2.14
illustrates: the small-grained initial structure (TiC in the example) absorbs
far more radiation than the final, smooth surface. Extended areas can be
polished by subsequent overlapping pulses that will melt only the rough por-
tions without affecting those already polished.
A rather different me cham ism is at work in excimer-Iaser polishing of
synthetic diamond films (Fig. 2.15): diamond does not melt, hence the pol-
ishing effect cannot rely on surface tension. Rather, it is based on the trans-
formation of a thin surface layer of diamond into graphite which subse-
quently evaporates [2.78]. Evaporation of ragged surfaces tends to smooth
them because the local vapor pressure is lower in the gaps between asperi-

Fig.2.IS . Scanning electron micrograph of a CYD-deposited polycrystalline diamond

film with a polished channel produced by linearly scanned 193 nm excimer laser pulses

37
ties than at their tips [2.79]. Furthermore, tips tend to absorb more radiation
and conduct less heat, hence they become hotter and evaporate faster. Best
polishing is obtained at oblique beam incidence [2.80]. Facets with small
angles to the beam normal absorb and disappear preferentially. Again, the
process self-terminates when the absorption of the smooth surface falls
below threshold.

2.3.3 Hole Drilling

Hole drilling, to be discussed in more detail in Sect.5.2.3, requires irradi-

ances capable of substantial surface evaporation. The speed of material re-
moval increases with the local irradiance, and Gaussian beams tend to form
round holes with diameters roughly equal to the beam (or focal spot) diame-
ter near the surface . However, as the hole grows deeper upon continued
irradiation, it usually narrows below the surface, rather than widening as
one might expect from the beam's diffraction angle. The reason for this
phenomenon is "channeling" of the beam by the steep walls of the hole. As
an example, Fig .2.16 shows hole profiles in Perspex drilled with CO2 -laser
pulses of various durations, focused with an f/5 lens onto the surface . Chan-
neling keeps the beam power concentrated near the axis of the hole and
enables its tip to propagate, until either the irradiation stops or absorption
and scattering losses in the evolving vapor jet make the process unstable
(note the frayed ends of longer holes in Fig .2.16) . Irregular or even curved
holes tend to result in inhomogenuous materials or for unstable beams
[2.81]. Optical breakdown in the vapor is promoted by the presence of
molten or even solid ejected particles, particularly in the case of short, in-
tense pulses [2.82].
In metals, beam channeling by holes yields the additional effect of
reducing or even eliminating reflection losses . The hole acts as a beam trap
(also known as a keyhole) in which the light energy is efficiently absorbed
in multiple-reflection events . Effective use of "keyholing" is made in pene-

Fig.2.16. Holes in Perspex, drilled

by COrlaser pulses of durations 10
ms to 5 s (from left to right), illustrat-
ing self-\:hanneling of the laser beam

38
tration welding by high-power CW CO 2-laser beams, to be discussed in
Sect.5.2.3. Here the beam energy is deposited deep within the molten metal
by means of a cavity kept open by the evaporation pressure, enabling effec-
tive thermal penetration depths of 1 cm or more for a higher-power beam.
The necessary beam powers are, however, in the kilowatt range.

2.3.4 Evaporation and Plasma Effects

The vapors formed by intense irradiation can play important roles in beam-
solid coupling, particularly for infrared beams. The range of irradiances
where evaporation is achieved stretches from some 103 W/ cm2 up to the
highest irradiances realized to date (10 15 W/cm2 or more). It is clear that
many physically distinct regimes are found in this enormous range. The fol-
lowing is a brief summary of phenomena related to beam-solid coupling.
At relatively moderate irradiance (below :::=: 106 W / cm2) the vapor is
tenuous and essentially transparent, but with increasing irradiance it tends
to become supersaturated as it evolves from the surface. Condensing drop-
lets of submicrometer size then lead to absorption and scattering. Apart
from this, the vapor cloud is a medium of refractive index different from its
surroundings and distorts the incident wavefront.
Between roughly 107 and 10 10 W/cm2 - depending on wavelength - the
vapor becomes partially ionized and absorbs a substantial fraction of the
laser energy. On the other hand, blackbody radiation emitted by the vapor
plasma tends to be absorbed by the solid more efficiently than the laser radi-
ation, particularly for infrared lasers. If the plasma stays close to the sur-
face, it may actually enhance the fraction of beam energy absorbed in the
solid. At irradiances somewhat higher than those producing ionization of
the hot vapor, ionization may even occur in the cold ambient gas, due to
optical breakdown. The breakdown plasma typically propagates as a super-
sonic absorption wave against the incident beam and shields the material
completely. This effect seriously limits the energy deliverable by intense in-
frared laser beams to targets at atmospheric pressure.
At even higher irradiance (above 109 +lO lO W /cm2) the plasma, owing
to its high temperature, becomes transparent and light is again transmitted
to the dense surface. The ablation pressure drives a shock wave into the
material which may alter its optical properties. For example, compression
of semiconductors affects the band structure, in particular the band gap
[2.83]. Fourfold coordinated nonmetals tend to transform into more densly
packed metallic phases at high pressure. Strong shocks ionize every nonme-
tal. In metals a significant drop of reflectance under picosecond irradiation
was predicted on the basis of a decrease in Te caused by loss of degeneracy
in the hot electron gas [2.84]. Finally, at the highest irradiances any sharp
boundary between the condensed material and the plasma disappears. Light

39
Table 2.1. Summary of self-induced optical coupling phenomena. (i: insulator, s:
semiconductor, m: metal, t: increase, ,: decrease)

Effect Material Spectral range Results in

Self-focussing i, s liw < Eg I t

Thermal free-<:arrier generation s hw « Eg ex t
Optical free-<:arrier generation s liw <: Eg ex t
Avalanche breakdown flw < Eg ex t
Free-electron reflectance m infrared R'
Heating, melting m any R'
Melting Si. Ge any R t, ex t
Crystallization a-Si, a-Ge liw -Eg R '. ex t
Surface corrugation s, III any (R ')
Hole drilling any any beam trapping
Vapor plasma any any enhanced coupling
or shielding

is absorbed at that surface where the electron density makes the plasma fre-
quency equal to the laser frequency. Additional absorption and reduced re-
flectance arises in the plasma from turbulent collective motion of the elec-
trons.
The physical conditions that produce this diversity of coupling phe-
nomena will be discussed in some detail in Chap.5.
To conclude this chapter, Table 2.1 presents a summary of self-induced
beam-solid coupling phenomena and indicates their main effects on the opti-
cal coupling parameters. It is this set of parameters that - together with
(2.10) - defines the "secondary" source mentioned at the outset. Unfortu-
nately. as we have discussed in this chapter. the dependencies indicated are
not always describable accurately by simple equations. Their practical
consequences will be considered in connection with the specific processing
steps discussed in the following chapters.

40
3. Heating by Laser Light

Heat treatments serve a wide range of purposes in today's material technol-

ogy. Examples include softening or hardening of metals, "annealing" of
crystals, dopant diffusion in semiconductors, compound formation in mix-
tures or thin-film couples, oxide-layer growth, polymerization of plastics
and many others. The use of lasers as heat sources in place of furnaces is
being adopted or explored in a growing number of heat-treatment processes.
What advantages do lasers have to recommend them for the job?
The main advantage, obviously, is that heating can be done in a local-
ized mode, both in space and in time. By matching the wavelength and the
beam power to the optical and thermal material properties, the amount of
heating can be chosen accurately to suit the needs of a process. If surface
heating is required and/ or volume heating is to be avoided, a large absorp-
tion coefficient and a short interaction time can be selected. Sharply delim-
ited areas can be heated to high temperature while the remainder of the
workpiece stays virtually cold. Rapid cooling of the heated material can be
achieved by using short pulses or rapidly moving beams. Apart from locali-
zation, laser beams are chemically "pure" and free of inertia, can be moved
easily and passed through windows to reach remote or inaccesible parts of a
workpiece.
This chapter presents fundamental facts and current examples of laser
solid-state thermal processing. Beam powers and pulse energies are gener-
ally moderate in this regime. The only effect of the irradiation is to raise
the temperature of the material, which then reacts according to its own
laws. Before discussing specific processes, we devote a section to the calcu-
lation of temperature distributions in the solid (excluding latent heat
effects) for a variety of irradiation conditions and material responses.

3.1 Temperature Distributions

3.1.1 Thermalization and Heat Transport

The primary product of absorbed laser light is, strictly speaking, not heat
but particle excess energy - excitation energy of bound electrons, kinetic
energy of free electrons, perhaps excess phonons. The partition of the ab-

41
sorbed energy among the degrees of freedom of the material is not thermal
at first. The degradation of the ordered and localized primary excitation en-
ergy into uniform heat involves three steps. The first step is spatial and tem-
poral randomization of the motion of excited particles, proceeding with the
collision time (i.e., momentum relaxation time) of the particles in question.
This time is shorter than even the shortest laser pulses, perhaps shorter than
an optical cycle. The next step, energy equipartition, tends to involve a
large number of elementary collisions and intermediate states, notably in
nonmetals. Several energy transfer mechanisms may be involved, each with
its own characteristic time constant. For example, hot carriers in nonmetals
lose energy first by phonon emission in the conduction band (Fig. 2.8), and
subsequently - with a different time constant - upon recombination. Non-
thermal phonon populations may be created in the process, which decay
with yet another time constant. Energy equipartition is particularly slow in
dielectrics and organic polymers. Intense optical excitation of such materi-
als by UV lasers has been found to result in photodesorption of molecules
without a significant temperature rise [3.1].
To describe thermal effects one usually ignores the intricacies of ele-
mentary relaxation channels and characterizes equipartition by an overall
energy relaxation time TE , as was done in (2.40. Typical orders of magni-
tude for TE are 10- 13 s in metals and between 10- 12 and perhaps 10-6 s in
nonmetals, depending on the material and on the irradiance. In semiconduc-
tors energy can be transported by hot carriers before it is given to the lattice
(Sect. 2.2.3). This will be relevant whenever the carrier lifetime is not negli-
gible compared to the laser-pulse duration.
The last step is heat flow. Once the laser energy is converted to heat it
still tends to be highly localized on a macroscopic scale. The mathematical
theory of heat conduction is based on the assumption that the heat flux
across a plane in a solid is proportional to the local temperature gradient

(3.1)

where K is the thermal conductivity of the material. Accepting this for the
moment; we can express the energy balance of a slab of material bounded
by planes at z and z + Ll z in terms of its volumetric heat capacity cp IV
c
Llt[~(z) -~(z+Llz)] = LlT ~LlZ. (3.2)

Here Ll T is the change in temperature brought about by a net heat flux

across the boundaries. Letting Llz -+ 0, the bracketted term on the Left-
Hand Side (LHS) can be expressed as (a~1 az)LlZ. Replacing the finite
differences by differentials and using (3.1) then yields

42
 ~ fK aT] = S? aT (3.3)
azt az v at
which is the usual fonn of the heat-flow equation in one dimension. If heat
is produced in the material, the power density of the heat source is added to
the LHS. Before discussing solutions let us briefly examine the physical val-
idity of (3.3) for the case of laser heating.
At issue is the validity of making Az arbitrarily small, since this
amounts to neglecting the finite mean-free path £ of the heat carriers (free
electrons or phonons). Harrington [3.2] has demonstrated that in a metal
most of the heat flux across a given plane is carried by electrons that had
their last collision with the lattice within several mean-free lengths from the
plane in question. This means that the heat flux follows the local tempera-
ture gradient only if the latter is constant over a distance of several mean-
free lengths, for otherwise a significant fraction of the heat is carried by
particles that "remember" a different temperature gradient. The criterion
that Az should not be made smaller than, say, 10£ affects the validity of
replacing the flux difference in (3.2) by a differential. By developing the
flux difference term into a Taylor series around z one finds that (3.3) is
only valid provided

az «I a 2..1·
5£1 a 2..1
3 2 (3.4)
3 az 2

If not, higher-order spatial derivatives of the temperature profile must be

retained to calculate the temperature distribution correctly. In metals typical
absorption lengths are of the order of 10 nm, and the same holds for the
mean-free paths of electrons at room temperature. The requirement of a
thennal gradient constant over several mean-free paths may therefore be
violated for short times. Eq.(3.3) then overestimates the heat flux away
from the surface and underestimates the surface temperature. The deviation
is, however, significant only for a very shallow surface region (lOnm or so)
and for short times. "Linear" heat conduction theory is thus good enough
for most laser applications in materials processing. Possible exceptions are
laser-induced thennal surface phenomena like thermionic emission or de-
sorption. Evidence of "nonthennal" electron distributions in tungsten irradi-
ated by ruby-laser pulses as long as 60 ns has been reported [3.3]. "Nonli-
near" heat conduction effects are also significant in plasma heating by
ultrashort laser pulses (Sect. 5.4) [3.4].

43
3.1.2 Solution of the Heat-Flow Equation

Let us now look for solutions of the heat-flow equation with allowance for
heat production by absorption of laser light. It turns out that a number of
rather severe simplifications must be made to arrive at analytical solutions
(more accurate numerical calculations will be discussed in Sect. 4.1.2).
First, we assume the power density of heat production to equal that of
the absorbed laser light, as given by (2.lO). Further, we take all material
parameters to be constants! (a requirement to be relaxed somewhat later
on). We consider a homogeneous material in the form of a slab between the
planes z = 0 and z = L (this includes the semi-infinite solid for which L =
(0). The material is taken to be at zero initial temperature - a uniform initial
temperature is simply added to the calculated temperature. The slab is as-
sumed to be thermally insulated, i.e., no heat flow across the boundaries is
allowed. This means that the temperature distribution T(x, y, z, t) must
satisfy the condition

aT
- = 0 for z = 0 and z = L at all times. (3.5)
az
The laser beam is incident onto the plane z = 0 and taken to be of cylindri-
cal symmetry. The heat-flow equation can now be written as

(3.6)

where K = KV I cp is the heat diffusivity, and Ia is the unreflected part of the

incident irradiance.
A mathematical method well adapted to the laser-heating problem is the
Green's function technique [3.5]. The Green's function g gives the tempera-
ture distribution produced by an "instantaneous" heat source of unit energy
and has dimensions [K/1]. Once g is known, the transient temperature dis-
tribution for a continuous heat source of power

PaCt') = HIa(x',y',t')dx'dy'
is found from
t

T(x,y, z) = foPa (t')g(x, y, z,x', y', z', t-t')dt' (3.7)

! Suitably averaged specific heats can be estimated from the data of Table A.4.

44
where x,y,z,t and x',y',z',t' are the coordinates of the field and source
points, respectively. We shall treat two kinds of lateral irradiance distribu-
tions here, the uniform and the Gaussian distributions. The former de-
scribes one-dimensional heating by a large-area beam where lateral heat
flow can be neglected while the latter applies to a focused beam of funda-
mental-mode radiation. Further, we consider the limiting case of a vanish-
ing absorption length (referred to as a surface source), as well as a finite ab-
sorption length (penetrating source). Green's functions for these source
geometries are given in the Appendix B. In the following we present a
number of solutions for the slab and for the semi-infinte solid, for use in
later discussions . We shall only consider sources of constant power, being
switched on at t = O. For notational convenience, we introduce the diffu-
sion length 0 = 2(Kt) 112.
For a uniform surface source of irradiance Ia = P a / S we have

(3.8)
n=-oo

In particular, for the semi-infinite solid

o
T(O,t) =Ia Kv1l'" (3.9)

The general expression for the uniform penetrating source is

(3.10)

Closed solutions are found for the semi-infinite solid

T(z, t) ~
=K { O' i erfc(z/ 0) - (1/ a)e-ca + (l!2a)e(a0!2) 2

x[ e-aZ erfc(ao!2-z/o) + eaZ erfC(a0i2+z/0)]} (3.11)

and

T(O, t) = K V 11"
Ia{O 1[ 1 - e(a0/2) 2 erfc(a0/2)] }.
-;;- (3.12)

45
 o.6.------.------.----.-----.

0.4

-;;C 0.3
to-

0.2

0.1

--
%L----~OS~---~---~lli~~
z/6
Fig.3.t. Normalized temperature distributions for uniform heat sources as a function
of normalized depth. Numbers indicate the ratio of the absorption length a-I to the dif-
fusion length 0 = :zv;t

Temperature profiles for uniform sources with various values of Ola, calcu-
lated from (3.8) (with L = 00) and from (3.11), are shown in Fig. 3.1. Note
that going from zero absorption length to an absorption length of one tenth
of the diffusion length lowers the surface temperature markedly, but barely
affects the profile deeper within the solid. The temperature at depths
beyond one absorption length is thus well described by the surface source
provided that a » 11 Ol (in the case of metals this holds for times exceeding
about 0.1 ns, i.e., whenever the present linear theory holds). In the other
extreme; for a « 11 Ol, the temperature distribution essentially follows the
exponential absorption profile.
The general expression for the Gaussian sUrface source is

a
T(r,z,t)=
Pa
3/2
J exp[- r2]
2 2T12
d~ 2· (3.13)
2K1I" 0 ~ +w ~ +w

46
A closed solution is found for the slab at steady state (t = 00)

00
(2nL - Z)2] [ 2nL - Z]
T(O,z,oo) = 4Kwv7r exp[ w2 erfc w .(3.14)
n=-oo

Simple expressions for the semi-infinite solid are

T(O, 0, t) = KW7r 3/2 arctan(Oiw) , (3.15)

Pa
T(O,O,oo) = 2Kwv7r ' (3.16)

(3.17)

In (3.17), which was first derived by Lax [3.6], 10 denotes the modified Bes-
sel function of order zero. Note that the Gaussian source, as any kind of lat-
erally confined source, leads to a finite steady-state temperature distribu-
tion, in which heat flow just balances the input from the source (the indefin-
itely increasing temperature (3.8) is due to a source of infinite power).
Comparing (3.15) with (3.16) shows that 90% of the steady-state tempera-
ture is reached after a time t ~ lOw2/K. Finally, the general expression for
the Gaussian penetrating source is

aP a fo [-
T(r, z, t) = 4 K exp 2
r2
2
a 2 {32]
+ -4- T2 2
(3d{3
2 (3.18)
7r 0 {3 +W {3 +W

for which only solutions in terms of tabulated functions are known [3.6].

3.1.3 Cooling

To include cooling of the material after the laser pulse, the integration in
(3.7) is extended beyond the pulse duration. Consider the uniform surface
source with a rectangular temporal pulse shape of duration tp and an irradi-
ance la = Pa / S. The temperature distribution during and after the pulse is
described by

47
ell
5
'5 08
Q;
CL
~ 06
"0
.~ 04
o
E
oc 0.2

" 6
normalized time
8 10

Fig.3.2. Normalized temperature as a function of normalized time tltp at various

depths z/2(Kt p )-1/2 for a pulse of rectangular temporal shape

T(z,t) = P a fogusdt' = (2Ia/K) (Kt) 112 ierfc(z/2v1;ct)

- [t > tp]v'K(t -tp)ierfc[z/2v'K(t -tp)] (3.19)

where the symbol [t > tp] equals one if t > tp and zero otherwise. Normal-
ized temperature-vs-time curves at various depths computed from (3.19) are
shown in Fig.3.2. Cooling curves for other source geometries and pulse
envelopes are obtained analogously.

3.1.4 Moving Sources

There are two types of situation in which the medium is moving with re-
spect to the heat source: Either the laser beam is scanned across the sample
surface, or the light absorbing zone, with a stationary beam, moves into the
target as a result of material removal. Formally, the motion of the medium
with respect to the source is allowed for by replacing the coordinates
(x,y,z,t) in (3.7) by "delayed" coordinates. For example, for a motion at
the velocity u parallel to the x axis, the field coordinates are ([x-u(t-t')],
y, Z, t). With this substitution any desired temperature distribution caused by
a moving beam can be derived by the procedures used above. We only con-
sider surface sources of constant power and velocity, heating a semi-infinte
solid (L = (0).

48
 10
0.0

~
::J 0.8
ni
.,0.
~

-...
E 06

."

.
N

E
0.4

0
c: 0.2

0 L-I I
-1.0 -0.8 -0.6 -04 -02 0 02 0.4 06 08 1.0
normalized distance
Fig33. Normalized surface temperature versus normalized distance from the beam
axis, (x-ut)!w, for a continuous beam scanned at constant velocity u along the x axis.
Numbers indicate values of the normalized velocity UW!2K [3.8J

For the Gaussian source scanned along the x axis we get

Pa J 00
[(X-ut+U{)2/ 4 K)2 + y2 Z2] d{)
T(x, y, z, t') = 71"3/2 K 0 exp - {)2 + w2 - {)2 ()2 + w2

(3.20)

Equivalent expressions have been given in the literature [3.6,7]; an elliptical

beam was treated by Nissim et al. [3.8]. For illustration, Fig. 3.3 shows sur-
face temperature profiles along the x axis in a moving frame for a circular
beam scanning at various velocities, and normalized to the steady-state tem-
perature (3.16) at the center of a stationary beam. The scaling of the sur-
face temperature with Pa/w shown by (3.16) turns out to be preserved at a
nonzero scanning velocity and is a characteristic feature of heating by a lat-
erally scanned beam.
The steady-state distribution for a uniform surface source moving in
the z direction at velocity u is obtained form (3.7) as

K
T(z , t) = I a -e-
Ku
lU1K (3.21)

r
where = z-ut is a coordinate parallel to z and originating at the source. It
is assumed here that the physical surface is always at z = r, due to evapora-
tion of the material in the region between 0 and r, as we shall discuss in
detail in Chap. 5.

49
 4
1.5

3
/-
--
Al 20 3

-- -- -===----
-- -- -- - -

---- -------
./ Si
x: ,/ 1.0
E x:
u / ~
~
.....
~ 2
x: U
Q.

0.5

1000 1500 2000

temperature (K)
Fig.3.4. Thermal conductivity K and specific heat cl' as a function of temperature for
single-crystalline intrinsic Si and polycrystalline AI2 U3

3.1.5 Variable Parameters

The analytical approach to heat flow, elegant and intuitive as its results may
be, rests on the rather unrealistic assumption of temperature-independent
material properties. The expressions derived above are adequate to repro-
duce the essential features of a temperature distribution, but more than
qualitative agreement with experiments cannot usually be expected. Both K
and cp , and hence K, are often quite strong functions of temperature, partic-
ularly in nonmetals. Figures 3.4,5 show K and cp as a function of tempera-
ture for sapphire and Si, and for Al and Cu, respectively. We shall discuss
numerical means to obtain accurate solutions of the heat-flow equation with
arbitrarily variable parameters in Sect. 4.1, but let us, for the moment, con-
sider methods to improve on the realism of the analytical approach without
sacrificing all its virtues. A classical method consists in allowing the con-
ductivity to depend on temperature while keeping V and K constant. Note
that the constancy of K implies proportionality between cp and K. As first
proposed by G. Kirchhoff, we introduce a new variable

T
() = _1 J
Ko T
K(T')dT' (3.22)
o

50
 4
1.5

Al

--
3
"...
- - - -I- ----Z"
./
x: 1.0

-
E x:
~ 01
;: 2 I I
-.
I
Q.
x: v
I Cu _
I
I "...
----- - t - - - - - - - 0.5

/.(
,I
00/ 500 1000 1500 2000
temperature (K)
Fig.3.5. Thermal conductivity K and specific heat cp as a function of temperature for
Cu and AI

where Ko is the conductivity at some reference temperature To. The heat

flow equation in terms of () reads

a()
at = KV2() + J(x, y, z, t)K/Ko . (3.23)

If J does not itself depend on temperature, then the form of (3.6) is pre-
served, and all solutions given above can be taken over by replacing T by ()
and K by Ko. The transformation from () back to T requires inversion of
(3.22), which can be done on a desk calculator. An alternative procedure is
to fit K(T) to an analytic expression for which the integral (3.22) yields a
closed-form solution. For example, the conductivity of Si is well repre-
sented by the form K(T) = a/(T-b) [3.8] (for Si the use of (3.22) is not
particularly suitable given that K varies with temperature more strongly than
K in the temperature range of interest).
We note in passing that the relation (3.22) may also serve to transform
the temperature profiles given above into a normalized, material-independ-
ent form. For instance, with reference to (3.19), if we denote the surface
temperature at the end of the pulse by

2Ia _r--:-
()' = ()(O , t P) = - y Kt
Ko
/7r
P
(3.24)

51
then we can define the dimensionless coordinates

()* = ()I ()'., z* = zl2v::t·

p , t* = tltp (3.25)

in terms of which (3.19) reads

()* (z* ,t*) = -v'rt* i erfc(z/vt*) - [t > t p] v'1I"(t* -1) i erfc(z* /"\I't* -1) .
(3.26)

This equation holds for any material, provided the temperature dependence
of V and K can be neglected.
The next step towards increased computational sophistication would be
to solve the integral (3.7) numerically. The concept of the instantaneous
heat source then offers the possibility of including further phenomena, such
as self-induced coupling effects or more realistic pulse shapes, by adjusting
the pertinent parameters after each step in the integration. This procedure
yields a reasonably realistic description of laser-beam heating with rela-
tively moderate computing expenditure. Still beyond the reach of this meth-
od are, however, phase transitions involving latent heat, such as melting and
solidification.

3.1.6 Impact of Absorption Phenomena on Temperature

The coupling parameters R and a, as discussed in Chap. 2, tend to undergo

dynamical changes during irradiation (Table 2.1). The question we want to
address here is how these changes affect sample heating. We shall take a
qualitative approach based on analytical formulas.
As an example, consider uniform surface heating (stationary beam) of
a material the reflectivity of which decreases linearly with temperature. Set-
ting R(T) = Ro-bT, and taking all other parameters to be constant, the sur-
face temperature is found to be [3.9]

l-Ro 1
T(O, t) = --------- (3.27)
b [1 + erf(B)]exp(B2) - 1
where

B = Ib.v,ct IK .

This expression increases faster than exponentially with time, rather than
just as Vt, predicted by (3.9) for a constant reflectivity. Similar behavior is
found in the case of an absorption length which decreases with temperature,

52
 -1--------
=,-1=_===========
tOT
- -- ---
I

laser fluence
Fig.3.6. Schematic increase of material temperature with incident laser f1uence. The
shape of the curve (a, b, c) determines the variation in final temperature, oT, caused by
a variation in f1uence oT

as observed in semiconductors at photon energies near the band gap [3.10].

Such accelerated heating illustrates an important property of many of the
effects listed in Table 2.1, namely the tendency to amplify by positive feed-
back. The problem also exists, if somewhat less pronounced, for scanned
CW beams [3.11].
In most applications of laser material processing the sample tempera-
ture is a critical parameter which must be kept within narrow limits in order
to ensure the desired result. A behavior like that of (3.27) is unacceptable in
view of the lateral beam profiles as well as the reproducibility of real lasers.
The crucial question is therefore, what temperature variations are induced
by a given amount of variation in beam energy or irradiance?
The answer depends on the relation between the temperature and the
pulse fluence. Figure 3.6 exhibits three schematic temperature-vs-fluence
curves. For curve (b) the temperature rises linearly with the fluence, for
curve (a) the rate of temperature rise decreases with fluence, and for curve
(c) the reverse is true. The final temperature is the same in all cases, but
obviously a given variation in beam fluence (either due to the lateral beam
profile or to energy fluctuations from pulse to pulse) causes a large varia-
tion in temperature for case (c), whereas for case (a) it does not. From a
practical point of view, clearly case (a) is the desirable one. Which of the
cases in Fig.3.6 represents a given situation depends on absorption phe-
nomena as well as on heat flow.
Consider a pulse of fluence F, incident on the surface of a sample of
thickness L, creating heat over the thickness 11 Ol. During the pulse of dura-

53
tion tp (thought to be fixed), heat diffuses over a dimension op = 2(Kt p)112 ;
let us assume that op < L (otherwise the following discussion applies with
op replaced by L). Two limiting cases may be distinguished:
(i) If a-I «~P' then the surface temperature at the end of the pulse is,
from (3.9)

(3.28)

In this case the slope T(F) depends on the behaviors of both Rand cp . Since
cp increases with temperature, T(F) tends to saturate, corresponding to case
(a) above, unless the reflectivity decreases during the interaction, as is often
the case in metals.
(ii) If a-I »~P' then heat is created over a large depth. Expanding
(3.12) in terms of (XO and neglecting higher-order terms yields

(3.29)

In this case clearly the behavior of a as a function of temperature or flu-

ence is crucial. A small but increasing absorption coefficient results in a
very unfavorable type (c) heating curve, at least as long as the condition (X-l
» op holds. This situation occurs in nonmetals at photon energies near or
below the band gap. Controlled heating is quite impossible in such cases.
Measures by which type (c) heating can be avoided in practical situa-
tions follow from the above considerations. Examples are an appropriate
choice of laser wavelength (perhaps by frequency doubling), the use of ab-
sorbing layers or antireflective coatings on highly reflective metals, or the
provision of an adequate initial absorption coefficient, by preheating, car-
rier injection or auxiliary short-wavelength illumination, in nonmetals. In
addition, the beam profile can be smoothed by suitable devices to reduce the
variance of. For example, Cullis et al. [3.12] described a curved quartz rod
beam guide through which they passed their ruby-laser pulse to obtain an es-
sentially square beam profile. Similar results were obtained in the author's
lab with a simple aluminum beam guide with a conical bore, the diameter
and angle of which are matched to the beam size. The device, which can be
fitted with a ground glass plate at its entrance, also brings about a slight
focusing effect. The smoothing action of such devices relies on spatial and
temporal randomization of the pulse energy by means of many different
optical path lengths. The smoothed profiles still contain transient irradiance
maxima, but these tend to be short-lived and localized enough to be aver-
aged out by heat flow.

54
3.2 Heat- Treatment Processes
3.2.1 Annealing with Laser Beams

The typical purpose of annealing is the promotion of a thermally activated

reaction like crystallization or bulk diffusion. In such reactions the reaction
frequency per atom or molecule follows the Arrhenius law

II(T) = 110 e-Q/kT (3.30)

where T is the absolute temperature, and Q is the activation energy per par-
ticle (e.g., the energy needed to break one kind of bond in order to allow the
formation of another, or the strain energy of a lattice deformation that lets
a diffusing atom "squeeze" itself across a lattice plane). The quantity 110 can
be thought of as an "attempt frequency", of the approximate order of the
thermal vibration frequency kTIh, while the exponential factor accounts for
the probability that an attempt is successful. Quite often Q is only approxi-
mately constant and different values must be used in different temperature
regimes. Expression (3.30) is valid for a reaction taking place under homo-
geneous and isothermal conditions. If the reaction goes on for a time t, the
total amount of reaction is proportional to

n = lit = 1I0t e- Q/kT . (3.31)

Now, conditions in laser processing are usually far from being uniform
and isothermal, and the simple law (3.31) thus appears inadequate. Fortu-
nately, it can be adapted to nonisothermal processing by simple reasoning
as follows: In solid-state processing the variation of temperature with time
is typically rather slow compared to the time constants of elementary
processes. It is thus reasonable to replace (3.31) by
t

n(t) = 110 f_ooe- Q/kT (!') dt' (3.32)

where T(t) describes the local temperature history. It is now a straightfor-

ward matter to determine the amount of reaction due to an irradiation event
by inserting the appropriate temperature function and solving the integral.
Yet the procedure can be further simplified. Consider a reaction taking
place in a strongly absorbing thin film irradiated by a square pulse of dura-
tion tp. If the film thickness is small compared to the thermal diffusion
length then its temperature can be taken to be uniform and equal to the sur-
face temperature. Using (3.19) and allowing for a finite ambient tempera-
ture To we may write

55
 T(t) = To + Ll T max [v'Ut; - [t > tplv't/tp -1 ] (3.33)

where LlT max = (2Ia/K)(Ktpl1r)1I2 is the maximum temperature rise, reach-

ed at t = tp' The expression resulting if (3.33) is inserted into (3.32) can be
shown to be well approximated by the form of (3.31) [3.13], i.e.,

(3.34)

where T max == To + Ll T max' and where teff is an "effective" reaction time

that depends on the temperature. It is given by

(3.35)

The approximations made in deriving (3.34) are based on the strong temper-
ature dependence of (3.30), which allows one to neglect reactions occurring
far from the maximum temperature. Similar reasoning can be applied in the
case of a laterally scanned beam [3.14F if the dwell time 2w I u exceeds the
time lOw2 I K required for the surface temperature to reach the steady-state
value (3.17). The temperature at a point on the surface through which the
beam moves at t = 0 is given by (3.17) with the coordinate r replaced by ut.
Inserting this into (3.32) again yields a result that can be represented by
(3.34), but with Ll T max now given by (3.17), and teff by

(3.36)

This result can be generalized to the case in which an extended area is

treated by a series of partially overlapping parallel scans [3.15]. Integrating
the amount of reaction by subsequent scans again yields a result of the form
(3.34), with Ll T max given by (3.17) and teff by

7rW2 2kLl Tmax [ T max ]2 (3.37)

teff =t -S- Q ~
se max

t being the total scan time and Sse the scanned area. Note that (3.37) is equi-
valent to (3.35), i.e., in the approximation used here, scanning a given area
with a focused beam is equivalent to irradiating the same area with a single
pulse of appropriate duration.

2 Gold and Gibbons considered the special case of Si with a temperature-dependent conductiv-
ity of the form K(T) =a/(T-b)

56
 Although we have neglected temperature gradients inside the material,
it is clear that the irradiated surface will be somewhat hotter than the
regions beneath, and reactions will usually start here and proceed inwards.
The main quantity of interest is the thickness of material reacted in an irra-
diation event. To calculate this we need, in addition to the above formulas,
a relation between the reaction rate p and the velocity of the interface separ-
ating reacted and unreacted material. This relation is characteristic of the
type of reaction. In interface-limited reactions the interface moves at a
velocity given by ap, a being a molecular diameter (Sect.4 .1.4). The total
thickness of material transformed during an irradiation is then

(3.38)

Examples of such reactions are epitaxial growth (crystallization) of amor-

phous semiconductor layers as well as certain compound-forming reactions
in layered binary thin-film structures. Diffusion-limited reactions, in con-
trast, show a dz ex vlt rather than a dz ex t dependence and can be de-
scribed by a law of the form

(3.39)

where B is a material constant. Reactions of this type are limited by the rate
of transport of unreacted atoms to the interface, and it is the diffusivity that
is responsible for their Arrhenius behavior. Most compound forming solid-
state reactions, including oxide growth, are diffusion- rather than interface
limited.

3.2.2 Thermal Stress

Material which is heated, expands. The transient and nonuniform tempera-

ture distributions resulting from laser heating cause different zones of the
material to expand by different amounts and thermal stress to develop.
While the stress is small it is absorbed by the elasticity of the material. If
the stress exceeds a certain level, the material yields and defect formation,
cracking or plastic flow occur. Thermal stress effects are probably the most
serious difficulty in the way of a successful beam-annealing application.
A mathematical description of thermal stresses caused by nonuniform
heating is complex even if strongly simplifying assumptions are made, and
is beyond the scope of this volume. A systematic development of the theory
of thermal stress with solutions for various geometries can be found in the
book by Boley and Weiner [3.16]. Under some simplifying assumptions per-
sistent deformation or damage can be assumed to occur if at some point the
local stress exceeds the yield stress of the material. The yield stress is often

57
anisotropic and, in general, it depends on the local temperature as well as
on the strain rate.
In single-crystalline Si yielding occurs preferentially by slip along
(111) planes. The yield stress is related to the activation energy for the for-
mation of slip dislocations and decreases in an exponential fashion with
temperature. The amount of damage created by stress beyond the yield
point ranges from isolated dislocations reducing carrier lifetimes to macro-
scopic deformation or peeling of films, depending on the maximum temper-
ature and on the degree of inhomogeneity of the temperature distribution. It
appears that short pulses are less liable to induce damage due to the finite
time required for the nucleation of slip dislocations. Rozgonyi and Baum-
gart [3.17] concluded that pulses shorter than 50 ms do not induce slip in
Si, even if the melting temperature is reached. On the other hand, powerful
pulses can lead to thermal stress damage extending far beyond the directly
irradiated region where damage, e.g., by cratering, is most obvious [3.18].
To model stress development during CW laser-induced re-growth in
single-crystalline Si, Correra and Bentini [3.19] calculated stress distribu-
tions produced by a scanned line-shaped heat source. They found that local
stresses can change rapidly from tensile to compressive as the beam moves
across the surface of the sample, and they were able to relate the topograph-
ical distribution of stress-induced defects to the irradiation condition pro-
ducing them. If heating is rapid, large temperature gradients are also cre-
ated accross the thickness of a wafer. Cline [3.20] estimated that thermal
deformation in Si wafers should occur for heating rates exceeding 8000
K/s. He also investigated deposited thin films of Si and found that thermal
stress due to nonuniform heating causes plastic deformation of the film for
lateral temperature differences as small as 10 K. Even totally uniform heat-
ing by a substantial amount can, of course. produce thermal deformation in
thin film structures if the thermal expansion coefficients of the film and the
substrate are different.
In metals, the energies for formation and propagation of dislocations
are much smaller than in Si. Extensive defect formation and deformation is
already observed far below the melting point and even for pulses in the ns
regime [3.21]. In single-crystalline specimens defects are found to be
strongly orientation-dependent [3.22,23]. In polycrystalline materials sub-
jected to repetitive heating cycles, as, e.g., in high-power laser mirrors, in-
tergranular slip and fatigue cracks are known to lead to gross structural
damage, without the material ever being molten.

3.2.3 Crystallization of Amorphous Semiconductor Layers

In modern semiconductor technology amorphous semiconductor layers, typ-
ically some 100 nm thick, arise as a result of ion-implantantion of dopants
into single-crystalline wafers. Alternatively, amorphous layers can be pro-

58
 T lOe)
600 575550 525 500 400 375 350 325
Epitaxial regrowth role 01 amorphous Si and Ge

100 )..
C \
·e
"-
.~ ~>\
... \~
~

\~
.c.
'j 10
e
\~)
(100)
...
0'
a:

\(110)

I~----~~~~~~~----~~----~~
1.1 1.2 1.3 1.5 1.6 1.7
I03/T(K)
Fig.3.7. Regrowth velocity as a function of temperature in self-ion implanted amor-
phous Si and Ge for various crystallographic directions [3.24J

duced by vapor deposition on suitable substrates. Upon heating, the amor-

phous material becomes thermodynamically unstable and crystallizes spon-
taneously. Crystallization is unidirectional only if it is epitaxial, i.e., if the
crystallized layer grows from one single seed - usually the substrate - rather
than from randomly formed nuclei.
The Arrhenius law holds remarkably well in this case. Figure 3.7 shows
epitaxial growth velocities in various crystallographic directions on a loga-
rithmic scale versus the inverse temperature in ion-implanted Si and Ge
wafers, as measured during isothermal furnace annealing [3.24]. Note that
the prefactors Po are sensititve to the orientation (they also depend on the
density and nature of impurities present in the semiconductor), whereas the
activation energies are not. Olson et al. [3.25] used a scanned laser to deter-
mine regrowth velocities in amorphous Si also at higher temperatures. Inter-
face velocities were inferred from time-resolved reflectivity measurements,
taking advantage of the different refractive indices of crystalline and amor-
phous Si. The experiment showed no deviation of the activation energy
from its low-temperature value up to temperature close to the melting point
of crystalline Si.3

3 This contrasts to thermodynamic considerations according to which amorphous Si should

melt at a temperature several hundred degrees lower than crystalline Si (Chap. 4).

59
 10000
1400

1200

u 1000 1000
°
l-
. 800
.1.l!!.:
•
TO: 50 0 e
4 K 10·· SI!C :4.
N
<I
600
100
400

200

o
o QI
w
ABSORBED LASER POWER I BEAM RADIUS, P/w (W/l'm)

Fig.3.8. Peak temperature and regrown layer thickness in self-ion implanted Si as a

function of absorbed power per unit beam radius [3.15]

To illustrate typical parameters in scanned-beam annealing, Fig.3.8

exhibits peak temperature and regrown layer thickness as a function of ab-
sorbed laser power per unit beam radius in self-implanted amorphous Si,
calculated from (3.38) using avo = 3.2.10 4 mls and Q = 2.35 eV [3.15].
The main advantage of scanned-beam annealing of semiconductors is that it
allows higher effective growth temperatures and shorter processing times
than conventional furnace annealing. The higher temperature generally
leads to better crystal quality, while the shorter annealing time virtually
eliminates the unwanted redistribution of dopant atoms by solid-state diffu-
sion [3.26]. Wafers are often preheated to a few hundred degrees before
laser scanning in order to improve absorption and controllability of the
process. Carefully choosing the beam power and scan speed is essential to
avoid local overheating which may result in slip dislocations [3.27], or even
in local melting causing surface corrugation. The concentration of electri-
cally activated dopant atoms like As can exceed the equilibrium solid-solu-
bility limit, but the excess dopants are deactivated if the material is sub-
sequently furnace annealed [3 .28]. Work with compound semiconductors
like GaAs has been less successful, mainly because of the tendency of such
materials to decompose at high temperatures, and because of thermal stress
problems [3.29]. A general weakness of the technique of solid-state, as
opposed to liquid-state, epitaxy (discussed in Chap.4), is its extreme sensi-
tivity to impurities like 0, N or C, which lead to defect formation or may
prevent single-crystal growth altogether even at relatively small concentra-
tions.
Scanned-beam annealing techniques have also been used to produce
single-crystalline semiconductor sheets on amorphous substrates from
which no epitaxial growth is possible. The interest in this application der-
ives from current efforts towards three-dimensional integration of VLSI cir-

60
cuits. Geis et al. [3.30] used a process they termed "graphoepitaxy", in
which fine artificial relief gratings were etched into fused silica substrates
to serve as templates for oriented crystal growth. They obtained large ori-
ented grains from CVD deposited amorphous Si films on such substrates
upon scanned Ar laser annealing. Related attempts have used predefined
amorphous islands to control crystal alignment during laser scanning [3.31].
The island boundaries also act as dislocation sinks and allow relief of ther-
mal misfit stress between the substrate and the film.
The maximum achievable grain size in this as well as in other anneal-
ing techniques is limited by random nucleation. The rate of random nuclea-
tion has a stronger temperature dependence than the epitaxial growth rate
(Sects. 4.1.4,5), and lower annealing temperatures should favor larger
grains. Stress and interface effects seem to limit grain sizes achievable by
uniform (furnace) heating to some 10 pm or less, however [3.32]. Localized
heating by laser scanning has the crucial advantage of allowing suppression
of nucleation by means of suitably taylored temperature profiles. For exam-
ple, heating of the amorphous material can be concentrated to the immedi-
ate neighborhood of the desired growth interface. Shaping the thermal gra-
dient influences the growth direction of unwanted crystallites. Biegelsen et
al. [3.33] reported that by replacing the circular beam profile by a crescent-
shaped one, unwanted crystallites could be forced to grow out of the beam
path, enabling continuous growth of one single crystallite by the scanned
beam. Reported crystallites dimensions by laser scanning can exceed 100
Itm [3.34].
An interesting aspect of solid-state regrowth is the role played by the
latent heat liberated upon crystallization. The latent heat contributes signifi-
cantly to heating in materials like Si and Ge, and it may in fact make the
transformation self-promoting, an effect known as explosive crystalliza-
tion. It is most readily observed in films on poorly heat-conducting sub-
strates. Figure 3.9 exhibits two different morphologies of crystallized amor-
phous Ge on a silica substrate, scanned with a focused Kr laser beam at
slow (left frame) and fast (right frame) scanning speed. Similar morpholo-
gies can be observed in Si. The track of small-grained polycrystalline mate-
rial shown on the left part of the figure results from normal solid-state crys-
tal growth, whereas the large-grained crescent-like structures on the right
result from a periodic explosive growth mode. The periodicity of the pattern
for a given material is determined by the power and scan speed of the beam
as well as by the ambient temperature [3.35]. Rapid explosive growth can
proceed at velocities exceeding 1 m/ s [3.36] and seems to involve a narrow
molten film separating crystalline and amorphous regions [3.37]. Slower
self-promoted crystal growth not involving a liquid layer has also been ob-
served. Which one of the various possible growth modes actually materi-
alizes depends on the beam power, the scan speed and the substrate temper-
ature as well as on the homogeneity of the amorphous film [3.38].

61
 Fig.3. 9. Optical transmission micrographs
showing two typical morphologies in scan-
ned laser annealed amorphous Ge on a glass
substrate. Left: scanning speed below 2
cm/s, right: above 2 cm/s. Courtesy G.
Badertscher, Bern.

3.2.4 Compound Synthesis

Scanned laser beams have also proven useful in the formation of compounds
from binary layered structures. The best-studied application to date is the
growth of metal silicides for use as ohmic or Schottky-barrier contacts to Si
devices [3.39]. A metallic film, of the order of 100 nm thick, is deposited
on a Si wafer and heated by a pulsed or CW scanned beam. Upon heating
the elements react spontaneously, starting from the original metal-Si inter-
face, by forming a binary compound. The composition of the compound
formed, as well as the kinetics of its growth, are characteristic of the metal.
It is known from isothermal furnace annealing that in systems with several
equilibrium compounds one particular compound always forms first, but
only empirical rules are available to predict the phase forming first. A well-
known rule, due to Walser and Bene [3.40], states that the first silicide to
grow is the one with the highest congruent melting point next to the deepest
eutectic. The rule has been generalized [3.41] to cover other alloy systems.
The kinetics in a number of cases follows (3.38), i.e., the compound
layer thickness grows linearly with time for a constant temperature. Exam-
ples of systems behaving this way are Cr-Si or Mo-Si [3.42]. However, most
compound forming solid-state reactions, including oxide growth, are diffu-
sion rather than interface limited and grow according to (3.39). As an exam-
ple of laser-induced diffusion limited compound formation, Fig.3.10 dis-
plays the absorbed laser power as a function of the substrate temperature
To, required to grow 10 and 100 nm thick layers of the silicide Pd2 Si from
a Pd layer on top of a Si substrate, subjected to partially overlapping scans
(calculated from (3.39) with B = 3'10- 3 m2 /s, Q = 1.5eV).
A third type of kinetic law for solid-state compound-forming reactions
is characterized by an extremely strong, almost threshold-like temperature
dependence. These reactions are limited by nucleation rather than growth
rates. Transformation starts with random grains or islands of the compound
phase, which subsequently grow together. Such behavior is found in several

62
 0.2.---,..---,---.--..--,..---, Fig.3.1O. Combinations of absorb-
ed laser power and substrate temper-
ature required to form 10 and 100
nm thick films of Pd2 Si by partially
overlapping laser scans [3.15J

a
UJ
m
II:
o
C/l
m
c(
°0~_J--~~-J-~2~00~-J-~300
TEMPERATURE (oCI

rare-earth/Si systems, as well as in many compounds that are not first-

growing phases, like Pdsi. Producing uniform layers by furnace annealing
tends to be difficult in these cases. Scanning laser annealing has been dem-
onstrated to produce uniform compound layers also in the case of Pdsi, but
at a larger beam power than required for the first growing Pd2 Si phase
[3.43].
Whatever the particular kinetic law, compound layers formed by a
scanning-laser beam in solid-state reactions tend to be smooth and of uni-
form microstructure. Typically they resemble furnace-grown layers rather
closely and usually consist of one single compound phase, in contrast to
pulse-laser formed ones, as discussed in Chap. 4 (there are exceptions, like
Pt, which forms a mixture of silicides during scanned laser annealing).
Homogeneous silicide layers can even be obtained with refractory metals
like Mo, Wand Nb which require high reaction temperatures. Repeated
scans may be needed, but heating the whole substrate, as in furnace anneal-
ing, is avoided. The localized heating achieved by scanned-laser annealing
also causes much faster cooling of the processed material than conventional
heat treatments. The formation of metastable compounds, not available
through furnace annealing, has been reported in various metal-metal and
metal-semiconductor systems after laser scanning [3.15]. Rapid cooling of
the material after irradiation enables high-temperature phases to persist at
room temperature without transformation to the thermodynamically stable
low-temperature phase (this effect is also what is aimed at in hardening, see
Sect. 3.2.5).
In addition to silicide growth, laser-induced compound synthesis is cur-
rently being explored in a number of innovative approaches. It is not always
clear, and in fact sometimes irrelevant, whether these processes occur as

63
pure solid-state reactions or involve melting and solidification. While it is
difficult to give a well-seasoned assessment of these approaches at the pre-
sent time, it is certainly worth mentioning some of these new developments.
Compound-Semiconductor Synthesis. Pulsed or CW irradiation of
multilayer thin films has been used to obtain crystalline compounds in sev-
eral III-V, II-VI and IV-VI systems [3.44]. Here, the average composition of
the multilayer is chosen to coincide with that of the desired compound. The
compounds are reported to be generally single-phase and stoichiometric.
Although the actual reaction temperature has not been determined, it seems
that the effect can be understood in terms of melting, mixing of the ele-
ments and subsequent solidification, a sequence we shall discuss in a bro-
ader context in Chap.4.
Oxide and Nitride Formation. Irradiation of easily oxidizing materi-
als like Si, Ti or V in oxygen-containing ambients results in strongly en-
hanced rates of oxide growth. Experiments were done with both CW [3.45]
and pulsed [3.46] lasers at wavelengths ranging from infrared to ultraviolet.
To what extent non-thermal (photon-induced) and nonequilibrium processes
are responsible for the enhanced oxidation rates is still subject to debate
[3.47]. The oxides are often found to be disordered or amorphous and -
e.g., in the case of Ti - multiphase and nonstoichiometric.
Nitrides and oxinitrides of reactive metals can also be obtained from
irradiation with pulsed or rapidly scanned CW beams in air [3.48]. The dis-
solution of oxygen into the metal upon rapid heating apparently causes a
local and transient oxygen depletion near the surface which gives nitrogen a
chance to react. More controllable nitride formation of semiconductor (Si,
Ge) or metal (Fe, Ti, Zr, Hf, ... ) substrates results if the irradiation is done
in pure ammonia or nitrogen atmospheres [3.49].
Pyrolytic Film Decomposition. Irradiation of metalloorganic films
by a scanned laser beam results in local thermal decomposition, as a result
of which a metallic track is left in the beam path [3.50]. The heat of reac-
tion (mainly due to exothermic pyrolysis of the organic carrier) appears to
playa role similar to that of the heat of crystallization in explosive recrys-
tallization (Sect. 3.2.2) and can give rise to periodic growth patterns remin-
iscent of those shown in Fig. 3.9.
Material Deposition by Gas or Liquid Phase Reactions. Laser
beams are being used to achieve deposition of metallic or semiconducting
materials by pyrolitic or photolytic decomposition of gaseous carrier media
such as metallorganics, silane, etc. The beam is propagated through the car-
rier gas and a substrate is placed either parallel close to the beam or, more
often, directly into the beam path. Decomposition of gas molecules, by
direct absorption of photons, by impact on the laser-heated substrate or by
thermal electrons from the latter, results in formation of chemically active
species which then deposit on the substrate. The details of this process are

64
still far from being completely understood. In some experiments liquid car-
riers have been employed. Here the reaction cell typically resembles an elec-
trochemical cell in which one electrode is irradiated by the laser beam.
Laser-enhanced reaction rates (with or without an applied voltage) seem to
be largely due to electrode heating and/ or improved transport in the liquid
by thermal convection. For detailed discussions on laser-induced liquid or
gas-phase reactions consult the reviews [3.51].
Pulsed-Laser Deposition. In this approach (which strictly does not
even belong here) the laser beam is not directed at the material being syn-
thesized, but serves as a kind of evaporation gun: A target material is vapor-
ized and partially ionized by energetic ns to ms laser pulses in a vacuum
chamber. Some of the vapor condenses on a substrate placed near the beam-
impingement area. We shall discuss pulsed-laser deposition in some detail at
the end of Chap. 5.
Novel and exotic techniques like these ensure, if nothing else, that the
interest of materials scientists in laser beams is not going to fade away
soon.

3.2.5 Transformation Hardening

It has been known since old ages that metals can be hardened by first heat-
ing them to some transformation temperature and then quenching them rap-
idly, e.g., by a plunge into cold water. That laser beams have been found
suitable for the job is not surprising given their ability to provide rapid and
controllable heating and subsequent cooling of a metal surface.
The physical mechanisms responsible for the hardening effect of a
heating and cooling cycle can be quite complex and depend on the composi-
tion and microstructure of the metal to be hardened. Conventional harden-
ing of carbon steel requires heating to a temperature above the <Y-,}, transi-
tion temperature (between 750° and 900°C, depending on the carbon con-
tent) where the soft pearlite phase transforms into austenite and carbon par-
ticles dissolve. Upon subsequent rapid cooling the austenite transforms into
metastable martensite, whereas excess carbon forms various kinds of pre-
cipitates which contribute importantly to the mechanical properties of the
hardened material. The minimum cooling rate to avoid formation of un-
wanted soft modifications is of the order of 10 3 K/ s in carbon steel. Even
more complex processes are involved in hardening of alloyed steels. A role
is also played by thermal stresses and hydrostatic pressure created during
the thermal cycle. They do not only influence the phase equilibrium at high
temperature but may also lead to plastic deformation and work hardening.
The standard lasers for hardening of ferrous alloys are CW CO2 lasers
with beam powers of up to several kW. Coatings to reduce metal reflection
are often used. To achieve uniform hardening over extended areas the beam

65
is often oscillated rapidly in a direction perpendicular to the scan direction,
in order to produce the thermal equivalent of a line-shaped source [3.52].
Typical beam dwell times range from O.Ol-:-ls. Hardenable depths turn out
to be of the order of two or three mm under typical conditions. There is no
evidence that scanned-laser beam hardening involves any unusual metallurg-
ical processes. The main difference to conventional techniques is that heat-
ing and cooling tends to be faster and is concentrated near the surface.
There is, on the other hand, also less time available for the material to reach
equilibrium at high temperature. This is an undesirable feature, but it can
largely be compensated for by using relatively homogeneous material with
finely dispersed carbon to begin with. The rapid heating also limits the max-
imum material depth that can be hardened in a laser scan, because it favors
surface melting and evaporation. Beneficial is that bulk heating and associ-
ated thermal distortions are avoided.
Modelling the process of laser hardening is readily done on the basis of
the theory presented in Sect. 3.1. The process parameters - laser power,
spot size and scanning speed - are determined by the requirement that the
volume to be hardened is heated to a temperature above the transformation
temperature but below the melting point, and that the cooling rate is suffi-
ciently large. Evaluation of (3.20) in the case of a scanned Gaussian beam,
or of (3.19) for the case of irradiation of an extented area, gives all the
necessary information. As an example, Fig. 3.11 shows combinations of ab-
sorbed irradiance and interaction time (dwell time or pulse duration) for
hardening of cast iron with a melting point of 1200 0 e [3.53]. Note that the
heated zone becomes very narrow at high irradiance because the rapid sur-
face heating limits the allowable interaction time. Deeper heating can be

8000ru-n~-.----------------------------'

Irradiance
W/cm 2
Surface temp - 0c

0.50

°0~--~0.~2----0~.~4----0~.~6-----0~.8-----1~.0----~1.2
Zone heated above austenitizing temperature - mm
Fig.3.1l. Surface temperature and material thickness heated above austenitizing tem-
perature (== 800°C) as a function of absorbed irradiance and interaction time, in laser
hardening of cast iron (3.5 %carbon, melting point I200°C) [3.53] (Courtesy UTRC)

66
achieved with lower beam power and slower scanning, but at the cost of a
smaller cooling rate. Insufficient cooling rates lead to partially tempered
rather than hardened material. This again limits the practically achievable
case depth to about 3 mm in steel.

67
4. Melting and Solidification

Arguably the most interesting material modifications by laser irradiation are

those involving melting and solidification. Melts have high atomic mobili-
ties and usually unlimited solubilities, enabling structures that are simply
not available from pure solid-state reactions. Examples range form near-
perfect crystals, intricately structured alloys and composites to glassy
metals. The basic sequence that produces this variety of structures under
transient laser irradiation is always the same: Melting of a surface layer and
redistribution of atoms in the melt, followed by more or less rapid solidifi-
cation. We shall summarize this basic sequence as laser remelting. There
are two parameters which largely determine the structure resulting from a
laser remelting process: the composition of the melt and the velocity of the
solid-liquid interface.
The experimental work available to date on laser remelting can roughly
be divided into three categories: (i) regrowth of ion- implanted substrates,
(ii) surface alloying, and (iii) melt quenching.
The wide-spread interest in the first category of activities can be traced
back to the pioneering work of Russian scientists [4.1] who observed that
irradiating ion-implanted semiconductors with Q-switch laser pulses could
restore crystalline order far more perfectly than furnace (solid-state) anneal-
ing - apparently because crystal growth occurred from a melt, rather that
directly from the amorphous phase. However, the technique has turned out
to be more than just another implementation of the old art of growing a
crystal from its melt. Solidification velocities observed for short pulses are
measured in mis, ten orders of magnitude or so faster than in conventional
liquid-phase epitaxy, and work in this regime has produced - if not actual
applications in the semiconductor industry - at least important new insights
into the ways crystals grow.
Whereas ion-implantation is limited to a few at. % of admixture to a
substrate material, sUrface alloying enables mixing of two (or more) ele-
ments at any desired composition. That the alloys tend to be nonuniform in
their microstructure does not impair their basic appeal, which is to allow the
surface properties of a material to be made independent of its bulk proper-
ties. Much of the earlier work in alloying has been concerned with semicon-
ductor substrates, whereas recent emphasis has shifted more towards metal-
lic substrates. Here an additional motivation is the considerable saving in ex-
pensive or scarce alloying elements achieved if bulk alloying is replaced by
surface alloying.

68
 The application of lasers in melt quenching with the aim of obtaining
glassy alloys is the most recent of the three groups of laser remelting proce-
dures, although the potential was realized long ago [4.2]. Surface remelting
with sub-1-tS pulses can yield cooling rates of 1010 K/s or more, many orders
of magnitude faster than conventional mechanical methods like melt spin-
ning. As a result, many materials are becoming available in glassy form that
were previously not thought to be glass-formers.
It should not be concealed that laser processing in the melt regime, for
all its appeal, also has its shortcomings - at least from an applications point
of view. Beam powers are much higher than those typical in solid-state pro-
cessing, particularly in experiments with short pulses. Self-induced coupling
phenomena, as discussed in Chap.2, tend to make controlled heating more
difficult. As a result, the "power-window" to conduct melting experiments
without material damage can be quite narrow. In addition, the mechanical
forces that arise during rapid melting can lead to destruction of delicate
samples like semiconductor structures or thin films. or produce intolerable
surface warping. Accurate control of processing parameters is therefore
critical for applications in the regime under discussion - more so than in all
others considered in this volume.
We start this chapter with a section discussing models and fundamental
aspects of laser remelting, with emphasis on its two key variables, the melt
composition and the interface velocity. Subsequent sections then focus on
the three approaches to laser remelting defined above.

4.1 Fundamentals
4.1.1 Regimes of Laser Remelting

There are two fundamental aspects to be considered in the sequence we call

laser remelting. The first is chemistry - in order to understand a remelted
structure we need to know the chemical properties of the elements con-
cerned as well as their concentration. Some of this information is given -
for the case of thermodynamic equilibrium - by the phase diagram. This is
not sufficient, however, since equilibrium is seldom achieved in a process
as rapid as laser remelting. The second aspect, then. is kinetics. which deals
with the paths and velocities at which phases form and equilibrium is ap-
proached. Thermodynamics and kinetics are the topics of Sects.4.1.3-5 (we
shall limit the discussion to binary systems, because they are simpler and
because only little quantitative work has been done on more complex sys-
tems).
In order to make efficient use of thermodynamic and kinetic informa-
tion, we first need a model description of the laser remelting process, which

69
quantifies the flow of particles and heat in the irradiated zone. Such model-
ling takes as its input the initial elemental distribution of the sample as well
as the parameters of the heat source and yields, as its main results, the tran-
sient melt-concentration distribution and the velocity of the melt-solid inter-
face . Let us first consider particle flow . To keep the mathematics simple,
we shall consider one-dimensional geometries only .
The initial configuration of the sample is characterized by some ele-
mental distribution as a function of depth beneath the surface. Concentra-
tion profiles of three types of samples often used in laser remelting are dis-
played in Fig.4 .1:
(a) The Gaussian function, describing approximately the profile obtained
by ion implantation of an "impurity" atom B into an elemental sub-
strate A.
(b) The rectangular distribution, occurring in the case of a deposited layer
B on top of a substrate A.
(c) The square-wave profile, describing a multilayer of alternating depos-
ited films of B and A on top of some inert substrate.
Upon irrdiation, the material heats up and melts, and the elemental dis-
tribution starts to change by diffusion . Diffusion in the short premelting
period can generally be neglected since the diffusivities of most materials
increase by several orders of magnitude upon melting . Diffusivities in the
range 10-5 to 10-4 cm 2 / s are typical in molten metals, and changes in the
original elemental distribution within the irradiated structure occur even if
the melt lifetime is only a fraction of one f-tS. The effect of diffusion for a
time t on the elemental distribution is also shown in Fig.4 .1 (shaded).

c
.Q
e
c
QJ
u
C
ou

(b)
depth -
Fig.4.l. Concentration profiles before and after diffusion for a time t: (a) Gaussian
profile (ion-implanted single-crystal), (b) rectangular profile (deposited film on a·
substrate), (c) square-wave profile (multilayer)

70
 Mathematically, particle diffusion is similar to diffusion of heat,
except that there are now several - at least two - diffusing particle species,
which differ in structure and mass and which need not have the same "in-
trinsic" diffusivities. The simplest case, which we assume here, is that the
total volume occupied by the elements is not changed by diffusion. Interdif-
fusion of two elements can then be described in terms of a single "mutual"
diffusivity D, which may depend on the composition. The differential equa-
tion for isothermal diffusion (known as Fick's second law) can be written as

ax
at = V'(DV'X) . (4.1)

If D is taken as a constant, (4.1) can be solved analytically by methods simi-

lar to those used in Sect.3.1 [4.3]. In particular, instantaneous concentration
profiles after diffusion for a time t in a semi-infinite melt (z > 0) are
obtained from

J [ (z - z')2
(Xl

1 (z + Z')2 ]
X(z, t) = zv';Dt 0 Xo (z') exp [- 4Dt ] + exp[- 4Dt ] dz' (4.2)

where Xo(z) is the initial distribution. This equation (which was used to
construct Fig.4.I) neglects the fact that the melt lifetime is not everywhere
the same. For more accurate calculations one may allow t to vary within the
melt according to the finite velocity of the solid-melt interface (Sect.4.1.2)
[4.4]. Alternatively, (4.1), along with the appropriate initial and boundary
conditions, can be solved numerically by the finite-difference technique.
The procedure is completely analogous to the one outlined for the heat flow
equation in Appendix C.
Figure 4.1 demonstrates that interdiffusion for a time t produces mix-
ing of the elements A and B over a characteristic length 2Vnt. The result-
ing distribution (shaded in the figure) can be characterized as inhomogene-
ous and diluted in case (a), inhomogeneous and concentrated in case (b),
and concentrated but homogeneous in case (c). It should be no surprise that
solidification in the three cases produces quite different structures. The sit-
uation may, somewhat simplified, be summarized by saying that
case (a) leads to epitaxial growth of A,
case (b) to nonuniform and heterogeneous A-B alloy formation, and
case (c) to uniform and often metastable phase formation.
In Sects.4.2-4 we shall consider the three regimes corresponding to the
three basic configurations of Fig.4.1 in some detail. Let us now turn to heat
flow.

71
4.1.2 Heat Flow and Latent Heat

For a crystal to grow, it (or, more accurately, the crystal-melt interface)

must be kept at a temperature below the equilibrium melting point T sl .
Since crystal growth liberates latent heat, this extra heat must be continu-
ously carried away by heat conduction to preserve the undercooling. Simi-
larly, heat must constantly be furnished to the interface to preserve superhe-
ating during melting. This requirement determines the velocity of the inter-
face. In the following we neglect the elemental inhomogeneity of the mate-
rial (it could approximately be accounted for by using suitably averaged
values for the thermophysical constants).
A planar interface moving at the velocity u can be thought of as a
moving source or sink of heat, creating a heat flux equal to -uLlHs1/V,
where LlHsI is the latent heat and V the molar volume (taking u positive for
solidification and negative for melting). Motion of the interface is sustained
by the local heat flow provided that

(4.3)

where Ks and KI are the conductivities of the solid and the liquid, respec-
tively, and Zj is the coordinate of the interface. The velocity is thus propor-
tional to the difference between the temperature gradients on the liquid and
the solid "faces" of the interface. The temperature of the interface, in turn,
is not determined by heat flow but by kinetics, as we shall discuss in Sect.
4.1.4. A slowly moving interface (driven by a small thermal gradient) will
be at a temperature close to T sl' while for faster movement the temperature
will be appreciably higher or lower. During very rapid cooling the "heat
flow" velocity may even exceed the maximum possible kinetic velocity. In
this case the melt will not crystallize but freeze into a glass.
To discuss the heat-flow aspects of melting and solidification at ordi-
nary velocities it is convenient to ignore kinetics, and to assume that the in-
terface is at a fixed temperature, usually taken to be the equilibrium melting
point. The interface is then an isothermal surface, the motion of which is
obtained by solving the appropriate heat-flow problem.
The simplest approach is, obviously, to neglect the latent heat. The
heat-flow problem then reduces to that treated in Sect. 3.1. In one-dimen-
sional heat flow the isothermal surface T = Tj is a plane with coordinates
Zj' tj given by T(zj' tj) = Tj . To illustrate the concept, Fig.4.2 depicts the
positions of various isothermal planes for the normalized distribution
(3.26). The isothermal planes (and hence the interface) move into the solid
during, and shortly after, the laser pulse; thereafter they return to the sur-
face. The velocities are related to the temperature by u = (JT/Jt)/(JT/Jz)
and are seen to become infinite in the end. This is a consequence of our

72
 0.2

1.0

0.8

:E
0.
<II
"0
"0 0.6
<II
.!::!
a
E
o
c
0.4

0.2

8
normalized time
Fig.4.2. Positions z* of various isothermal surfaces as a function of time t*, calcu-
lated from (3.26). The parameter gives the value of the normalized temperature (J*

assumption of a thermally insulated surface where the temperature gradient

vanishes by definition.
If latent heat is included, solution of the heat-flow equation becomes
considerably more difficult. Analytical treatments have yielded solutions
only for a few special cases not of much interest for our problem [4.5], and
one has to resort to numerical methods. The main difference between the
present heat-flow problem and that without a phase change is that now the
state of the material is no longer uniquely defined by the temperature, since
near T sl it may be either solid or liquid. In the following we shall, for sim-
plicity, take the molar volume V to be constant and the interface tempera-
ture to equal T sl. A convenient quantity that uniquely defines the state of
the material is the difference in total enthalpy between the ambient temper-
ature and temperature T

T
.:l R(T) = J cp (T ')dT' + [T > T sd.:l Rs1 . (4.4)
To

73
The symbol [T > T sl] equals one above and zero below the chosen transi-
tion temperature. The use of a step function here implies that we assume
the kinetics to be "infinitely fast", i.e., the material instantly switches its
state at a fixed temperature, no matter what the interface velocity may be.
This assumption has implicitly been made in most calculations of interface
velocity published to date [4.6-8] and we adopt it for this discussion (for
calculations that allow for finite transformation kinetics, see [4.9]). Pro-
ceeding then, we can write the heat-flow equation as (retaining heat floW
along the z-axis only)

a.::lH
- =V [ - a [ aT) ]
-
at az Kaz- +J(z ' t) (4.5)

where J is the power density of the heat source. Eq.(4.5), together with
(4.4), may be regarded as a generalisation of (3.6), valid for arbitrary tem-
peratures and suitable for numerical evaluation. A simple numerical method
of solution is outlined in Appendix A.3.
Let us now consider a laser remelting cycle as modelled with the nu-
merical method. Temperature distributions and interface motion turn out to
be deeply influenced by the turnover of latent heat. This is illustrated by
Figs.4.3,4 for the case of a semi-infinite solid, irradiated by a rectangular
pulse. Figure 4.3 displays temperature profiles at various times during and
after irradiation, while Fig.4.4 gives the positions of the interface during
the melting and solidification cycle for various values of the latent heat.
Surface absorption and constant thermal parameters were chosen for this
calculation to allow a comparison with Fig.4.2 for the case without phase
change; the same normalized coordinates are used in the graphs. The calcu-
lation demonstrates that the latent heat has a rather strong impact on the in-
terface velocity during cooling. A latent heat of only 116 of the heat content
of the solid at the melting point already doubles the lifetime of the melt
(Table A.4 gives actual values of the ratio of latent heat to the heat content
of the solid at T s1 )' Furthermore there is, perhaps contrary to intuition, vir-
tually no temperature gradient in the melt during solidification. This means
that the second term on the RHS of (4.3) is negligible compared to the first
one, and the solidification velocity, for a given latent heat, is determined by
the conductivity of the solid phase alone. This fact has been used to influ-
ence the solidification velocity in Si, where the conductivity varies strongly
with temperature (Fig. 3.4), by varying the substrate temperature [4.10]
The solidification velocity obviously depends on the fluence and the
duration of the heat pulse, too. A more energetic or longer pulse results in a
larger melt depth, and since the stored heat is larger the thermal gradients
become shallower. In strongly absorbing media and for not too short pulses
(1/ IX « 0) the thermal gradient near the surface is roughly T(O)/o ex: t- 1I2
(Fig.3.1) and the velocity should scale with the pulse duration like tp -112.

74
 2.5,----.-----y-----r---,---,----.----.----.

2.0

..!ii
j::
1.5
~
.3
0
Q;
a.
E
~
.,u
u
1.0
::>
u
~

OS

°0L---L---L---~--~---L---L--~~
0.2 0.4 0.6
normalized depth
Fig.4.3. Numerically calculated temperature profiles at various times during melting
and solidification. Time and distance are expressed in terms of the same normalized
coordinates as used in Fig.4.2

This is borne out in Fig.4.S, which illustrates the impact of the substrate
temperature (i.e., Ks) as well as the pulse duration on the interface velocity
in crystalline Si irradiated with pulses in the nanosecond regime [4.7]. The
basic reliability of the numerically calculated interface velocities in Si has
been verified by direct measurements in which the transient conductance of
suitably shaped samples under laser irradiation was monitored [4.11]. These
measurements confirmed, in particular, that very large crystal-growth velo-
cities, of the order of several mis, are reached in materials irradiated by ns
laser pulses.
The same type of calculation can, of course, be applied to melting by
scanned CW beams, except that cylindrical or three-dimensional Cartesian
coordinates must be used. Figure 4.6 shows calculated melt depths in Si
scanned by an Ar laser beam for a range of beam powers and scanning
speeds [4.12]. Note that the sensitivity of the melt depth to the laser power
decreases towards larger scanning speeds, suggesting that melting should be
easier to control by scanning an intenser beam more rapidly.

75
 0.6

.J:.
C.
a.
1:) 0.4
1:)
a.
.!:!
C
E
....
o
c
0.2

normalized time
Fig.4.4. Position of the I-s interface as a function of time for various ratios of the
latent heat to the enthalpy at the melting point,AH~11 A H(T~I)' For curves a-f this ratio
is 1/3000, 1/6, 1/3, 1/2, 2/3 and 516, respectively. Curve e corresponds to Fig.4.3.
The pulse energy is kept constant

Having gained some insight into the macroscopic mass and heat flow
during laser remelting, let us now consider how this information relates to
the microscopic processes that determine the final structure.

4.1.3 Thermodynamics

Thermodynamic equilibrium between a crystal and its melt requires that

both phases have equal free energies. The Gibbs free energy of a phase at
absolute temperature T is defined as G = H-TS, Hand S being the enth-
alpy and entropy, respectively. The equilibrium melting point T sl is that
temperature at which the liquid and the solid have equal free energies at 1
atmosphere. Any deviation of temperature from Tsl causes one phase to
become unstable, Le., to have a higher free energy with respect to the
other. The difference in free energy between the phases

(4.6)

is the chemical driving force for promotion of the interface. Here, .dHsl
(the latent heat) and .dS sl are the enthalpy and entropy differences between
the solid and the liquid, and .d T = T sl - T is the undercooting. The substitu-
tion .dS sl = .dHsI/T sl made above strictly holds only a equilibrium, but is

76
i
o
20 -

dx
....
VI
II:
X
I- 10
CI
z
;j
....Z
Z
c(

~ 5
I-

~
>-
t:
u
o..J
....
>
....
u
~
2
....
II:
I-

.
~
VI
..J

5 10 50 100
PULSE DURATION (n5)

Fig.4.5. Calculated interface velocity in Si (crystalline substrate with a 100 nrn thick
amorphous surface layer), irradiated by ruby laser pulses, as a function of pulse dura-
tion for various substrate temperatures [4.7]

1.2

1.0 fO 25
emf· em/
E
:::t 0.8
:I:
t-
o.. 0.6
\.LI
0
0.4
~
\.LI
:::iE
0.2

0.0
4 5 6 7 8 9 10 11
lASER POWER. W
Fig.4.6. Calculated melt depth as a function of laser power in single-<:rystaIline Si
scanned at different velocities [4.12]

77
used as an approximation for T #- T sl' too. Note that with the sign conven-
tion of (4.6) we have .dGsl < 0 in the undercooled melt.
Equation (4.6) holds for elemental as well as for mixed phases, but the
free energy of a mixture at a given temperature is a function of composi-
tion. Considering a binary mixture A-B, where X denotes the molar frac-
tion of element B, we have

G(X) = (l_X)GA + XGB + GM = (l_X)gA + XgB . (4.7)

Here, GA and GB are the free energies of the pure elements A and B, GM is
the free energy of mixing (or formation), and gA and gB are the partial
molar free energies (or chemical potentials) of the components in the mix-
ture. A partial molar quantitity is the increase in the respective total quan-
tity of the mixture upon addition of an infinitesimal amount of pure sub-
stance (per mole added). The total and partial free energies are related by

G = gA + X(oG/oX) = gB -O-X)(oG/oX). (4.8)

This means that the chemical potentials at some composition Xl are found
by drawing the tangent to the G(X) curve and extending it to the respective
elemental axes. A schematic free energy curve of a particular phase in a
binary mixture is shown in Fig.4.7 (the detailed shape of the curve depends
on its free energy of mixing).
Equilibrium between two phases requires that, in addition to pressure
and temperature, the chemical potentials of all components are the same in
both phases. Figure 4.8 sketches free-energy curves for a liquid and two
solid phases at the temp~rature '"G below the melting points of both solids.
The solid compositions Xsl and Xs2 in equilibrium with each other are ob-
tained by drawing the common tangent to the respective curves. The driving
force for solidification at the temperature T I of a melt of composition X is

(4.9)

and can be read off the figure . .dGsl(l) and .dGsl (2) are the driving forces
for the initial precipitation of solids 1 and 2, respectively, while .dGsl is the
driving force for the formation of the equilibrium configuration which is a
mixture of both solids. Note that, as a rule, the driving force for any single
solid phase decreases with increasing compositional difference between it
and the melt.
The above concepts are concepts of equilibrium thermodynamics, and
the reader may perhaps wonder about their relevance in laser processing,
where large thermal gradients and irreversible flow of heat tend to create a
state of strong nonequilibrium: Nevertheless, a solid-liquid interface can
be a T sl - and hence at rest - even while heat is flowing through it, so the

78
 T = Tl

A B
composition
Fig.4.7. Schematic free energy-versus-composition curve. GA and GB are the free
energies of pure elements A and B at temperature T I' gA and gB are the chemical
potentials at composition Xl

phase there must be in some sort of equilibrium. This is called local equili-
brium. While the heat flow by itself is irreversible, the entropy it produces
is generally not available for the phase transition, a local phenomenon
taking place at a narrow interface [4.13]. Equilibrium in this local, re-
stricted sense must be understood whenever equilibrium concepts are ap-
plied to a transient process. Experience shows that most laser-induced mate-
rial phenomena, even for pulses in the ns range, can well be understood
from local-equilibrium thermodynamics. Exceptions are the phenomena of
impurity trapping and glass formation, to be treated in Sects.4.2.2 and 4.4.
Systematic application of the common tangent rule yields the equili-
brium phase diagram. Even when global (in contrast to local) equilibrium is
never reached in an alloying process, the phase diagram can provide impor-
tant clues about the phase to be expected. While phase diagrams come in
greatly varied shapes and forms, we may distinguish four main types which
exhibit typical behavior upon melting and solidification. The four types,
roughly in sequence of progressively more positive heat of mixing, can be
characterized as follows (examples are listed in Table A.S):

79
Fig.4.8. Schematic free energy curves_of a li<J!!id and two solids at a temperature
below the melting points of the latter. Xs1 and Xs2 are solid compositions in equili-
brium with each other. Also shown are the driving forces for initial precipitation of
solids 1 and 2 from a melt of concentration XI_ as well as for solidification of the latter
into the equilibrium configuration

(1) Partial or complete solid solution formers,

(2) compound formers,
(3) systems with immiscible solids (simple-eutectics), and
(4) systems with immicible melts.
The last group of systems is rather unfavorable for alloying by melting and
solidification. whereas the others differ mainly in the microstructure of the
alloys they form. Type-l systems produce homogeneous (single-phase) al-
loys at any composition, while Type-2 and Type-3 systems usually produce
heterogeneous grained alloys with grain sizes inversely related to the cool-
ing rate. The basic behavior of a system, as predicted by the phase diagram,
is often modified by the presence of metastable phases, particularly if cool-
ing rates are high and the resolidified structure never comes close to global
thermodynamic equilibrium.
So much for thermodynamics. The question about the actual velocity of
phase formation is not answered by thermodynamics, but requires a micro-
scopic model of the phase transition.

80
4.1.4 Interface Kinetics

The kinetic picture of equilibrium between a crystal and its melt is a situa-
tion in which the same number of atoms cross the interface per unit area
and unit time in either direction, i.e., the crystal "melts" and "freezes" at
the same rate. Both rates are assumed to be thermally activated, with the
activation energy for freezing being larger than that for melting by the
latent heat per particle. Net movement of the I-s interface results from an
imbalance of the two rates. The velocity of a plane interface can be ex-
pressed as

u = aAPj [ 1 - exp[ LlRil ]] (4.10)

Here a is the interatomic spacing in the growth direction, A is the accom-

modation probability for atoms in the growing phase, and Pj is a thermally
activated jump frequency. The accommodation probability for growth can
be interpreted as the fraction of sites at the crystal boundary, which are
next to a step or ledge, where attachment of a new atom is energetically fav-
orable. For most metals A is close to unity at all temperatures, while for
semiconductors, halogens and metalloids at small undercooling it is small
and anisotropic -larger for non-closely packed interface directions than for
closely packed ones, making growth along a (100) direction in a cubic crys-
tal faster than along (111) [4.14]. Figure 4.9 exhibits a schematic of u as a
function of the interface temperature (for A = 1). Note that for small und-
ercooling (LlGsl «RT) the velocity is proportional to the undercooling.
Since in laser remelting the interface velocity is dictated by heat flow, see
(4.3), the actual undercooling at the interface can be estimated by equating
the "kinetic" and the "heat-flow" velocities. For very large undercooling the
bracketed term in (4.10) tends towards unity and the velocity becomes small
and proportional to the thermally activated jump frequency. Heat flow then
plays no role in determining the interface velocity. This is the regime of
thermally activated growth discussed in Sect.3.2.
The activation energy determining Pj can be estimated from the activa-
tion energy for self-diffusion or for viscous flow in the melt. In pure metals
this is small, and atoms seem to remain relatively mobile down to tempera-
tures far below ambient. Turnbull and his associates have suggested that
growth in pure metals is not limited by mobility but only by the impingment
rate on the interface [4.15]. An upper limit for Pj is

(4.11)

where Us is the sound velocity in the melt; maximum growth velocities

would thus be expected to be in the 10 3 mls range.

81
 temperature
Fig.4.9. Schematic variation of the liquid-solid interface velocity with temperature

In solidification of molecular or covalently bound materials like Si02

(as well as in solid-state crystallization of amorphous Si or Ge) atomic rear-
rangements are sluggish even at elevated temperature. Crystallization here
requires coordinated motion of whole groups of atoms, referred to as acti-
vated complexes, rather than just jumps of single atoms. The corresponding
rearrangement rate has been expressed as [4.15]

p.
J
= nPf0 e-AG'/RT (4.12)

where ~G' is the free energy of formation of the activated complex, and n
is the number of individual rearrangements resulting from a single activa-
tion. Since pairs of dangling half-bonds, once formed, may migrate and
enable several atoms to regroup before they recombine, n may be larger
than unity. Intermediate cases are materials like Si and Ge which form cova-
lent crystals but metallic melts. The change in atomic coordination as well
as the 10 % change in density required for crystallization make l-s interface
motion slower than in pure metals, but it still remains much faster than in
nonmetallic melts. Growth velocities for a given undercooling depend on the
crystal orientation, and are largest for (100) and smallest for (111) growth,
due to differences in the accommodation probability [4.16]. Si(100) is ob-
served to grow epitaxially up to a velocity around 20 mis, but it turns amor-
phous at larger velocities. We will come back to this phenomenon in
Sect.4.2.
Growth of a compound crystal from a binary melt is more complicated,
particularly if the crystal and the melt have different compositions. Jackson
[4.17], considering growth of a solid solution, applied reaction-rate theory

82
to both species independently, with each component growing at a velocity
determined by its own chemical potential change and proportional to its
concentration in the melt. The total interface velocity is then just the sum
of the component velocities, weighted with their concentrations in the crys-
tal. However, growth of an ordered compound crystal requires strictly coor-
dinated, rather than independent, component growth processes. The usual
assumption is that the general form of (4.10) (with a suitable driving force)
is still valid. The necessary redistribution of melt atoms at the interface can
be allowed for in an approximate way by replacing the jump frequency by
the frequency for diffusive transport in the melt

p.
J
== D/a2 (4.13)

where D is the melt diffusivity at the interface. For metallic compounds Pj

would be of the order of lOll S-I, significantly smaller than for pure
metals. If the melt composition differs from the composition of the crystal,
growth is further limited by long-range mass transport, which is not ac-
counted for by (4.13). These limitations may explain why metallic glass for-
mation by melt quenching (Sect.4.4) is achieved for alloys but not for pure
metals.

4.1.5 Nucleation

The situation discussed so far is that of a crystal in contact with its melt, as
it occurs in a surface-melted semiconductor wafer. If only one phase is pre-
sent initially, then the other must, under normal circumstances, be nucle-
ated before any transition can occur. Nucleation involves the formation of a
new surface between the phases, which is energetically expensive and oc-
curs only if the driving force exceeds a certain critical value.
How, and when, does a crystal melt under laser irradiation? The ques-
tion is not as pointless as it may sound at first. Initiation of melting with
moderate heating rates is a nucleation and growth process. Melt nucleation
in solids internally heated by light absorption has been observed long ago
[4.18]. The amount of superheating required is largest for materials with
highly viscous melts, such as quartz. However, significant superheating may
be expected in any material at the prodigious heating rates available with
short laser pulses (10 15 K/s or more), although the question has attracted lit-
tle attention so far. There appears to be a limit for the superheating of
solids, given by the criterion that the resistance of the crystal to shear van-
ishes, causing it to melt continuously, without prior nucleation. Based on
Frenkel's [4.19] theory of melting, Martyniuk [4.20] predicted the limiting
relative superheating -~T/Tsl in electrically exploded metal wires to be be-
tween 15 and 30%. Even larger superheating has recently been reported in

83
GaAs irradiated by 150 fs pulses, where melting was concluded to occur be-
tween 500 and 1000 K above the equilibrium melting point [4.21].
Crystal nucleation, in contrast to melt nucleation, is of great practical
significance in melt solidification. It is undesirable in laser annealing be-
cause it results in poly- rather than single-crystalline material.
The theory of homogeneous crystal nucleation is based on the idea that
in an undercooled melt crystal-like clusters form by way of fluctuations.
Growth of clusters is driven by the free energy of crystallization ~ Gsl ' but
opposed by the crystal-melt interface energy. Assuming the latter to be
equal to the macroscopic interface tension u, and taking the cluster to be
spherical, the two opposing forces are found to balance each other once the
cluster has reached a critical radius. The free energy of forming such a crit-
ical cluster, or nucleus, is

(161r/3)U3 y2
(4.14)
~Gs12

Once a nucleus has reached the critical size it no longer has to rely on fluc-
tuations but can grow in the undercooled melt in the normal way. The
steady-state rate of formation of nuclei is usually written as

(4.15)

where N is the melt particle density and AN is a constant. The very strong
temperature dependence of Y below T sl leads to threshold-like behavior.
The "threshold" temperature (referred to as the homogeneous nucleation
temperature TN) was estimated to be about 0.8T sl in pure metals, with the
corresponding critical nucleus containing about 200 atoms [4.18].
The nucleation rate (4.15) holds under isothermal conditions. In melts
cooled rapidly from a temperature near T sl to a temperature below TN
measurable nucleation is found to occur only after a finite delay or time lag.
This may be interpreted as the time required to establish an equilibrium pop-
ulation of clusters corresponding to the new temperature [4.22, 23]. The
time-lag can be expressed as

RTuY
(4.16)

and should be minimum roughly at the temperature where the steady-state

nucleation rate is at its maximum. It obviously cannot be shorter than the
number of atoms in a critical nucleus divided by the jump frequency. Peer-
cy and coworkers [4.24], investigating pulsed laser annealing in Sb-imp-
lanted AI, estimated the time required for AISb nucleation from molten Al

84
to be between 5 and 25 ns. If this result is typical for alloys, it may help to
explain the surprising ability of ns pulses for binary metallic glass forma-
tion (Sect.4.4).
In many practical situations nucleation is found to occur at much
smaller undercooling than predicted by (4.15), due to heterogeneous nuclea-
tion. In the presence of a substance which is wetted by the nucleating mate-
rial, the cluster surface exposed to the liquid for a given cluster volume is
decreased, and so is the critical nucleus size. Suitable substances include
impurity particles in the melt as well as material boundaries, e.g., a film-
substrate interface. Nucleation of melting usually occurs heterogeneously at
the surface of the heated solid, rather than in the interior. On general
grounds, impurity-mediated heterogeneous nucleation may be expected to
become less important at large driving forces where homogeneous nuclei,
once forming, quickly outnumber the limited supply of heterogeneous nuc-
leation centers.

4.2 Regrowth of Ion-Implanted Substrates

Ion implantation produces shallow buried regions contammg implanted

atoms some tens to hundreds of nm beneath the surface. The distribution
profile resembles a Gaussian with a depth (the mean ion range) and a width
(the RMS deviation) determined by the ion kinetic energy and the host ma-
terial. Maximum concentrations are typically in the at. % range, limited by
surface sputtering. Somewhat similar elemental distributions have also been
produced by deposition of ultrathin buried layers.

4.2.1 Semiconductor Substrates

Doping semiconductor wafers by ion implantation, while offering better

control of the dose and the concentration profiles than the conventional dif-
fusion technique, also causes extensive lattice damage. In Si room-tempera-
ture implants of ions heavier than Zn at doses exceeding about 10 14 cm-2
result in amorphization of the implanted layer, while lighter ions, lower
doses or higher implant temperatures create at least extended defects. Sub-
sequent annealing is required to remove the damage and to bring the dopant
atoms to electrically active lattice locations, but furnace annealing often
requires the wafers to be subjected to temperatures as high as 1100°C or is
even found unable to fully restore the lattice. As alternative techniques,
rapid solid-state annealing by scanned beams (Sect. 3.2.3) as well as surface
remelting by sub-ItS laser or electron-beam pulses (often termed pulsed an-
nealing) have therefore attracted enormous interest.

85
 6(xx)
i
4(XX)
i
oi
~

1/1

C
::J
o
"
u
9.
~

o...J
W 3 \ implanlE'd .=
>
(!) D4f'lNg .:"
z
ii 2 iafler laser pulse uplo 18Vcm2
w
l-
I-
0«
~ aligned speclra (100)
:.:::
':;;i
a:l

r2.21/m2
AAtI\A;~6 "6. ..III.
......¥AA ___.~.... '.
O~~l____~1_ _ _ _~_ _~~_ _~~~.
08 10 12
ENERGY(MeV)
Fig.4.10. 2.0 MeV He+ backscattering spectra of an ion-implanted Si(lOO) wafer
with a 450 nm thick amorphous surface layer, before and after irradiation with 20 ns
Ruby laser pulses of various fluences [4.25]

Lattice regrowth in pulsed remelting of implantated substrates has a

sharp threshold, which depends on the thickness of the amorphous layer.
Figure 4.10 exhibits MeV He+ channeling spectra from a Si(100) wafer
with a 450 nm thick amorphous layer, irradiated with 50 ns ruby laser
pulses [4.25]. The lattice is restored to the quality of the virgin wafer at a
fluence of 2.2 J /cm 2 , while no improvement is seen for fluences up to 1.8
J/cm2 . The structure for fluences between about 0.5 a,nd 1.8 J/cm2 turns
out to be polycrystalline, while for even lower fluences it is still amorphous.
Equivalent observations are made also at other wavelengths, e.g., in the UV
[4.26]. They are readily understood if one assumes the regrowth mechanism
to be melting and subsequent liquid-phase epitaxy. For the amorphous layer
to crystallize within the short interaction time it must be molten (the melt
threshold for ruby-laser pulses is about 0.5 J / cm2 in amorphous Si), but for
it to regrow epitaxially the melt front must have penetrated into the single-
crystalline matrix beyond the damaged layer. The regrowth threshold also
depends on the structure of the damaged layer. Only slightly damaged or
polycrystalline layers have higher thresholds than amorphous ones, because
of their smaller absorption coefficient (Table A.1) and their larger heat cap-
acity.
Pulsed laser remelting under the proper conditions produces regrown
layers free of extended defects in both Si (100) and (111), implanted with
the common Group-III and-V dopants. However, fluences sufficient for

86
complete regrowth in (100) cut Si wafers yield heavily twinned material in
Si (111), where only a somewhat larger fluence, other parameters being
equal, gives a comparable crystal quality [4.27]. The reason for this puz-
zling discrepancy turns out to be directly related to the interface velocity
[4.28,29]. Interface velocities in Si for pulse durations around 50 ns are of
the order of several m/s; and they decrease somewhat with increasing flu-
ence (Sect.4.1.2). It must be concluded that these velocities are close to the
limit at which defect-free Si crystals can grow, and the limit is apparently
lower for growth along the (111) than along the (100) direction. This trend
can be ascribed to the anisotropy of the accommodation coefficient A, as
discussed in Sect.4.1.4. The limiting velocity for defect-free growth in Si
(111) has been determined to be near 6 mls [4.28]. In (100) material defect-
free growth is observed up to a velocity of 18 mis, while for an even larger
interface velocity the nature of the transition alters dramatically - the Si
turns amorphous. The same happens in Si(1ll) at only 14 m/s.
The production of amorphous Si by rapid cooling of the melt was first
demonstrated in 1979 [4.30,31]. The short pulses used to melt the surface
of single-crystal wafers to a depth of only a few tens of nm left behind
amorphous patches distinguished from their crystalline surroundings by a
somewhat higher reflectance. The reason why Si turns amorphous is obvi-
ously that cooling is too rapid for the crystal to grow. However, the situa-
tion is different from that of glass formation by melt quenching, to be dis-
cussed in Sect.4.4. Amorphous Si is a four-fold coordinated semiconductor
while molten Si is a metal with a coordination number near 11. Whereas
glass formation can be thought of as a mere freezing of the melt, the trans-
formation from liquid to amorphous Si seems to require a first-order phase
transition [4.15,32].
Figure 4.11 represents a free-energy diagram of Si (showing free-ener-
gy differences with respect to the crystalline phase), where amorphous Si
(a) is treated as an independent phase with an equilibiritim melting point Tal
[4.33]. The latent heat of the I-a transition was reported to be 37 kII mol, as
compared to 50.3 kJ/mol for the l-c transition. Consider what happens if
the melt (1) is cooled below the equilibrium melting point of the crystal
(T Ie). Ll GIc , and hence the crystal growth velocity, increase, but below Tal
nucleation and growth of amorphous Si competes with crystallization. That
the amorphous rather than the crystalline phase (which has a still larger
driving force) grows from the melt only indicates that the former is kineti-
cally favored [4.34]. The apparent maximum crystal-growth velocity is thus
no intrinsic limit but simply the value of the growth velocity at that underco-
oling at which the amorphous phase overtakes the crystalline one [4.35].
Melting of the amorphous phase, as observed in pulsed-laser remelting, can
be understood from similar kinetic arguments: Once the amorphous phase is
heated to a temperature between Tal and Tc1' melting occurs simply because
it is faster than crystallization (melt-in velocities in Si of several 100 ml s

87
 8

0; 6
(5

..,....E 4
~
CI
<l
2

c
0

-2

-4
500

FigA.ll. Free energy of amorphous (a) and liquid (f) Si relative to that of crystalline
(c) Si as a function of temperature. The dashed area indicates the uncertainty in the
thermodynamic data for the amorphous phase [4.33]

during irradiation with ps pulses was reported [4.36]). On the other hand,
kinetic hindrance of melt nucleation may explain the observation of solid-
state crystallization in amorphous Si up to temperatures close to T cl (Sect.
3.2.3).
Apart from silicon, extensive work has also been done with other semi-
conductor materials, in particular Ge and GaAs. Ge is found to behave very
similarly to Si in most respects. The threshold for surface melting in amor-
phous Ge for ruby-laser pulses is less than 0.2 J / cm2 , and complete re-
growth is observed in (l00) material at 1.0 J / cm2 [4.37]. The lower thres-
holds are explained by a somewhat larger absorption coefficient than Si, as
well as by a lower melting point. The melting point of amorphous Ge was
estimated to be 969 K [4.32], as compared to 1210 K for the crystal.
Amorphization of Ge by laser irradiation has so far not been demonstrated.
The case of compound semiconductors like GaAs is somewhat more
complicated, mainly due to their tendency to lose the more volatile compo-
nent by preferential evaporation at elevated temperature. Furnace annealing
is usually done by using surface encapsulation, usually with layers of Si02
or Si 3 N4 , to prevent evaporation losses, but such layers complicate the cou-
pling of optical beams and have often been omitted in laser remelting stu-
dies. GaAs forms a metallic melt [4.38]. Single-crystal regrowth is achieved
with ruby-laser pulses of about 1 J / cm2 [4.39], but the regrown material
tends to contain dislocations and other defects. These disappear at higher

88
fluences, but at the cost of increasing As loss, resulting in excess Ga ag-
glomerates on the surface which must be removed by etching [4.40]. The
concentration of electrically active dopants was found to be higher than in
furnace-annealed material, but the carrier mobilities tend to be lower [4.41].
Attempts to activate low-dose implants (which do not amorphize the
material) have met with only limited success. Similarly mixed results have
been obtained with other III-V or II-IV compound semiconductors like GaP,
InP, or Cdse. For a review, see [4.42].
We have so far mainly dealt with lattice regrowth in the ion-implanted
amorphous semiconductors. Let us now consider what happens to the im-
planted species during the rapid crystallization process.

4.2.2 Segregation and Trapping

It is well known that impurities in an undercooled melt tend to be segregated

from the advancing crystal boundary - this is the basis of zone refining.
The amount of segregation depends on the interfacial distribution coeffi-
cient which gives the ratio of the impurity concentration at the crystal
boundary to that in the melt adjacent to it

(4.17)

Here both Xs and XI are assumed to be small (typically a few at. % or less)
whence k is practically independent of composition. At near-equilibrium,
i.e, for slow growth, the concentrations Xs and XI are those given by the
solidus and liquidus lines of the phase d~gram; the!! ratio is referred to as
the equilibrium distribution coefficent k. Usually k is smaller than unity,
indicating that the impurity is less soluble in the crystal than in the melt.
Consider now what happens to an implanted impurity distribution in a
semiconductor irradiated by a short pulse. Figure 4.12 exhibits a sequence
of numerically obtained distribution profiles [4.7]. The distribution of imp-
lantation is roughly Gaussian in shape, with the position and width of the
peak determined by the mass and energy of the implanted ions. As the semi-
conductor is molten (t = 0), the distribution begins to spread rapidly by dif-
fusion. However, as the interface (dashed vertical line) begins to return to
the surface, only part of the impurity present there (10%, corresponding to
k = 0.1 in the example) is absorbed by the crystal, while the remainder ac-
cumulates in the melt ahead of the interface. The melt gets more and more
enriched in the impurity, and when solidification is completed (t = SOns) a
considerable amount of it ends up close to the surface.
What this description ignores is the fact that in the end the melt is
likely to be a concentrated mixture of two species. The assumption XI « 1
will break down, and the distribution coefficient will not remain independ-

89
z
o
>-
«
a:
>-
z
w
u
z
o
u

1000 0 1000
THICKNESS(A)

Fig.4.12. Numerically calculated impurity profiles in a sample with a Gaussian initial

impurity distribution. The frames show instananeous profiles on both sides of the I-s
interface (dashed line) during solidification. Parameters used: u = 2 mIs, D = 10-4
cm2 /s, k = 0.1 [4.7J

ent of the composition. Moreover. except for a few binary systems which
form continuous solid solutions. there will always be a maximum value for
Xs beyond which compound phases will form, rather than a solid solution.
Also, interface instability and cellular growth are often observed near the
surface. In the next section we shall deal with this and other solidification
phenomena in concentrated mixtures. Let us now focus on that part of the
final distribution in Fig.4.12 where Xs is still small.
Dopant redistribution during laser remelting and the resulting concen-
tration levels of dopants in Si have extensively been investigated, and con-
ventional models have not always been found adequate to explain the re-
sults. As an example, Fig.4.13 plots measured concentration profiles in In-
implanted Si before (open circles) and after (closed circles) regrowth by a
15 ns ruby-laser pulse. Also shown are fitted profiles in which the distribu-
tion coefficient was treated as a fitting parameter. Some surface accumula-
tion of In is evident, but, more importantly, the best-fit-value of k turns out
to be 0.15, three orders of magnitude above the equilibrium value! The
concentration of substitutional In atoms is 5· 10 19 cm-3 , exceeding the
equilibrium solid solubility limit by a factor of 60. Similar results have been
obtained for all Group-III and -V dopants, each element showing charac-
teristic enhancement factors for k and the maximum solubility [4.43].
Growth direction ,also plays a role - k values for growth along (111) tend to
be systematically larger than for growth along (100) [4.44]. Increased solu-
bility under ns laser remelting was also observed for Pt [4.8], but not for

90
 Fig.4.13. Rutherford backscattering
, o implanted spectra of In distribution in ion-im-
.Iaser annealed
planted and laser annealed Si, showing
supersaturated solid solution. The peak
o
at the surface (closed dots) is indicative
• 0
of segregation. The dashed line gives the
.
'b '/,
.~o profile calculated using_the equilibrium
c o segregation coefficient k, the solid line
o o is a fit to the measured profile [4.43]
~
....
C
Cb
u
C
ou
limit

' . . ".lC: O.()()()1S

1017 L..--L.._.L.---.I:>'--..l---L_-'-----'
o 0.1 0.2 0.3
depth (J.fTl)

other transition metals like Cr, Cu, Fe, W, Zn which are nonsubstitutional
in Si [4.45].
The formation of supersaturated crystals from rapidly cooled melts is,
in itself, not surprising and not in contradiction to local equilibrium thermo-
dynamics. Driving an interface at a larger velocity increases its undercool-
ing (Sect. 4. 1.4) which usually leads to a larger solubility. However, the case
of Si is special in that its equilibrium solubility curve for most Group-III and
-V dopants is of the retrograde type. In retrograde systems the solubility,
rather than increasing with the undercooling, has a maximum beyond which
it decreases with further undercoooling. Supersaturation beyond a retro-
grade maximum, as observed in the laser remelted Si, indicates a deviation
from local equilibrium at the interface. The effect is, following Baker and
Cahn [4.46], known as solute trapping.
How does solute trapping work? Segregation, as apparent from Fig.
4.12, requires the rejected solute atoms to move ahead of the advancing in-
terface by diffusion. The idea of solute trapping is simply that the interface
moves faster than the solute atoms, which are thus engulfed by the interface
and get buried in the crystallizing solvent. The interface velocity u at which
this should occur can be estimated from the melt diffusivity. The minimum
time an impurity atom needs to diffuse out of a monolayer of thickness a
ahead of the interface is a2 /D, while the crystal grows by the same amount
in a time a/u. Impurity trapping should occur when a2 /D = a/u, or u =
D/a. Setting, for an order-of-magnitude estimate, a = 2.10- 8 cm and D =
10-5 cm2 / s yields a critical interface velocity of 5 m/ s, in rough agreement
with the experiment. Impurity trapping is, incidentally, also observed during
solid-state regrowth of ion-implanted Si [4.47]. Here velocities are many

91
orders of magnitude slower than in pulsed remelting, but the difference is
made up by the smaller diffusivity in the amorphous phase.
Supersaturation destabilizes the crystal lattice due to the stress created
by a large number of substitutional foreign atoms. For example, boron
which has a smaller covalent radius than Si, causes the Si lattice to contract
by an amount proportional to the local boron content, and the Si cracks for
concentrations exceeding 4 at. % [4.48]. Sb, which has a larger radius than
Si, causes a lattice expansion instead [4.49]. Thermodynamically, the desta-
bilization of the lattice is expressed by an increase in the free energy of the
crystal. There is a thermodynamic limit for supersaturation, given by the
criterion that the free energies of the solid and the liquid become equal, i.e.,
that the driving force for crystallization vanishes. The corresponding con-
centrations are those at which the free-energy curves of the solid and the
liquid cross (Fig.4.8). Plotted as a function of temperature on the phase
diagram, these concentrations form a curve located somewhere between the
liquidus and solidus lines, known as the To curve. The To curve indicates
the theoretical maximum of supersaturation achievable by growth from the
melt. Estimates indicate that the values observed for Group-III and -v dop-
ants in laser annealed Si are indeed close to the theoretical limit [4.50].
It is quite obvious that the simple binary growth model mentioned in
Sect.4.1.4 with independent component growth rates is unable to describe
solute trapping, in which one species is buried in the other. A number of
kinetic models of solute trapping have been put forward [4.50-53]. These
models are based on specific assumptions about the growth process and the
structure of the interface, amounting to various forms of relationship be-
tween the component growth rates. Jackson and coworkers essentially ad-
ded a term proportional to u' Xl to the solute freezing rate [4.51], while
Wood took the activation energy for solute remelting to increase with
increasing interface velocity [4.52]. Figure 4.14 illustrates the variation of k
with interface velocity for various dopants in Si, calculated from Wood's
model [4.52]. Aziz's calculation allowed the solute to be trapped and to
remelt at rates determined by the interface velocity and the diffusivity
[4.53]. His segregation constant for a continuous growth process is of the
form
-
+k
(3
k(u) = - - (4.18)
{3 + 1

where (3 == ua/D. Note that k reaches unity only if the interface velocity u
becomes large compared to the "diffusive" velocity 01 a, as intuitively
expected from the qualitative argument given earlier. The finding that for
(111) growth in Si k is larger than for (100) growth is consistent with this
picture, as a larger undercooling is required for a given interface velocity in
the (111) case. Hence the melt diffusivity should be smaller, favoring trap-

92
 0.8
>-
;z
UJ
U
;;:
...
w
0.6
o
u
5
~
C>
l:l! 0.4
13
III

0.2

2 4 6 8 fO
MEI.T-FRONT VEI.OCITY (m/S,)

Fig.4 .14. Calculated variation of the segregation coefficient with the I-s interface vel-
ocity. Some experimental data are shown for comparison [4.52]

ping. Segregation and trapping are even observed if the amorphous, rather
that the crystalline phase grows form the melt, with similar values of k(u)
[4.54].

4.2.3 Metallic Substrates

Stimulated by the intriguing results obtained with Si, many researchers have
also applied ion-implantation followed by surface remelting (by laser or
electron beams) to metal single-crystals, such as Al [4.55], eu [4.56], or Ni
[4.57]. The ultimate goal here is not doping but the improvement of corro-
sion or wear resistance of metallic surfaces. The results obtained so far,
although not as complete as those for Si, tend to reveal qualitatively the
same phenomena of epitaxial regrowth, impurity segregation and trapping
as found in semiconductors. Important differences between metals and sem-
iconductors exist with respect to the coupling parameters as well as the ther-
mal data (Appendix A). Metals have high reflectances which tend to
decrease with temperature, and choosing the correct fluence to melt without
damaging them is often more critical than in Si. Further, the absorption
lengths in metals are extremely short and lead to very large volumetric heat-
ing rates and large thermal gradients. Regrowth velocities for equal pulse
durations are also appreciably larger than in Si because thermal conductivi-

93
ties are larger and latent heats are smaller. For example, calculated inter-
face velocities in Si and Al irradiated by 50 ns electron-beam pulses (for
which differences in the absorption length are insignificant) are 1.7 mls in
Si and 8.3 mls in Al [4.58].
Regrowth after pulsed melting in pure as well as ion-implanted single
crystals has been studied in some detail for the case of Al [4.55]. Electron
rather than laser-beam pulses have often been used in this work, but the
metallurgical results are not sensitive to the method of energy deposition.
The most conspicuous difference to regrowth in Si is that regrown metal
crystals contain numerous defects and never quite reach the quality of the
virgin crystal, even in pure specimens (for this reason the term laser an-
nealing is rather misleading in the case of metals). Metals have high equili-
brium densities of vacancies at elevated temperature. Large densities of vac-
ancies also appear to be introduced during solidification. Upon rapid cool-
ing the vacancies are quenched-in but remain mobile enough at ambient
temperature to coalesce into dislocation lines or loops in ways that depend
on the implanted species [4.59]. In addition, thermal stresses present during
and after resolidification can lead to slip deformations. As an illustration,
Fig.4.15 shows slip bands in (110) Al after melting by a 20 ns laser pulse
[4.60]. Slip occurs along the (111) planes, and large densities of dislocations
are present between the slip traces.
Differences between metals and semiconductors also exist with re-
spect to segregation and trapping. Extended solid solubility by trapping
tends to be pronounced even more than in Si, mainly due to the larger re-
growth velocities. Picraux et al. [4.58] have compared segregation of imp-
lanted Sb in Al and Si, which both show retrograde solubility for Sb, under
identical conditions. Enhancements of the distribution coefficient were
about a factor of 10 in Si, but a factor of 250 in AI. Residual segregation
was observed in Si, but complete trapping (corresponding to k = 1) in AI.
The difference was explained by the larger interface velocity in AI, as well
as by the fact that the diffusity of Sb in liquid Al is only about one tenth of
that in liquid Si. Strongly supersaturated solutions in Al have also been
obtained with a number of other implanted species, including Cu, Cr, Ni,
Mo, Sn and Zn.
Complete trapping means that the impurity profile present in the melt
at the instant of solidification is frozen-in unaltered aQd thus available for
quantitative analysis after solidification. Such analysis has yielded interest-
ing insights into the transient diffusion process acting in the short-lived
melt prior to solidification. The simple diffusion equation (4.1) is strictly
valid only for diffusion under isothermal conditions. while in pulsed-melting
diffusion is far from isothermal. Heat and mass flow occur simultaneously
within the same volume, and they interact [4.61]. In a binary mixture kept
in a constant thermal gradient one observes the spontaneous buildup of a

94
Pulsed Laser Irradiation of Aluminum

TEM Observed

3.5 J / cm 2 , 20 ns
Fig.4.15. Electron micrograph taken normal to the surface, and schematic cross sec-
tion showing slip bands in (110) single-crystalline AI after laser melting [4.60]

compositional gradient (typically with the heavier component accumulated

at the cooler end), such that at steady state the two gradients are related by

VX
VT = X(l - X)ST . (4 .19)

Since the total particle flux vanishes at steady state, there must be a contri-
bution to the particle flux caused by the thermal gradient. This is known as
the thermodiffusion, or Soret effect. i The Soret coefficient ST is typically
of the order of 10- 5 to 10- 3 K-i and may depend on temperature [4.62].
Neglecting this dependence, the corrected diffusion equation can be written
as
ax
at = VD[VX + XO-X)ST VT] . (4.20)

i There is also a contribution to heat flow from the concentration gradient (Dufor effect), but
it is not significant in liquids.

95
 L
(a)

Cu

1.5 fbI

c: 1.0
N
::I:
0

c;
0.5

00
depth Inm)
Fig.4.16a,li. Measured (dots) and calculated (lines) concentration profiles of ion-im-
planted Cu and Zn in AI after laser irradiation. Dashed and solid lines indicate profiles
calculated with and without allowance for thermodiffusion, respectively. The value of
ST used in (a) is 2.5.10- 2 K-l, that in (b) is \.10-2 K-l [4.63J

Thermodiffusion, while in principle a second-order effect, can be signifi-

cant in laser remelting where thermal gradients are large. Figure 4.16 ex-
hibits measured and calculated Cu and Zn profiles in ion-implanted and
laser-irradiated Al [4.63]. Calculated profiles were obtained from (4.20)
with and without allowance for thermodiffusion, with ST treated as a fitting
parameter (the Soret coefficients determined by Dona Daile Rose and
Miotello are net values, averaged over the whole thermal cycle [4.63]).

4.3 Surface Alloying

The basic sample structure in surface alloying is an overlayer on a substrate

(Fig.4.1 b). Alloying is achieved by melting the overlayer as well as a com-
parable thickness of the substrate. This geometry allows a higher concentra-
tion of admixture at a depth within a substrate larger than is achievable by
ion implantation. Rather than with "host" and "impurity" species, we deal
here with mixtures of substrate and overlayer material in comparable
concentrations. In this situation (as opposed to that of the previous section)

96
the type of phase diagram of the substrate-over layer system is important in
explaining the microstructure of the alloy. However, due to incomplete in-
terdiffusion, the alloyed layers tend to be inhomogeneous even on a macro-
scopic scale (in contrast to those of the next section). This basic inhomo-
geneity further influences the microstructure of the alloys.

4.3.1 Semiconductor Substrates

Many transition-metal silicides are metallic and yield Ohmic contacts of low
resistance and good thermal stability in semiconductor devices. Conven-
tional techniques of silicide formation employ furnace annealing of thin
vapor-deposited metal films on Si, in which stoichiometric compounds grow
by solid-state reactions (Sect. 3.2). The laser-remelting approach avoids
exposing the entire wafer to the high temperature required for the solid-
state reaction. Pulses in the ns regime have been used in most of this work.
The results tend to be rather different from those of solid-state compound
formation.
The elemental distribution resulting from pulsed surface alloying can,
in a first approximation, be understood from simple diffusion considera-
tions. As a typical example, Fig.4.17 shows Rutherford backscattering spec-
tra of two Pd films, 200 and 48 nm thick, on a Si wafer after irradiation by
equivalent laser pulses [4.64]. The profile obtained with the thick film,
where some unreacted material is left at the surface, closely resembles a
diffusion profile. The thinner film is completely consumed by the alloying
process, and the resulting profile is steplike. This indicates that in the first
case diffusion is incomplete and limited by the melt lifetime, whereas in the

r
01
01

::
01
01
0> ---virgin °1
°1
C
.;: -Ie ser i rradieted ,"0,
~
C
u
III
"'g"
.0

0.6 1.0 1.4 1.8 0.6 1.0 1.4 1.8

energy (MeV)
FigA.17. 2.0 MeV He+ backscauering profiles of vacuum-deposited Pd films on
Si(lll). Dashed lines: As-deposited, solid lines: after laser alloying with a 1.6 J/cm 2 ,
18 ns Nd laser pulse (the arrows give the surface position of the respective atoms)
[4.64]

97
second case it is complete and yields a final concent'ration determined by
the melt depth. The threshold for the alloying process follows from the
requirement that the melt depth exceeds the deposited film thickness. It de-
pends on the thickness of the film as well as on the melting points of film
and substrate material. For element combinations with a deep eutectic in the
phase diagram, the threshold is lowered because melting sets in at the eutec-
tic temperature.
Closer examination of the alloyed structures reveals a variety of com-
plex microstructures not apparent from Fig.4.17. X-ray diffraction often
shows not only one but several compound phases, often including meta-
stable ones. For example, in the Pd-Si alloyed layers the compounds Pds Si,
Pd4 Si, Pd3 Si, P~ Si, and Pdsi can be detected simultaneously, with concen-
trations depending somewhat on the original film thickness and on the laser
fluence. The electrical resistivity of such mixtures can be a factor of 5 to 10
higher than that of a single-phase layer formed by solid-state reaction
[4.65]. This does not necessarily render them useless as Ohmic contacts,
however. Device-quality laser formed Ohmic and Schottky-barrier contacts
have been demonstrated [4.66]. Satisfactory Ohmic contacts have also been
obtained on n-type GaAs by laser alloying of deposited Ge layers [4.67].
The presence of several compound phases is not surprising in view of
the fact that the local melt concentration prior to solidification ranges from
pure Si to almost pure metal - covering, as it were, the whole phase diag-
ram. However, the alloyed layers often also exhibit a complex lateral
microstructure, consisting of "cells" of nearly pure Si surrounded by "walls"
of silicide. The occurrence of cellular growth, which has been observed for
a variety of metal layers on Si (including Pd, Pt, Ni, Fe, Co and Rb) is due
to the effect of Constitutional Supercooling (CS).

4.3.2 Constitutional Supercooling

Under conditions of CS, an initially plane interface tends to develop into a

corrugated array of columns or even more complex shapes [4.68]. The
effect depends on a subtle interplay between macroscopic diffusion of heat
and solute, and microscopic interface dynamics. To understand it, consider
crystal growth from a diluted mixture in the presence of segregation. Segre-
gation causes a buildup of impurity in the melt ahead of the liquid-solid (1-s)
interface. Enriching a melt usually lowers its liquidus temperature and thus
increases its undercooling, as indicated in Fig.4.18a. Under the conditions
shown (liquidus temperature decreasing with increasing impurity content;
gradient in liquidus temperature larger than gradient in actual temperature)
the undercooling, and thus the "nominal" growth velocity, increase with dis-
tance away from the I-s interface. Crystal growth under such conditions is
inherently instable: Random protrusion in the crystal boundary, due to fluc-

98
 liquidus
tempera lure

...::Ja..
"0... actual
a.. C.-_..L--- temperature
Q.

..Ea..
a) distance b}
Fig.4.18 . (a) Schematic temperature profiles in the presence of Constitutional Super-
cooling. (b) Schematic sequence showing column growth from random fluctuations

Fig.4.19. Electron micrographs showing

p·20J-lm, (a) plain view and (b) cross section of cells
in Sb implanted and laser annealed Si [4.71]

tuations, experience at their tip a larger undercooling and grow faster than
their surroundings (Fig.4 .18b). Eventually they develop into columns. The
remaining melt is enriched in the impurity and trapped between adjacent
columns. It eventually freezes out in the form of walls surrounding columns
of relatively pure substrate material. An example of the resulting structure
in the case of Sb implanted Si is shown in Fig.4 .19.
A quantitative theory of CS developed by Mullins and Sekerka [4.69]
predicts the critical impurity concentration for cell formation, as well as the
cell size , to decrease as a function of interface velocity. The theory, which
in its original form assumes local equilibrium at the interface, has been ad-
apted by Narayan [4 .70,71] to allow for a velocity-dependent distribution
coefficient. In this modified form the theory appears to give correct predic-

99
 101
E
:I.
ILl
N
in 100
...J
...J
ILl
U

10- 1

10-2 L-~J...,i..;.J.JJ.l...--'-...J.....l...u..u.:.L--'-L..J..J..l.'-"'---.L.....J....J...1.u.'.J..L.L...--L-1...JL......u.o..J
10- 4 10- 3 10- 2 10- 1 100
SOLIDIFICATION VELOCITY (m 5-1)

Fig.4.20. Calculated cell size at the onset of instability due to Constitutional Superco-
oling in Si containing In. The three curves are obtained by using for the distribution
coefficient the equilibrium value k, the measured value 0.15 (Fig.4.l3), and a function
of interface velocity, f(v), respectively. The dot shows an experimental point [4.70]

tions of the critical impurity content as well as the resulting cell size in
laser-annealed implanted Si. As an illustration, Fig.4.20 shows calculated
cell sizes as a function of interface velocity in In-implanted Si.
Constitutional supercooling depends on the composition gradient in a
melt, but the gradient need not be due to segregation - it arises naturally
from interdiffusion of two elements in laser alloying. It is fairly obvious
that instability is likely to occur whenever the melting point of the substrate
material, from which solidification starts, is higher than that of the overla-
yer or of any intermediate phase. The concept is sketched in Fig.4.21: The
concentration profile, e.g., that shown in Fig.4.1b, is combined with the
phase diagram to yield a depth profile of the liquidus temperature. The oc-
currence of CS can now roughly be predicted by adding transient tempera-
ture profiles to the plot: Whenever the undercooling increases away from
the momentary position of the interface, growth instability is expected
[4.72]. The scheme predicts CS in all cases where cells have been observed.
No CS is predicted, e.g., for Mg on Si, where indeed none is observed
[4.73].
The type of instability illustrated by Fig.4.21 is somewhat different
from that due to segregation alone, and it mayor may not result in patterns
as simple as those of Fig.4 .19. Since a large volume of melt tends to be sim-
ultaneously undercooled in the present case, nucleation and growth of com-
pounds, or of the pure overlayer material, may occur in the melt ahead of
the "main" interface, perhaps concurrently with cell development by the lat-

100
 + =
C1I
c:
, «II
:; 0
~
:;,
:;::;
0.... T[ ·iii 0
C1I
a.
0
a. Gi
a.
E E E
~ 0
u ~

a a Zo 0 z2 z, Zo
concentration depth depth
Fig.4.21. Schematic illustrating the occurrence of constitutional supercooling during
regrowth of samples of the type of Fig.4.1b. To. T 1 • and T2 are subsequent tempera-
ture profiles in the melt. The hatched region is undercooled [4.72]

30 '" Zr P"'"
300.lt Pd

..
Si (100)
~-

planar view

5000 A
~ I

crass section
Fig.4.22. Cell formation in a Pd-Si sample irradiated by a Q -switch laser pulse [4.72]

ter. The result can be a complex or even apparently random mixture of

phases. The case of Pd on Si is illustrated in Fig.4.22. Here the Si columns
are epitaxial with the substrate. while the cell walls consist of a mixture of
silicide phases . In the case of Pt on Si partially amorphous regions have
been detected. along with a mixture of stable and metastable compound
phases [4.65.73]. Co and Ni. on the other hand, form single phase disili-
cides which. even though in a cellular arrangement, can be coherent with
the Si matrix. Two different cell structures in Co and Si were described by
van Gurp et al. [4 .75]. Superimposed on a network of small (lOOnm) cells
due to CS were larger (::::: l/Lm) cells, which van Gurp et al. interpreted as
Benard cells. caused by a convective instability in the melt. As a source of

101
 30 ns laser pulse Fig.4.23. Frozen-in vapor bubbles in a W-Si

D
sample irradiated by a Q-switch laser pulse [4 .72]

480 W
Si (100)

Cross

- .
. ",, !it:..d:~• .~
sec tional

TEM
1 5000 AI
the instability they suggested a gradient in surface tension , arising from lat-
eral variations in melt temperature caused by their sharply focused laser
pulse. Yet another instability is observed in connection with refractory over-
layers : They may crystallize while part of the melt below is still superheated
[4.72]. Figure 4.23 illustrates the case of W on Si: Beneath a layer of poly-
crystalline WSi2 (etched away in the micrograph) there are myriads of tiny
globular voids - frozen-in vapor bubbles an indication that the melting point
of W is above the equilibrium boiling point of Si . Each bubble is connected
to a stem of W-rich material, which apparently acted asa centre for hetero-
geneous nucleation of vapor in the superheated Si melt.
Interface instability is not limited to compound-forming systems but
also observed for solid-solution formers and eutectic systems. Heteroepitaxy
of Ge on Si (100) by laser alloying was demonstrated [4.76, 77], but cell for-
mation and misfit dislocations prevented device quality heterojunction
growth, The simple-eutectic systems Au-Si, Au-Ge, Ag-Si, and Ge-AI were
investigated by Lau et al. [4 .78]. A sample result for the first system is
shown in Fig.4.24 (a·thin amorphous Si layer on top of the AuJayer served

102
 .
_ 1.5 MeV"He+
30
SiA
(100) 25~
Si
a ----
~I
Virgin
26 ,
300A T \'\
i<'l
Q24
X
1I "--4
10

"'020 Loser
~
.J
~
I"
Q)
>= , (X4) ~J' '"I"
0> 16 ~~ +Si J I"
c: Mixture I ,
.... \ AuS2 Si 18 I"
~ •
.1
Q)
~ 12 Eu t e c t i c . J ,"
c "Composition, JI I"
u
IJ) ; Amorphous ,
t; 8 • Phose I ,"
c
CD 1
, 1\ \"
4
~~~-~ t
/1

"'t -'
~~
__ 8-
,:
,"

I",-ao_
\

o
0.45 0.60 0.75 0.90 1.05 1.20 1.35
Energy (MeV)
Fig.4.24. Backscattering spectra of a Au-Si sample before (dashed) and after (dots)
laser alloying. The composition of the buried mixed layer is close to eutectic
(Au 82 Si 18 ) [4.78]

to facilitate absorption of the laser pulse). The main result is that the mixed
layer between the Au film and the substrate, while varying in thickness with
the laser fluence, always has a uniform composition close to the eutectic
composition (18at. % Si), as long as not all of the Au is consumed. Analo-
gous observations were made for the other systems. The interpretation is
that a eutectic melt starts to form at the interface as soon as the eutectic
temperature (370°C) is reached, i.e., while both the Au film and the Si sub-
strate are still solid. Only at a much higher fluence does the elemental distri-
bution start to resemble a diffusion tail, as in Fig.4.17. The structure of the
eutectic alloy was amorphous at low fluence and crystalline at higher flu-
ences (we come back to metastable Au-Si alloys in Sect. 4.4).
An obvious way to prevent growth instabilities is to eliminate the
concentration gradient in the melt, e.g. by premixing· or pre-reacting the
sample, and to choose the composition within a one-phase region of the
phase diagram. Tung et al. [4.79] used this approach to obtain epitaxial Ni
and Co disilicides on Si. Homogeneous melts with compositions in a two-
phase region, on the other hand, are known to solidify in a compositionally
modulated pattern of periodic stripes called lamellae [4.80]. The stripes are
formed by alternating regions consisting of the two phases compositionally
adjacent to the melt composition. Submicrometer lamella formation has
been demonstrated in CW Ar-Iaser melted Co-Si eutectic films [4.81,82].

103
4.3.3 Metallic Substrates

The ultimate aim of surface alloying of metals is to improve the mechanical

strength or the environmental stability of low-cost metals without having to
invest large amounts of expensive transition metals as in bulk alloying.
While the basic experimental approach is the same as in semiconductor
alloying, the emphasis is quite different. Details of the microstructure are
less important than the average composition of the surface alloy and its
hardness, corrosion resistance and macroscopic integrity. The elemental
distribution created by pulsed-laser alloying of metals is basically the same
as in semiconducting substrates, and so are many of the microstructural fea-
tures associated with segregation and CS.
However, while many experiments have used pulsed laser beams, the
thrust for industrial applications is clearly towards using rapidly scanned
high-power CW laser beams. With powers in the kW range and dwell times
of tens of ms, the melt depths achieved can reach one mm. Mixing of the
elements over such distances cannot rely on diffusion alone, but requires
convection. The scanned beam creates lateral temperature gradients away
. from the beam axis as well as along the scan direction. The temperature gra-
dients induce surface-tension gradients which drive convection currents, as
sketched in Fig.4.25. An analysis by Gladush et al. [4.83] shows that the
convection influences heat and mass flow in a 1 mm deep melt pool for ab-
sorbed laser powers exceeding as little as 10 W/ cm2 . At a more typical 50
kW/cm 2 they estimate convection current velocities of several mis, corres-
ponding to Reynold's numbers much larger than 1. Such convection cur-
rents produce far more efficient mixing than possible by diffusion. Rem-
nants of macroscopic convective cells are often still evident in the resolidif-
ied structure. The same convection currents are also responsible for the
characteristic surface rippling of CW-Iaser remelted surfaces [4.83-84].
Current methods of overlayer deposition include vapor deposition,
sputtering, plating (electro- or electroless), powder or paste coating, spray-
ing and rolling-on of foils. An interesting alternative to predeposition of
overlayers is to add the alloy metal during scanned-beam irradiation of the

Fig.4.25. Convection currents during CW-Iaser alloying. The scan direction is perpen-
dicular to the plane of the figure (schematic) [4.83J

104
substrate. This can be done by using a powder jet or a wire feeder nozzle
close to the beam impingment area. For example, Ayers and Tucker [4.86]
used this appraoch to inject TiC and WC particles into surface-molten steel.
Snow et al. [4.87] described a technique termed "layer glazing", in which
thick (up to one cm) alloys are built up layer-by-Iayer in repeated scans. The
deposited material experiences a much larger cooling rate than achievable in
casting, and its composition can be continuously varied by changing the
feedstock. In a typical demonstration of this technique, Snow and co-work-
ers deposited Ni-AI-Mo alloys on a circle along the circumference of a
rotating cylinder, building up a disk about 15 cm in diameter, that was sub-
sequently machined into a turbine wheel.
While the experimental techniques of laser surface alloying reached an
impressive level of versatility, there are a number of physical constraints on
what combinations of metals can be alloyed. The first is the type of phase
diagram (Table A.5). The most unfavorable are Type 4 systems which form
immiscible melts. Experiments with Ag-Ni [4.57,87] have shown that even
in such systems thin substitutional alloys can be achieved, but only if ns
pulses are employed and if the overlayer (Ag in this case) is no thicker than
some 10 nm. Thicker overlayers and longer irradiation times lead to phase
separation. Generally speaking, successful alloying of Type 4 systems at
large concentrations can be expected only if very high melt temperatures -
above the liquid miscibility gap - can be sustained for times long enough to
enable interdiffusion, followed by sufficiently rapid quenching to suppress
melt separation. Here laser remelting offers potential advantages over
conventional alloy forming processes.
Systems of Type 1-3 (Table A.S) do not pose particular obstacles to
alloy formation from the thermodynamic point of view. Of particular inter-
est are Type-2 and -3 systems in which the rapid cooling inherent in laser
remelting enables the formation of small-grained - and therefore hard -
alloys not available by other techniques. This is not the case for Type-l sys-
tems which tend to form homogeneous substitutional alloys independently
of the method of preparation.
Another physical contraint on the element combinations suitable for
laser alloying is evaporation. Melting points and vapor pressures vary wide-
ly among the elements (Table A.6). Combinations of volatile and refractory
metals necessarily lead to problems with evaporation losses or bubble for-
mation. Serious bubble formation was observed, e.g., in laser-remelted cast
Fe alloys containing phosphorous [4.89]. The amount of evaporation can
apparently be influenced by formation of low-vapor pressure intermetallic
phases. For example, strong Zr evaporation losses were observed in laser
alloyed Zr layers on AI, Ti and V substrates, but less on Fe and not at all on
Ni [4.90].
A significant number of alloy systems have been investigated by sur-
face alloying to date and extensive bibliographical reviews have appeared in

105
Q

Fig.4.26a,b. Micro-section of a steel substrate, CO 2 -laser remelted in sequential par-

allel scans at 1 m/s and 50 kW/cm2 . Zones: 1 - remelted, 2 - heat-affected, 3 - U11dis-
turbed. (a) Overview, (b) enlargement of the boundary between zones 1 and 2. Courte-
sy R. Dekumbis, EPFL Lausanne

the literature [4 .91]. Particular attention has been devoted to Fe-based sub-
strates, surface alloyed to elements such as Ni [4.85], Mo, V [4.92], Cr
[4, 92,93] and others . This work has mainly relied on scanned CW CO2
lasers . A general result is that laser surface alloying of low-grade steel sub-
strates can yield corrosion behavior equivalent to that of bulk stainless steel,
but at a small fraction of the precious metal expenditure.
Detailed microstructural investigations have been performed on pre-
alloyed steel substrates, laser remelted with or without additional overlayers
[4 .89]. Such investigations tend to reveal the presence of three well-defined
microstructural zones beneath the laser track - (1) the remelted layer itself,
(2) a region of heat-affected material, followed by (3) the undisturbed sub-
strate material. A typical example is shown in Fig.4.26 for the case of a
bare tool-steel. The sharp boundary between layers (1) and (2) marks the
maximum penetration of the liquid-solid interface . It often coincides (in the
case of a foreign overlayer) with a compositional discontinuity. Layer (2)
consists of substrate material that merely underwent a solid-state thermal
cycle. Its boundary with the substrate (3) represents the isothermal surface
corresponding to the transformation temperature (Sect. 3.2.5). Depending
on the material composition and the local cooling rate, the material (2) can
turn out to be hardened or softened with respect to the substrate. Both the
remelted and the heat-affected layers exhibit characteristic microstructures.

106
In Fig.4.26, layer (1) consists of austenite (white), interwoven by a fine net-
work of dendritic carbides (black), as is typical for rapidly solidified carbon
steels. Zone (2) here contains mainly newly formed martensite, along with
scraps of carbide (rodlike, near the lower edge of the figure) that have sur-
vived unaltered from the original material (3) - ferrite with interdispersed
carbide.
Among the non-ferrous alloy systems that have been subjected to laser
remelting there are examples from all the alloy types of Table A.5 [4.91].
Yet much remains.to be done in this field, given the number of possible
combinations of interest. Technological metals like AI, Cu. Ni or Ti have
mainly served as substrates. Somewhat randomly selected examples of sys-
tems investigated are Cr and Pb on Al [4.95], Au, Sn and Ta on Ni, Ni-Cr
on Cu [4.96] and Pd on Ti [4.97]. The high reflectance of many of these
metals poses practical problems, particularly for infrared lasers and for par-
tially overlapping scans. Highly heat conductive substrate materials, like Cu
or AI. make alloying by CW beams more difficult than others, like Fe, Ni,
or Ti. but produce higher cooling rates. Electron or repetitively pulsed laser
beams may prove advantageous in such cases.

4.4 Melt Quenching

Laser remelting of structures of the type shown in Fig.4.1c yields a binary

melt of homogeneous composition if the individual layers are thin enough.
As an example. Fig.4.27 exhibits MeV He+ backscattering spectra from a
Pt20 Si80 multilayer - prepared by sequential vapor deposition - on a sap-
phire substrate before and after remelting by a 50 ns laser pulse. Similar
results are obtained with co-deposed or pre-alloyed iayers, as well as by the
use of homogeneous bulk alloys in place of a film/ substrate configuration.
The essential difference from alloying of an overlayer B with a substrate-A
is that now there are fixed supplies of A and B, such that the equilibrium
configuration of the alloy is clearly defined by the phase diagram. However,
the thrust of the experiments to be considered here is not towards equili-
brium alloys, but towards glassy phases.

4.4.1 Glass Formation

Both nucleation and growth of crystals have significant rates only between
the melting point and the glass temperature Tg. Cooling a melt through this
range without nucleation and growth produces a glass, a frozen undercooled
liquid. Substances like Si02 fail to crystallize if cooled faster than 1 K/h or
so, while for metallic melts cooling rates between 106 and 10 12 K/ s or more
107
 ~
-E4
5 30ns
u
,...., o Loser
o
~3
AI~ pulse
g Virgin
PI

~ PI :Si = 1:4 ~
~2
a:: \
w
I-
··
~
u l Si ·
··
(f)
::..:: ~
~
O~_ _~_ _~~~~~~~~~
·
0.5 0.75 1.0 1.25
ENERGY (MeV)
Fig.4.27. 1.5 MeV He+ backscattering spectra of a Pt-Si film (80 at. % Si) on sap-
phire. before (solid line) and after (dots) laser irradiation [4.107]

are required for the same purpose. Recently. a new class of multicomponent
alloys has been discovered, which solidify in glassy form at cooling rates as
low as 10 K/s or less [4.97b). These enormous differences reflect different
intrinsic nucleation and growth rates, arising for the reasons mentioned in
Sect.4.1. Metallic glass formation is performed today on an industrial scale
mostly by melt spinning, while for some easy-glass-forming multicompo-
nent alloys even ordinary casting techniques work.
The available experimental work on laser melt quenching tends to fall
into one of two categories: Fundamental investigations, mainly using short
pulses, that explore glass formation in new materials, and more application-
oriented work, typically employing rapidly scanned continuous beams and
aiming at surface hardening of machine alloys. Cooling rates typical for
scanned beams are comparable to those of melt spinning, and the results
obtained with them have not been shown to be qualitatively different from
those of mechanical quenching. We shall mainly emphasize the pulse work
here.
Cooling rates in pulsed-laser quenching depend on the pulse duration
and on the conductivity of the heat sink, but are quite insensitive to the type
of material being quenched. Figures 4.28a and b exhibit numerically calcu-
lated cooling rates in films of Au50 Ti50 alloy, deposited on sapphire or
tungsten substrates, and irradiated by pulses of 50 ns and 50 ps FWHM dur-
ation, respectively (the film thickness must be adapted to the pulse duration
for complete melting without evaporation). The cooling rates relevant for
melt quenching are those below the melting point (dashed line) and are seen
to be of the order of 109 K/s for the ns pulses and 10 12 for the ps pulses.

108
 a

x:
-2(XX)
~
.2
o
Q;
a.
E
!:
1500

i'
t \
/ \pUlse

time Insl

b
2500 .,.-5E12
~
\
\
\
\
\
\
\
\
- -----,----------
\
\
1500 ',lE12
,
:\
.
: \ pulse
"'-
', ...... ..............
SAPPHIRE

.......... ___ TUN3STEN

, \
~1L-~L-~~a~5~~~~~~-L-L~~~~~~~~2~~
~-----

time Ins)
Fig.4.28. Numerically calculated surface temperatures in (a) 150 nm and (b) 60 nm
thick Au 50 Ti50 films on tungsten and sapphire substrates, irradiated by (a) 50 ns and
(b) 50 ps laser pulses. The dashed line indicates the melting point of the films.
Exponential numbers indicate instantaneous cooling rates in K/s.

The use of refractory metal instead of sapphire substrates enhances the

cooling rate by about a factor of three, due to a larger thermal conductivity.
In Fig.4.29 phase diagrams of a number of systems quenched with 50 ns
pulses are shown, with dark areas indicating the ranges where glass forma-
tion has been observed. The upper boundaries of the dark areas indicate the

109
 1600 1600

1400

1200

.
~
0
; 800 - -

r
t
} 600

'00

200 200

0.,. 20 <0 80 eo 50 AIJ 20 <0 60 eo 50

at.,. 51 ol"l.Si at..", Si at.'- Si

01.-" Si

aL"f... T. 0'-1. T. 0'-1. To

Fig.4.29. Phase diagrams of some binary systems indicating experimentally esta-

blished ranges of glass formation by ns-Iaser melt quenching of samples of the type of
Fig.4.27 (dark areas) . Upper boundaries of dark areas give crystallization temperatures
of glassy films upon slow heating . Dotted areas are regions of diffusionless crystalliza-
tion.

temperature at which crystallization of the glass upon slow heating was de-
tected. Most of the glassy phases represented in the figure are not available
at lower cooling rates.
Most of present knowledge about what is somewhat vaguely called
Glass Forming Ability (GFA) has emerged from the experimental back-

110
ground of mechanical quenching, and hence pertains to cooling rates of the
order of 106 K/s. Since short laser pulses produce cooling rates several
orders of magnitude higher, it is not ol>vious that the same criteria should
predict GFA by the two techniques. Kinetically, it is clear that the larger the
reduced glass temperature Tgr = TgIT sl' the better the GFA. Good metallic-
glass formers at 106 K/s have Tgr values around 0.5 to 0.6 [4.98]. A related
criterion is the relative melting point depression of the mixture, Il. T mp =
(T av - T sl )/Tav' Tav being the weighted average of the component melting
temperatures. Il.Tmp values of 0.2 or more indicate good GFA [4.99]. Obvi-
ously, both criteria favor compositions close to a deep eutectic. Neither is
very successful in predicting GFA by laser quenching, as an inspection of
Fig.4.29 shows. The same is true for several other empirical criteria of
GFA considered in the literature [4.100].
Yet these are obvious regularities in the glass ranges shown in Fig.4.29
- they are all in two-phase regions of the respective phase diagrams, which
are all of Type 2 or 3 (Table A.5). Continuous solid-solution formers do not
yield glasses by ns quenching. An interesting limiting case is provided by
systems like Cr-Ti (bottom right) which form solid solutions only at high
temperature. Here the range of glass formation is very sensitive to the cool-
ing rate - quenching with 50 ns pulses yields glasses between 45 and 65
at. % Ti for films on sapphire, but up to 85 at. % Ti for films on tungsten.
These observations suggest that the requirement of long-range mass trans-
port is the main hindrance to crystallization at cooling rates around 1010
K/s.
Diffusionless crystallization can become possible at large undercooling
even in two-phase regions, however, due to formation of either supersatu-
rated equilibrium phases or of "new" metastable compounds. Supersaturated
solid solutions may form in systems with limited equilibrium solubility by
means of trapping. The thermodynamic limits of trapping can be derived
from free-energy considerations and quantified by means of To curves
(Sect. 4. 1.3). An example is the Ag-Cu system, which has a Type-3 (Table
A.5) phase diagram but forms a continuous solid solution even at small und-
ercooling [4.101]. Laser quenching with ns pulses in Ag-CU produces the
metastable solid solution instead of a glass. An example of a system form-
ing a "new" metastable compound upon quenching is Ag-Si (Fig.4.24), in
which glass formation is achieved only within a small window of composi-
tion near the Si-rich end of the phase diagram.
Complications with metastable crystals notwithstanding, it turns out
that glass formation can be predicted quite reliably from standard thermo-
dynamic data even at cooling rates around 1010 K/s. The idea is to con-
struct Temperature-Time Transformation (TTT) plots for crystallization of
every crystal phase competing with glass formation [4.102]. The crystal-
lized volume fraction in a homogeneous melt cooled at t = 0 from T sl to a

111
 Fig.4.30. Schematic time-tempera-
ture-transformation cu rve for for-
mation of a 10- 6 volumc f raction of
lsi a crystal phase from an undercooled
melt, according to equati on (4 .21) .
"- Thc dashed line indicates the critical
C1J
.... "- cooling rate for glass formation .
::J
0
....
C1J
a.
E
~
19

time

temperature T below the nucleation temperature can be approximated by

(assuming spherical crystallites)

E(T,t) == (1r/3)Y(T)u(T)3t4 (for E « 1) (4.21)

where Y and u are expressions for the steady-state nucleation rate and the
interface velocity, respectively, as given in Sects.4.1.4,5. Transient effects
like the nucleation time-lag are neglected here. Glass formation is assumed
to occur if E remains below some detection limit, usually taken to be 10-6 .
A schematic TTT curve indicating the region of detectable crystallization,
as well as the limiting cooling rate to avoid it (dashed line), is shown in Fig.
4.30. The scheme can be extended to crystallization in a two-phase region
by using the concentration-dependent driving forces introduced in Fig.4.8
[4.103]. The interface energy (J required to calculate the nucleation rate can
be estimated from the latent heat of melting, while the jump frequency can
be related to the melt viscosity [4.104]. To predict ranges of glass forma-
tion, critical cooling rates are calculated for all possible crystal phases.
"New" metastable compounds are treated analogously (this requires that
suitable free-energy parameters can be estimated for them). Figure 4.31 dis-
plays critical cooling rates calculated for the systems Au-Si and Ag-Si, tog-
ether with the glass-forming range at the cooling rate used in the experi-
ments [4 .105]. The procedure correctly predicts the observed ranges of
glass formation by laser quenching (5 . 109 K/s) as well as those observed in
mechanical quenching (== 106 K/s). The basic physical reason for the differ-
ence in GF A shown by the two systems is that the heat of mixing of liquid
Au-Si is strongly negative, stabilizing the melt and hence the glass, while
that of Ag-Si is positive at most compositions. It is clear that a strongly neg-
ative heat of mixing of the melt is a generally favorable condition for glass
formation, independently of the method of quenching.

112
 LOGle LOGle
H ...........u.................................~...................,..........................................'t H ...........................................~~.........u...................'"'""""u.......T
13 13
12 12
11 11
_~ 113 GFR _~ 113
I~ 9 I~ 9
,: 8 ,: 8
~u 7 ~u 7

'" 6 "" 6
3' 5 o " Au ! s
'\ A = Au 3Si '\ 0 = Ag
3 0 =Si 3 A = Ag 3Si
2 2 0 = Si
1
e~~~~~~~~~~~~+ e~~~~~~~~~~~~
13 1 2 3 i 5 6 7 8 9 113 13 1 2 3 i 5 6 7 8 9 113
Inole F'rnction Si X1B-1 mole F'rnction Si Xle- 1

FigA.31. Calculated critical cooling rates as a function of composition for the sys-
tems Au-Si and Ag-Si, as determined by precipitation of the pure metal (0), the pure Si
(<», and a metastable silicide (/l). Also included are observed glass forming ranges at
5.10 9 K/s [4.105]

4.4.2 Silicon-Based Systems

Si and other Group-IV elements are known to facilitate glass formation in

alloys. The first glassy metallic alloy ever to be produced by melt quenching
was a eutectic mixture of Au and Si (l8at. % Si), quenched by P. Duwez and
co-workers in 1960 by means of a shock tube device at about 10 6 K/s.
The formation of pure amorphous Si by pulsed laser irradiation was
discussed in Sect. 4.2.1. Glassy Si-based alloys have been obtained by laser
quenching with Q-switch pulses in a number of systems [4.105-107]. The
crystallization temperature of the glassy films, as Fig.4.29 demonstrates,
varies widely from system to system and even within systems. Au-rich glas-
sy Au-Si films show precipitation of Au even at temperatures around -30°C
and must be kept cooled during and after irradiation, whereas Nb-Si glasses
remain stable up to 600 °C. Crystallization does not in all cases produce
equilibrium phases. Metastable silicides form, e.g., in the Au-Si and Pt-Si
systems from the glass, and the equilibrium configuration is reached only
after annealing to a temperature 100 0 0r 200°C higher than the crystalliza-
tion temperature indicated in the figure.
Glassy Si-metal alloys offer a number of interesting features in their
electronic properties. In metal-rich glasses of Au-Si, the Si atoms are found
to be in a metallic state (pure solid metallic Si exists only at pressures above
120kbar). This is manifested in the electrical conductivity as well as in the
reflectance spectra of the films [4.109]. Above about 30 at. % Si the metallic
character of Si is lost, although the alloy still behaves as a true disordered
metal. This holds for compositions up to about 85 at. % Si, where the materi-

113
al abruptly changes from metallic to semiconducting. The critical composi-
tion of 85 at. % is in accordance with Mott's criterion for a metal-nonmetal
transition in this system. Altogether, the resistivity of the laser-quenched
glassy Au-Si alloys can be varied via the composition over four orders of
magnitude. Similar behavior is also found in other Si-metal systems.

4.4.3 Metal-Based Systems

By a metal-based system we mean a system in which nonmetals (including
Si) are either absent, or present only as traces to facilitate glass formation,
as often used in mechanical quenching. Glass formation in a number of
metal-based systems has been achieved with scanned high-power CO 2 -laser
beams. Examples include alloys such as FesoB16Si4 [4.110], Fe40Ni40P14B6
[4.111], (FeCr 12 )so(C,Bho' NixNb lOO _x (x = 30--;.-60) [4.112], Pd7SCu6Si16
and Fe 61 CrlOMoS P 16 C s [4.113]. These alloys are "easy glass formers" well-
known from mechanical quenching. The samples are usually either crystal-
line bulk alloys or powders applied to a steel base. Laser powers employed
range from several hundred W to a few kW, and scanning rates are in the
range of 0.1 to 1 m/ s. With typical spot sizes of between 0.1 and 1 mm,
dwell times are of the order of 1 ms. With these parameters, cooling rates
are of the order of lOS to 10 6 K/ s, i.e., the same as those typical in mechan-
ical quenching. The thickness of material amorphized is given by the melt
depth and is typically between a few tens and a few hundreds of pm. The
technique has interesting potential for the surface hardening of machine
parts [4.114] or the production of corrosion-resistant coatings, and it may
also yield acceptable throughput for large-scale fabrication. Certain prob-
lems do exist, however, such as the tendency of amorphized zones to recrys-
tallize during subsequent partially overlapping scans, the cracking of amor-
phous layers on substrates due to thermal stress [4.115], as well as the ten-
dency of overlayers to lose their GFA due to admixture of substrate materi-
al.
Glass formation at compositions beyond conventional ranges of GFA,
or in systems not previously considered to be glass formers, has been achi-
eved by ns- or ps-pulse laser quenching in a number of cases. Among these
are iron-rich Fe-B alloys [4.116], Ni-Nb; Mo-Ni, Mo-Co, and Nb-Co alloys
[4.117], Au-Bi alloys [4.118] and even pure Ga, the only metal so far re-
ported to be quenched to a glassy phase [4.119] (an amorphous phase ob-
served in nominally pure Al after irradiation by 15 ns pulses turned out to
be impurity-stabilized [4.120]). Particularly interesting kinds of metastable
phases have been discovered in Au-Ti and Cr-Ti alloys that were deliber-
ately laser quenched at cooling rates just insufficient for glass formation.
The metastable crystalline phases thus obtained revert spontaneously to an
amorphous structure upon subsequent furnace annealing [4.121].

114
5. Evaporation and Plasma Formation

The subtle and varied structural modifications considered in the previous

chapters represent only a narrow window of irradiance or fluence, in which
material is heated or melted but (essentially) not vaporized. In the follow-
ing, material effects pertaining to the open-ended interval of irradiances ex-
ceeding the evaporation threshold will be considered. The physics involved
is quite different from that considered in Chap. 4, and so is our point of
view: Rather than studying what are essentially relaxation phenomena oc-
curring after irradiation, we shall again deal with true light-matter interac-
tions and consider relaxation effects only occasionally. The material effects
in this regime are also rather less subtle than those discussed so far: Instead
of microscopic structural rearrangements we shall be concerned with the
transport of macroscopic masses over macroscopic distances, with ablation
and compression of material and with internal energies large compared to
chemical activation energies.
The main topic of this chapter, material evaporation, is the most tradi-
tional branch of laser processing. Related research activities started in the
early sixties, soon after the first high-power ruby lasers became available.
The early work was mainly motivated by the prospect of using laser beams
as machining tools. Most of the phenomena governing evaporation by laser
beams were understood by 1975, although, of course, important insights
and refinements have been made in the meantime and continue to be made.
Present research activities are mainly centered in two fields: improvements
in the understanding and control of machining processes, and studies of the
interaction mechanisms of ultra-intense radiation with matter.
Speaking of applications, machining - in all its variations from cutting
through welding to milling and drilling - was the first area in which laser
processing crossed the threshold to widespread industrial use. More prospec-
tive areas of application are shock treatments of materials and material test-
ing under extreme conditions. Somewhat elusive remains the production of
ultra-dense plasmas capable of sustaining thermonuclear fusion for energy
production. Dense and hot laser-produced plasmas promise other uses, how-
ever, e.g., as intense sources of X-ray radiation.
This chapter is organized loosely following a sequence of increasing
irradiance. We start again with a section discussing fundamentals. Section
5.2 then deals with "normal" material evaporation in which the laser beam
just happens to furnish the energy, along with its main applications in

115
machining. Section 5.3 considers irradiances sufficient to cause substantial
ionization of the vapor and the surrounding atmosphere, while Sect. 5.4
covers the highest available irradiances at which the very distinction be-
tween the condensed material and its vapor disappears. In the final Section
5.5 we return to solid earth, as it were, and discuss what has now become
the most promising application of laser-produced vapors: their use as depo-
sition sources for the production of thin films.

5.1 Fundamentals

5.1.1 The Thermodynamics and Kinetics of Evaporation

Since evaporation usually occurs from a liquid, let us start by considering

the conditions for a liquid-vapor equilibrium. Phase equilibrium between a
melt and its vapor requires equality of the free energies, but vapors are
easily compressible and the equilibrium conditions now depend on the pres-
sure. The change of free energy with temperature can be written as

dG aG aG dp dp
dT = aT + ap dT == - S + V dT (5.1)

where we have used the thermodynamic relation dG = Vdp-SdT. Eq.(5.1)

must hold for every phase individually. At equilibrium the temperature and
the pressure inside the liquid and the vapor must be the same, from which
follows that

Sv -SI
dp/dT = V - V :::: (5.2)
v I

where dV lv = (Vv-V,) is the volume change upon evaporation and where

we have replaced the entropy change dS lv = Sv -SI by dH lv /T lv ' the ratio
of the latent heat of evaporation to the liquid-vapor equilibrium tempera-
ture. Eq.(5.2) is the Clausius-Clapeyron equation which describes the de-
pendence of vapor pressure on temperature, or, if written in the form

(5.3)

also specifies the dependence of the equilibrium temperature on pressure. I

1 The values of the quantities T lv' dH lv and d Vis will henceforth be understood to be those
at 1 atmosphere, unless noted otherwise.

116
 /
/

Tly Ter
temperature
Fig.5.1. Schematic p-T diagram showing equilibrium ranges of existence of the solid.
liquid and vapor phases

The transition between a liquid and its vapor at ordinary temperature is a

first-order phase transition, like that between a crystal and its melt. The
latent heat of evaporation is typically of the order of 2+4 eV per atom at
normal pressure, as compared to 0.1 to 0.5 eV per atom for melting (Table
AA).
The equilibrium ranges of existence of the solid, liquid and vapor
phases of an elemental substance are illustrated by the schematic pressure
diagram in Fig.5.1. The liquid-vapor equilibrium curve, unlike the one sep-
arating the solid and liquid states, ends in a critical point beyond which
there is no thermodynamic distinction between the liquid and the vapor. The
critical temperature T cr for metals is typically of the order of 3 to 4 times
Tty. At the critical point the latent heat as well as the volume change go to
zero like (Tcr - T) 1/2, and hence the densities of liquid and vapor merge
smoothly. Several material properties behave anomalously at the critical
point; this is related to the fact that here (dpl aVh vanishes. For example.
density fluctuations become very large and the electrical conductivity drops
sharply. Beyond Tcr "evaporation" proceeds as a continuous decrease in

117
density, without a phase transition and without thermodynamical instability.
An interesting problem exists with respect to the evaporation of met-
als' which involves the transition from a conducting to an insulating state.
The transition is the inverse of Mott's transition and occurs at some transi-
tion density Pmi intermediate between the normal liquid and vapor densities.
During subcritical evaporation the density changes discontinuously at the
liquid-vapor boundary and "jumps" over Pmi' hence the liquid-vapor and the
metal-insulator transitions occur simultaneously. But what, if the material is
heated beyond the critical point and the liquid density decreases continu-
ously to the vapor level? It is often assumed that Pmi coincides with the criti-
cal density Per [5.1], although there is no reason a-priori that this must be
so. It has been hypothesized [5.2], that for certain metals Pmi > Per' i.e., the
metal-insulator transition should already occur in the condensed liquid
below Ter' This would produce a transparent surface layer and strongly in-
fluence light absorption in the evaporating liquid. Experimental evidence
for such an effect is, however, rather indirect and pertains only to the case
of Hg [5.3]. Moreover, ionization of the vapor, which tends to obscure ab-
sorption phenomena in the dense phase, is possible at vapor pressures as
low as a few atmospheres, far away from the critical regime.
Let us now return to normal evaporation. An expression for the equili-
brium vapor pressure as a function of temperature is obtained by integra-
tion of (5.2)

-p(T) = poexp[T -T
LlHlv RTT lv
Iv ] [ LlGlv ]
= poexp - RT (5.4)

where Po is the ambient pressure. Eq.(5.4) is an approximation valid in the

range T lv < T < T er of interest for the present discussion. It is based on
the empirical fact that the ratio of the latent heat to the difference in com-
pressibility between the vapor and the liquid is approximately constant in
the mentioned temperature range [5.2]. An equation of the form (5.4), but
with different parameters, is also valid at temperatures below T lv (some
vapor pressure data are given in Table A.6)
In nonelemental systems the liquid-vapor eqUilibrium obeys the same
rules as that between a liquid and a crystal. Nonelemental vapors can always
be regarded as ideal solutions, with the partial pressures given by pi = pXj
where Xj is the concentration of species j in the vapor. The partial pres-
sures satisfy

(5.5)

which generalizes (5.2). Here .ahlJ is the "partial heat of evaporation" of

species j, (the enthalpy change per mole of solution contributed by evapora-
tion of that species alone), and .aVJ is the corresponding change in the par-

118
tial molar volume. If the liquid is an ideal solution as well, then both quanti-
ties vary linearly with the melt composition.
As far as kinetics is concerned, there are two distinct modes of evapo-
ration from a melt under laser irradiation: volume evaporation and surface
evaporation. Surface evaporation is the normal case for metals, while
volume evaporation may play a role in weakly absorbing media. Let us
briefly consider the main aspects of the two modes of evaporation. The rate
of surface evaporation can be obtained from considerations similar to those
pertaining to the I-s interface velocity, but since we are mainly interested in
the condensed phase (or what remains of it) we shall take a simplified ap-
proach here. Under equilibrium conditions the rate of particles evaporating
from an open liquid surface is equal to the rate of particles condensing on it
from their saturated vapor. The rate of condensation, and hence also evapo-
ration, is obtained by integration over the Maxwellian velocity distribution
of the vapor particles [5.4]. This yields a molar evaporation flux [moles/
m2 s] of

(5.6)

Here A has the meaning of a "sticking coefficient", giving the fraction of

vapor particles hitting the surface that adhere to it. For metals A is gener-
ally close to unity. If the actual pressure p in a liquid at temperature T is
less than p (T), the liquid is superheated. However, the evaporation rate of a
liquid should not depend on whether or not it is surrounded by its equili-
brium vapor, and (5.6) is also valid for evaporation into vacuum or into an
ambient atmosphere, provided p is not too large. The I-v interface velocity
is then
- -
U =jyVd = Mjy/Pd (5.7)

where Vd and Pd are the molar volume, and the density of the melt, respec-
tively. If a substantial vapor pressure develops, the interface velocity is re-
duced due to backflow and condensation of particles. as discussed in Sect.
5.2.
The superheated liquid also tends to boil. Boiling, or volume evapora-
tion, requires nucleation and growth of vapor bubbles. The volumetric rate
of steady-state homogeneous vapor nucleation can be put in a form equiva-
lent to (4.15)

(5.8)

where the jump frequency can be considered as purely impingement-lim-

ited, as in (4.11) . .1gN is the free energy of forming a critical vapor nuc-

119
leus, obtained from (4.14) by replacing LlGsl and (J by the corresponding
quantities for the liquid-vapor transition. In rapidly heated melts, vapor nuc-
leation starts only after a time-lag similar to that observed in crystal nuclea-
tion [5.5]. Typical magnitudes for molten metals at moderate superheating
are critical bubble radii of the order of 10 + 100 nm and time-lags of the
order of 10-8 s. The strong temperature dependence of (5.8) again results in
a sharp increase of the homogeneous nucleation rate above a relatively well-
defined nucleation temperature TN' However, boiling may develop at a
much lower temperature as a result of heterogeneous nucleation in the pres-
ence of foreign particles (as Fig.4.23 nicely demonstrates).
Since a large change in density is associated with vapor nucleation, the
strong increase in the nucleation rate above TN causes violent expansion of
the superheated liquid, called a vapor explosion. The energy liberated in a
vapor explosion (the work done by the vapor) may reach a significant frac-
tion of LlH\v (it may, in fact, exceed the specific energy of common chemi-
cal explosives [5.6]). In order for the liquid to reach TN heating must be fast
enough to prevent boiling by heterogeneous nucleation at small superheat-
ing. Martyniuk [5.6] estimated that heating rates of the order of 10 9 K/s are
required to keep heterogeneous nucleation in liquid metals insignificant.
Such heating rates are, of course, easily achieved with pulsed lasers in ab-
sorbing materials.
The composition of a binary vapor is, in general, different from that of
the liquid in equilibrium with it, in accordance with the vapor-liquid phase
diagram. In open evaporation from a surface the vapor is continuously re-
moved, and only the freshly formed vapor is in contact with the melt. It is
obvious that in this situation the melt is enriched in the less volatile species
until either the more volatile component is lost or a congruently evaporating
(or "azeotropic") liquid mixture is reached. However, this process depends
on the presence of diffusional equilibrium in both the liquid and the vapor.
If evaporation is fast, only a surface layer of the liquid is depleted in the
volatile species, and the rate of evaporation of the latter drops quickly. At
the same time backflow from the vapor (to be discussed in the next section)
will mostly consist of atoms of the volatile species. The situation is reminis-
cent of the one discussed in connection with segregation in Sect.4.2.2 - in
the absence of a diffusional equilibrium the vapor composition approaches
the composition of the liquid. If backflow is neglected, this will happen
after a time = D/ku2 , D being the diffusivity of the volatile component in
the melt, and k the distribution coefficient at the vapor-liquid interface. Re-
membering our discussion of trapping (Sect.4.2.2), we may again expect
that k will tend towards unity if u approaches the "diffusive" velocity D/a,
since particles of the less volatile species will be dragged along by the more
volatile ones as the evaporating boundary overtakes diffusion in the melt. In
a laser-irradiated solid where the melt is continuously formed ahead of the
evaporating surface, the melt composition will itself be equal to the compos-

120
l110n of the solid. Moreover, under powerful irradiation the atoms can
acquire thermal energies far in excess of the binding energies, and differ-
ences in the partial enthalpies of the chemical species become unimportant.
Hence strongly superheated solids will evaporate without a change in com-
position, and even the very presence of a liquid phase can usually be neg-
lected. This feature is crucial in laser pulse-induced material deposition
(Sect. 5 .5).
We mention in passing that besides thermal (heat-driven) evaporation
there are also nonthermal evaporation phenomena. Ultraviolet laser beams
can induce optical excitation of molecules in some insulators, notably
organic polymers, at rates rivaling or exceeding thermalisation. Since exci-
tation tends to lower binding energies, excited particles may "evaporate", or
desorb, before the material temperature rises appreciably [5.7]. This pro-
cess, known as "photolytic desorption", explains why certain materials can
be UV-ablated "cold" and at rates exceeding those expected for normal ther-
mal evaporation [5.8].

5.1.2 Hydrodynamics

Hydrodynamics provides a point of view totally different from that of kinet-

ics, useful particularly at intensities well above the evaporation threshold.
The concept is to simply ignore the structure of both the material and also
the liquid-vapor interface. The former is treated as a structureless fluid,
characterized only by its density, pressure and energy content, and the lat-
ter as a mere discontinuity in the thermodynamic variables. We shall, as
before, use the enthalpy to characterize the material in each phase. From
the thermodynamic identity H = U +pV, where U = JCydT is the internal
energy, the enthalpy H = Jc p dT can be written as

(5.9)

where 'Y == cp / cy is the adiabtic index. This procedure neglects the differ-
ence in temperature dependence of the specific heats, and for real sub-
stances averaged values for 'Y must be used. For a monatomic ideal gas 'Y =
5/3, and for a diatomic one 'Y = 7/5.
The familiar conservation laws of mechanics are now applied to relate
the corresponding quantities on both sides of the boundary. Denoting with a
subscript d quantities of the dense phase (no distinction is made between
solid and liquid) and with v those of the vapor, we require that the flow of
mass, momentum and energy across the boundary be continuous. For one-
dimensional gas flow and neglecting viscosity we have, respectively, [5.9]

121
 Pd Vd = Py Vy , (5.10)
Pd + Pd Vd 2 = Py + Py Vy 2 , (5.11)

Hd/M + Vd2/2 = Hy/M + v/12. (5.12)

Here the velocities 2 are measured with respect to the boundary (the velocity
of the boundary in laboratory coordinates is u = -vd), and we have used the
mass density P = M/V instead of the molar volume. If, e.g., Pd and Hd are
known in advance,then the system (5.10-12) contains 6 unknowns. Using
(5.9) for Hy , we need 2 more equations which have to be chosen according
to the problem under consideration. Often the vapor is treated as an ideal
gas for which the ideal gas law

pV =pMlp = RT (5.13)

provides a connection between pressure, density and temperature.

A most important aspect of evaporation by a laser beam is the interac-
tion of the evolving hot vapor with the beam itself. This interaction leads to
strong nonlinearities in the coupling of an intense beam with a target, in ad-
dition to those that may arise in the condensed material.

5.1.3 Ionization of the Vapor

The equilibrium degree of ionization of a gas of density N and temperature

T follows from the Saha equation

-N 2
kT[me
=2N -gl - -]3/2 e -E J /kT (5.14)
e g2 21(fl2

where gl and go are the statistical weights of the ionized and neutral states,
respectively, and E J is the ionization potentiaL In a partially ionized gas,
light is absorbed by thermally excited atoms (bound-free absorption), as
well as by ions (bremsstrahlung absorption). The total absorption coeffi-
cient for flw « E J can be estimated from the Kramers-Unsold formula
which holds for hydrogen-like gases [5.10]

(5.15)

2 In this chapter we use the symbol v for particle velocities and u for interface velocities.

122
More accurate calculations of absorption in hot vapors have been per-
formed, e.g., for 10.6 IJ.m radiation in Al vapor [5.11] and in metal oxide
vapors [5.12].
With increasing irradiance the temperature and the enthalpy of the
vapor increase. Light absorption further heats the vapor, which leads to
even more absorption. This positive feedback favors the creation of a
plasma in front of an evaporating target even at light fluxes far below the
threshold for breakdown in a cold gas (see below). Typical irradiances at
which this phenomenon is observed are 10 8 W / cm2 for Nd lasers and 107
W/cm 2 for CO 2 lasers [5.13]. Associated vapor pressures are of the order
of tens of atmospheres, corresponding to only moderate superheating of the
liquid.
Once the gas is fully ionized, light absorption is dominated by brems-
strahlung absorption. For hot plasmas (kT » hw) the absorption coeffi-
cient can be written as [5.14]

(5.16)

This absorption coefficient varies like >..2, and Fig.5.2 displays its depend-
ence on temperature in fully ionized hydrogen plasmas at various densities
[5.15]. Above the cutoff density (= 1021 cm- 3 ) where the plasma frequency
exceeds the laser frequency, n, becomes very small (Sect. 2.1), and the
plasma behaves essentially like a metal. Below the cutoff density the absorp-
tion coefficient decreases strongly with increasing temperature, a feature
that has important implications for laser-produced plasmas, as we shall dis-
cuss in Sect. 5.4.
The hot vapor represents an easily ionizable medium, due to the ther-
mal excitation of the atoms. However, even cold gases become ionized and
absorbing in the presence of high irradiances, due to optical breakdown.

5.1.4 Gas Breakdown

Laser-induced gas breakdown involves the same mechanisms as breakdown

in solid dielectrics. As discussed in Sect.2.2.4, ionization for hw < Er can
occur by multiphoton absorption or by avalanche (impact) ionization.
Breakdown thresholds are usually determined by avalanche ionization, ex-
cept for ultrashort pulses, or at gas pressures so low that the mean-free path
between electron-atom collisions exceeds the focal diameter. Avalanche
breakdown requires the presence of some "priming" free electrons, as in
solid dielectrics. For the average energy gain of a free electron (2.44) still
holds, but inelastic energy losses are negligible in the gas (oE/E is of the
order of the ratio of the electron mass to the atom mass). Furthermore, Te is

123
Fig.5.2. Calculated bremsstrahlung absorption coefficient as a function of tempera-
ture in a fully ionized neutral hydrogen plasma, for}" = 1.06 /lm and various electron
densities. [5.15]

now the electron-atom collision time, which is inversely proportional to the

gas pressure p. At normal pressures and optical frequencies we have w »
11 Te and, from (2.44), the breakdown threshold is expected to be propor-
tional to Te and hence to lip. A minimum in the threshold should occur at w
= 11 Te' which turns out to be at some hundred atmospheres for optical fre-
quencies. Both predictions are borne out by the experiment [5.16]. For shar-
ply focussed beams the breakdown threshold increases due to diffusion of
electrons out of the focal volume. and an inverse relationship between the
breakdown threshold and the focal volume results [5.17]. A detailed study of
avalanche breakdown in air was presented by Kroll and Watson [5.18].
A key difference between avalanche breakdown in solid dielectrics and
in cold gases is that priming electrons are very much scarcer in the latter. In
normal air at STP the equilibrium density of ions is only about 103 cm- 3 ,
and the natural ionization rate (by cosmic radiation, etc.) is no more than
some 10 cm- 3 S-I. The probability of finding a free electron within the
volume of a focal spot is thus negligible. The first electrons of the ava-
lanche have to be delivered either by absorbing impurities of low-ionization
potential, like dust particles, or by multiphoton ionization of gas atoms. In

124
particular, gas-breakdown thresholds are found to be dramatically lowered
by the presence of an absorbing target in the beam path. Here adsorbed
impurities, which are evaporated and ionized far below the melting point of
the target material, are almost always present. In practice, air breakdown
thresholds in the vicinity of targets are of the order of 107 W/ cm2 for CO2
lasers and 10 9 W/cm 2 for Nd lasers, roughly two orders of magnitude lower
than the corresponding thresholds in clean air.

5.2 Evaporation at Moderate Irradiance Levels

The subject of this section is the regime of normal evaporation, in which

the laser beam merely heats the condensed phase. This means, in particular,
that the vapor is cool enough to be essentially transparent. Important laser
applications in machining are often performed in this regime.

5.2.1 Beam Heating and Evaporation

A lower bound to irradiances of interest here may be estimated from the en-
ergy flux required for the surface to reach T tv while the irradiation lasts,
namely

(5.17)

Here dH(T) is the enthalpy difference of the material between ambient tem-
perature and temperature T, and d is the larger of the absorption length ex-I
and the thermal diffusion length 2(Ktp ) 112. For a typical metal, taking 50
kJ/cm 3 for the RHS of (5.17) gives, e.g., for

1s 1 ms 1 /LS 1 ns
I(1-R) >32 kW/cm 2 > 1 MW/cm 2 >32 MW/cm2 >1 GW/cm2

The incident irradiance required in the case of metals would be a factor of

10 or so higher because of reflection. Since for the shorter pulses these irra-
diances approach or exceed those causing vapor ionization or even gas
breakdown, the regime of normal evaporation is mainly relevant for pulse
durations or dwell times well exceeding 1 /Ls. Nevertheless, most of the
discussion which follows still applies even in the presence of vapor ioniza-
tion. although the connection between incident irradiance and absorbed heat
flux then becomes more complicated.

125
 While it is understood that evaporation usually occurs from a liquid, we
shall neglect the solid-liquid transition in the following. This is justified
since the latent heat of melting is but a small fraction of the total heat con-
tent of a material at the evaporation temperature, as apparent from the ent-
halpy values listed in Table A.4. The heat-flow problem, as far as the dense
phase is concerned, is thus reduced to that of Sect. 3.1, but there is an
important new aspect: evaporation cools the surface. Every particle escap-
ing from the dense phase carries along its binding energy (given by dH lv /
NA ) plus some excess energy. On the other hand, as we shall discuss later
on, there is also some back flow of particles from the vapor by which the
cooling effect is reduced. If we neglect the excess energy as well as the
backflow for the moment, the heat flux carried away by the vapor is given
by jv dH lv ' A simple way to allow for this heat flux is to regard it as due to a
"surface heat sink", to be treated in the same way as a heat source. Simi-
larly, boiling could be described in terms of a "volume sink". However,
volume evaporation causes strong expansion, not readily treatable by the
formalism of Sect. 3.1. We shall limit this discussion to surface evaporation.
It is convenient to distinguish, as in Sect. 3.1, between surface heating
(surface source) and volume heating (penetrating source). The first ap-
proach was shown to be adequate whenever the absorption length a-I is
small compared to the width 2(Kt) 1/2 of the heated layer. This generally
holds in metals as well as in nonmetals after breakdown. If heating is
described in terms of a surface source, cooling by surface evaporation is
allowed for by taking the net source flux as the difference

(5.18)

During the initial stages of beam heating, the temperature profile is de-
scribed by (3.8). Once the surface approaches T 1v ' material begins to evapo-
rate and the absorbing surface - and hence the surface source - starts to
move into the material with a velocity given by (5.7). For constant irradi-
ance a steady state with a constant surface velocity will eventually be estab-
lished. The temperature distribution inside the condensed material is then
given by (3.21), with la replaced by CP. The regime of steady-state evapora-
tion by an extended source is adiabatic in the sense that the heat conducted
away from the surface is not lost but eventually recovered by the advancing
interface. In this regime the volumetric energy content of the blown-off ma-
terial per unit time equals the absorbed irradiance, i.e., the steady-state in-
terface velocity must satisfy

Vd
U = I a dH(T d) + dH (5.19)
lv

126
 3

copper

~73-------10~4-------1075-------10~6-------10~7-------10~8------~

absorbed flux (W I cm2 )

Fig.5.3. Reduced surface temperature T/T lv during adiabatic evaporation as a func-
tion of absorbed irradiance for a Cu target. The dashed line is calculated from (5.20);
the solid line also allows for backflow and excess energy of the vapor, according to
(5.24) and (5.25)

where Td is the surface temperature. The surface temperature in the adia-

batic regime can be calculated by equating (5.7) to (5.19), which gives

(5.20)

This shows that, as a result of the heat loss through evaporation, the depend-
ence of temperature on irradiance is essentially logarithmic, rather than
linear. The dashed line in Fig.5.3 gives the surface temperature as a func-
tion of Ia calculated from (5.20), using parameter values for copper (the
solid line corresponds to a refined model discussed below).
The time required for establishing the steady state may be estimated
from the time it takes the distribution (3.8) to reach the width Klu of (3.21).
This gives a build-up time for constant Ia of approximately Klu2. In the typi-
cal metal considered at the beginning of this section, u would be 2 mm/s for
Ia = 10 kWI cm2 , and 20 cml s for 1 MWI cm2 . The steady state would be
reached after about 1 ms in the latter case, but only after as long as 10 s in
the former. Evaporation at small irradiances is thus, in practice, always far
from adiabatic, and most of the laser energy is spent in heating the solid,
rather than in evaporating it. Adiabatic evaporation also requires that the
heat flow is one-dimensional. This can be considered to hold if the width

127
Klu of the heated layer is small compared to the beam diameter. Thus evapo-
ration by the 1 MWI cm2 beam will be close to adiabatic if its diameter is at
least 1 mm. However, under conditions of fully developed evaporation, ma-
terial is often found to be extracted by the beam much faster than expected
from (5.19), due to ejection of melt, as will be discussed in Sect.5.2.3. Less
energy is then invested in the blown-off material. and the adiabatic regime
is reached earlier than the above estimates indicate.
The case of a penetrating source is more difficult to handle. If a steady
state is ever reached, the surface velocity still obeys (5.19), but the relation
between Ia and Td is different from that for the surface source. Since the
laser now heats a volume whereas evaporation only cools the surface, tem-
perature maxima inside the material may develop. Approximate expressions
for the temperature distributions have been obtained by Dabby and Paek
[5.19], who found that the amount of superheating inside the dense phase
(the surface was taken at T\v) increases with the absorption length and with
the ratio dH\v I dH(T\v). Such superheating beneath the surface favors
volume evaporation and may result in explosive removal of the superheated
region before it is fully vaporized. The surface then recedes in jumps rather
than continuously. Expression (5.19) still holds on the average, although
again with an enthalpy value on the RHS smaller than that for pure evapora-
tion.
We have neglected heat losses by thermal radiation in the above ana-
lysis. The thermal flux emitted by a hot surface at temperature T is given by

cl>rad = EUS B (T - To)4 (5.21)

where E < 1 is the emissivity of the surface, and To is the ambient tempera-
ture. Radiation losses by the condensed material can be allowed for in the
same way as surface cooling by evaporation, but they are usually insignifi-
cant. At T ::::: 3000 K, cl>rad is less than 300 WI cm2 , negligible compared to
laser irradiances usually used in the evaporation regime.
The above analysis of heat flow in the condensed phase holds for pure
near-equilibrium evaporation. However, evaporation by laser beams is usu-
ally neither pure - part of the blow-off may consist of melt - nor at near-
equilibrium. To get a refined picture we must next consider some properties
of the evolving vapor.

5.2.2 Vapor Expansion and Recoil

The vapor particles escaping from a hot surface have a Maxwellian velocity
distribution corresponding to the surface temperature, but their velocity
vectors all point away from the surface. This anisotropic velocity distribu-
tion is transformed into an isotropic one by collisions among the vapor par-
ticles. This happens within a few mean-free paths (typically of the order of

128
a few /Lm) from the surface, a region known as the Knudsen layer. Some of
the particles experience large-angle collisions and are scattered back to the
surface. Beyond the Knudsen layer the vapor has reached a new internal
equilibrium with a temperature different from the surface temperature. To
obtain the properties of the vapor we treat the surface as a discontinuity and
apply the conservation laws (5.10-12), with subscripts d denoting properties
of the liquid surface and subscripts v those of the vapor just beyond the
Knudsen layer. Furthermore, we treat the vapor as an ideal gas according to
(5.13) and take the expansion velocity Vv equal to the local sound velocity,
i.e.,

(5.22)

where 'Y is the adiabatic index of the vapor. Thus we have 5 equations by
which all properties of the vapor can be expressed in terms of, say, the sur-
face temperature Td. This calculation was performed by Anisimov [5.20]
who, taking 'Y = 5/3, A = 1 and recognizing that u « vv' found the fol-
lowing result

(5.23)

where p (T) = 15 (T)M/RT denotes the saturated vapor density at tempera-

ture T. This vapor is thus significantly cooler and less dense than the vapor
in equilibrium with the surface. An analysis of the velocity redistribution in
the Knudsen layer gave, under the same assumptions, that about 18% of all
evaporating particles return to the surface. The net evaporation rate is thus
somewhat smaller than the equilibrium value (5.6)

(5.24)

The effective energy invested in the vapor exceeds the latent heat by the
enthalpy and the kinetic energy of the vapor, or .:lH1: = .:lHlv + Hv +
Mvv2/2. Using (5.9) we find that

(5.25)

The solid line in Fig.5.3 represents the surface temperature in adiabatic

evaporation with allowance for backflow, and with .:lHlv replaced by the
effective value .:lHI:. Obviously the approximations made in (5.20) intro-
duce only small errors as far as the surface temperature is concerned.
The simplest assumption about the further fate of the evolving vapor
beyond the Knudsen layer is that of an adiabatic gas expansion. The pres-
sure and temperature rapidly decrease away from the surface while the par-

129
ticle velocity increases [5.9]. If the vapor pressure significantly exceeds the
ambient pressure, the flow velocity eventually becomes supersonic, while
the pressure in the flow falls below the ambient pressure [5.21]. The super-
sonic low-pressure gas flow transforms into a subsonic flow at ambient
pressure in a shock front which forms at a distance typically of the order of
1 cm from the target. Under stationary conditions (pulse duration long com-
pared to the characteristic times of vapor expansion), the shock front is
immobile with respect to the target and can be observed [5.22,21]. How-
ever, it follows from the above numerical results that the vapor is supersatu-
rated, i.e., that Pv < P(Tv)' whenever the surface temperature Td is smaller
than about dH\v /3R, as is always the case in the present regime. Some of
the vapor will therefore condense upon further expansion until a state of
saturation is reached, and hence the expansion is not strictly adiabatic. A
detailed discussion of vapor expansion and condensation can be found in the
book by Anisimov et al. [5.24].
The dense phase is subject to recoil forces excerted by the evolving
vapor. The actual pressure Pd is found from (5.23,24) to be about half the
saturated vapor pressure

(5.26)

The total force on the irradiated body is the integral of Pd over the irradi-
ated area, while the momentum is the time integral of the force. For a pulse
of duration tp the total recoil momentum can be estimated from

(5.27)

where w is the beam radius. Incidentally, this momentum exceeds that by

the radiation pressure, equal to F(1 +R)tp / c, by many orders of magnitude.
The recoil momentum can be measured, e.g., by utilizing a target in
the form of a small ballistic pendulum. Such measurements tend to reveal an
optimum level of irradiance, characteristic of the material, where the mo-
mentum transfer per unit laser pulse energy is largest [5.2,25]. That there
should be an optimum is not surprising: At too Iowan irradiance a regime
of adiabatic evaporation is never reached, and most of the pulse energy is
spent in heating the condensed phase. As the irradiance increases, the sur-
face temperature and the molar energy of the vapor also increase, reducing
its efficiency in imparting momentum to the target (a larger number of less
energetic particles carry more momentum than a lesser number of more
energetic ones). In addition, more energy also goes into internal degrees of
freedom of the vapor particles (dissociation, ionization) which do not con-
tribute to the recoil. However, the dependence of momentum transfer on in-
cident irradiance tends to reflect the behavior of the degree of absorption
more than evaporation kinetics per se, as we shall discuss in the next sec-

130
tion. We shall also consider further - and even more violent - recoil effects
in Sects.5.3,4.
The primary target of the vapor recoil is, of course, the evaporating
melt layer. The interaction affects the dynamics of evaporation and its
response to transients in the irradiation [5.26]. Net forces onto the melt
layer result only from pressure gradients, but lateral pressure gradients are
always present, if only because a beam has a finite cross section. These
forces invariably result in lateral displacement of liquid, which may range
from mere changes in the surface topography of the irradiated material
(Sect. 2.3) to the formation of holes and deep welds.

5.2.3 Drilling, Welding, Cutting

Hole drilling is done with stationary beams, while welding and cutting con-
sists essentially in moving a hole across the material by a scanned beam.
Related techniques are turning and milling, in which the laser beam replaces
the cutting tools in a lathe or a milling machine. Technical details and im-
plementation data on a variety of laser machining processes have been given
in [5.27]. Here we shall focus on the basic physical processes.

Hole Drilling
Let us consider hole drilling first. Drilling by pure evaporation is observed
in sublimating materials or in metals at low irradiance. At higher irradi-
ances, drilling velocities are often found to be much higher (perhaps by
factors of 2 to 5) than can be accounted for by evaporation alone. The
reason is that much of the material extracted leaves the hole as melt rather
than as vapor. Surface-tension forces play a role in laser drilling of thin
films for optical recording (Sect.2.3.2) [5.28], while melt extraction from
deeper holes in bulk metals relies on the evaporation pressure. A simple
model to help visualize this process is to think of the evaporating surface as
a piston that exerts a pressure Pd onto the melt, squirting it out of the hole
radially (Fig. 5.4). The thickness of the melt layer can be estimated from
(3.21) as

(5.28)

In the following we treat the melt as a nonviscous, incompressible fluid. If,

in a rough approximation, we replace the true pressure distribution by a
"top hat" profile with a pressure Pd inside the beam radius wand ambient
pressure Po outside, then the radial velocity of the melt follows from the
volume work

131
Fig.5.4. "Piston" mechanism of melt ejection by the evaporation recoil pressure .
Small arrows symbolize the evaporation pressure distribution

(S.29)

where .1p == Pd -Po is the working overpressure. The liquid escapes through
an opening area of the order of 21rwo m along the circumference of the
piston . The rate of melt ejection per unit beam area is found to be [S .29]

(S.30)

where the rate of material removal by evaporation alone has been neglected.
The inverse relationship between the melt ejection rate and the beam radius
makes the "piston " mechanism mainly relevant for focussed radiation .
There is a threshold for melt ejection, determined by the surface tension of
the melt , which must be overcome by the radial pressure force . Liquid
metals have surface tens ions of the order of 1 J/m 2 which gives threshold
overpressures of the order of 10 4 Pa :::: 0 .1 atm for w :::: 0 .1 mm. However,
for short pulses (1 fts or less) the minimum radial melt velocity required to
carry the melt out of the beam spot while the pulse lasts (wit ) results in a
significantly higher threshold than surface tension . Initial phases of melt
ejection by a focused Gaussian Nd-Iaser pulse incident on a copper target
are shown in Fig.S.Sa : On the smooth molten surface of the first frame the
evaporation pressure creates a shallow dimple which quickly develops into a
doughnut-shaped billow expanding radially. The maximum radial velocity
reached is close to SO mi s, corresponding, according to (S.29) , to a .1p of
about 100 atm (the irradiance used here is, in fact, far above threshold, and
the formation of a surface plasma, to be discussed in the next section, oc-
curs between the third and the fourth frame of Fig.S .5a).

132
 (a)

Fig.S.S. Illustration of melt ejection and hole drilling by 100 JLm diameter Nd-Iaser
pulses of rectangular temporal shape and variable duration: (a) early stages of melt
ejection at 140 MW/cm 2 in copper, after 330, 380, 520 and 720 ns; (b) hole profiles
in steel at 20 MW / cm 2 , irradiated for 0.5, I, 2, 5and 10 JLS (always from left to right)

Melt ejection results in a much larger velocity of the absorbing boun-

dary than evaporation alone. The total velocity is the sum

(5.31)

The melt fraction in the total mass extracted by a focused laser beam in a
metal can exceed 90% at irradiances close to threshold. The effective ener-
gy content of the extracted material (:: Ia ijm) then falls substantially below
the value :: [LlH(T 1v ) + LlH1v ] for pure evaporation, and approaches the
heat content of the melt. At higher irradiance the efficiency of the process
gradually decreases, due to a smaller specific recoil as well as a smaller melt
thickness Gv scales linearly with the vapor pressure, jm only with its fourth
root) . A series of micrographs illustrating hole drilling by a focused Nd-
laser pulse are shown in Fig . 5.5b. Note the "washed-out" appearance of the
hole profiles, indicative of the action of laminar melt flow along the hole
walls . Once the drilled hole grows deeper than about its radius, the irradi-
ance required to sustain a steady flow of melt increases somewhat because
of heat lost to the walls. The melt fraction decreases accordingly.

133
 Melt can also be ejected from the irradiated zone by violent boiling.
Boiling requires significant superheating of the melt, as well as pulses long
enough to allow bubble nucleation. Melt ejection by boiling does not depend
on focused radiation, but requires materials with a significant absorption
length. The amount of melt ejected clearly depends on the superheating
reached and thus on the heating rate, but quantitative estimates are difficult.
The purity of the material and its content of dissolved gases would be
expected to have a strong influence on the nucleation kinetics. Melt frac-
tions of 60 -:-80 % of the total mass ejected have been observed in electrically
exploded metal wires [5.30]. However, metals, due to their short absorption
length, are unlikely candidates for laser-induced vapor explosions. Explo-
sive ejection of material by laser irradiation was reported in the case of cer-
amics by Gagliano and Paek [5.31]. Using irradiances of a few 10 7 WI
cm2 , they found the amount of material ejection to vary strongly with irra-
diance and observed material ejection to continue even after the end of the
pulse. Hole drilling by the explosive mechanism is, in general, more diffi-
cult to control than drilling by melt ejection, and the holes tend to be less
regularly shaped [5.32]. Ionization of the vapor may occur before the nucle-
ation temperature is reached in the liquid. The associated increase in pres-
sure tends to suppress boiling by reducing the amount of superheating in the
melt. Drilling then occurs by normal surface evaporation. Even highly tran-
sparent materials like sapphire can be drilled very well by intense pulses that
cause optical breakdown of the material near the beginning of the pulse
[5.33].

Penetration Welding
Let us now consider penetration welding. It is typically done by means of
CW CO2 lasers with powers of the order of several kW. Welding depths in
steel can reach 1 cm or more, depending on beam power and scan speed,
and are obviously unrelated to the light absorption length. Evidence shows
that the focused beam forms a narrow cylindrical vapor cavity in the melt
which "traps" the beam and moves through the workpiece as the beam is
scanned. Figure 5.6 exhibts a schematic of the vapor cavity and the associ-
ated melt volume in a homogeneous slab of metal. The actual welding oc-
curs in the solidifying melt puddle trailing the vapor cavity. The cavity
formed by a beam scanned at constant velocity can be regarded as a hole at
steady state: while for a stationary beam the hole grows indefinitely (in
principle), the material extracted by the scanned beam is constantly replen-
ished as the beam moves forward. However, a stable steady state is only
found within certain limits of the parameters beam power, spot radius, and
scan speed.
Quantitative models of the complex heat and mass flows in penetration
welding are available [5.34]. Here we shall follow a somewhat simplified
analytical treatment by Klemens [5.35] which well illustrates the essential

134
 u

Fig.S.6. Schematic of vapor cavity and melt zone during penetration welding with a
scanned high-power beam

features. He took the surface of the cavity to be at T lv and treated the melt
flow in a horizontal plane. The width om of the melt layer in front of the
cavity (Fig. 5 .6) is estimated by requiring that the heat flux across the layer,
K(Tlv-Tsl)/om' heats and melts material at exactly the rate u[LlH(T sI )+
LlHsl ]/V required for propagation at the scan speed u. This gives

(5.32)

The melt is forced to flow around the cavity by the evaporation pressure at
its front. Klemens assumed that a fraction (3 of the material intercepted by
the cavity is evaporated, hence

(5.33)

The momentum transported by the flow of vapor creates the excess pressure

(5.34)

The melt again acquires a velocity related to the overpressure by urn =

(2Llp/ Pd)1I2. The melt volume flowing around the cavity on each side per
unit time and unit cavity length, urn om' is a fraction (1-(3) of the total
volume, uw, swept out by the corresponding half-cavity of radius w. Com-
bining this requirement with (5.32-34) yields

_ ~ T lv -LlHlv/c pV""'--
(3 - T -T P) 2Pd (5.35)
K Iv sl

135
 PENETRA TION VS POWER IN 304 STAINLE SS STEEL

I I I I I

.........
.............. ..-
1.6
..-
1.4 ///
// 1.3em/s
E
~
1.2
,,1'/.
/
/
"'"
1.0
Z
Q ,. /
/

,./
...... ,,4.3em/s

--
f-
<{
0:: 0.8
/
/
..- -
..-"-

,, .....-----
f-
w / ____
,-1/

-- --
8.~~

z 0.61-
W ~ /
a..
/ ..-
0.4r / .......... ~ 12.7 em/s -
/ ..-"-- /

0.21- ....- ..... ..- ~

-
0 I y'1 I I I I I
0 2 4 6 8 10 12 14 16 18

POWER (KW)

Fig.5.7. Weld penetration in 304 stainless steel as a function of CO 2 -laser power at

various scan speeds [5.36]

Since p) Pd is of the order of 10-4 near T 1v ' {3 is small. Choosing typical

parameters (steel, w = 1 mm, u = 3 cm/s), Klemens found {3 = 0.02, ~p =
16 Pa and urn = 4 m/s. The absorbed beam power to create a stable 1 cm
long cavity at this scanning speed is obtained as about 4 kW, with 3 kW
spent on melting and only some 100 W on evaporation (the remainder is lost
through heat conduction into the solid). Given that the cavity can be ex-
pected to trap most of the incident beam power, the process is quite effi-
cient.
In reality the processes during high-power laser penetration welding
are quite a bit more complex that the above analysis suggests. The pressure
inside the cavity varies over its cross section as well as over its depth. Kle-
mens showed that, as a result of vapor viscosity and surface tension, actual
cavity shapes are cone-like with constricted exit holes, rather than cylindri-
cal. Moreover, the hot vapor inside the cavity becomes partially ionized and
absorbing as the beam power increases. This, together with the increasing
vapor fraction {3, causes the penetration depth to saturate as a function of
beam power. This is reflected in Fig.5. 7, which shows measured penetra-
tion depths in steel as a function of beam power for various scan speeds
[5.36].

136
Cutting
If the length of the vapor cavity reaches or exceeds the thickness of the irra-
diated metal slab, it can be used for cutting. However, cutting sheet metal
by laser beams is usually - and far more efficiently - done with a jet of gas
assisting the laser [5.36]. The gas jet, rather than evaporation pressure, pro-
vides the momentum to expel the melt, so the laser must only melt the mate-
rial. Gas flows with Mach numbers around 0.2 appear to be most efficient
in removing the melt [5.37]. In the case of steel and other reactive metals
oxygen gas is often used. Exothermic oxidation reactions then contribute to
the power available for cutting [5.38]. Although the gas flow also cools the
interaction zone, the cutting speed for a given laser power can be doubled
by an oxygen jet. Sheet metals can be cut at acceptable speeds with beam
powers as small as 200 W. Apart from metals, a large variety of industrial
materials, ranging from ceramics to leather, are nowadays routinely and
successfully cut with scanned CO2 -laser beams.

5.3 Absorption Waves

We shall now move on to irradiances at which the interaction of the beam

with the evolving vapor plays a dominant role. In this regime, the vapor be-
comes ionized and absorbs part or all of the incident radiation. The energy
is converted into internal energy of the plasma, radiated away as thermal
radiation or consumed in hydrodynamic motion. The properties of the eva-
porating surface become less important and may be treated as mere boun-
dary conditions for the processes in the hot ionized vapor.
Since the density and the temperature of the vapor decrease away from
the target, the plasma is most likely to form close to the evaporating sur-
face. The temperature and the degree of optical thickness reached in the
vapor depend on the incident irradiance. The irradiance received by the con-
densed surface is accordingly reduced. Once an absorbing gas plasma has
formed, an intriguing phenomenon is observed: The plasma cloud rapidly
expands away from the surface. but remains essentially confined to the light
channel formed by the beam. This has little to do with ordinary gas expan-
sion, but results from the dynamics of interaction of the light with the
plasma. The propagating plasma is generally referred to as a Laser-Sup-
ported Absorption Wave (LSAW). Figure 5.8 exhibits a series of high-speed
photographs illustrating ignition and expansion of LSAWs upon irradiation
of (a) an aluminium and (b) an alumina target with CO2 -laser pulses [5.39].
The dynamics of formation and propagation of such LSAWs significantly
influences the interaction of the beam with the target.

137
(a)

ro
[m

Fig 5.8. See caption on the opposite page

LSA Ws come in widely varied shapes and features, depending on

experimental conditions such as irradiance. wavelength, focal-spot size,
target material and ambient atmosphere. Although, in practice, the distinc-
tion is not always easy to make. LSA Ws can be roughly divided into two
classes, depending on whether they propagate at subsonic or at supersonic
speed with respect to the gas (vapor or ambient atmosphere). The two clas-
ses further differ in the range of irradiance in which they are typically ob-
served, as well as in the optical density and internal energy reached by the
plasma . We shall first consider the weakly absorbing subsonic variety of
which Fig.5 .S displays two examples. Such plasmas are generally known as
Laser-Supported Combustion Waves (LSCWs) .

5.3.1 Laser-Supported Combustion Waves

As the name suggests, there is a basic similarity between a LSCW and a

chemical combustion wave which propagates in a combustible gas mixture .
The source of energy driving the latter is an exothermic chemical reaction
with a reaction rate that typically follows an Arrhenius law. The chemical
combustion wave propagates as layers of cold gas adjacent to the hot reac-
tion zone are heated by conduction or thermal radiation until they start
burning and producing heat themselves . In the LSCW the role of the chemi-
cal reaction is played by light absorption, which, according to (5.15), also
behaves in an Arrhenius-like manner . Propagation is limited to the channel

138
 (b)

o 0 .32 0 .95 1.66 1.74 1.78 1.82

~
em

1.86 2.14 2 .42 2.73 2.85 2.89 2.93

2 .97 3.01 3.05 3.09 3. 17 3.29 3.4 1

Fig.S .8. High-speed camera frames of LSAW 's developed from (a) 2024 aluminiulll
alloy and (b) alumina targets, irradiated by 5 ms . 1.5 MW / cm 2 CO 2 -laser pulses
(incident from above). Several decoupling and reignition events can be seen. Numbers
below frames give the time in ms [5.39J

formed by the laser beam, but it relies on the same mechanisms of heat
transport as the chemical combustion. In conlrast to the supersonic propaga-
tion modes to be considered later, an LSCW tends LO be optically thin to the
laser radiation, i.e., it absorbs only a fraction of the beam flux. For this
reason LSCWs are often observed to propagate away from the point of igni-
tion in two directions simultaneously - towards as well as away from the

139
 a)

I. d

b)
Fig.S.9. (a) Flow lines (arrows) and isothermal contours of a laser-supported combus-
tion wave; (b) schematic of gas heating in the LSCW

laser beam - provided no target blocks the latter path. Otherwise a plasma
stationary with respect to the target is formed .
The theory of LSCW propagation was formulated by Raizer [5.40] and
has since been refined by a number of workers. The basic structure of an
LSCW is shown in Fig.5.9a. In the wave's own frame of reference, cold gas
enters at the front , is heated by conduction and light absorption and leaves
the wave at the rear end as well as laterally. The gas within the wave gains
energy from the laser beam and loses energy by radiation and heat conduc-
tion to the surrounding gas . Mathematical modelling of the LSCW is com-
plicated because the problem is three-dimensionaL and because the absorp-
tion and radiation properties of weakly absorbing plasmas depend on the gas
composition, the temperature and the wavelength in a rather detailed fash-
ion . Let us consider here a simplified one-dimensional problem , treated by
Boni and Su [5.41], which well illustrates the qualitative features of
LSCWs.
We regard the LSCW as a disk of thickness d, within which gas is
heated from a temperature To to T 1 (Fig. 5. 9b). The energy balance of the
wave may be written as

(5.36)

140
where a is the absorption coefficient at the laser frequency, Jloss is the volu-
metric energy loss rate of the plasma due to conduction and thermal radia-
tion, and Keff is an effective thermal conductivity allowing for diffusive as
well as radiative energy transport

(5.37)

For radiative transport to be treated in terms of a conductivity, the

medium must be optically thick, i.e., its linear dimensions must exceed the
absorption length for the important part of its thermal radiation. Further-
more, there must be local equilibrium between the medium and the radia-
tion. The blackbody radiation density at any point is then characteristic of
the local temperature. This is the case provided the temperature varies negli-
gibly over a distance of one radiation mean-free path [5.10]. Under these
conditions the radiative conductivity is

(5.38)

where RR is an averaged radiation mean free path known as the Rossland

mean-free path,

(5.39)

The weighting function inside the integral is maximum at flw ::::: 4kT, indi-
cating that the high-energy photons are dominant in the energy transfer
process. The condition of thermal equilibrium is usually satisfied for
LSCWs, whereas the condition of optical thickness only holds for the
vacuum-ultraviolet part of the thermal emission. In their model, Bani and
Su approximated the true absorption spectrum of the plasma by a two-band
spectrum with a large constant absorption coefficient above some limiting
photon energy 00.9 eV for air) and a negligible one below. The high-fre-
quency part of the thermal radiation thus contributes to the energy trans-
port. The low-frequency part which dominates at the temperatures typical
for LSCWs (of the order of 1 -;.. 2 eV), has an absorption length exceeding
typical plasma dimensions and is treated as a loss. The mass of gas heated in
the wave to the final temperature T 1 is now given by

(5.40)

141
where Uo is the propagation velocity of the LSCW. Eliminating the wave
thickness d from (5.36,40) yields for the wave velocity

(5.41)

The LSCW velocity thus scales essentially with the square root of the irradi-
ance and vanishes at a critical irradiance equal to J loss I IX, at which the laser
beam just balances the energy loss rate. The LSCW then becomes station-
ary with respect to the surrounding gas [5.42]. Such stationary plasmas,
known as plasmotrons, have been sustained (after spark ignition) with CW
CO 2 lasers at powers below 100 W in Ar or Xe gas at one atmosphere
[5.43]. At irradiances well above the critical value, LSCW velocities in the
range of a few 10 to a few 100 m/s are found to be typical [5.44].
The apparent simplicity of (5.41) is delusive: A realistic evaluation of
K eff and J loss, which both depend sensitively on the nature and temperature
of the gas as well as on the geometry of the problem, is quite complicated.
Allowance for just two frequency bands is, of course, an oversimplifica-
tion. To obtain more than just qualitative agreement with experiment, a
numerical treatment of the hydrodynamic and radiation-transport problems
is required. In the case of LSCWs formed in a target vapor, one must also
allow for the fact that the wave propagates in a medium which is already hot
and in motion [5.45]. The range of irradiance where the LSCW model ap-
plies extends, roughly speaking, from tens of kW I cm2 to tens of MW I cm2 ,
and covers the range where intense evaporation of targets occurs within the
pulse duration. However, these fluxes are too low to ionize cold gases and
hence vapor-LSCWs, as a rule, do not propagate beyond the region oc-
cupied by hot vapor.
Let us now return to the evaporating target and consider how LSCWs
interfere with the beam-material interaction. One might think of an LSCW
which extracts energy from the beam and dissipates it by expansion or radi-
ation into 411", as a mere parasite in the beam-target interaction. Experience
shows. however, that at least under certain circumstances LSCWs can actu-
ally enhance the amount of beam energy deposited in a solid.

5.3.2 Plasma-Enhanced Coupling

As mentioned in Sect.2.3.3, it is well established that the fraction of beam

flux absorbed in metals increases above a certain threshold irradiance. The
increase tends to be particularly large (up to a factor of ten or more) for in-
frared lasers which are poorly absorbed by cold metals. There is clear evi-

142
dence that such an effect can be produced by formation of an LSCW in the
evolving vapor.
How does this phenomenon of plasma-enhanced coupling work? A typ-
ical sequence of events might be as follows. First, material is heated to the
point of evaporation by normal absorption. The vapor close to the surface is
partially ionized and begins itself to absorb significantly [5 .13]. A station-
ary plasma layer close to the evaporating surface forms, often coincident
with initiation of a propagating LSCW moving up-beam. The temperature
of the near-surface plasma is typically in the range of 1-:-3 eV [5.46]. Ener-
gy from the plasma can now be transferred to the dense phase by any of
three mechanisms : (i) normal electron heat conduction, (ii) short-wave-
length thermal plasma radiation which is efficiently absorbed by the metal
surface, and (iii) condensation of vapor forced back to the surface by the
plasma pressure [5.47]. These mechanisms provide an additional heat flux to
the dense material, which mayor may not exceed the loss of light flux due
to plasma absorption. One would speak of plasma-enhanced coupling in the
former and plasma shielding in the latter case. Most experiments show that
the enhanced coupling phenomenon is strongest right at its threshold and
decreases at higher irradiance [5.48], suggesting that the plasma tends to
decouple from the surface. The energy transferred from the plasma to the
target is often found to be distributed over an area significantly larger than
that of the optical beam spot, due to lateral expansion of the plasma [5.49].
Let us consider a slightly simple-minded model of the process,
sketched in Fig. 5.10, in order to illustrate some of its features (a somewhat
more complete model, yielding almost identical conclusions, has been
discussed by Nielsen [5.50)). The plasma is represented as a disc of thick-
ness d. The dense target absorbs a fraction (1- R) of the irradiance I)
transmitted through the plasma. A fraction (3 of the light flux (I-I) ab-

Fig.S.IO. Schematic of a partially absorbing surface plasma with energy transfer to

the dense phase

143
sorbed by the plasma is assumed to be transferred to the target by any of the
mechanisms mentioned above. The total flux deposited in the target is writ-
ten as

(5.42)

The left-hand identity serves to define the effective beam-target coupling

coefficient A, whereas the right-hand equation expresses energy conserva-
tion at the target-plasma interface. The amount of absorption in the plasma
depends on its density and its temperature. Of these, the density varies
strongly with the flux absorbed in the target, whereas the temperature does
not. In a weakly ionized plasma we have a ex Pv ' (5.15), whereas in a fully
ionized one, which we shall assume here, we have a ex Pv 2 , (5.16). The
total absorption in the plasma will also depend somewhat on the kinetic en-
ergy of the vapor which affects the length over which the density decreases,
but the dependence is not strong and we just neglect it here. Thus we can
write for the flux transmitted to the target

(5.43)

where, for a given laser wavelength, Bo is a constant; according to (5.16),

Bo scales like A2. Inserting (5.43) into (5.42) yields an implicit equation for
A, i.e.,

-B A2
A-(l-R-J3)e ( -13 =0 (5.44)

where B( = Bo (I1vv ~Hlv *)2. Figure 5.11 plots the effective coupling coeffi-
cient calculated from (5.44) as a function of B( ex F for (l-R) = 0.05 and
for various values of 13. Note that there is an abrupt threshold for plasma-
enhanced coupling if 13 is large (even though no threshold is built into the
assumptions, the vapor always being taken to be fully ionized), while the
transition becomes gradual at smaller 13. The transition in all cases occurs
while the plasma is still transmitting most of the laser light. The case 13 = 0
leads to reduced coupling (shielding) at larger irradiance.
Good thermal coupling (large 13) evidently requires the plasma to stay in
close proximity to the surface. Any decoupling (drop in 13) would immedi-
ately reduce the rate of evaporation and hence the density of the vapor and
the absorbing power of the plasma. The close circular relationship between
the plasma-target energy-transfer rate, the evaporation rate and the plasma
density may lead either to a self-balanced steady state (corresponding to a
non-propagating LSCW) or to oscillatory behavior (repeated ignition and
decoupling of propagating LSCWs), always assuming the incident irradi-
ance to be constant. The latter behavior is evident in Fig.5.8 where decou-
144
 "I' I
0.7

0.6
'EQ.o

-
g 0.5
~
0
u
0.5

0.4 0.35
01
.~
0.
::J 0.3
0
u
Q.o
0.2
0.2
~

- ~
u
~ 0.1
~ 0.1
001
I I
0
5 10 20 50 100 200 500
S,
Fig.5.11. Effective coupling coefficient calculated from (5.44) for various values of
the energy transfer factor {3. B( is proportional to the square of the incident irradiance

pling occurs, e.g., in series (a), frames 2.72-;.-4.33 ms, and reignition is visi-
ble at 4.54 ms (there is no correlation of these events with features of the
incident pulse envelope, as given by Stegman et al. [5.39]). Repeated decou-
pIing and reignition is also obvious in series (b) of Fig.5.8. The target sur-
face temperature monitored under interaction showed strong maxima during
periods of coupling and minima during periods of decoupling.
The behavior of the LSCW, as well as the resulting degree of coupling
or shielding, apparently depends on a variety of parameters - the laser flux
and wavelength, the target material as well as the ionization potential and
pressure of the ambient gas [5.39,51]. Large plasma-enhanced beam cou-
pling to metals (up to 30 -;.- 50 % of incident beam power absorbed in the con-
densed phase) have only been reported for infrared laser beams, notably for
Nd lasers [5.52] and for CO2 lasers [5.53].
With further increasing irradiance the temperature, pressure and velo-
city of the absorption wave increase. At the same time the wave becomes
more absorbing and consumes a larger fraction of the beam flux. Compres-
sion of the gas engulfed by the LSAW contributes more and more to pre-
heating and ionization. Eventually compression, rather than heat conduc-
tion, becomes the dominating propagation mechanism. The velocity in this
regime turns out to be supersonic with respect to the gas ahead of the wave,
hence the wave is a shock wave. Waves of this kind are known as Laser-
Supported Detonation Waves (LSDWs).

145
S.3.3 Laser-Supported Detonation Waves

The absorption wave due to a laser-supported detonation is, in contrast to

LSCWs, not confined to the vapor but can form and propagate in cold
gases , too . In fact, LSDWs are typically observed to form ahead of a target
before the latter has even started to evaporate, due to surface-mediated opti-
cal breakdown of the ambient gas , as discussed in Sect. S.l.4. Breakdown
plasmas tend to be strongly absorbing and can, once ignited, readily be sus-
tained and propagated as LSDWs . The name of LSDW alludes to its similar-
ity to a strong chemical explosion which propagates by shock compression
of a combustible mixture . In the LSDW light absorption plays the role of
the chemical reaction . Typical plasma temperatures reached in LSDWs are
of the order of 10 eV, at which most gases are multiply ionized. Absorption
lengths for infrared-laser beams are typically no more than a few tens of
~m, small compared to usual beam dimensions.
This being so, the main characteristics of an LSDW - temperature,
pressure, and propagation velocity - can be found by neglecting its internal
structure altogether. The wave is treated as a mere hydrodynamic discontin-
uity at which the incident light flux is deposited (Fig.S.12). To find its char-
acteristics we apply the conservation laws (5.10-12). Denoting quantities in
front of and behind the wave with subscripts 0 and 1, respectively, and neg-
lecting the pressure and the enthalpy of the undisturbed gas we have

Povo = PI vI ' (S.45)

povo2 = PI + PI vl 2 , (S .46)
vo2 /2 = HI 1M + v l 2 12 - II PI vI (5.47)

Po, Po=o, Ho=o

Fig.5.12. Hydrodynamic variables for calculating the characteristics of a laser-sup-

ported detonation wave. Subscript 0 identifies the undisturbed ambient gas, subscrip 1
the hot gas behind the wave (moving frame of coordinates)

146
Here the laser irradiance is assumed to be fully absorbed at the discontinuity
and thus appears as additional enthalpy of the gas behind the wave. Further,
we take the velocity of the gas behind the shock wave to be at its maximum
possible value - the local sound velocity

(5.48)

Together with the equation of state (5.9) for HI we have 5 equations which
allow us to eliminate all quantities pertaining to the gas behind the wave,
and to express the wave velocity in terms of the irradiance. This yields for
the wave velocity -vo == u [5.54]

(5.49)

Substituting , for illustration, I = 10 l O WI cm 2 = 10 14 WI m2 , Po = 1. 3 kgl

m 3 (air at STP) and')' = 4/3 into (5.49) yields u :::: 50 km/s. The gas tem-
perature can be estimated from the internal energy behind the front

HI Mu2 ')'
VI - - - - - - - - - (5.50)
')' (')'2 - 1)(')' + 1) .

In the above example we get V 11M:::: 1.8 109 J/kg, corresponding to an

equilibrium gas temperature of about 120.000 K. The compression has the
value
(5.51)

which, in our example. is 7/4. Finally, the pressure behind the wave is
found to be

= Pa u2 (5.52)
PI ')' +1

or approximately 1.4-109 Pa :::: 14000 atm. The LSDW, as these numbers

illustrate, is certainly a rather violent phenomenon.
The above calculation neglects lateral expansion of the gas, which re-
presents a power loss to the LSDW and reduces its temperature and velo-
city. According to Raizer [5.54], the loss to lateral gas expansion can be ap-
proximately allowed for by replacing the incident flux I in (5.49) by a
reduced flux 1/(1 +d/w), where d oc a-I is the thickness of the wave and w
is the beam radius.
Apart from the detonation mechanism, other propagation modes for
supersonic LSA Ws have been considered. One of them is the breakdown
wave [5.54] which applies for light fluxes exceeding the breakdown thresh-

147
old in cold gas (without a target). Propagation of the breakdown wave relies
on the time dependence of avalanche breakdown: The avalanche first devel-
ops in the region of highest flux (usually the focal point), and somewhat
later at points of lower flux and longer avalanche build-up time. The velo-
city of this "wave" is inversely proportional to the beam-aperature angle and
hence can be arbitrarly large. Since plasmas in front of absorbing targets
are formed at irradiances much below those required for breakdown in cold
gases, the breakdown wave is of little relevance for laser-target interac-
tions. Yet other possible LSAW propagation modes rely on the far-ultravi-
olet thermal radiation of the hot plasma (flw of the order of tens of eV), for
which the plasma itself is transparent, but which is strongly absorbed in a
cold ambient gas [5.54,55]. The experimental distinction between an
LSDW and the radiation-assisted LSA W modes (which have a far more
complicated mathematical description) is not obvious, however. We shall
henceforth refer to a supersonic LSAW as an LSDW.
Let us now return to the evaporating target and consider how LSDWs
interfere with the beam-material interaction. Their role turns out to be quite
different from that played by LSCWs. Enhanced coupling may be present
initially, but it is quickly replaced by complete shielding of the target as the
strongly absorbing plasma propagates away from the surface.

5.3.4 Effects of LSDWs on the Beam-Material Interaction

Both the coupling and the shielding phases of target-mediated LSDW forma-
tion have been studied mainly in connection with high-power CO2 lasers for
which the threshold is about 107 W/ cm2 . The period of enhanced coupling
after ignition of an LSDW lasts typically less than one {J.S and is dominated
by mechanical rather than thermal energy transfer.
The LSDW, as shown above, creates a pressure of hundreds or thous-
ands of atmospheres. The gas expands behind the wave and the particle vel-
ocity must vanish on the surface of the target. Assuming the expansion to
be isentropic and one-dimensional, Pirri [5.56] obtained for the actual pres-
sure acting on the target surface

(5.53)

where PI is given by (5.52). This should be valid until the LSDW has moved
away from the target further than about its diameter. The pressure increases
with increasing irradiance, but the time during which the pressure acts on
the target decreases because the LSDW moves faster. The total momentum,
in the one-dimensional limit, turns out to be proportional to the total pulse
energy and independent of irradiance (provided the latter is well above the
threshold for LSDW ignition). A pictorial impression of this regime is given

148
 -
~ 0.4
E velocity scal.
~
(a)
"- 0.2 5·1()5 em/s

'ltlJWfr
l~!tlnfff~
~
0.4 'h;.
E
u
v.2 (b)

,~,

..........,', ... , .. t , .."...............

,"'~~

,.."
,._

0.6

HIP
,d{ i11]1'" ~~H7.~ •

U.~llllffr~
(c)
E
u 0.4

02 "","-! ~ X~'\'\ l~
~~~~~"
........
""'"
..
"~~,,-,,,,,,,,\,,~,,,.~
~~ ...... "
,/.,.~
.
, ..-..........
0.2 0.4 0.6 0.6 1.0 0 0.2 0.4 0.6 0.6 1.0
z (em) z (em)
Fig.5.13. Calculated pressure contours and flow fields in a LSDW sustained by a 10
MW/cm 2 CO 2 -laser beam: (a) after 0.66 p.s, (b) after 1.63 p.s, (c) after 2.19 p.s. Pres-
sure contour scale (in 10 6 erg/cm3 ): l:l.8, 2:3.6, 3:7.2, 4:14.4, 5:28.8, 6:57.6, 7:115.
[5.57]

by Fig. 5.13, which shows numerically determined pressure contours and

flow vectors in an LSDW formed at a surface by a 10 MW/cm 2 CO2 -laser
pulse [5.57]. Equations (5.49,52,53) yield P2 :::: 35 atm, in fair agreement
with the surface pressure predicted by Fig.5.13 while the gas motion is one-
dimensional. Direct measurements by means of piezoresistive gauges [5.58]
also confirmed the peak pressures predicted by theory.
LSDW -related pressure pulses produce stress waves in the target mate-
rial which can lead to deformation or fracture. Effects in various materials
have been investigated by a number of researchers [5.59]. The impulses de-
livered to reflecting materials by infrared lasers through LSDW ignition by
far exceed those expected from normal evaporation at the same irradiance
(even stronger impulses are produced in the regimes to be considered in the
next section). However, the period of LSDW-enhanced coupling usually
lasts only for a small fraction of the pulse duration. As soon as the LSDW is
decoupled, the target is completely shielded and receives no further energy
from the laser beam while the LSDW lasts. The fraction of pulse energy
transmitted to the target thus depends on the lifetime of the LSDW.
In focused beams the flux I received by an LSDW moving away from
the focal region decreases in proportion to the inverse square of the aper-
ture angle and the distance travelled by the wave. Eventually the flux inci-

149
 PEAK LASER POWER [kW) Fig.5.14. Lifetime of the LSDW cre-
5 10 20 50 100 200 ated by a TEA CO 2 -laser pulse (500ns
spike, followed by a 4 Ils tail) focused by
a f 15 lens onto a target in various ambi-

~'~
ent gases. The point AB indicates the air-
VI
C breakdown threshold without a target.
+ 0;1) Solid lines were calculated from LSDW
z 500 + + ~cjSI- :A.AB theory. [5.61]
0
i= ~OA"
«
0::
:J
0
200 0 8
0 ~.o-
lJJ
> 100
0
f/
«
~ ~eliUm
--I t f
I
0 50
\J)
..J

20
0.2 0.5 2 5 10
LASER FLUENCE [kJ Icm 2)

dent on the LSDW is insufficient to further sustain its losses, the wave
decays and light is again transmitted to the target. The "lifetime" of the
LSDW thus depends on the aperture angle of the light cone. The minimum
"keep-alive" flux is found to be somewhat smaller than that required to
ignite the wave [5.60,61]. The expanding shock wave leads to rarefaction of
the gas in front of the target which may persist for tens of ,.,.S and which
tends to suppress reignition of subsequent LSDWs [5.23]. The fraction of
pulse energy lost to shielding can be minimized by using large-aperture
focusing optics and ambient gases with a large sound velocity [5.61]. LSDW
lifetimes in various gases as a function of laser fluence are shown in Fig.
5.14. The dependence of the hole-drilling efficiency on pulse duration and
irradiance in this regime has also been discussed by Hamilton and Pashby
[5.62].

5.4 Phenomena at Very High Irradiance

This section is devoted to regimes accessible only with light fluxes exceed-
ing some 109 or 10 10 W/cm2 . The interest in this area has been stimulated
mainly by the prospect of producing plasmas hot and dense enough to achi-
eve thermonuclear fusion. Other potential appliactions include the use of
plasmas as pulsed sources of fast ions or of X rays. Stimulated X-ray emis-
sion at wavelength as short as 4 nm has been realized utilizing sophisticated
geometrical arrangements of high-density laser-produced plasmas [5.63].

150
The regime under consideration, while fascinating, is somewhat peripheral
to the main theme of this monograph, so we will merely sketch the main
ideas, in order to provide, as it were, a glance across the fence delimiting
regimes of current interest in laser material processing.
In the previous section we have been concerned with gaseous plasmas
with an optical density increasing with irradiance. Once matter is fully ion-
ized this trend is reversed since the degree of ionization cannot increase any
further. The absorption coefficient of a fully ionized plasma, given in
(5.16), scales like p2T-3/2, hence a sufficiently hot plasma will be transpar-
ent to the laser radiation. Consequently, radiation is again transmitted to the
dense phase. The temperature at which this happens depends on the atomic
number of the material. Most of the relevant work has been concerned with
the lightest elements, which are of interest in fusion, and where full ioniza-
tion is relatively easy to achieve.
At irradiances typical for the present regime, even nonmetals can be
considered strongly absorbing, since breakdown will occur within a small
fraction of the pulse duration. At the highest irradiances (10 15 W/cm2 or
more) ionization occurs by multiphoton absorption, which becomes equiva-
lent to the tunnel effect, essentially within one light cycle. The specific en-
ergy acquired by the material exceeds the heat of evaporation by many
times, and even the dense phase can be treated as an ideal gas. However,
strong non equilibrium between the electrons and the ions will prevail while
such a pulse lasts.
We are thus confronted with the problem of a dense, strongly absorb-
ing material, in the first few tens of nm of which energy at a rate of per-
haps 1020 W/ cm3 is liberated. Part of this energy, once randomized, is con-
ducted into the bulk of the material, while part is converted into directed
kinetic energy by thermal expansion of the heated layer. It has been custo-
mary to distinguish two regimes in which most experiments have been done:
First, a regime characteristic of ns (Q-switch laser) pulses, which is domi-
nated by the expansion and ablation of material, and second, a regime, char-
acteristic of ps (mode-locked laser) pulses, in which heat conduction domi-
nates, as hydrodynamic motion during the pulse duration is negligible.
Let us first consider the former regime. The main difference to that
considered in Sect.5.2 is that the thermal pressure of the heated layer,
orders of magnitude greater, is now sufficient to cause significant compres-
sion of the underlying target material. In the simplest case we have to deal
with three different zones (Fig.5.15): The undisturbed solid (denoted with
the subscript 0), the compressed layer (subscript 1) and the vapor (subscript
2). We can apply the conservation laws (5.10-12) to the boundaries between
the three regions, but for a full description, as before, additional assump-
tions must be made. Various models have been considered which differ
mainly in the treatment of light absorption in regions 1 and 2. A model ap-
propriate for intermediate fluxes (between some 109 and 10 14 W/cm 2 for

151
Fig.5.15. Three-zone model of irradiated material in the self-regulating plasma
regime. Subscripts 0, 1, 2 identify the undisturbed solid. the compressed layer and the
expanding vapor plasma, respectively. Laboratory coordinates

light targets) is based on the idea. first proposed by Krokhin [5.64], of a

self-regulating plasma.

5.4.1 The Self-Regulating Plasma

The basic idea is simple and has been hinted at in Sect. 5 .3: Absorption in
the gaseous plasma in region 2 will decrease the rate of evaporation and
hence the plasma density. The plasma is thus expected to transmit just
enough radiation to the dense surface in order to sustain itself. A very
simple but intuitively appealing model is that of a "plasma between walls"
[5.64]. The plasma is thought to have a constant thickness d. The plasma
temperature due to absorption of laser radiation (with constant irradiance I)
then increases according to

(5.54)

where ex is the absorption coefficient at the laser wavelength. From (5 .16),

it can be written as

ex -- B2 p 22T 2-3/2 (5.55)

with a constant B2 . The plasma density, in turn, depends on the irradiance

transmitted

aP2 _ Ie-ad
(5.56)
at - dc p T2 .

152
The system (S.S4-S6) has the solution

T2 = (aB2)217 (ItM/cp )417 ;

P2 = (aB22)-117 (ItM/c p d2 )317 (S.S7)

which is easily verified by insertion. The optical thickness of the plasma,

ad, turns out to be equal to In(7/3) ::::< 0.8S, independent of time and irradi-
ance. This means that 317 of the pulse energy is spent in evaporation and
417 in plasma heating. This simple model demonstrates, if nothing else, that
effective self-regulation of the plasma is possible.
A somewhat more realistic model. also based on the idea of self-regula-
tion of the plasma but allowing for shock compression of the solid was dev-
eloped by Caruso and Gratton [S.6S] (Fig.S.1S). The particle velocity in the
plasma region, assumed to contain ions of charge Z only, is taken as the
isothermal sound velocity

v2 = V(1 + Z)RT 2 /M (S.S8)

and the physical thickness of the plasma is d = J~ v2dt. Based on the ab-
sorption law (S.SS), the optical thickness is again found to be constant and
equal to 114. The density and pressure of the plasma turn out to be

(S.S9)

where B3 = B2 [(1 +Z)R/M]3/2. The conservation laws for mass and mom-
entum at the shock front read (in laboratory coordinates)

Pou s =PI(uS -vI)' (S.60)

Po u S2 = PI (us -v l )2 + PI ' (S.61)

where Us is the shock velocity and where the pressure Po of the undisturbed
solid has been neglected. The pressure PI inside the shocked region is ap-
proximately equal to P2' which, in view of its weak time dependence, can be
assumed constant. Solving (S.60, 61) with PI = P2 ;:; P yields

(S.62)

(S.63)

153
 The surface of the solid retreats due to compression of the underlying
material (at velocity VI)' as well as due to evaporation. The velocity due to
evaporation alone is negligible and thus

(5.64)

where on the RHS use has been made of (5.62,63) as well as the fact that P2
« Po :::;;; PI· The energy given to the solid consists essentially of the com-
pression work, =::: pVI' which itself can be shown to be a small fraction (of
the order of (P2 / Po) 112 « 1) of the pulse energy. Nevertheless, this small
fraction is sufficient for ionizing the solid provided

Caruso and Gratton estimated this to be satisfied for I > lOll W/cm2 for a
hydrogen isotope target, for which the model is expected to hold for fluxes
between 109 and 10 14 W/cm2 in the case of ruby-laser light. As an illustra-
tion of this regime, Fig. 5 .16 shows density, velocity and temperature pro-
files from a numerical simulation by Mulser [5.66] (based on a more
detailed model) for a hydrogen target shaped as a thin foil, irradiated by a
ruby-laser pulse. The scaling at shorter wavelengths was discussed by Ng et
al. [5.67]. Propagation of shock waves through solid material and their capa-
bility to produce spalling upon reflection at the back surface of irradiated
slabs has also been investigated [5.68]. Gupta [5.69] used laser-induced spal-
lation to measure quantitatively the adhesive strength of a planar interface
between a substrate and its coating.
If the irradiance is further increased the pressure and density, and
hence the temperature of the plasma increase, according to (5.59). The as-
sociated reduction in the absorption coefficient is balanced by an increase
of the physical length of the plasma only as long as the plasma motion re-
mains one dimensional, i.e., as long as d is smaller than about one focal-
spot diameter. At the same time, the plasma density cannot increase beyond
the cutoff density, at which the plasma turns opaque to the laser beam. The
cutoff density is

(5.65)

where w is the laser frequency. Since the plasma temperature increases with
its density, the maximum achievable temperature is reached when P2 = P* .

154
 ~. J
2
t =0 ns

IP
50 0
-x(~ml

4
t=O 5ns
p T
~2 P
v

2
v t=1·5ns
007cm 5-')

O~--------~~--~~-----------

t =2·0 ns T

0·5

O~------~~==~---------------

Fig.5.16. Calculated density p, particle velocity v and temperature T in a solid

hydrogen foil irradiated by a ruby-laser pulse of 10 12 W/cm 2 (incident from the right).
The shock wave velocity is 2.7 .106 cm/s [5.65]

From (5.58,59) the maximum temperature is found to scale with the 2/3
power of the incident flux

M [I )2/3
RT 2 (max) = 1 + Z p* (5.66)

155
 Po ",0. Po. TO

--- v,

Fig.S.17. Target structure in the deflagration regime . Subscripts 0, I, and 2 have the
same meaning as in Fig.S.IS . uD is the velocity of the deflagration front, Us that of the
shock front (Laboratory coordinates)

Once the plasma is hot enough to be transparent to the laser pulse,

volume absorption in the self-regulating plasma is replaced by sheet absorp-
tion at the interface between regions 1 and 2 in Fig.S.1S, where the density
equals the cutoff density. The resulting structure can be described as a laser
driven deflagration wave .

S.4.2 Laser-Driven Deflagration Wave

A deflagration wave, like a detonation wave, is a discontinuity, but it is a

rarefraction wave which moves at subsonic speed [S.70). It is always pre-
ceded by an ordinary shock wave. The situation can be described by two
sets of conservation laws for the two interfaces separating the regions 0 , I ,
and 2 (Fig.S.17). We have for the shock wave (neglecting Po)

Pou s = PI (us -vI) , (S .67)

Pous 2 =PI +PI(U I -v I )2, (S .68)
HolM + u S 2 12 = HI 1M + (us -V I )2 12 . (S.69)

For the deflagration wave, where the flux I is absorbed, we have

PI (u o -VI) = P2 (uo +v2 ) , (S.70)

PI +PI(UO-v l ) =P2 + P2(U O +v2 )2 , (S.71)
HIIM + (u o -v l )2 12 = H2/M + (uo +v2)2 - II P2 (uo +v2) . (S .72)

Here P2 is taken to equal the cutoff density P* . Assuming further that the
compression by the shock wave is at its maximum possible value , PI I Po =
(),+I)/(),-I) [S .9], neglecting the ionization energy and taking v2 equal to
the local sound velocity, we can express the velocities of the shock wave and
the deflagration wave in terms of the irradiance [S.14]

156
 Us = j("( + 1)(1" -1)
"('
p*
Po
[I" + 1 ~)1/6
1" - 1 2p*2

(5.73)
2
uD = Us "( + 1

where "( and "(' denote the effective adiabatic coefficients of the solid and
the plasma, respectively. If, as an example, a Nd-Iaser pulse of 1 = 10 14
W/cm2 is incident on a deuterium target ("( = 7/5, "(' = 5/3) we have p* =
3.4.10 3 g/ cm3, and the model gives Us :::::: 107 cm/s, uD = 0.83 us. The ion
temperature in the plasma is again found to scale with 12/3 and reaches kT 2
:::::: 3.5 keY in the example.

5.4.3 Inertial Confinement

Let us now consider the situation for pulses in the ps regime where strong
heating of the dense material occurs before hydrodynamic expansion has
even started. The plasmas produced in this regime have essentially the same
density as the solid.
Suppose a powerful pulse with a zero rise-time is switched on at t = 0
and pumps energy into the electrons of the material. The electron-ion colli-
sion time in a fully ionized plasma varies like T e 2/3 and is of the order of
10- 13 s for keY electrons. The electron-ion energy relaxation time is larger
than this by a factor of the order of the ion mass divided by the electron
mass, and hence it is likely to be longer than a mode-locking pulse. The ions
will thus stay essentially cold at first and the only path for energy dissipa-
tion is heat conduction into the bulk material. Heat conduction is as fast as
thermalization, while hydrodynamic motion is much slower - a characteris-
tic time for the latter is the time it takes a sound wave to traverse the heated
layer.
The electronic heat conductivity in a plasma is strongly temperature de-
pendent, K oc T e 5/2 [5.71]. The resulting temperature distribution is a "heat
wave", characterized by a sharply rising leading edge moving into the solid,
followed by a plateau where the temperature is high and nearly uniform.
The coordinate zH of the leading edge of the heat wave at time t after the
pulse is switched on, is given by [5.72]

(5.74)

where I is measured in erg/ cm2 sand Ne in cm- 3 . The average electron tem-
perature behind the heat wave is

(5.75)

157
 While the heat wave develops, energy is fed to the ions and the heated
layer begins to expand. The time during which the heat-wave regime lasts
can be estimated as the time it takes the rarefraction wave, moving at the
sound velocity, to catch up with the heat wave. This happens after a time

(5.76)

where A IS the atomic number. Setting A = 1 and Ne = 5.1022 cm-3 gives,

e.g., for

I [W/cm2] 1012 1014

to [s] 10-15 10-13 10-11 10-9

The above formulas hold for constant irradiance; for pulses much shorter
than to a somewhat different scaling is found [5.73].
We have so far neglected the problem of absorption of that light which
reaches the region of the cutoff density. For a plasma at or near solid-state
density almost total reflection of the light should be expected. Instead,
strong absorption is observed experimentally. This is thought to be due to
nonlinear optical effects in which electromagnetic and electrostatic plasma
waves are excited. These, in turn, cause turbulence in the motion of the
electrons and lead to far more absorption than expected from the normal
electron-ion collisions. The effect has been described in terms of a pheno-
menological effective collision time [5.74]

(5.77)

where 7e is the normal electron-ion collision time and 7p characterizes the

damping of the induced plasma waves. Particularly efficient coupling to
longitudinal electrostatic plasma oscillations is observed if the beam is inci-
dent obliquely to the density gradient created by the hydrodynamic motion.
since there is then a nonzero component of the electric field vector in the
direction of the plasma oscillation. The effect is known as resonance ab-
sorption because there is an optimum incidence angle [5.75]. Damping of
the electrostatic waves tends to produce "hot" (nonthermal) electrons and
ions which decouple from the heat wave and limit the degree of inertial con-
finement achievable. Detailed discussions on these topics can be found in
the literature [5.76].

158
5.5 Pulsed Laser Deposition

Having discussed the formation and evolution of laser plasmas in some

detail, we shall now return to the solid state for this final section, closing
full circle our parade of laser-induced phase transitions.
The basic arrangement in Pulsed Laser Deposition (PLD) is depicted in
Fig. 5 .18. Some target material - usually placed inside a chamber with a
controlled atmosphere - is irradiated obliquely by a focussed high-power
laser pulse . A vapor plume forms and expands perpendicularly away from
the target. Several cm from the impact spot a substrate is placed that inter-
sects the expanding plume. Some of the vapor condenses and forms a de-
posit on the substrate. It is this deposit that we shall now consider.
PLD is basically just a physical vapor deposition scheme in which a
laser happens to furnish the energy . Although the first attempts at PLD
were reported as early as 1965 - as soon as high-power ruby lasers became
available - the technique has attracted widespread interest only since the
late eighties, mainly due to spectacular successes with high-T c supercon-
ductor materials . The list of materials deposited today by PLD includes such
diverse items as optical coatings, compound semiconductors, oxide super-
conductors and bioceramics . Before discussing the PLD process itself, let
us briefly look into the basics of film growth from a vapor .

Fig.5.l8. Experimental arrangement in pulsed laser deposition

159
5.5.1 Solid Film Growth from a Vapor

Vapor condensation is a nucleation-and-growth phenomenon like crystalli-

zation from a melt (Sect.4.l) or evaporation (Sect.S.l). Neglecting the
fleeting liquid that may form from the vapor first, the overall driving force
for condensation of the solids is the free energy difference between solid
and vapor, LlGsv = Gs-Gv ' By means of the Clausius-Clapeyron law this
can be expressed in terms of the pressure p or the molar fluxj, see (S.4, 6):

LlGsv = - RTln[p/p (T)] = - RTln[j/j(T)] (S.78)

In the present context of film condensation from a vapor, p is to be inter-

preted as the actual (local) vapor pressure, T as the substrate temperature, p
as the equilibrium vapor press!!re at temperature T, while j is the actual
(molar) condensation rate and j the reevaporation rate, equal to the equili-
brium condensation rate. We are, of course, interested in the regime where
the vapor is undercooled or supersaturated. (p > p ) and hence LlGsv < O. In
this regime the quantity -LlGsv is referred to as the supersaturation.
The first step of condensation is nucleation. The theory of crystal nuc-
leation from a vapor is quite similar to the one for nucleation from a melt,
as described in Sect. 4 .l. 5. Here, as there, the nucleation rate depends upon
the size of a "critical nucleus" which is determined by a balance between
volume and surface forces (4.14). The difference to nucleation from a melt
is that now the interfacial energies are much larger. Hence, homogeneous
nucleation is more difficult (in fact, almost negligible) and "wetting" effects
lowering the nucleation barrier are crucial. In the present context we are
thus interested in heterogeneously nucleated condensation with the substrate
as the seed.
To describe heterogeneous nucleation, the single interfacial energy (1
used in connection with homogenous nucleation from a melt (Sect.4.l.5)
needs to be replaced by the three interfacial energies (1sm' (1sv and (1mv be-
tween the three phases present - the substrate (subscript s), the condensing
material (m) and the vapor (v). The minimum-energy shape of a nucleus is
no longer a sphere but a "cap" - the flatter the smaller the contact-angle ex
given by (Fig.S.19)

(S.79)

The size of the critical nucleus, for reasons discussed in Chap.4, depends
on the driving force, and therefore. by virtue of (S.78), on the deposition
rate as well as on the substrate temperature. The relatively large nuclei char-
acteristic of small supersaturation create isolated patches (islands) of film
on the substrate which subsequently grow together. As the supersaturation

160
 v

Fig.5.19. Cap-shaped nucleus showing contact angle 0:. (s: substrate, v: vapor, m:
deposit)

increases, the critical nucleus shrinks until its height reaches one atomic
diameter 3 and its shape is that of a two-dimensional layer. The transition
from an island-type, or "3-D", to a layer-by-Iayer, or "2-D", type of nuclea-
tion is favored by good wetting between film and substrate (small a), as
realized with atomically clean substrates chemically identical with the
deposit. Only for very large supersaturation is the layer-by-Iayer nucleation
regime also accessible with incompletely wetted foreign substrates [5.78].
Apart from nucleation, film formation also involves growth, mainly
lateral growth. Since arriving vapor atoms (called "adatoms") must, on aver-
age, diffuse several atomic distances before sticking to a stable position
within the newly forming film, lateral growth depends on a sufficient sur-
face mobility of the adatoms. Surface diffusion is a thermally activated pro-
cess determined by the substrate temperature. High temperature favors
rapid and defect-free crystal growth, whereas at low temperature or large
supersaturation crystal growth may be overwhelmed by impingement, re-
sUlting in disordered or even amorphous structures.
A useful, if necessarily simplified, framework to characterize the in-
fluence of the two main parameters substrate temperature T and impinge-
ment flux j on film deposition is due to Kaschiev [5.78]. He defined a quan-
tity termed N99 as the average number of mono/ayers required for the
growing film to reach continuity (actually defined as 99% of coverage).
The value of N99 reflects both nucleation and growth rates, and Fig. 5.20
shows its dependence on the substrate temperature and impingment flux.
Theory predicts that below a "critical" value near N99 = 2 film formation
occurs layer by layer, whereas above it involves formation and coalescence
of progressively thicker islands. One would thus choose the former regime
to produce continuous monolayers and the latter to achieve single-crystal-
line film growth.

3 At very large supersaturation the critical nucleus predicted by classical theory may turn out
to be smaller than one atom. In this situation atomistic models which use atomic potential
energies instead of bulk free energies are better suited [5.77]. Conceptually, however, the con-
clusions are the same.

161
 substrate temperature T lOCI
400 300 200 100
20 (amorphous
10

layer growth
particle flux
lcni 2 s1 )
10
10

3.0

Fig.5.20. Growth diagram illustrating regimes of layer-by-Iayer growth and island

growth depending on substrate temperature T and impingement particle flux [5.78,79]

5.5.2 The PLD Process

In order to apply the above considerations quantitatively to a PLD experi-

ment, one would have to know the relevant parameters of the vapor which
define the quantities p and j in (5.78). This, of course, is no trivial matter.
As we have discussed in most of this chapter, parameters like particle den-
sity and kinetic energy, degree of ionisation, temperature and pressure in
an evolving laser plume tend to vary widely depending on the absorbed laser
fluence, pulse duration and wavelength. Moreover, the vapor usually con-
sists of several kinds of particles normally not in thermodynamic equili-
brium.
No quantitative general theory relating all vapor properties of interest
to the laser pulse and material parameters is available. The best we can do
(as so often) is to pull together what is known from theory and experiment
in order to obtain at least a qualitative picture of the complex phenomena
leading to PLD.
Whatever the details of the interaction, the laser impact creates a high-
pressure vapor plume. The plume, somewhat similar to a rocket exhaust,
expands away from the impact zone with a strongly forward directed velo-
city distribution, often approximated by a distribution of the form cos n «()), ()
being the angle to the target normal and n a number between about 5 and 12
[5.80]. While the laser pulse lasts, it interacts with the plume in various
ways depending on irradiance and wavelength, as discussed in previous sec-
tions. As a result, the plume may heat up far beyond the surface tempera-

162
ture. Single- or multiple-charged and energetic ions may be produced and
shock waves may develop.
After the laser pulse the vapor plume expands further, cools, and even-
tually dissipates. Somewhere in this sequence the plume front meets the sub-
strate which precipitates condensation. In order to determine the vapor pro-
perties at this point, the plume is usually treated as an adiabatically expand-
ing ideal gas (Sect. 5.2.2). Simplified calculations of plume expansion and
resulting film deposition parameters have been presented [5.81,82]. In their
analytical treatment, Anisimov et al. [5.81] assumed a self-similar adiabatic
expansion mode and predicted a film profile of the form

Mo k2 r2 -3/2
her) = --[1 + k2-] (5.80)
2prz2 z2

Here Mo is the total initial mass of the plume (a function of the laser pulse
energy), p is the film solid density, z is the distance between the target and
the substrate (Fig.5.18) and k is a (tabulated) function of the adiabatic index
of the vapor, the focal spot radius and the pulse duration, roughly with
values between 2 and 10. For typical parameters (SOns laser pulse, one to
several J/cm 2 of absorbed fluence, z = Scm, Mo = 1...1OJtg, p = 2.5
g/ cm2, k = 2) the predicted film thickness deposited per pulse is 1...10 A
thick at the center with a radius at half-thickness of 2 cm. The particle flux
j/NA in this process ranges between 1026 and 20 22 cm-2s- 1 and the supersa-
turation dGsv reaches some 10 5 J / mol. By suitably dosing the pulse energy,
as this estimate indicates, individual monolayers can readily be deposited.
These are impressive numbers. Indeed, the plumes created by short
laser pulses differ significantly from the near-equilibrium vapors present in
conventional (crucible-type or electron-gun driven) evaporation systems.
Firstly, PLD vapors are hotter - the surface reaches temperatures well in
excess of the equilibrium evaporation temperature of the target material. As
a consequence, the vapors tend to be fully dissociated even for molecular
targets [5.83]. More important, however, is that their elemental composition
is equal to that of the target - no matter what the vapor-pressure curves of
the constituents (for the explanation of this fact see the end of Sect. 5 .1.1).
This feature, which is largely responsible for the popularity of PLD among
materials scientists, enables them to deposit stoichiometric compound semi-
conductors like HgCdTe - which consist of species with vapor pressures
differing by several orders of magnitude - simply from a target of the same
material.
Another property of PLD vapors, as opposed to equilibrium ones, is
the presence of fast (sometimes falsely called superthermal) atoms, as well
as excited particles including ions. The high atom velocities are a natural
consequence of the high gas pressures reached and are usually found to be
Maxwellian distributed around stream velocities of some tens of km/s, in

163
accordance with gas-dynamic considerations. Deviations from Maxwellian
behavior is only found for UV laser pulses, indicative that here evaporation
is partly due to photolytic desorption [5.84].
Ion velocities are even higher, apparently due to acceleration in electric
fields building up in the plume. The fraction and energy of energetic parti-
cles increases with the irradiance. Typical ion fractions for UV lasers are
less than 5 % for irradiances below 100 MW / cm2 but approach 100 % at 10
GW / cm2 ; infrared lasers reach full ionisation even earlier. Ion kinetic ener-
gies in the range of 10..;- 1000 eV, one to two orders of magnitude higher
than for neutrals, seem to be typical [5.82]. There are, of course, also large
numbers of free electrons present in the ionized vapor, but little is known
about their properties and impact on film formation. As to the ions, their in-
fluence on film formation is mainly beneficial - they seem to provide extra
energy to the growing surface, enhancing surface mobility and hence im-
proving crystal quality and film density even on cold substrates. High-ener-
gy ions. on the other hand, have been blamed for damaging the growing
film by defect formation and sputtering.
Yet another species populating PLD plumes is unconditionally unpopu-
lar - melt droplets. Tiny droplets may arise due to condensation in the sup-
ersaturated vapor (Sect.5.2.2) [5.85], but the main contribution seems to be
melt ejected from the target surface, either by boiling or by evaporation
recoil (Sect.5.2.3). Droplets are detrimental to the quality of the film and
unsuitable for applications. The fraction of droplets in the plume can be
reduced by appropriately selecting the fluence [5.86], but this may not
always be practical. Fortunately, droplets move at much lower speeds than
gas particles and can be filtered out of the plume by means of mechanical
velocity filters [5.87]. Alternatively, the use of a second laser beam to eva-
porate droplets has been proposed [5.88].
Otherwise, PLD is experimentally quite straightforward. Apart from
the laser, a simple target chamber with an optical window and fixtures for
the substrate (perhaps heatable) and the target (preferentially rotatable to
expose a fresh spot for every laser shot) is all that is required. A good vacu-
um is not necessary - at deposition rates of 104 ..;- 106 Ai s, impurity incor-
poration is always negligible compared to deposition. PLD also works well
under relatively high pressures (l..;-100mTorr) of reactive gases.
What materials can be deposited by PLD? As it appears, in principle
almost any - for detailed reviews see [5.83,88]. Early reports emphasized
optical coatings based on oxides, cha1cogenides and metals, which were
lauded for being denser and smoother than thermally evaporated ones
[5.88]. The inherent monolayer resolution of PLD - supported by advances
in laser technology - was soon put to use in the fabrication of layered struc-
tures like tailored bandgap (superlattice) semiconductors [5.90], X-ray mir-
rors [5.91] or Bi-CdTe pairs with barrier layers as thin as 20 A [5.91], just
to mention a few examples. However, PLD has turned out to be at its best

164
in the deposition of complex multicomponent ceramic crystals, in which
stoichiometry is crucial but difficult to achieve by standard deposition tech-
niques. The most notable exampls are the oxide ceramic superconductors
like YBa2 CU3 0 7 (YBCO) and others, but also ferroelectrics like BaTi03 or
even (Pb, La)(Zr, Ti)03' ferrites and many others [5.89]. The laser of
choice in this work was the excimer, the radiation of which is efficiently ab-
sorbed in the ceramic target.
Singh and Narayan [5.82] investigated PLD of YBCO materials by ns-
excimer laser pulses in detail. Plasma temperatures of 10 000 K and particle
velocities of several km/ s (scaling approximately as the cube root of laser
pulse energy) are characteristic of this regime. Because lighter atoms move
faster than heavier ones (although the dependence is weaker than the in-
verse-square-root mass dependence expected for non-interacting particles),
slight spatial variations of film composition tend to build up, with lighter
species enriched at the center of the deposit, even though the evaporation
itself is stoichiometric. The effect becomes smaller at larger target-sub-
strate distances and higher background pressures since collisions tend to
equalize species velocities.
The stoichiometric evaporation achieved by PLD thus does not auto-
matically guarantee exactly stoichiometric films in all cases. Deviations
may also occur due to differing elemental sticking coefficients, reevapora-
tion of volatile elements or preferential sputtering by fast ions. In order to
enhance incorporation of oxygen and stabilize the desired tetragonal YBCO
phase between laser pulses, oxygen background gas or even oxygen jets
directed at the substrate have been employed [5.88,93]. Apart from oxides,
a similar kind of "reactive PLD" has been demonstrated for nitride forma-
tion in a N2 atmosphere [5.94]. Yet another approach to promote incorpora-
tion of a volatile element is the use of a second target rich in that element
[5.95].
The list of experimental variants of the PLD method as well as the
range of materials successfully deposited is certain to grow in the future.
The technique is in its infancy, at least as far as its quantitative physical
understanding is concerned. There remains plenty to be done.

165
A. Appendix

A.1 Selected Material Data

In the following we provide a short compilation of optical, thermophysical

and thermodynamic material properties often used in treating light-material
interaction phenomena. The data are meant as a quick reference for the
convenience of the reader. Original sources should be consulted when accu-
rate data for quantitive calculations are needed.

Table A.I. Optical absorption lengths (11 a) and reflectances (R) of semiconductors
(c: crystalline, a: amorphous, I: liquid), insulators and evaporated metal films at room
temperature for various wavelengths. Data from [A.l] and other sources. Semiconduc-
tor data depend on purity and the method of preparation, and are meant as typical
values only

Wavelength: 0.2S~m O.S~m 1.06~m 10.6~m

lIa R 1/a R l/a R Ita R

GaAs (c) 6nm 0.6 100nm 0.39 70~m 0.31 >lcm 0.28
Ge (c) 7nm 0.42 ISnm 0.49 200~m 0.38 >lcm 0.36
Ge (a) lOnm 0.48 SOnm 0.47 l~m 0.42 >lcm 0.34
Si (c) 6nm 0.61 SOOnm 0.36 200~m 0.33 Imm 0.30
Si (a) lOnm 0.7S 100nm 0.48 l~m 0.3S >lcm 0.32
Si (I) 8um 0.72 13nm 0.72
KCl >lcm O.OS >lcm 0.04 >lcm 0.04 >lcm 0.03
Si02 >lcm 0.06 >lcm 0.04 >lcm 0.04 40~m 0.2

Ag 20nm 0.30 14nm 0.98 12um 0.99 12nm 0.99

Al 8nm 0.92 7nm 0.92 lOnm 0.94 12nm 0.98
Au 18nm 0.33 22nm 0.48 13nm 0.98 14nm 0.98
Cu (0.1) 14nm 0.62 13nm 0.98 13nm 0.99
Ni (O.IS) 12nm 0.62 ISnm 0.67 37nm 0.97
W 7nm O.SI Bum 0.49 23nm 0.S8 20nm 0.98

166
Table A.2. Optical breakdown threshold irradiance IB for various wavelengths (A) and
pulse durations (t p) in insulating and semiconducting materials (B: bulk, S: surface).
Data selected from a review article by Smith [A.2] where references to the original lit-
erature can be found

Substance A [/Lm] tp IB [W/cm2]

Al20 3 B 1.06 4.7 ns 1.8 x10 11

B 1.06 30 ps 7.7 x10 11
S 1.06 4.7 ns 4.6 x1O IO
Diamond B 10.6 s X10 6
Silica B 1.06 4.7 ns 1.1 X10 11
B 1.06 30 ps 1.3 x10 12
B 0.53 21 ps 1.4 x10 12
B 0.355 20 ps 2+9 X10 12
S 1.06 125 ps 8 X1010
glass:BK-7 S 1.06 125 ps 7.4 X1010
glass:BSC-2 B 1.06 125 ps 7.4 X1OIO
S 1.06 4.7 ns 5+8.8 x1O IO
glass:Pyrex S 1.06 125 ps 6.4 X1010
PMMA plastic S 1.06 12 ns 1.6 X1010

NaCI B 10.6 600 ns 0.46+1.6 xlO lO

B 10.6 80 ns 1.55 X1010
B 1.06 4.7 ns 2.2 X1010
B 1.06 30 ps 9 X 1010
B 1.06 15 ps 6.3 X 1011
B 0.53 21 ps 6.3 xl0 11
ZnSe B 10.6 92 ns 4+5 X10 8
B 1.06 4.7 ns 1.2 X 109
S 10.6 92 ns 3.5 X 108
H20 (Jiq) B 1.06 30 ps 1.6 x10 12
D20 (Jiq) B 1.06 30 ps 8 xlO 11
GaSa S 10.6 0.2 s 1.3 X10 3
GaP S 10.6 0.2 s 3.6 X10 3
Ge S 10.6 0.2 s 8.5 X10 3
S 10.6 1.4 ns 3 X 10 9
Si S 10.6 0.2 s 1+2 X104

167
Table A.3. Melting points and absorptances [%] of some elements at various tempera-
tures. (l_R)(fe) values calculated from (2.46), (l-R)(exp) from literature data. Based on
data from [A. I, 6-8), and other sources

Metal Tsl (l_R)fe (l-R)exp

[K) 300K Tsl (sol) Tsl (Jiq) Tsl +500K A = Il-tm, 300K
(lowest) (practical)

Ag 1234 0.4 2.0 4.3 5.4 0.6 I

Al 933 1.1 4.4 10.0 13 5.6 6
Au 1336 0.6 3.5 8.3 lO.3 1.4 2
eu 1356 0.5 2.8 5.8 7.3 1.0 2
Sn(gray) 505 0.1 0.2 0.4 0.5 54
W 3655 0.8 17.1 18.5 42 45
Zn 693 2.1 6.0 13.5 13.4 43

A.2 Green's Functions for Solving the Heat-Flow Equation

The starting point of our analysis is the Green's function for a point-like
heat source. For the boundary conditions (3.5) it is given by [A.9]

_ V [(X - x')2 + (y _ y')2 )

go - 7r3/2 C {33 exp - {32
P
00

x L [ exp [- (2nL-Z-z')2]
{32
+
[(2nL-Z +Z')2]]
exp -
{32
(A2.I)
n=-oo

with {3 = 2(Klt-t'I)1/2. From (A2.1) Green's functions for any desired

source geometry can be constructed. We are interested in a distributed heat
source with an energy density decreasing exponentially inside the material.
A particularly simple case is that of a vanishing absorption length a-I = 0.
We refer to this geometry, which is described by (A2.I) with z' = 0, as a
surface source. For future reference we denote the sum in (A2.1) for z' =
° as TI
00
(2nL - Z)2]
exp [ {32 . (A2.2)
n=-oo

168
 Table A.4. Thermodynamic data of selected elements, from [A.l, 3,4] and other sources (M: molar mass; N:
number density; Ps ,Pi: mass densities of solid and liquid, respectively; Tsi: solid-liquid equilibrium tempera-
ture; T iv: liquid-vapor equilibrium temperature; dH = Jc p dT; dHsi ,dHtv: latent heats)

M N Ps PI Tsl TIv dH dHsl dH dHLV

(300K) (300K) (T sl ) 300K-+Tsl Tsl-+T lv
[g/mole] [1O-22 cm-3 ] [g/cm3] [g/cm 3 ] [K] [K) [kJ/mole] [kJ/moJe) [kJ/mole) [kJ/mole)

Ag 107.9 5.85 10.50 9.30 1234 2485 23.3 12.0 37.2 254.2
Al 27.0 6.02 2.70 2.37 933 2793 17.2 10.7 52.4 284.1
Au 197.0 5.91 19.3 17.3 1338 3130 29.2 12.6 52.6 310.6
Cd 112.4 4.60 8.6 8.0 594 1040 8.2 6.1 13.3 99.4
Co 58.9 8.99 8.9 7.7 1768 3201 54.0 15.9 49.8 389.6
Cr 52.0 8.33 7.1 6.5 2130 2945 62.5 14.7 32.1 305.4
Cu 63.5 8.48 8.96 8.0 1358 2836 29.5 13.0 46.4 304.8
Fe 55.8 8.48 7.87 7.0 1809 3135 65.0 19.4 55.5 353.8
Ge 72.6 8.43 5.32 7.0 1809 3107 25.8 34.8 58.0 284.7
Mg 24.3 4.30 1.74 1.57 922 1363 17.7 9.2 14.8 131.9
Ni 58.7 9.13 8.9 7.8 1726 3187 47.4 17.6 56.3 378.8
Pb 207.2 3.29 11.7 10.6 601 2023 8.5 4.8 40.4 178.0
Pp 106.4 6.77 12.0 lO.7 1825 3237 46.4 17.2 53.2 372.6
Pt 195.1 6.62 21.5 18.9 2043 4100 53.5 21.9 77.5 5lO.8
Sb 121.8 3.27 6.68 6.5 904 1860 16.6 20.1 30.1 195.5
Si 28.1 4.98 2.33 2.53 1685 3514 36.0 50.3 136 385.6
Sn 118.7 3.70 7.30 6.96 505 2876 5.9 7.1 72.5 230.2
Ta 180.9 5.53 16.6 15.0 3287 5731 95.3 31.5 95.5 753.4
Ti 47.9 6.68 4.5 4.2 1943 3562 54.1 19.3 52.9 422.9
W 183.9 6.32 19.3 17.7 3680 5628 102.6 35.3 77.2 774.6
~
Zn 65.4 6.55 7.14 6.66 693 1180 10.8 6.7 15.3 114.8
CJ)
to
Table A.5 Selected binary systems grouped according to the types showing simi-
lar alloying behavior

(i) Extended or complete solid-solution formers

Ag-Au Ag-Pd Au-Cu Au-Pd Co-Fe Co-Mn Co-Ni Co-Pd Co-Pt
Cr-Fe Cr-Ni Cr-Pt Cr-V Cu-Ni Cu-Pd Cu-Pt Fe-Mn Fe-Ni
Fe-Pd Fe-Pt Fe-V Ge-Si Mn-Ni Mn-Pt Mn-V Ni-Pd Ni-Pt
Ti-W V-W

(ii) Compound formers

Ag-Al Ag-Pt Ag-Mg Ag-Sn Ag-Ti Ag-Zn AI-Au AI-Co Al-Cr
AI-Cu AI-Fe AI-Mg AI-Mn AI-Ni AI-Pd AI-Pt Al-Ti AI-W
AI-Zn Au-Cr Au-Fe Au-Mg Au-Mn Au-Pb Au-Sn Au-Ti Au-V
Au-Zn Co-Cr Co-Cu Co-Mg Co-Si Co-Ti Co-V Co-W Co-Zn
Cr-Ge Cr-Mg Cr-Mn Cr-Pd Cr-Si Cr-Zn Cu-Fe Cu-Ge Cu-Mg
Cu-Si Cu-Sn Cu-Ti Cu-Zn Fe-Ge Fe-Si Fe-Ti Fe-W Fe-Zn
Ge-Mg Gc-Mn Ge-Ni Ge-Pd Ge-Pt Ge-Ti Gc-V Mg-Ni Mg-Pb
Mg-Pd Mg-Pt Mg-Si Mg-Sn Mg-Zn Mn-Pd Mn-Si Mn-Sn Mn-Ti
Mn-Zn Ni-Si Ni-Sn Ni-Ti Ni-V Ni-W Ni-Zn Pb-Pd Pb-Pt
Pb-Ti Pb-V Pd-Si Pd-Sn Pd-Ti Pd-Zn Pt-Si Pt-Sn Pt-Ti
Pt-V Pt-Zn Si-Ti Si-V Si-W Si-Zr Sn-Ti Sn-V Sn-Zr
Ti-Zn V-Zn Zn-Zr

(iii) Immiscible solids (simple eutectics)

Ag-Cu Ag-Ge Ag-Pd Ag-Si AI-Be Al-Ge AI-Si AI-Sn Au-Co
Au-Ge Au-Si Fe-Mg Ge-Pb Ge-Sn Ge-Zn Mg-Ti Pd-Sn Pd-V
Pd-W Pt-W Si-Sn Si-Zn Sn-Zn

(iv) Immiscible melts

Ag-Cr Ag-Fe Ag-Ni Ag-V Ag-W AI-Cd AI-Pd Au-Ru Au-W
Co-Pd Cr-Pb Cr-Sn Cu-Mo Cu-Pb Cu-V Cu-W Fe-Pb Fe-Sn
Mg-V Mg-W Mn-Pb Ni-Pb Pb-Si Pb-Zn Sn-W Zn-W

170
Table A.6. Vapor pressure of selected elements at various temperatures. Data from
[A.3, 5] and other sources

Vapor pressure [atm] at temperature [0C]

1000 1500 2000 2500 3000 3500

Ag 0.01 0.3 3.5 18 70

AI 0.7 1.2 8 38
Au 0.12 1.3 7
Cd 0.02 9 35
Cr 0.05 1.1 10
Cu 0.03 0.6 5 25
Fe 0.3 3 18
Ge 0.15 2.5 12
Mg 0.4 15
Ni 0.01 0.3 3.5 20
Ph 0.2 3.2 20
Pt 0.02 0.25
Si 0.03 0.4 2.5
Sn 0.03 0.5 2.5 10
Ti 0.01 0.3 3
Zn 2.5 70

In a material with a finite absorption length, light absorption represents a

penetrating source, for which the appropriate Green's function is

L
gl =a f go e--o:z'dz'
·0
aV ((X-X')2 + (y_y')2 a 2{32]
2 exp - 2 - 4 T2 (A2.3)
27r cp {3 (3

with

00

T2 = L exp[a(2nL - z)]
n=-oo

X [ erfc [
(2n + I)L - z
{3
a{3]
+ 2 - erfc
[2nL - z
{3 + 2
a{3]]

171
 + exp[- a(2nL - z)] (A2.4)

(2n + I)L-z a{3] [2nL-Z a{3]]

x [ erfc [- {3 + 2 - erfc - (3 + 2 .

Note that for the semi-infinite solid L = 00, and thus only the terms n = 0
in the sums T 1 and T2 need to be retained. To describe heating by an
extended laser beam the Green's functions given above must be integrated
over the lateral beam distribution. If f(r) is the radial irradiance profile we
have (with r2 = x2 + y2)

00 271"
fa fa gj(r,r')f(r')r'dr'd¢ (A2.S)

where gj is one of the expressions (A2.1 or 3). (Note that the distance be-
tween a field point and a source point in cylindrical coordinates is [(x-x')2
+ (y_y')2]1/2 = (r2 +r'2-2rr'cos¢)1/2). We only consider two cases here,
the uniform and the Gaussian source. Other beam profiles can be treated
by the same procedure.
The uniform source, f(r) = constant, is somewhat unphysical since it
is not normalizable. We override the difficulty by introducing an infinite
surface Sao, with the understanding that for an evaluation of (3.7) Pa/Sao is
replaced by the source irradiance. Thus, we obtain for the surface source

v
(A2.6)

and for the penetrating source

(A2.7)

For the Gaussian sources we set f(r) = exp(-r2/w2) and get for the
surface source

(A2.8)

and for the penetrating source

V r2 a2 {32
ggp =
271"c ({32 +w2) exp [- {32 + w2 + --)'T
4 2.
(A2.9)
P

172
 To obtain a temperature distribution, the appropriate Green's function,
together with a function describing the temporal pulse envelope Pa (t), is in-
serted into (3.7) and the integral is solved.

A.3 Numerical Solution of the Heat-Flow Equation

Equations (4.4,5), along with the appropriate initial and boundary condi-
tions, completely specify our problem. In a sense, they already represent a
computional scheme. For an actual computer evaluation, the differential
equation (4.5) must merely be transformed into a finite-difference equation.
Here, instead of continuous functions of space and time, such as T(z, t) or
ilH(z, t), one has discrete functions defined at points on a lattice into which
the irradiated solid is divided. Since we are considering one-dimensional
heat flow only (a generalization to three dimensions does not add anything
new), the lattice is one-dimensional and may be visualized as dividing the
solid into parallel layers with uniform internal temperature and material pro-
perties. Lattice points are separated by the finite length az, and function
values are evaluated only after discrete time intervals ilt. For convenience,
we introduce the volumetric enthalpy W == aH/V (V is taken as a constant).
A finite-difference version of (4.4,5) can now be written as

Win+ 1 = Win + (at/ az 2 )[K n-l,n

i i _ Ti )
(T n-l n

(A3.1)
and
T~ = T(W~). (A3.2)

Here X~ denotes the value of the quantity X at the lattice site n at time i
(i.e., i time increments at after the chosen time origin). Further, K~ n+l
specifies the conductivity at time i between the lattice points nand n + 1.
Eq.(A3.2) is an inverted version of (4.4) in the form of a table or an ana-
lytic approximation.
The procedure stipulated by (A3.1,2) works as follows. An initial dis-
tribution ~, corresponding to the desired initial temperature distribution
T~, is specified and inserted into (A3.1). If there is heat production, i.e., if
J~ is nonzero, then a new distribution W~ is calculated, with W! > ~.
Next, the enthalpy distribution W! is converted into a temperature distribu-
tion T! by means of (A3.2), and inserted back into (A3.1). The procedure is
repeated as many times as desired, yielding a new instantaneous tempera-

173
ture profile T~ after each time step. The boundary condition for an insu-
lated slab is enforced in the calculation by adding two virtual lattice points
at n = 0 and n = N + 1, where n = 1 and n = N are the outmost lattice
points of the slab and setting W~ = W~ and W~ + 1 = W~a' The interface is
conveniently defined as that lattice site which has absorbed or liberated 50 %
of the latent heat of melting (the interface temperature is specified by
(A3.2». The position of the interface is recorded after each time step.
There is almost complete freedom to readjust the material parameters or the
heat source characteristics between computing steps, in order to allow, e.g.,
for a temperature-dependent conductivity, for temporal variations of the
irradiance, or for absorption effects.
The advantage of such a code (termed an explicit code by the experts)
is that it is straightforward and allows maximum flexibility in the choice of
the physical parameters. The price paid for these advantages is that the in-
crements dt and dz must satisfy the stringent condition

(A3.3)

everywhere and at all times. This means, in practice, that rather small time
steps must be chosen if a good spatial resolution is desired. The reason for
this condition has nothing to do with rounding errors as sometimes stated,
but lies rather in the fact that the finite-difference equation (A3.1) has solu-
tions which do not satisfy the original differential equation (4.5). Only
those solutions of (A3.1) which also satisfy (A3.3) remain always real and
finite.

A.4 Units and Symbols

Systeme International (SI) units are used in formulas, except when noted
otherwise. Several familiar non-SI units have been retained in quoting
numerical data. The values of some non-SI units employed in the book,
expressed in SI-units, are

1 eV 1.602.10- 19 J
1 at = 9.806,10 4 Pa
1 g/ cm3 = 10 3 kg/m3
1 W/cm2 10--4 W/m2

The main symbols and frequently used subscripts are listed in the following.
Equilibrium quantities are, where necessary, distinguished from actual ones
by an overbar.

174
A.4.1 Constants

c vacuum velocity of light (2.99S·10 8 m/s)

e electron charge (1.602·1O- 19 C)
h Planck's constant/211" (1.054.10-34 J·s = 6.5S3 .1O-16 eV ·s)
k Boltzmann's constant [1.3S0·1O- 23 J/K !:::(1I1l605)eV/K]
me free electron rest mass (9 . lOS . 10-31 kg)
NA Avogadro's number (6.023 ·10 26 mole- l )
R Gas constant (S.314·10 3J/K)
EO dielectric constant (S.S54·1O- 12 C/Vm)
uSB Stefan-Boltzmann constant (5.667 . 10-8 W/m 2 K4)

A.4.2 Variables

A particle accommodation factor

a molecular diameter or lattice constant [m]
cp specific heat at constant pressure [J I mole K]
CV specific heat at constant volume [J I mole K]
D diffusivity[m2/s]
Damb ambipolar diffusivity [m2Is]
E energy [J]
Ep Fermi energy [J]
Eg band-gap energy [J]
E electrical field strength [V1m]
E emissivity
F fluence [Jlm 2 ]
G Gibbs free energy (free enthalpy) [J/mole]
g partial Gibbs free energy (free enthalpy) [J/mole]
g Green's function
H magnetic field strength [Aim]
H enthalpy [J/mole]
h partial enthalpy [J/mole]
I irradiance [W/m2]
J power density [WI m3]
j molar flux [moles/m 2 s]
K thermal conductivity [W/m K]
k distribution coefficient
L slab thickness [m]
£ mean free path [m]
M molar mass [kg/mole]
m particle mass [kg]
N particle density [m-3]
n complex refractive index

175
 real and imaginary parts of n
power [W]
pressure [Pa]
reflection coefficient
coordinate, radius [m]
entropy [J/mole K]
Soret coefficient [K-I]
temperature [K]
time [s]
pulse duration [s]
velocity [m/s]
sound velocity [m/s]
molar volume [m 3 /mole]
v velocity [m/s]
w beam radius [m]
X molarl atomic fraction
x coordinate [m]
y nucleation rate [m-3 S-I]
y coordinate [m]
z ion charge number
z coordinate [m]
a absorption coefficient [m- I ]
r damping constant [S-I]
'Y = c p I C v adiabatic index
o diffusion length [m]
complex dielectric function
real and imaginary parts of {'
relative coordinate along z-axis [m]
normalized temperature
K thermal diffusivity [m2 Is]
A wavelength [m]
jump frequency [S-I]
mass density [kg/m3]
cross section [m2]
(J surface free energy [J/m2]
(J electrical conductivity [A/Vm]
T time constant [s]
<I> heat flux [WI m 2]
volume fraction
W angular frequency [s-I]
wp plasma frequency [S-I]
[ ... ] equals one if the bracketed condition holds, zero otherwise

176
A.4.3 Subscripts

a absorbed
cr critical
d dense phase (solid or liquid)
e electron
h hole
I liquid
r recombination
s solid
v vapor

177
References

Chapter 1

1.1 For recent surveys see, e.g.,

Laser Materials Processing -Industrial and Micrelectronics Applications, ed.
by E. Beyer, SPIE Proc. 2207 (1994)
Laser Assisted Fabrication of Thin Films and Microstructures, ed. by I. Boyd,
SPIE Proc. 2045 (1994)
Laser Processing in Manufacturing, ed. by R.C. Crafer, PJ. Oakley (Chapman
& Hall, New York 1993)
Laser Materials Processing, ed. by P. Denney, I. Miyamoto, B. Mordike (Laser
Inst. Am., Orlando, FL 1994)
S.M. Metev, V.P. Veiko: Laser-Assisted Microtechnology, Springer Ser. Mater.
Sci., Vo1.19 (Springer, Berlin, Heidelberg 1994)
P.F. Barbara, W.H. Knox, G.A. Mourou, A.H. Zewail (eds.): Ultrafast Phe-
nomena IX, Springer Ser. Chern. Phys., Vol.60 (Springer, Berlin, Heidelberg
1994)
1.2 W. Koechner: Solid-State Laser Engineering, 3rd edn., Springer Ser. Opt. Sci.,
VoU (Springer, Berlin, Heidelberg 1992)
Solid State Lasers and New Materials. SPIE Proc. 1839 (Washington 1992)
1.3 W. Witteman: CO2 Lasers, Springer Ser. Opt. Sci., Vol.53 (Springer, Berlin,
Heidelberg 1987)
1.4 Ch.K. Rhodes (ed.): Excimer Lasers, 2nd edn., Topics Appl. Phys., Vo1.30
(Springer, Berlin, Heidelberg 1984)
1.5 M. von Allmen: In Laser Annealing of Semiconductors, ed. by J.M. Poate,
J.W. Mayer (Academic, New York 1982) pp.43-74
1.6 See the magazines Laser Focus (Advanced Technology, Littleton, Mass.) and
Lasers and Applications (High-Tech Publications, Torrance, Calif.)
1.7 D.C. Winburn, G. Gomez: Practical Laser Safety (Dekker, New York 1985)
1.8 D. Bauerle: Chemical Processing with Lasers, Springer Ser. Mater. Sci., Vol.1
(Springer Berlin, Heidelberg 1986)
1.9 Optical Thermal Response of Laser-Irradiated Tissue, ed. by AJ. Welch, M.
van Gernert (Plenum, New York 1995)

Chapter 2

2.1 H. Kogelnik: Bell Syst. Tech. J, 44, 455-494 (1965)

H.K.V. Lotsch: Optik 30, 1-14, 181-201,217-233,563-576 (1969170)
2.2 F. Wooten: Optical Properties of Solids (Academic, New York 1972)
2.3 P. Grosse: Freie Elektronen in Festkorpem (Springer, Berlin, Heidelberg 1979)
2.4 C.F. Bohren, D.R. Huffman: Absorption and Scattering of Light by Small Par-
ticles (Wiley, New York 1983)

179
2.5 S.A. Akhmanov, A.P. Sukhorukov, R.V. Khokhlov: Sov. Phys. - JETP 23,
1025- 1033 (1966)
S.A. Akhmanov, R.V. Khokhlov, A.P. Sukhorukov: In Laser Handbook, ed. by
F.T. Arecchi; E.O. Schulz-Dubois (North-Holland, Amsterdam 1972) pp.
1151-1228
2.6 O. Svelto: Progress in Optics 12, 3-51 (North-Holland, Amsterdam 1974)
2.7 Y.R. Shen: Rev. Mod. Phys. 48, 1-32 (1976)
2.8 F.W. Dabby, J .R. Whinnery: Appl. Phys. Lett. 13,284-286 (1968)
2.9 A.I. Osipov, V.Ya. Panchenko, A.A. Filippov: SOY. J. Quantum Electron. 15,
465-470 (1985)
2.10 E. Yablonowitch, N. Bloembergen: Phys. Lett. 14,907-910 (1972)
2.11 H.Y. Fan: Semiconductors and Semimetals 3, Chap. 9 (Academic, New York
1967)
2.12 K.G. Svantesson, N.G. Nilsson: Phys. Scr. 18,405-409 (1978)
2.13 A. Bhattacharyya, B.G. Streetman: Solid State Commun. 36,671-675 (1980)
2.14 Y.P. Varshni: Physics 34, 149-154 (1967)
C.D. Thurmond: J. Electrochem. Soc. 122, 1133-1141 (1975)
2.15 LW. Boyd, T.D. Binnie, LB. Wilson, MJ. Colles: J. Appl. Phys. 55, 3061-3063
(1984)
2.16 W.B. Gauster, J.C.Bushnell: J. Appl. Phys. 41, 3850-3853 (1970)
2.17 R.G. Ulbrich: Solid-State Electron. 21, 51-59 (1978)
2.18 EJ. Yoffa: Appl. Phys. Lett. 36, 37-38; Phys. Rev. B 21,2415-2425 (1980)
2.19 A. Lietoila, J.F. Gibbons: J. Appl. Phys. 53, 3207-3213 (1982); Appl. Phys.
Lett. 40, 624-626 (1982)
2.20 A. Haug: Solid-State Electron, 21, 1281-1284 (1978)
2.21 A. Elci, A.L. Smirl, C.Y. Leung, M.O. Scully: Solid-State Electron. 21, 151-158
(1978)
2.22 M. Rasolt, H. Kurz: Phys. Rev. Lett. 54, 722-724 (1985)
2.23 V. Heine, J.A. van Vechten: Phys. Rev. B 13,1622-1626 (1976)
J.C. Inkson: J. Phys. C 9,1177-1183 (1976)
2.24 J.A. van Vechten, R. Tsu, R.W. Saris: Phys. Lett. A 74,422-426 (1979)
J.A. van Vechten: In Semiconductor Processes Probed by Ultrafast Spectros-
copy, ed. by R.R. Alfano (Academic, New York 1984) Vol.2, pp.95-169
2.25 A. Compaan, A.Aydinly, M.C. Lee, H.W. Lo: MRS Proc. 4, 43-48 (Elsevier,
New York 1982)
2.26 G.E. Jellison, Jr., D.H. Lowndes, R.F. Wood: MRS Proc. 13, 35-42 (Elsevier,
New York 1983)
2.27 D.H. Auston, J.A. Golovshenko, A.L. Simons, R.E. Slusher, R.P. Smith, C.M.
Murko, T.N.C. Venkatesan: Appl. Phys. Lett. 34, 777-779 (1979)
2.28 W.P. Dumke: Phys. Lett. A 78, 477-480 (1980)
2.29 B. Stritzker, A .. Pospieszczyk, I.A. Tagle: Phys. Rev. Lett. 47, 356-358, 1676-
1677 (1981)
2.30 B.C. Larson, C.W. White, T.S. Noggle, D. Mills: Phys. Rev. Lett. 48, 337-340
(1982)
2.31 N. Baltzer, M. von Allmen, M.W. Sigrist: Appl. Phys. Lett. 43,826-828 (1983)
2.32 A.M. Malvezzi, H. Kurz, N. Bloembergen: Appl. Phys. A36, 143-146 (1985)
2.33 M.C. Downer, C.V. Shank: Phys. Rev. Lett. 56, 761-764 (1986)
2.34 C.V. Shank, R. Yen, C. Hirlimann: Phys. Rev. Lett. 50,454-457 (1983)
2.35 WJ. Siekhaus, I.H. Kinney, D. Milam, L.L. Chase: Appl. Phys. A 39, 163-166
(1986)

180
2.36 P. Braunlich, A. Schmid, P. Kelly: Appl. Phys. Lett. 26, 150-153 (1975)
2.37 L.B. Glebov, a.M. Efimov, G.T. Petrovskii, P.N. Rogovtsev: SOY. J. Quantum
Electron. 15, 1367-1370 (1985)
2.38 N. Bloembergen: IEEE J. QE-lO, 375-386 (1974)
2.39 H.L. Holway: J. Appl. Phys. 45,677-683 (1974)
H.L. Holway, D.W. Fradin: J. Appl. Phys. 46,279-291 (1975)
2.40 M. Bass, H. Barrett: IEEE J. QE-8, 338-342 (1972)
2.41 M. Bass, D.W. Fradin: IEEE J. QE-9, 890-896 (1973)
2.42 N. Bloembergen: Appl. Opt. 12,661-664 (1973)
2.43 D. Ryter, M. von Allmen: IEEE J. QE-17, 2015-2017 (1981)
2.44 W.L. Smith: Opt. Eng. 17,489-503(1978)
2.45 J. Bass: In Landolt-Bornstein, New Ser., Vo1.15a, ed. by K. Hellwege, J.L.
Olsen (Springer, Berlin, Heidelberg 1982) pp.5-137
2.46 A.V. Grosse: Rev. Hautes Temp. Refract. 3, 115 -146 (1966)
2.47 M. Sparks, E. Loh, Jr.: J. Opt. Soc. Am. 69, 847-858 (1979)
R.E. Lindquist, A.W. Ewald: Phys. Rev. 135, A191-194 (1964)
L.V. Nomerovannaya, M.M. Kirillova, M.M. Noskov: SOY. Phys. - JETP 33,
405-409 (1971)
2.48 V.A. Batanov, F.V. Bunkin, A.M. Prokhorov, V.B. Fedorov: SOY. Phys. -JETP
36,311-322 (1973)
2.49 O.N. Krokhin: In Laser Handbook, ed. by F.T. Arecchi, E.O. Schulz-DuBois
(North-Holland, Amsterdam 1972) Chap. 7
2.50 D.C. Emmony, R.P. Howson, LJ. Willis: Appl. Phys. Lett. 23, 598-600 (1973)
2.51 J.F. Young, J.E. Sipe, J.S. Preston, H.M. van Driel: Appl. Phys. Lett. 41, 261-
264 (1982)
2.52 P.A. Temple, MJ. Soileau: IEEE J. QE-17, 2067-2071 (1981)
2.53 J.E. Sipe, J.F. Young, I.S. Preston, H.M. van Driel: Phys. Rev. B 27,
1141-1154 (1983); ibid. 30, 2001-2015 (1984)
2.54 Z. Guosheng, P.M. Fauchet, A.E. Siegman: Phys. Rev. B 26,5366-5381 (1982)
2.55 DJ. Ehrlich, S.R. Brueck, 1.Y. Tsao: Appl. Phys. Lett. 41, 630-632 (1982)
2.56 F. Keilmann, Y.H. Bai: Appl. Phys. A 29,9-18 (1982)
2.57 HJ. Leamy, G.A. Rozgonyi, T.T. Sheng, G.K. Celler: Appl. Phys. Lett. 32,
535- 538 (1978)
2.58 I.W. Boyd, S.C. Moss, T.F. Boggess, A.L. Smirl: Appl. Phys. Lett. 45 80-82
(1984)
2.59 J.C. Koo, R.E. Slusher: Appl. Phys. Lett. 28, 614-616 (1976)
2.60 M. von Allmen, W. Luthy, K. Affolter: Appl. Phys. Lett. 33, 824-825 (1978)
2.61 W.G. Hawkins, D.K. Biegelsen: Appl. Phys. Lett. 42, 358-360 (1983)
2.62 M. Combescot, J. Bok, C. Benoit a la Guillaume: Phys. Rev. B 29, 6393-6395
(1984)
2.63 T.R. Anthony, H.E. Cline: J. Appl. Phys. 48, 3888-3894 (1977)
2.64 K. Affolter, W. Luthy, M. Wittmer: Appl. Phys. Lett. 36, 559-561 (1980)
2.65 G. Gorodetsky, J. Kanicki, T. Kazyaka, R.L. Melcher: Appl. Phys. Lett. 46,
547-549 (1985)
2.66 N. Postacioglu, P. Kapadia, J. Dowden: J. Phys. D 24, 1288-1292 (1991)
2.67 C. Hill: In Laser Annealing of Semiconductors, ed. by J. M. Poate, J.W. Mayer
(Academic, New York 1982) pp.479-557
2.68 F. Keilmann: Phys. Rev. LeU. 51, 2097-2100 (1983)
2.69 J.F. Ready: IEEE 1. QE-12, 137-142 (1976)
2.70 DJ. Broer, L. Vriens: Phys. A 32,107-123 (1983)

181
2.71 A. Blatter, C. Ortiz: J. App!. Phys. 73, 8552-8560 (1993)
2.72 C. Ortiz, A. Blatter: Thin Solid Films 218, 209-218 (1992)
2.73 D. Tuckerman, R.L. Schmitt: Proc. 1985 Multilevel Interconnect Conf. (IEEE,
New York 1985) pp.24-31
2.74 A. Bachli, A. Blatter: Surface Coating Technology 45,393-397 (1991)
2.75 D.H. Lowndes, M. DeSilva, M.I. Godhole, A.I. Pedraza, D.B. Geohegan: Mat.
Res. Soc. Proc. 285,191-196 (1993)
2.76 R.I. Baseman, T.-S. Kuan, M.O. Aboelfotoh, J.C. Andreshak, R.E. Turene,
R.A. Previti-Kelly, J.G. Ryan: Mat. Res. Soc. Proc. 236, 361-369 (1992)
2.77 A. Bachli, A. Blatter, M. Maillat, H.E. Hintermann: Surface Modification Tech-
nologies V, ed. by T.S. Sudarshan, J.F. Braza, The Institute of Materials
821-833 (1992)
2.78 M. Rothschild, C. Arnone, D.I. Ehrlich: J. Vac. Sci. Techno!. B 4, 310-314
(1986)
A. Blatter, U. Bogli, L.L. Bouilov, N.!. Chapliev, V.!. Konov, S.M. Pimenov,
A.A. Smolin, B.V. Spytsin: Proc. Electrochem. Soc. 91, 357-364 (1991)
2.79 B.I. Palmer, R.G. Gordon: Thin Solid Films 158, 313-341 (1988)
2.80 V.N. Tokarev, J.L.B. Wilson, M.G. Jubber, P. John, D.K. Milne: Diamond and
Related Materials (Submitted June 1994)
P. Tosin, A. Bachli, A. Blatter: Proc. Symp. on New Diamond and Related Ma-
terials, Eight CIMTEC (June 1994)
2.81 W.W. Duley, W.A. Young: J. Appl. Phys. 44, 4236-4237 (1973)
2.82 E.V. Dan'shchikov, F.V. Lebedev, A.V. Ryazanov: SOy. J. Quantum Electron.
14,960-964 (1984)
2.83 R.W. Keyes: Semiconductors and Semimetals 4, 327-341 (Academic, New
York 1968)
2.84 G.P. Banfi, P.G. Gobbi: Plasma Phys. 21, 845-859 (1979)

Chapter 3

3.1 R. Srinivasan, B. Braren: Chern. Rev. 89, 1303 (1989)

3.2 R.E. Harrington: J. Appl. Phys. 38, 3266-3270 (1967)
3.3 A.V. Benkov, A.V. Zinov'ev, T. Usmanov, S.T. Azizov: SOy. J. Quantum Elec-
tron. 15, 643-649 (1985)
3.4 H. Salzmann: Phys. Lett. A 41,363-364 (1972)
R.I. Bickerton: Nuclear Fusion 13.457-458 (1973)
3.5 H.S. Carlslaw, J.C. Jaeger: Conduction of Heat ill Solids, 2nd. edn. (Oxford
Univ. Press, Oxford 1959)
3.6 M. Lax: Appl. Phys. Lett. 33, 786-788 (1978)
M. Lax: J. Appl. Phys. 48,3919-3923 (1977)
3.7 H.E. Cline, T.R. Anthony: J. App!. Phys. 48,3895-3900 (1977)
3.8 Y.!. Nissim, A. Lietoila, R.B. Gold, J.F. Gibbons: J. App!. Phys. 51, 274-279
(1980)
3.9 M. Sparks, E. Loh, Jr.: J. Opt. Soc. Am. 69, 847-858 (1979)
3.10 M. von Allmen, W. Luthy, M.R.T. Siregar, K. Affolter, M.-A. Nicolet: AlP
Proc. 50, 43-47 (Am. lnst. Phys., New York 1979)
3.11 E. Liarokapis, Y.S. Raptis: J. Appl. Phys. 57, 5123-5126 (1985)
3.12 A.G. Cullis, H.C. Webber, P. Bailey: J. Phys. E 12,688 (1979)
3.13 Z.L. Liau, B.Y. Tsaur, J.w. Mayer: App!. Phys. Lett. 34, 221-223 (1979)

182
3.14 R.B. Gold, J.F. Gibbons: J. AppJ. Phys. 51,1256-1258 (1980)
3.15 J.F. Gibbons, T.W. Sigmon: In Laser Annealing of Semiconductors, ed. by
J.M. Poate, J.W. Mayer (Academic, New York 1982) pp.325-382
3.16 B.O. Boley, J.H. Weiner: Theory of Thermal Stresses, (Wiley, New York 1960)
3.17 G.A. Rozgonyi, H. Baumgart: J. Physique 41, C4-5, 85-88 (1980)
3.18 Y. Matsuoka: J. Phys. D 9, 215-224 (1976)
3.19 L. Correra, G.G. Bentini: J. AppJ. Phys. 54, 4330-4337 (1983)
3.20 H.E. Cline: J. AppJ. Phys. 54, 2683-2691 (1983)
3.21 D.M. Follstaedt, S.T. Picraux, P.S. Peercy, W.R. Wampler: AppJ. Phys. Lett.
39, 327-329 (1981)
3.22 F. Haessner, W. Seitz: J. Mat. Sci. 6,16-18 (1971)
3.23 L. Buene, D. Jacobson, D.C. Nakahara, J.M. Poate, C.W. Draper, J.D. Hir-
vonen: MRS Proc. 1,583-590 (Elsevier, New York 1981)
3.24 S.S. Lau: J. Vac. Sci. TechnoJ. 15, 1656-1661 (1978)
3.25 G.L. Olson, S.A. Kokorowski, J.A. Roth, L.D. Hess: Mat. Res. Soc. Proc. 13,
141-154 (Elsevier, New York 1983)
3.26 A. Gat, J.F. Gibbons, T.J. Magee, J. Peng, V.R. Deline, P. Williams, C.A.
Evans, Jr.: AppJ. Phys. Lett. 32,276-278 (1978)
3.27 R.H. Uebbing, P. Wagner, H. Baumgart, HJ. Queisser: AppJ. Phys. Lett. 37,
1078-1079 (1980)
3.28 A. Lietoila, J.F. Gibbons, T.W. Sigmon: AppJ. Phys. Lett. 36, 765-768 (1980)
3.29 J.S. Williams: In Laser Annealing of Semiconductors, ed. by J.M. Poate, J.W.
Mayer (Academic, New York 1982) pp.383-435
3.30 M.W. Geis, D.C. Flanders, H.1. Smith: AppJ. Phys. Lett. 35, 71-74 (1979)
3.31 J.F. Gibbons, K.F. Lee, TJ. Magee, J. Peng, R. Ormond: AppJ. Phys. Lett. 34,
831-833 (1979)
3.32 J.A. Roth, G.L. Olson, L.D. Hess: Mat. Res. Soc. Symp. Proc. 23, 431-442
(Elsevier, New York 1984)
3.33 D.K. Biegelsen, N.M. Johnson, DJ. Bertelink, M.D. Moyer: MRS Proc. 1,
487-502 (Elsevier, New York 1981)
3.34 P. Zorabedian, T.1. Kamins, C.1. Drowley: J. AppJ. Phys. 57, 5262-5267 (1985)
3.35 HJ. Zeiger, J.C.C. Fan, B.J. Palm, R.P. Gale, R.L. Chapman: In Laser and
Electron Beam Processing of Materials, ed. by C.W. White, P.S. Peercy
(Academic, New York 1980) pp.234-240
3.36 R.L. Chapman, J.e. Fan, HJ. Zeiger, R.P. Gale: AppJ. Phys. Lett. 37, 292-295
(1980)
3.37 G.H. Gilmer, H.J. Leamy: In Laser and Electron Beam Processing of Materi-
als, ed. by e.W. White, P.S. Peercy (Academic, New York 1980) pp.227-233
3.38 G. Auvert, D. Bensahel, A. Perio, V.T. Nguyen, G.A. Rozgonyi: AppJ. Phys.
Lett. 39, 724-726 (1981)
3.39 E. D'Anna, G. Leggieri, A. Luches: Thin Solid Films 218,95-108 (1992)
3.40 R.M. Walser, R.W. Bene: AppJ. Phys. Lett. 28, 624-626 (1976)
3.41 R. Pretorius, T.K. Marais, C.C. Theron: Mat. Science Eng. 10, 1-83 (1993)
3.42 M.-A. Nicolet, S.S. Lau: VLSI Electronics: Microstructure Sci. 6, SuppJ.A,
329-464 (Academic, New York 1983)
3.43 T. Shibata, J.F. Gibbons, T.W. Sigmon: AppJ. Phys. Lett. 36, 566-568 (1980)
3.44 R. Andrew, L. Baufay, A. Pigeolet, L.D. Laude: J. AppJ. Phys. 53, 4862 (1982)
C. Antoniadis, M.C. Joliet: Thin Solid Films 115, 75 (1984)
M.C. Joliet, C. Antoniadis, R. Andrew, L.D. Laude: AppJ. Phys. Lett. 46, 266
(1985)

183
3.45 V.A. Bobyrev, F.V. Bunkin, N.A. Kirichenko, B.S. Luk'yanchuk, A.V. Sima-
kin: SOy. J. Quantum Electron.: 12,429-434 (1982)
R. Merlin, T.A. Perry: Appl. Phys. Lett. 45, 852-853 (1984)
T. Szorenyi, L. Baufay, M.C. Joliet, F. Hanus, R. Andrew, I. Hevesi: Appl.
Phys. A 39,251-255 (1986)
3.46 A.G. Akimov, A.P. Gagarin, V.G. Dagurov, V.W. Makin, S.D. Pudkov: SOY.
Phys. Tech. Phys. 25, 1439-1441 (1980); .
S.D. Pudkov: SOY. Phys. Techn. Phys. 25, 1439-1441 (1980);
T.E. Orlowski, H. Richter: Appl. Phys. Lett. 45, 241-243 (1984)
M. Thuillard, M. von Allmen: Appl. Phys. Lett. 47, 936-938 (1985); ibid., 48,
1045 (1986)
3.47 I. Ursu, L. Nanu, LN. Michailescu: Appl. Phys. Lett. 49, 109 (1986)
M. Thuillard, M. von Allmen: E-MRS Proc. 11, 137-142 (Editions Physique,
Les Ulis 1986)
3.48 1. Ursu, LN. Mihailescu, A.M. Prokorov, V.I. Konov (eds.): Laser Processing
and Diagnostics (Editions Physics, Les Ulis 1986) p.223
3.49 E. D'Anna, G. Leggieri, A. Luches: Thin Solid Films 218,219-230 (1992)
3.50 N. Bottka, PJ. Walsh, R.Z. Dalbey: J. Appl. Phys. 54,1104 (1983)
GJ. Fisanik, M.E. Gross, J.B. Hopkins, M.D. Fennell, KJ. Schnoes. A. Katzir:
J. Appl. Phys. 57,1139-1142 (1985)
3.51 D. Bauerle (ed.): Laser Processing and Diagnostics, Springer, Ser. Chern.
Phys., Vol. 39 (Springer, Berlin, Heidelberg 1984)
3.52 F.D. Seaman, D.S. Gnanamutu: Metal Progress (August 1975) pp.67-74
3.53 C.M. Banas:ln Physical Processes in Laser-Materials Interaction, ed. by M.
Bertolotti, Nato ASI Ser. B 84, 143-162 (Plenum, New York 1983)

Chapter 4

4.1 G.A. Kachurin, N.B. Pridachin, L.S. Smirnov: SOY. Phys. Semicond. 9, 946
(1975)
LB. Khaibullin, E.I. Shtyrkov, M.M. Zaripv, R.M. Bayazitov, M.F. Galyatudi-
nov: Rad. Eff. 36, 225 (1978)
4.2 W.A. Elliot, F.P. Gagliono, G. Krauss: Metallurg. Trans. 4, 2031-2037 (1972)
4.3 J. Crank: The Mathematics of Diffusion, 2nd edn. (Oxford Univ. Press, Oxford
1975)
4.4 Z.L. Liau, B.Y. Tsaur, S.S. Lau, I. Golecki, J.W. Mayer: Am. Inst. Phys. Proc.
50,105-110 (AlP, New York 1979)
4.5 H.S. Carlslaw, J.C. Jaeger: Conduction of Heat in Solids, 2nd. edn. (Oxford
Univ. Press, Oxford 1959)
4.6 D.H. Auston, LA. Golovshenko, A.L. Simons, R.E. Slusher, R.P. Smith, C.M.
Murko, T.N.C. Venkatesan: Appl. Phys. Lett. 34, 777-779 (1979)
4.7 P. Baeri, S.U. Campisano: In Laser Annealing of Semiconductors, ed. by J.M.
Poate, J. W. Mayer (Academic, New York 1982) pp.41-75
4.8 S.C. Hsu, S. Chakravorty, R. Mehrabian: Met. Trans. 9b, 221-228 (1987)
4.9 M. von Allmen: Mat. Res. Soc. Symp. Proc. 13, 691-702 (Elsevier, New York
1983)
A. Bachli, A. Blatter: Refr. Met. and Hard Mat. 11, 113-119 (1992)

184
4.10 A.G. Cullis, H.C. Webber, J.M. Ponte, A.L. Simons: Appl. Phys. Lett. 36,
320-323 (1980)
4.11 GJ. Galvin, M.O. Thompson, J.W. Mayer, R.B. Hammond, N. Poulter, P.S.
Peercy: Phys. Rev. Lett. 48, 33-36 (1982)
4.12 P. Schvan, R.E. Thomas: J. Appl. Phys. 57, 4738-4741 (1985)
4.13 J.C. Baker, J.W. Cahn: In Solidification (ASM, Metals Park, Ohio 1971)
pp.23-58
4.14 K.A. Jackson: In Treatise in Solid State Chemistry, ed. by N.B. Hannay
(Plenum, New York 1975) Vo1.5, Chap. 5
4.l5 F. Spaepen, D. Turnbull: Am. Inst. Phys. Con/. Proc. 50, 73-83 (AlP, New
York 1979)
4.16 F.F. Abraham, J.Q. Broughton: Phys. Rev. Lett. 56, 734-737 (1986)
4.17 K.A. Jackson: Can. J. Phys. 36, 683-891 (1958)
4.18 B. Chalmers: Principles of Solidification (Krieger, Huntington, Reprint 1977)
4.19 J. Frenkel: Kinetic Theory of Liquids (Oxford Univ. Press, Oxford 1946)
4.20 M.M. Martynyuk: Russ. J. Phys. Chern. 53, 1080-1081 (1979)
4.21 P. Hermes, B. Danielzik, N. Fabricius, D. von der Linde, J. Kuhl, J. Heppner,
B. Stritzker, A. Pospieszcyk: Appl. Phys. A 39,9-11 (1986)
4.22 D. Kashchiev: Surf. Sci. 14,209-220 (1969)
4.23 K.F. Kelton, A.L. Greer, C.V. Thompson: J. Chern. Phys. 79, 6261-6267 (1983)
4.24 P.S. Peercy, D.M. Follstaedt, S.T. Picraux, W.R. Wampler: Mat. Rex. Soc.
Symp. Proc. 4, 401-406 (Elsevier, New York 1982)
4.25 W.F. Tseng, J.W. Mayer, u.S. Campisano, G. Foti, E. Rimini: Appl. Phys.
Lett. 32, 824-826 (1978)
4.26 J. Narayan, C.W. White, M.J. Aziz, B. Stritzker, A. Walthuis: J. Appl. Phys.
57, 564-567 (1985)
4.27 G. Foti, E. Rimini, W.F. Tseng, J.W. Mayer: Appl. Phys. 15,365-369 (1978)
4.28 A.G. Cullis, H. Webber, N.G. Chew, 1.M. Poate, P. Baeri: Phys. Rev. Lett. 49,
219 (1982)
4.29 GJ. Galvin, J.W. Mayer: Appl. Phys. Lett. 46, 644-646 (1985)
4.30 P.L. Liu, R. Yen, N. Bloembergen, R.T. Hodgson: Appl. Phys. Lett. 34,
864-866 (1979)
4.31 R. Tsu, R.T. Hodgson, T.Y. Tan, I.E. Baglin: Phys. Rev. Lett. 42, 1356-1358
(1979)
4.32 B.C. Bagley, M.S. Chen: Am. Inst. Phys. Proc. 50, 97-101 (AlP, New York
1979)
4.33 E.P. Donovan, F. Spaepen, D. Turnbull: Appl. Phys. Lett. 42, 698-700 (1983)
4.34 F. Spaepen, D. Turnbull: In Laser Annealing of Semiconductors, ed. by 1.M.
Poate, 1.W. Mayer (Academic, New York 1982) pp.15-42
4.35 P.H. Bucksbaum, J. Bokor: Phys. Lett. 53,182-185 (1984)
4.36 P.H. BucksbaUlTI, J. Bokor: Mat. Res. Soc. Symp. Proc. 13 (Elsevier, New York
1983) pp.51-56
4.37 G. Foti, G. della Mea, E. Jannitti, G. Majni: Phys. Lett. A 68, 368-370 (1978)
4.38 T.N.C. Venkatesan, D.H. Auston, 1.A. Golovchenko, C.M. Surko: InLaseralld
Electron Beam Processing of Materials (Am. Inst. Phys., New York 1979)
pp.629-633
4.39 S.U. Campisano, G. Foti, E. Rimini, F.H. Eisen, W.F. Tseng, M.-A. Nicolet,
1.L. Tandon: J. Appl. Phys. 51, 295-298 (1980)
4.40 P.A. Barnes, H.J. Leamy, J.M. Poate, S.D. Ferris, J.S. Williams, G.K. Celler:
Appl. Phys. Lett. 33, 965 (1978)

185
4.41 E.D. Davies, E.F. Kennedy, R.G. Ryan, J.P. Lorenzo: Mat. Res. Soc. Proc. 1,
247-253 (North Holland, New York 1981)
4.42 J.S. Williams: In Laser Annealing of Semiconductors, ed. by J.M. Poate, J.W.
Mayer (Academic, New York 1982) pp.383-435
4.43 C.W. White, S.R. Wilson, B.R. Appleton, F.W. Young, Jr.: J. App!. Phys. 51,
738-749 (1980)
4.44 J.M. Poate: Mat. Res. Soc. Symp. Proc. 4, 121-130 (Elsevier, New York 1982)
4.45 C.W. White, J. Narayan, B.R. Appleton, S.R. Wilson: J. App!. Phys. 50,
2967-2069 (1979)
4.46 J.e. Baker, J.w. Cahn: Acta Met. 7, 575-578 (1969)
4.47 S.U. Campisano, G. Foti, P. Baeri, M.G. Grimaldi, E. Rimini: App!. Phys. Lett.
37,719-721 (1980)
4.48 B.C. Larson, C.W. White, B.R. Appleton, App!. Phys. Lett. 32, 801-803 (1978)
4.49 B.R. Appleton, B.C. Larson, C.W. White, 1. Narayan, S.R. Wilson, P.P.
Pronko: Am. Inst. Phys. Conf. Proc. 50, 291-298 (AlP, New York 1979)
4.50 J .W. Cahn, S.R. Coriell, W.J. Boettinger: In Laser and Electron Beam Proces-
sing of Materials, ed. by C.W. White, P.S. Peercy (Academic, New York
1980) pp.89-103
4.51 K.A. Jackson, G.H. Gilmer, H.J. Leamy: In Laser and Electron Beam Proces-
sing of Materials, ed. by C.W. White, P.S. Peercy (Academic, New York
1980) pp.104-110
4.52 R.F. Wood: App!. Phys. Lett. 37, 302 (1980); and Phys. Rev. B. 25, 2786-2811
(1982)
4.53 M.L Aziz: J. App!. Phys. 53, 1158-1168 (1982); and App!. Phys. Lett. 43
552-554 (1983)
4.54 S.U. Campisano, D.C. Jacobson, J.M. Poate, A.G. Cullis, N.G. Chew: App!.
Phys. Lett. 46,846-848 (1985)
4.55 S.T. Picraux, D.M. Follstaedt: In Surface Modification and Alloying: Alumi-
nium, ed. by J.M. Poate, G. Foti, D.C. Jacobson (Plenum, New York 1983)
pp.288-321
4.56 J .K. Hirvonen, J .M. Poate, A. Greenwald, R. Little: App!. Phys. Lett. 36,
564-566 (1980)
4.57 L. Buene, J.M. Poate, D.C. Jacobson, C.W. Draper, J.K. Hirvonen: App!. Phys.
Lett. 37, 385-387 (1980)
L. Buene, D.C. Jacobson, S. Nakahara, J.M. Poate, C.W. Draper, J.K. Hir-
vonen: Mat. Ser. Soc. Proc. 1,583-590 (1982)
4.58 S.T. Picraux, D.M. Follstaedt, J.A. Knapp. W.R. Wampler, E. Rimini: Mat.
Res. Soc. Symp. Proc. 1,575-582 (Elsevier, New York 1981)
4.59 D.M. Follstaedt, W.R. Wampler: App!. Phys. Lett. 38,140-142 (1981)
4.60 D.M. Follstaedt, S.T. Picraux, P.S. Peercy, W.R. Wampler: App!. Phys. Lett.
39,327-329 (1981)
4.61 S.R. De Groot, P. Mazur: Nonequilibrium Thermodynamics (North Holland,
Amsterdam 1962)
4.62 B.N. Bhat, R.W. Swalin: Acta Met. 20, 1387-1396 (1972)
M. Balourdet, J. Malmejac, P. Desre: Phys. Lett. A 56,51-52 (1976)
4.63 L.F. Dona dalle Rose, A. Miotello: Mat. Res. Soc. Symp. Proc. 14 (Elsevier,
New York 1982) pp.425-430
4.64 M. von Allmen, M. Wittmer: App!. Phys. Lett. 34,68-70 (1979)
4.65 M. Wittmer, M. von Allmen: J. App!. Phys. 50, 4786-4790 (1979)
4.66 M. Wittmer, W. Luthy, B. Studer, H. Melchior: Solid-State Electron. 24,
141-145 (1981)

186
4.67 G. Badertscher, R.P. Salathe, W. LUthy: Electron. Lett. 16, 113 (1980)
4.68 See, e.g., E. Ben-Jacob, P. Garik: Nature 343, 523-530 (1990)
W. Kurz, DJ. Fisher: Fundamentals of Solidification, 3rd edn. (Trans. Tech.
Publ. 1989)
4.69 W.W. Mullins, R.F. Sekerka: J. Appl. Phys. 35, 444-451 (1964);
R.F. Sekerka: J. Appl. Phys. 36, 264-268 (1965)
4.70 J. Narayan: J. Appl. Phys. 52, 1289-1293 (1981)
4.71 J. Narayan, H. Naramoto, C.W. White: J. Appl. Phys. 53, 912-915 (1982)
4.72 M. von Allrnen, S.S. Lau, T.T. Sheng, M. Wittmer: In Laser and Electron
Beam Processing of Materia/s, ed. by C.W. White, P.S. Peercy (Academic,
New York 1980) pp.524-529
4.73 M. Wittmer, W. Luthy, M. von Allmen: Phys. Lett. A 74, 127-130 (1979)
4.74 J.M. Poate, HJ. Leamy, T.T.Sheng, G.K. Celler: Appl. Phys. Lett. 33, 918-920
(1978)
4.75 GJ. van Gurp, G.EJ. Eggermont, Y. Tamminga, W.T. Stacy, J.R.M. Gijsbers:
Appl. Phys. Lett. 35, 237 (1979)
4.76 S.S. Lau, F.W. Tseng, M.-A. Nicolet, J.W. Mayer, J.A. Minnucci, A.R. Kirk-
patrick: Appl. Phys. Lett. 33, 235 (1978)
4.77 H.J. Leamy, CJ. Doherty, K.C.R. Chiu, J.M. Poate, T.T. Sheng, G.K. Celler:
In Laser and Electron Beam Processing of Materials, ed. by C.W. White, P.S.
Peercy (Academic, New York 1980) pp.581-587
4.78 S.S. Lau, B.Y. Tsaur, M. von Allmen, J.W. Mayer, B. Stritzker, C.W. White,
B. Appleton: Nucl. Inter. Meth. 182/183,97-105 (1981)
4.79 R.T. Tung, J.M. Gibson, D.C. Jacobson, J.M. Poate: Appl. Phys. Lett. 43, 476
(1983)
4.80 H.E. Cline: J. Appl. Phys. 53, 5898-5903 (1982)
4.81 M.A. Bosch, A.H. Dayem, T.R. Harrison, R.A. Lemons: Appl. Phys. Lett. 41,
363-364 (1982)
4.82 B.M. Ditchek, T. Emma: Appl. Phys. Lett. 45, 955 (1984)
4.83 G.G. Gladush, L.S. Krasitskaya, E.B. Levchenko, A.L. Chernyakov: Sov. J.
Quantum Electron. 12,408-412 (1982)
4.84 T.R. Antony, H.E. Cline: J. Appl. Phys. 48, 3888-3894 (1977)
4.85 T. Chande, J. Mazumder: Appl. Phys. Lett. 41, 42 (1982)
4.86 J.D. Ayers, T.R. Tucker: Thin Solid Films 73,201-207 (1980)
4.87 D.B. Snow, E.M. Breinan, B.H. Kear: In Superal/oys (Am. Soc. Metals, Metals
Park, Ohio 1980)
4.88 C.W. Draper, L. Buene, J.M. Poate, D.C. Jacobson: Appl. Opt. 20, 1730-1732
(1981)
4.89 H.W. Bergmann, B.L. Mordike: Z. Metallkde 71, 658-665 (1980)
H.W. Bergmann, B.L. Mordike: Z. Werkstofftech. 14,228-237 (1983)
4.90 C.W. Draper, .FJ.A. den Broeder, D.C. Jacobson, E.N. Kaufmann, M.L.
McDonald, J.M. Van den Berg: Mat. Res. Soc. Symp. Proc. 4, 419-424 (El-
sevier, New York 1982)
4.91 C.W. Draper: Appl. Opt. 20, 3093-3096 (1981)
C.W. Draper, C.A. Ewing: J. Mat. Sci. 19,3815-3825 (1984)
C.W. Draper, I.M. Poate: Int. Met. Rev. 30, 85-108 (1985)
4.92 V.S. Kovalenko, V.I. Volgin: Fiz. Khim. Obrab. Mater. 3, 28 (1978)
4.93 V.P. Greco: Plating Surf. Finishing 68,56 (1981)
4.94 P.G. Moore, E. Mccafferty: J. Electrochem. Soc. 128, 1391-1393 (1981)

187
4.95 A.K. Jain, V.N. Kulkarni, K.B. Nambiar, D.K. Sood, S.C. Shanna, P. Maz-
zoldi: Rad. Eff. 63, 175-181 (1982)
4.96 C.W. Draper, D.C. Jacobson, J.M. Gibson, J.M. Poate, J.M. Vandenberg, A.G.
Cullis: Mat. Res. Soc. Symp. Proc. 4, 413-418 (Elsevier, New York 1982)
4.97 C.W. Draper, L.S. Meyer, D.C. Jacobson, L. Buene, J.M. Poate: Thin Solid
Films 75, 237-240 (1981)
A. Inoue: Mater. Trans. JIM 36, 866-875 (1995)
4.98 P. Chaudhari, B.C. Giessen, D. Turnbull: Scientific American 242,84-86 (April
1980)
4.99 I.W. Donald, H.A. Davies: J. Noncryst. Solids 30, 77-85 (1978)
4.100 K. Affolter, M. von Allmen: Appl. Phys. A 33,93-96 (1984)
4.101 T.B. Massalsky: Proc. 4th Int. Conf. Rapidly Quenched Metals, Sendai (1981)
pp.203-208
4.102 D.R. Uhlmann: J. Non-Cryst. Solids 7,337-348 (1972);
H.A. Davies: Phys. Chern. Glasses 17, 159-173 (1976)
4.103 N. Saunders, A.P. Miodownik: Ber. Bunsenges. Phys. Chern. 87, 830-834
(1983)
4.104 D. Turnbull: Contemp. Phys. 10,473-483 (1969)
4.105 U. Kambli, M. von Allmen, N. Saunders, A.P. Miodownik: Appl. Phys. A 36,
189-192 (1985)
4.106 M. von Allmen, S.S. Lau, M. Maenpaa, B.Y. Tsaur: Appl. Phys. Lett. 36,
205-207 (1980)
4.107 M. von Allmen, S.S. Lau, M. Maenpaa, B.Y. Tsaur: Appl. Phys. Lett. 37, 84-86
(1980)
4.108 K. Affolter, M. von Allmen, H.P. Weber, M. Wittmer: J. Non-Crystal Solids
55, 387-393 (1983)
4.109 E. Huber, M. von A1lmen: Phys. Rev. B 28, 2979-2984 (1983); and B 31,
3338-3342 (1985);
U. Kambli, E. Huber, M. von Allmen: Phys. Rev. B 33,8643-8648 (1986)
4.110 G.G. Borodina, CH.V. Kopetskii, V.S. Kraposhin, N.V. Edneral: Sov. Phys.
Dokl. 26, 761-763 (1981)
4.111 R. Becker, G. Sepold, P.L. Ryder: Scripta Metall. 14, 1283-1285 (1980)
4.112 H.W. Bergmann, B.L. Mordike: J. Mat. Sci. 16,863-869 (1981)
4.113 H. Yoshioka, K. Asami, K. Hashimoto: Scripta Metall. 18, 1215-1218 (1984)
4.114 E.M. Breinan, B.H. Kear, C.M. Banas: Physics Today 44-50 (Nov. 1976)
4.115 S. Yatsuya, T.B. Massalski: Mat. Sci. Eng. 54,101-111 (1982)
4.116 C.J. Lin, F. Spaepen: Appl. Phys. Lett. 41, 721 (1982)
4.117 C.J. Lin, F. Spaepen, D. Turnbull: J. Noncryst. Sol. 61162, 707-712 (1984)
4.118 A. Wolthuis, B. Stritzker: J. Physique 44, C5, 489-491 (1983)
4.119 J. Frohlingsdorf, B. Stritzker: MRS Conf. Proc. 51, 271-276 (Mat. Res. Soc.,
Pittsburg, PA 1986)
4.120 P. Mazzoldi, G. Della Mea, G. Battaglin, A. Miotello, M. Servidori, D. Bacci,
E. Jannitti: Phys. Rev. Lett. 44, 88-91 (1980);
D.M. Follstaedt, P.S. Peercy, W.R. Wampler: Phys. Rev. Lett. 46, 1250 (1981);
P. Mazzoldi, G. Della Mea, G. Battaglin, A. Miotello, M. Servidori, D. Bacci,
E. Jannitti: Phys. Rev. Lett. 46, 1251 (1981)
4.121 A. Blatter, M. von Allmen: Phys. Rev. Lett. 54, 2103-2106 (1985)

188
Chapter 5

5.1 R.G. Ross, D.A. Greenwood: In Progress in Materials Science, Vo1.l4., ed. by
B. Chalmers, W. Hume-Rothery (Pergamon, London 1969)
5.2 V.A. Batanov, F.V. Bunkin, A.M. Prokhorov, V.B. Fedorov: SOy. Phys. JETP
36,311-322 (1973)
5.3 F. Hensel: In Properties of Liquid Metals, ed. by S. Takeuchi (Taylor and
Francis, London 1973) pp.357-363
5.4 L.D. Landau, E.M. Lifshitz: Statistical Physics (Pergamon, London 1958)
5.5 A. Kantrowitz: J. Chern. Phys. 19, 1097-1100 (1951)
5.6 M.M. Martynyuk: SOy. Phys. Tech. Phys. 19,793-797 (1974)
5.7 B. Luk'yanchuk, N. Bityurin, S. Anisimov, D. Bauerle: App!. Phys. A 57,
367-374 (1993)
5.8 R. Srinivasan, B. Braren: Chern. Rev. 89, 1303 (1989)
5.9 L.D. Landau, E.M. Lifshitz: Fluid Mechanics (Pergamon, London 1959)
5.10 Y.B. Zeldovich, Y.P. Raizer: Physics of Shock Waves and High-Temperature
Hydrodynamic Phenomena, VoU (Academic, New York 1966)
5.11 M.C. Fowler, D.C. Smith: J. App!. Phys. 46, l38-150 (1975)
5.12 A.A. Boni, F.Y. Su: J. App!. Phys. 44, 4086-4094 (1973)
5.l3 V.1. Bergel'son, A.P. Golub, T.V. Loseva, LV. Nemchinov, T.1. Orlova, S.P.
Popov, V. Svettsov: SOY. J. Quantum Electron. 4, 704-706 (1974)
E.V. Dan'shchikov, V.A. Dymshakov, F.V. Lebedev, A.V. Ryazanov: SOY. J.
Quantum Electron. 15, 1231-1237 (1985)
5.14 T.P. Hughes: Plasmas and Laser Beams (Hilger, Bristol 1975)
5.15 H.Hora, H. Wilhelm: Nuc!. Fusion 10,111-116 (1970)
5.16 D.H. Gill, A.A. Dougal: Phys. Rev. Lett. 15, 845-847 (1965)
5.17 C. DeMichelis: IEEE J. QE-5, 188-202 (1969)
5.18 N. Kroll, K.M. Watson: Phys. Rev. A 5,1883-1905 (1972)
5.19 F.W. Dabby, D.C. Paek: IEEEJ. QE-8, 106-111 (1972)
5.20 S.l. Anisimov: SOY. Phys. JETP 27,182-183 (1968)
5.21 V.1. Igoshin, V.1. Kurochkin: SOY. J. Quantum Electron. 14, 1049-1052 (1984)
5.22 V.A. Batanov, F.V. Bunkin, A.M. Prokhorov, V.B. Fedorov: JETP Lett. 11,
69-72 (1972)
5.23 W.E. Maher, R.B. Hall: J. App\. Phys. 47, 2486-2493 (1976)
5.24 S.l. Anisimov, Y.A. Imas, G.S. Romanov, Y.D. Khodyko: Effects of High-
Power Radiation on Metals (Nat. Technical Information Service, Springfield,
VA 1970)
5.25 D.W. Gregg, S.J. Thomas: J. App!. Phys. 37, 2787-2789 (1966)
5.26 V.1. Mazhukin, A.A. Samokhin: SOY. J. Quantum Electron. 14, 1608-1611
(1984)
5.27 W.W. Duley: CO 2 Lasers, Effects and Applications (Academic, New York
1976)
M. Bass (ed.): Laser Materials Processing (North Holland, Amsterdam 1983)
5.28 DJ. Broer, L. Vriens: Phys. A 32, 107-123 (1983)
5.29 M. von Allmen: J. App!. Phys. 47,5460-5463 (1976)
5.30 M.M. Martynyuk: SOY. Phys. Tech. Phys. 21, 430-433 (1976)
5.31 R.P. Gagliano, D.C. Paek: App!. Opt. 13,274-279 (1974)
5.32 W.W. Duley, J.N. Gonsalves: Can. J. Phys. 50, 216-221 (1972)
5.33 E. Kocher, L. Tschudi, J. Steffen, G. Herziger: IEEE J. QE-8, 120-125 (1972)

189
5.34 T. Chande, J. Mazumder: J. App!. Phys. 56, 1981-1986 (1984)
1. Dowden, M; Davis, P. Kapadia: J. App!. Phys. 51, 4474-4479 (1985)
5.35 P.G. Klemens: J. App!. Phys. 41, 2165-2114 (1976)
5.36 E.V. Locke, R.A. HeIla: IEEE J. 10, 179-185 (1974)
5.37 J.A. Robin, P. Nordin: J. App!. Phys. 46, 2538-2543 (1975)
5.38 W.W. Duley, J.N. Gonsalves: Optics and Laser Technology, 78-81 (April 1974)
5.39 R.L. Stegman, J.T. Schriempf, L.R. Hettche: J. App!. Phys. 44, 3675-3681
(1973)
5.40 Yu.P. Raizer: SOy. Phys. JETP 31,1148-1154 (1970)
5.41 A.A. Boni, F.Y. Su: Phys. Fluids 11,340-342 (1974)
5.42 Yu.P. Raizer: SOy. Phys. Usp. 23, 789 (1980); and SOy. Phys. Quantum Elec-
tron. 14,40-45 (1984)
5.43 N.A. Generalov, V.P. Zimakov, GJ. Kozlov, V.A. Masyukov, Yu.P. Raizer:
SOy. Phys. JETP 34,763-769 (1972)
5.44 F.V. Bunkin, V.1. Konov, A.M. Prokhorov, V.B. Federov: JETP Lett. 9,
371-374 (1969)
E.L. Klosterman, S.R. Byron: J. App!. Phys. 45, 4751-4759 (1914)
5.45 J.P. Jackson, P.E. Nielsen: AIAAJ. 12, 1498-1501 (1974)
5.46 V.K. Goncharov, L.Y. Min'ko, E.S. Tyunina, A.N. Chumakov: SOY. J. Quant.
Electron, 3, 29-32 (1973)
A.P. Gagarin, V.V. Druzhinin, N.A. Raba, S.V. Maslenikov: SOy. Tech. Phys.
Lett. 1, 149-150 (1975)
5.47 J.P. Jackson, EJ. Jumper: Laser Digest (Air Force Weapons Lab., Kirtland
AFB, NM 1975)
5.48 S. Marcus, I.E. Lowder, D.L. Mooney: 1. App!. Phys. 41,2966-2968 (1976)
5.49 J.A. McKay, I.T. Schriempf: App!. Phys. Lett. 31, 369-371 (1977)
J.E. Robin: J. App!. Phys. 49, 5306-5310 (1978)
5.50 P.E. Nielsen: J. App!. Phys. 49, 5306-5310 (1978)
5.51 N.N. Rykalin, A.A. Uglov, M.M. Nizametdinov: SOY. Phys. JETP 42, 367-377
(1975)
B.M. Zhiryakov, N.1. Popov, A.A. Samokhin: SOY. Phys. JETP 48, 247-252
(1978)
5.52 M. von Allmen, P. Blaser, K. Affolter, E. Sturmer: IEEE J. QE-14, 85-88
(1978)
5.53 W.E. Maher, R.B. Hall: J. App!. Phys. 51,1338-1344 (1980)
5.54 Yu.P. Raizer: SOY. Phys. JETP 21, 1009-1017 (1965)
5.55 V.1. Fisher, V.M. Kharash: SOy. Phys. JETP 55, 439-443 (1982)
5.56 A. Pirri: Phys. Fluids 16,1435-1440 (1973)
5.57 P.E. Nielsen: J. App!. Phys. 46, 4501-4505 (1975)
5.58 B.S. Holmes, D.C. Erlich: J. App!. Phys. 48, 2396-2403 (1977)
5.59 J.D. O'Keefe, C.H. Skeen, C.M. York: J. App!. Phys. 44, 4622-4626 (1973)
C.G. Hoffman: 1. Appl. Phys. 45, 2125 (1974)
J.F. Ready: IEEE J. QE-14, 79-84 (1978)
5.60 E. Sturmer, M. von Allmen: ZAMP 28, 1177-1182 (1977)
5.61 E. Sturmer, M. von Allmen: J. App!. Phys. 49, 5648-5654 (1978)
5.62 D.C. Hamilton, I.R. Pashby: Optics and Laser Technology 183-188 (August
1919)
5.63 R.C. Elton: X-Ray Lasers (Academic, Boston 1990)
5.64 O.N. Krokhin: SOY. Phys. Tech. Phys. 9, 1024-1026 (1965)
5.65 A. Caruso, R. Gratton: Plasma Phys. 10,867-877 (1968)

190
5.66 P. Mulser: Z. Naturforsch. 25a, 282-295 (1970)
5.67 A. Ng, D. Pasini, P. Celliers, D. Perfeniuk, L. Da Silva, 1. Kwan: Appl. Phys.
Lett. 45, 1046 (1984)
5.68 V. Gupta, A.S. Argon, D.M. Parks, J.A. Cornie: J. Mech. Phys. Solids 40,
141-180 (1992)
5.69 B. Steverding, H.P. Dudel: J. App\. Phys. 47, 1940-1945 (1976)
F. Cottet, M. Hallouin, J.P. Romain, R. Fabbro, B. Fara1, H. Pepin: Appl. Phys.
Lett. 47, 678 (1985)
5.70 C. Fauquignon, F. F1oux: Phys. Fluids 13,386-391 (1970)
J.L. Bobin: Phys. Fluids 14,2341-2354 (1971)
5.71 L. Spitzer: Physics of Fully Ionized Gases, 2nd edn. (Wiley, New York 1962)
5.72 A. Caruso: In Laser Interactions and Related Plasma Phenomena, ed. by H.J.
Schwarz, H. Hora (Plenum, New York 1971) pp.289-305
5.73 A. Caruso, R. Gratton: Plasma Phys. 11,839-847 (1969)
5.74 N.G. Denisov: SOy. Phys. JETP 4, 544-553 (1957)
5.75 V.L. Ginzburg: The Propagation of Electromagnetic Waves in Plasmas, 2nd
edn. (Pergamon, London 1970)
5.76 J. Dawson, P. Kaw, B. Green: Phys. Fluid 12, 815-882 (1969)
D.V. Giovanielli, R.P. Godwin: Am. J. Phys. 43, 808-817 (1975)
Y.V. Afanas'ev, N.N. Demchenko, O.N. Krokhin, V.B. Rosanov: SOy. Phys.
JETP 45,90-94 (1977)
W.M. Mannheimer, D.G.Colombant: Phys. Fluids 21, 1818-1827 (1978)
R.P. Godwin: Appl. Opt. 18,3555-3561 (1979)
T.P. Donaldson, J.E. Balmer, J.A. Zimmermann: J. Phys. D 13. 12221-1233
(1980)
5.77 D. Walton: In Nucleation, ed. by A.C. Zettlemoyer (Dekker, New York 1969)
5.78 S. Stoyanov, D. Kashchiev: Current Topics in Materials Science Vol.7 (North-
Holland, Amsterdam 1981) pp.69-141
5.79 S. Metev, K. Meteva: App\. Surf. Sci. 43, 402-408 (1989)
5.80 R. Kelly, R.W. Dreyfuss: Nuc\' lnstrum. Methods B 32,341-348 (1988)
5.81 S.l. Anisimov, D. Bauerle, B.S. Luk'yanchuk: Phys. Rev. 43, 12076-12081
(1993)
5.82 R.K. Singh, J. Narayan: Phys. Rev. B 41,8843-8859 (1990)
5.83 J.T. Cheungh, H. Sankur: CRC Crit. Rev. Solid State Mat. Sci. 15, 63-109
(1988)
5.84 B. Luk'yanchuk, N. Bityurin, S. Anisimov, D. Bauerle: App\. Phys. A 57,
367-374 (1993)
5.85 S.V. Gaponov, A.A. Gudkov, A.A. Fraerman: SOY. Phys. - Tech. Phys. 27,
1130-1133 (1982)
5.86 D. Bhattacharya, R.K. Singh, P.H. Holloway: J. App\. Phys. 70, 5433-5439
(1991)
5.87 W.P. Barr: 1. Phys. E 2,2 (1969)
5.88 H. Sankur: Thin Solid Films 218, 161-169 (1992)
5.89 G. Hubler: MRS Bulletin (February 1992) pp.26-58
5.90 J.T. Cheung, J. Madden: J. Vac. Sci. Techno\. B 5, 705 (1987)
5.91 S. Metev: Laser Processing and Diagnostics II, ed. by D. Bauerle, R.-L.
Kompa, L. Laude, Proc. E-MRS 11, 143-152 (Editions de Physique, Les UIis
1986)
5.92 S.V. Gaponov, B.M. Luskin, N.N. Salashchenko: SOY. Phys. Semicond. 14,
873-880 (1980)

191
5.93 H.S. Kwok: Thin Solid Films 218, 277-290 (1992)
5.94 V. Craciun, D, Craciun, I.W. Boyd: Mat. Sci. Eng. B 18,178-180 (1993)
5.95 A. Sajjadi, I.W. Boyd: Appl. Phys. Lett. 63, 1-3 (1993)

Appendix

A.l American Institute of Physics (AlP) Handbook, 3rd. edn. (McGraw-Hill, New
York 1972)
A.2 W.L. Smith: Opt. Eng. 17,489-503 (1978)
A.3 J. Bass: In Landolt-Bornstein, New Ser., Vol.15a, ed.by K. Hellwege, J.L.
Olsen (Springer, Berlin, Heidelberg 1982) pp.5-137
A.4 A.V. Grosse: Rev. Hautes Temp. Refract. 3, 115-146 (1966)
A.5 M. Sparks, E. Loh, Jr.: J. Opt. Soc. Am. 69, 847-858 (1979)
R.E. Lindquist, A.W. Ewald: Phys. Rev. 135, A191-194 (1964)
L.V. Nomerovannaya, M.M. Kirillova, M.M. Noskov: Sov.Phys. - JETP 33,
405-409 (1971)
A.6 Handbook of Chemistry and Physics, ed. by R. Weast (CRC, Cleveland 1981)
pp.55-60
A.7 Metals Reference Book, 4th edn., ed. by C.J. Smithells (Butterworth, London
1967) Vols.II-IV
A.8 A.N. Nesmeyanov: Vapor Pressure of the Chemical Elements (Elsevier,
Amsterdam 1963)
A.9 H.S. Carls)aw, J.C. Jaeger: Conduction of Heat in Solids, 2nd. edn. (Oxford
Univ. Press, Oxford 1959)

192
Subject Index

Absorptance 8,168 Film deposition 159-165

Absorption Free carrier 9,19-31,40
- coefficient 7-12,20,22,27,54,122- Fresnel reflectance 7
124,166
- cross section 20 Gaussian beam 6,7, 17
-length 7 Glass formation 87,107-114
- waves 39,137-150 Graphoepitaxy 61
Alloying 68, 80, 96-98, 104-107, 170 Green's function 44, 168ff
Annealing 55,60,68,85,93
Avalanche breakdown, see Optical Hardening 65-67, 106
breakdown Heat conduction 24,42, 43, 95
Heat wave 157
Boiling 119-120,134 Hole drilling 131-134
Bremsstrahlung absorption 123,124
Impact ionization 16,28-30
Cellular growth, see Constitutional su- Inertial confinement 157
percooling Interband transition 9,12,21,27,31
Channelling (of laser beam) 38 Interface
Conductivity - liquid-solid 72-76, 81-83, 87, 89,
- radiative 141 91-93,95,98,111,119
- thermal 42, 50, 51 - liquid-vapor 119, 127
Constitutional supercooling 98-100
Convection 101,104,105 Kerr effect 18,19
Critical point 32,117,118 Keyhole 38
Cutoff density 123,154,158 Knudsen layer 129,130
Cutting 137
Laser
Dielectric function 8-13,18 - -driven de flag ration wave 156, 157
Diffraction length 17 - -supported combustion wave 138-143
Diffusion 70,71,84-9092,94-96,103, - detonation wave 146-150
120 Lorentz 9
Drude 13
Maxwell-Garnett 12
Electron beam 2,93, 107 Melting 72-76,80,83,85
Electrostriction 19 Melt quenching 68,69,107-114
Evaporation 32,35,88,105,116-121, Mott (metal-insulator) transi-
125-128 . tion 27,29,124,151
- recoil 128-133,159 Multiphoton absorption 27,29,124,151
Explosive crystallization 61, 64

193
Nucleation Rossland mean free path 141
- of crystal 61,62,83-85, 100, 111, 160,
161 Segregation 89,90,92,94,98,120
- of melt 83-85 Self-focussing 16-20,40
- of vapor 102, 119, 134 Self-trapping 19
Solidification 68,72-76,79,84,90
Optical breakdown 20,28-31,40, Soret effect, see Thermodiffusion
123-125 Stress 57,58,114,149
Oscillator strength 9 Supersaturation (of vapor) 129,160-163

Photolytic desorption 121, 164 To -curve 92, 111

Plasma Thermal (blackbody) radiation 39, 128
- annealing hypothesis 26 Thermal runaway 15,21,31,52-54
- enhanced coupling 32,39,142-145, Thermodiffusion 95
148,149 Trapping
- shielding 39,145,148-150 - (of solute atoms) 89-94, 120
- waves, plasmons 25, 158 - (of laser beam) 19.38,39,134,136
Plasmotron 142
Vapor
Radiation pressure 130 - condensation 39, 130, 159-165
Recombination 23-28,42 - deposition 159-165
Refractive index 6,8, 10, 13, 17, 18,20 - expansion 128-131
Reflectance 7,10-14,26,27,31,32,40, - explosion 120,134
52-54,93,166,168 - pressure 116,118-120,130,160,162,
Remelting 69,86,94 163
- numerical simulation of 72-76
Resonance 9,10,12,18 Welding 134-136

Printing: Mercedesdruck, Berlin

Binding: Buchbinderei Liideritz & Bauer, Berlin

194
Springer Series in Materials Science
Advisors: M.S. Dresselhaus . H. Kamimura· K.A. MUller
Editors: U. Gonser· R. M. Osgood, Jr.. M. B. Panish . H. Sakaki
Managing Editor: H. K. V. Lotsch

Chemical Processing with Lasers* 14 Graphite Intercalation Compounds I

By D. Bauerle Structure and Dynamics
Editors: H. Zabel and S. A. Solin
2 Laser-Beam Interactions with Materials
Physical Principles and Applications 15 Crystal Chemistry of
By M. von Allmen and A. Blatter 2nd Edition High-Tc Superconducting Copper Oxides
3 Laser Processing of Thin Films By B. Raveau, C. Michel, M. Hervieu,
and Microstructures and D. Groult
Oxidation, Deposition and Etching 16 Hydrogen in Semiconductors
of Insulators By S.J. Pearton, M. Stavola,andJ. W.Corbett
By. I. W. Boyd
17 Ordering at Surfaces and Interfaces
4 Microclusters
Editors: A. Yoshimori, T. Shinjo,
Editors: S. Sugano, Y. Nishina, and S. Ohnishi and H. Watanabe
5 Graphite Fibers and Filaments
18 Graphite Intercalation Compounds II
By M. S. Dresselhaus, G. Dresselhaus,
Editors: S. A. Solin and H. Zabel
K. Sugihara, I. L. Spain, and H. A. Goldberg
6 Elemental and Molecular Clusters 19 Laser-Assisted Microtechnology
Editors: G. Benedek, T. P. Martin, By S. M. Metev and V. P. Veiko
and G. Pacchioni 20 Microcluster Physics
7 Molecular Beam Epitaxy By S. Sugano
Fundamentals and Current Status
21 The Metal-Hydrogen System
By M. A. Herman and H. Sitter 2nd Edition
By Y. Fukai
8 Physical Chemistry of, in and on Silicon
22 Ion Implantation in Diamond, Graphite
By G. F. Cerofolini and L. Meda
and Related Materials
9 Tritium and Helium-3 in Metals By M. S. Dresselhaus and R. Kalish
By R. Lasser
23 The Real Structure
IO Computer Simulation of High-Tc Superconductors
of Ion-Solid Interactions Editor: V. Sh. Shekhtman
By W..Eckstein
24 Metal Impurities
II Mechanisms of High
in Silicon-Device Fabrication
Temperature Superconductivity
By K. Graff
Editors: H. Kamimura and A. Oshiyama
25 Optical Properties of Metal Clusters
12 Dislocation Dynamics and Plasticity
By U. Kreibig and M. Vollmer
By T. Suzuki, S. Takeuchi, and H. Yoshinaga
13 Semiconductor Silicon 26 Gas Source Molecular Beam Epitaxy
Materials Science and Technology Growth and Properties of Phosphorus
Editors: G. Harbeke and M. J. Schulz Containing III-V Heterostructures
By M. B. Panish and H. Temkin

* The 2nd edition is available as a textbook with tbe title: Laser Processillg alld Chemistry
 Springer
and the
environ ment
At Springer we firmly believe that an
international science publisher has a
special obligation to the environment,
and our corporate policies consistently
reflect this conviction.
We also expect our business partners -
paper mills, printers, packaging
manufacturers, etc. - to commit
themselves to using materials and
production processes that do not harm
the environment. The paper in this
book is made from low- or no-chlorine
pulp and is acid free, in conformance
with international standards for paper
permanency.

Springer