Adv Electronics Electron Physics v61 Volume 61 Adv

ADVANCES IN ELECTRONICS AND
ELECTRON PHYSICS
VOLUME 61
TO THISVOLUME
CONTRIBUTORS
H. BREMMER
D. S. BUGNOLO
M. CAILLER
GILBERTDE MEY
J . P. GANACHAUD
A. G . MILNES
D. ROPTIN
Advances in
Electronics and
Electron Physics
EDITEDB Y
PETER W. HAWKES
Laboratoire d’Optique Electronique
du Centre National de la Recherche Scientifique
Toulouse, France
VOLUME 61
1983
ACADEMIC PRESS
A Subsidiary of Harcourt Brace Jovanovich, Publishers
New York London
Paris San Diego San Francisco Sao Paulo Sydney Tokyo Toronto
COPYRIGHT @ 1983,BY ACADEMIC PRESS, INC.
ALL RIGHTS RESERVED.
NO PART OF THIS PUBLICATION MAY BE REPRODUCED O R
TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC
O R MECHANICAL, INCLUDING PHOTOCOPY, RECORDING, OR ANY
INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT
PERMISSION IN WRITING FROM THE PUBLISHER.
ACADEMIC PRESS,INC.
111 Fifth Avenue, New York, New York 10003
United Kingdom Edition published by

ACADEMIC PRESS, INC. ( L O N D O N ) LTD.
24/28 Oval Road, London N W l 7DX
OF CONGRESS
LIBRARY CARDNUMBER:49-1504
CATALOG
ISBN 0-1 2-014661 -4
PRINTED IN THE UNITED STATES OF AMERICA
83 84 85 86 9 8 7 65 4 3 2 1
CONTENTS
CONTRIBUTORS
TO VOLUME 61 . . . . . . . . . . . . . . . . . . . vii
FOREWORD. . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Potential Calculations in Hall Plates

GILBERTDE MEY
I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
I1. Fundamental Equations for a Hall-Plate Medium . . . . . . . . 3
I11. The Van Der Pauw Method . . . . . . . . . . . . . . . . . . 7
IV . Influence of the Geometry on Hall-Mobility Measurements . . . 9
V . Conformal Mapping Techniques . . . . . . . . . . . . . . . . 10
VI . Relaxation Methods . . . . . . . . . . . . . . . . . . . . . 15
VII . The Boundary-Element Method for Potential Calculations in
Hall Plates . . . . . . . . . . . . . . . . . . . . . . . . . 18
VIII . Improvement o f t h e Boundary-Element Method . . . . . . . . 38
IX . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 48
Appendix 1 . The Three-Dimensional Hall Effect . . . . . . . . 49
Appendix 2 . On the Existence of Solutions of Integral Equations 52
Appendix 3. Green’s Theorem . . . . . . . . . . . . . . . . 54
Appendix 4 . The Hall-Effect Photovoltaic Cell . . . . . . . . . 57
Appendix 5 . Contribution of the Hall-Plate Current to the
Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . 58
Appendix 6 . Literature . . . . . . . . . . . . . . . . . . . 59
References . . . . . . . . . . . . . . . . . . . . . . . . . 59
Impurity and Defect Levels (Experimental) in Gallium Arsenide

A . G . MILNES
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 64
I1. Possible Native Defects and Complexes . . . . . . . . . . . . 65
I11. Traps (and Nomenclature) from DLTS Studies . . . . . . . . . 76
IV . Levels Produced by Irradiation . . . . . . . . . . . . . . . . 81
V. Semi-Insulating Gallium Arsenide with and without Chromium . . 91
VI . Effects Produced by Transition Metals . . . . . . . . . . . . . 100
VII . Group I Impurities: Li, Cu. Ag. Au . . . . . . . . . . . . . . 108
VIII . Shallow Acceptors: Be. Mg. Zn. Cd . . . . . . . . . . . . . . 116
IX . Group IV Elements as Dopants: C. Si. Ge. Sn. Pb . . . . . . . 118
X. Oxygen in GaAs . . . . . . . . . . . . . . . . . . . . . . . 123
v
vi CONTENTS
XI . Group VI Shallow Donors: S. Se. Te . . . . . . . . . . . . . i27

XI1 . Other Impurities (Mo. Ru. Pd. W. Pt. Tm. Nd) . . . . . . . . . 128
XI11. Minority-Carrier Recombination. Generation. Lifetime. and
Diffusion Length . . . . . . . . . . . . . . . . . . . . . . 131
XIV . Concluding Discussion . . . . . . . . . . . . . . . . . . . . 141
References . . . . . . . . . . . . . . . . . . . . . . . . . 142
Quantitative Auger Electron Spectroscopy

M . CAILLER.J . P . GANACHAUD. AND D . ROPTIN
I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . 162
11. General Definitions . . . . . . . . . . . . . . . . . . . . . 167
I11. Dielectric Theory of Inelastic Collisions of Electrons in a Solid . . 173
IV . Elastic Collisions . . . . . . . . . . . . . . . . . . . . . . 185
V . Auger Transitions in a Solid . . . . . . . . . . . . . . . . . 187
VI . Quantitative Description of Auger Emission . . . . . . . . . . 213
VII . Auger Quantitative Analysis . . . . . . . . . . . . . . . . . 244
VIII . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 289
References . . . . . . . . . . . . . . . . . . . . . . . . . 289
The Wigner Distribution Matrix for the Electric Field in a Stochastic Dielectric
with Computer Simulation
D . S . BUGNOLO AND H . BREMMER
I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . 300
I1. The Differential Equation for the Electric Field Correlations . . . 303
111. Derivation of the Equations for the Wigner Distribution
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 313
IV . Related Equations for the Wigner Distribution Function . . . . . 325
V . Asymptotic Equations for the Wigner Distribution Function . . . 330
VI . Equations for Some Special Cases . . . . . . . . . . . . . . . 335
VII . A Brief Review of Other Theoretical Methods . . . . . . . . . 345
VIII . The Coherent Wigner Function . . . . . . . . . . . . . . . . 347
IX . Computer Simulation of the Stochastic Transport Equation for
the Wigner Function in a Time-Invariant Stochastic Dielectric 354
X . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . 382
Appendix 1. A Listing of Experimental Program Number Two
for the Case of an Exponential Space Correlation Function . . . 383
Appendix 2 . A Sample of a Computer Simulation . . . . . . . . 386
References . . . . . . . . . . . . . . . . . . . . . . . . . 388
AUTHORINDEX . . . . . . . . . . . . . . . . . . . . . . . . . . 391
INDEX.
SUBJECT . . . . . . . . . . . . . . . . . . . . . . . . . 402
CONTRIBUTORS TO VOLUME 61
Numbers in parentheses indicate the pages on which the authors’ contributions begin.
H. BREMMER,31 Bosuillaan, Flatgebou’w Houdringe, Bilthoven, The

Netherlands (299)
D. S. BUGNOLO,*Department of Electrical and Computer Engineering,
Florida Institute of Technology, Melbourne, Florida 32901 (299)
M. CAILLER,Laboratoire d e Physique du MCtal, Ecole Nationale
SupCrieure de MCcanique, 44072 Nantes Cedex, France (161)
GILBERTDE MEY, Laboratory of Electronics, Ghent State University,
B-9000 Ghent, Belgium (1)
J. P. GANACHAUD, Laboratoire de Physique du Solide, Institut de Physique
de I’UniversitC de Nantes, Nantes, France (161)
A. G. MILNES,Department of Electrical Engineering, Carnegie-Mellon
University, Pittsburgh, Pennsylvania 15213 (63)
D. ROPTIN,Laboratoire de Physique du MCtal, Ecole Nationale SupCrieure
de MCcanique, 44072 Nantes Cedex, France (161)
*Present address: Department of Electrical and Computer Engineering, University of

Alabama in Huntsville, Huntsville, Alabama 35899.
vii
This Page Intentionally Left Blank
FOREWORD
The boundary between electron physics and solid state physics is by

no means sharp, as several of the contributions to the present volume
show: The articles on Auger electron spectroscopy and on gallium arsenide
bring this out very clearly, and it is one of the functions of a series such
as this that such topics, which span more than one discipline, can be ex-
plored at length. The other two contributions are more theoretical in na-
ture; the article on Hall plates is of course still concerned with semicon-
ductors, though here it is the calculation of the potential distribution that
is of interest. The other theoretical article, dealing with the Wigner dis-
tribution matrix in a stochastic dielectic, covers one aspect of a topic that
is currently attracting much attention; we plan to publish reviews of other
aspects of the Wigner distribution in forthcoming volumes.
It only remains for me to thank very warmly all the contributors to this
volume. The customary list of forthcoming articles is given below, and
as usual I encourage potential contributors to contact me, even if their
plans are still in the preliminary stage.
Critical Reviews:
Electron Scattering and Nuclear Structure G. A. Peterson
Large Molecules in Space M. and G. Winnewisser
Electron Storage Rings D. Trines
Radiation Damage in Semiconductors N. D. Wilsey and J. W.
Corbett
Visualization of Single Heavy Atoms with the Electron
Microscope J. S. Wall
Light Valve Technology J. Grinberg
Electrical Structure of the Middle Atmosphere L . C. Hale
Diagnosis and Therapy Using Microwaves M. Gautherie and A. Priou
Low-Energy Atomic Beam Spectroscopy E. M. Horl and E. Semerad
History of Photoemission W. E . Spicer
Power Switching Transistors P. L. Hower
Radiation Technology L. S. Birks
Infrared Detector Arrays D. Long and W. Scott
The Technical Development of the Shortwave Radio E . Sivowitch
CW Beam Annealing Process and Application for
Superconducting Alloy Fabrication J. F. Gibbons
Polarized Ion Sources H. F. Glavish
Ultrasensitive Detection K. H . Purser
The Interactions of Measurement Principles, Interfaces,
and Microcomputers in Intelligent Instruments W. G. Wolber
Fine-Line Pattern Definition and Etching for VLSI Roy A. Colclaser
ix
X FORE WORD
Waveguide and Coaxial Probes for Nondestructive

Testing of Materials F. E. Gardiol
The Measurement of Core Electron Energy Levels R. N. Lee and C. Anderson
Millimeter Radar Robert D. Hayes
Recent Advances in the Theory of Surface Electronic
Structure Henry Krakauer
Long-Life High-Current-Density Cathodes Robert T. Longo
Microwaves in Semiconductor Electronics J. L. Allen
Applications of Quadrupole Mass Spectrometers I. Berecz, S. Bohatka, and
G. Langer
Advances in Materials for Thick-Film Hybrid
Microcircuits J. Sergent
Guided-Wave Circuit Technology M. K. Barnoski
Fast-Wave Tube Devices J. M. Baird
Spin Effects in Electron-Atom Collision Processes H. Kleinpoppen
Recent Advances in and Basic Studies of Photoemitters H. Timan
Solid State Imaging Devices E. H. Snow
Structure of Intermetallic and Interstitial Compounds A. C. Switendick
Smart Sensors W. G. Wolber
Structure Calculations in Electron Microscopy D. van Dyck
Voltage Measurements in the Scanning Electron
Microscope A. Gopinath
Supplementary Volumes:
Microwave Field-Effect Transistors J. Frey
Magnetic Reconnection P. J. Baum and A. Bratenahl
Volume 62:
Predictions of Deep Impurity-Level Energies in
Semiconductors P. Vogl
Spin-Polarized Electrons in Solid-state Physics H. C. Siegmann, F. Meier,
M.Erbudak, and M.
Landolt
Recent Advances in the Electron Microscopy of D. B. Williams and D.E.
Materials Newbury
PETERW. HAWKES
ADVANCES IN ELECTRONICS AND
ELECTRON PHYSICS
VOLUME 61
ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS VOL . 61 .
Potential Calculations in Hall Plates

GILBERT DE MEY
Laboratory of Electronics
Ghent State Wniversiry
Ghenr. Belgium
I . Introduction ........................................ 2
11. Fundamental Equations for a Hall-Plate Medium ........................... 3
A . General Equations for a Semiconductor ............ 3
B. Approximations for a Thin Semiconducting Layer ... 4
C . Constitutive Relations with an Externally Applied Magnetic Field .......... 5
D . Boundary Conditions with Externally Applied M q n e 6
Ill . The Van Der Pauw Method ........................................... 7
IV. Influence of the Geometry on Hall-Mobility Measurements ................... 9
V . Conformal Mapping Techniques ..................... 10
A . Basic Ideas ......................................................... 10
B. Approximate Analysis of the Cross-Shaped Geometry .................... 11
C . Exact Analysis of the Cross-Shaped Sample ............................. 13
D . Properties of the Cross-Shaped Hall Plate . . .................... 14
V1. Relaxation Methods .................................... 15
VII . The Boundary-Element Method for Potential Calculations in Hall Plates ........ 18
A . Introduction ...................... ............................... 18
B. Integral Equation for the Potential Distribution in a Hall Plate . 18
C . Numerical Solution of the Integral Equation .......................... 20
D . Application to a Rectangular Hall Plate ..................... 21
E . Zeroth-Order Approximation .............................. 22
F. Numerical Calculation of the Current through a Contact . . . . . . . . . . . . . . . . . . 24
G . Application to the Cross-Shaped Geometry .............................. 25
H . Direct Calculation of the Geometry Correction .................. 21
1. Application to a Rectangular Hall Generator .................... 29
J . Application to a Cross-Shaped Geometry ............... 30
K . Application to Some Other Geometries ................. 33
VIII . Improvement of the Bounda 38
A . Introduction ................................................. 38
B. Integral Equation ...... .................... 39
C . Calculation of the Funct .......... 40
D . Application to a Rectangular Hall Generator ................... 43
E. Application to a Cross-Shaped Form ................................... 45
F . Application to Some Other Geometries ................................. 46
1X. Conclusion ............................................................ 48
Appendix 1. The Three-Dimensional Hall Effect .................... 49
Appendix 2. On the Existence of Solutions of Int ations 52
Appendix 3. Green's Theorem ............................................ 54
1
.
Copyright P 1983 by Academic Press Inc.
All rights oi reproduction in any form reserved .
ISBN n-12-014661-4
2 GILBERT DE MEY
Appendix 4. The Hall-Effect PhotovoltaicCell.. ............................ 57

Appendix 5. Contribution of the Hall-PiateCurrent to the Magnetic Field ...... 58
Appendub. Literature .................................................. 59
References ............................................................. 59
I. INTRODUCTION
Hall plates are thin semiconducting layers placed in a magnetic field.

Owing to the Lorentz force, the current density J and the electric field E are
no longer parallel vectors. This means that a current in a given direction
automatically generates a potential gradient in the perpendicular direction.
With suitable contacts (the so-called Hall contacts), a Hall voltage can then
be measured. In a first approximation one can state that the Hall voltage is
proportional to the applied magnetic field, the externally supplied current,
and the mobility of the charge carriers. Knowledge of the current and of the
Hall voltage yields the product pHBof the mobility and the magnetic
field B. This indicates two major applications of Hallcffect components. If
the mobility is known, magnetic field strengths can be measured. On the
other hand, if the magnetic field is known, the mobility can be calculated. The
latter is mainly used for the investigation of semiconductors because mobility
is a fundamental material parameter.
A first series of applications is based on the measurement or detection of
magnetic fields. Measurements of the magnetic fields in particle accelerators
have been carried out with Hall probes provided with a special geometry in
order to ensure a linear characteristic (Haeusler and Lippmann, 1968).
Accuracies better than 0.1% have been realized. For alternating magnetic
fields, Hall plates can be used at frequencies up to f l0,OOO Hz (Bonfig and
Karamalikis, 1972a,b). For higher frequencies emf measurements are recom-
mended to measure magnetic fields. The Hall probe can also be used to detect
the presence of a magnetic field. This phenomenon is used in some types of
push buttons. On each button a small permanent magnet is provided, and
the pushing is sensed by a Hall plate. At this writing Hall plates combined
with additional electronic circuitry are available in integrated-circuit form.
A Japanese company has produced a cassette recorder in which a Hall probe
reads the magnetic tape. The principal advantage here is that dc signals can
be read directly from a tape, whereas classical reading heads generate signals
proportional to the magnetic flux rate d4/dt. Magnetic bubble memories
have also been fitted out with Hall-effect readers (Thompson et al., 1975).
A survey of Hallcffect applications can be found in an article written by
Bulman (1966) that mentions microwave-power measurements, the use of
Hall probes as gyrators, insulators, function generators, ampere meters, etc.
Even a brushless dc motor has been constructed using the Hall effect (Kobus
POTENTIAL CALCULATIONS IN HALL PLATES 3
and Quichaud, 1970). Finally, Hall plates can also be applied as transducers
for mechanical displacements (Davidson and Gourlay, 1966; Nalecz and
Warsza, 1966).
A second series of applications, to which this article is mainly devoted,
involves the measurement of mobilities. A Hall measurement carried out in a
known magnetic field yields the value of p H . This constant is an important
parameter for investigating the quality of semiconducting materials.
Combined with the resistivity, it also enables us to calculate the carrier
concentration. Knowledge of these data is necessary for the construction of
components such as diodes, solar cells, and transistors starting from a
semiconducting slice.
The present article describes the Hall effect and its mathematical
representation. The well-known Van Der Pauw method for Hall-mobility
measurements is then discussed. The influence of the geometry on the Hall
voltage is pointed out using physical considerations. This explains why the
potential distribution in a Hall plate should be known in order to evaluate
the so-called geometry correction. Then several techniques for potential
calculations in Hall plates, such as conformal mapping, finite differences,
and the boundary-element method, are outlined and compared.
11. FUNDAMENTAL
EQUATIONS
FOR A HALL-PLATE
MEDIUM
A . General Equations f o r a Semiconductor

The fundamental equations for an n-type semiconductor (assuming low
injection) are the following:
( a n / a t ) - 4 - 1 V . J , = (an/at)gen - [(P - ~ o ) / ~ p ]
(1)
(a~/at) + q-' V *J p = (ap/af)gen- [(P - ~ o ) / ~ p ]
(substitute [(n - no)/r,,] for a p-type layer);

Jn = nqpnE + qDn Vn, Jp = PqPpE - 4Dp V P (2)
V.E = -Vz+ = (q/c,C)(p - n +ND - NA) (3)
where n is the electron concentration, p is the hole concentration, J , is the
+
electron current density, J,, is the hole current density, J = J , J p is the
total current density, ND is the donor concentration, N A is the acceptor
+
concentration, E is the electric field, is the electric poential, p, = qD,/kT
is the electron mobility, p p = qD,,/kT is the hole mobility, T,, is the electron
relaxation time, T~ is the hole relaxation time, no is the equilibrium electron
concentration (in the p layer), po is the equilibrium hole concentration
4 GILBERT DE MEY
(in the n layer), L is the dielectric constant, and (a/dt),,, is the generation rate
(e.g., due to incident light). Equations (1)-(3) are nonlinear for the unknowns
n, p , and 4; however, for Hall generators several reasonable assumptions can
be advanced so that the final problem becomes linear.
B. Approximations f o r a Thin Semiconducting Layer

For thin-film semiconducting layers with contacts sufficiently distant
from each other (order of magnitude in millimeters), one usually assumes
that the layer is sufficiently doped to ensure that the contribution of the
minority carriers becomes negligible. We shall work later with an n-type
semiconductor; however, the same treatment can be carried out for a p-type
semiconductor.
One also assumes that no space charges are built up in the conductor. It
can be shown that an occasional space charge only has an influence over a
distance comparable to the Debye length. For an n-type layer, the Debye
length is given by (Many et al., 1965)
LD = [(totkT)/q2ND)]”Z (4)
Normally, LD varies around 100-1000 A, so that a space charge can only be
felt over a distance much smaller than the distance between the electrodes.
Practically, space charges can only be realized at junctions or nonohmic
contacts. Because Hall generators are provided with ohmic contacts, the
space charge is zero everywhere. Hence the right-hand member of (3) should
vanish; for an n-type material this gives rise to
n = ND and p<< n (5)

The Poisson equation (3) is then reduced to the simpler Laplace equation
-V.E = V24= 0 (6)
From Eq. (5), it also follows that V n should be zero. This means that the
current density J , only consists of the drift component qp,,nE. The hole
current J , can be put at zero because both p and V p are negligible. One
obtains for the current density
J = J , = NDqp,E = CJE (7)
A thin semiconducting layer can be seen simply as a sheet with a constant

conductivity CJ.
For sufficiently high doping concentrations ND, the minority-carrier
concentration p can be put equal to its equilibrium value p o , hence p - po =
0. In the time-independent case, Eq. (1) reduces to
if we suppose that the generation term (an/dt),,, vanishes, which will be the
case if the layer is not illuminated or irradiated.
Equations (6) and (8) constitute the fundamental equations for a semi-
conducting layer, J and E being related by Eq. (7). From these equations, the
boundary conditions can easily be deduced. At a metallic contact the poten-
tial 4 should be equal to the applied voltage. At a free boundary the current
density must be tangential:
J-U,= 0 (9)
where u,, is the normal unit vector. Owing to Eq. (7), the boundary condition
(9) is equivalent to
v(p*u, = 0 (10)
The potential problem in a thin semiconducting laver is reduced to the
solution of the Laplace equation in a given geometry, 4 or Vq5.u, being
known on each point along the boundary. This is a classical potential
problem with mixed boundary conditions.
C. Constitutive Relations with an Externally Applied Magnetic Field

The fundamental equations (1) and (2) are still valid in the presence of an
externally applied magnetic field. Only the constitutive relations (2) have to
be extended in the following way (Smith et al., 1967; Madelung, 1970):
where the index i denotes the ith component in a rectangular coordinate

system. A similar expression can be written for the hole current density
Jp,i.The tensorial mobility pij in the presence of a magnetic field B is found
by the well-known Jones-Zener expansion, and it turns out that pij has to
be replaced by
Applying Eqs. (1 1) and (12) on a flat n-type semiconductor layer, one obtains
(De Mey, 1975)
J = oE + qDVn - opH(E X B) - qC(HD(Vn X B) (13)
Using the approximation (9,as has also been done in the foregoing section,
the diffusion components in Eq. (13) can be dropped, which yields
J = CE - opH(E x B) (14)
This relation is used later for potential calculations in Hall plates. It has also
been assumed that the semiconductor is isotropic, which explains why only
one pH coefficient remains in Eq. (14).
6 GILBERT DE MEY
Equation (14) is derived from the Jones-Zener expansion (12) neglecting

terms of order pAB2 that describe the physical magnetoresistivity. Hence
Eq. (14) can be inverted to
E = pJ + ppH(J x B) (15)
where p = l/o denotes the resistivity. Equation (15) is also correct to terms
of order pHB.
The fundamental equations are still V J = 0 and V x E = 0, but for a
flat semiconducting Hall plate in a uniform magnetic field B, one can show
the following:
V * J = C T V . E - I T ~ H V . ( EBX) = a V . E - a p H ( V x E ) . B = o V * E = O
V x J = O V x E - opc,V x (E x B) = -opH(B.V)E + opH(V.E)B = 0
-
in which (B V)E vanishes since B is directed perpendicular to the Hall plate,
whereas E is parallel to it. To solve the problem one can use either V x E =
V - E = 0 or V x J = V - J = 0; E can be derived from a potential 9. For
reasons which are explained in Section II,D, the current density J can also
be derived from a potential function i,b in the following way:
J = U, x Vi,b (16)
where u, is the unity vector directed perpendicular to the Hall plate. From
V x J = 0 it can be easily proved that i,b also satisfies the Laplace equation.
D. Boundary Conditions with an Externally Applied Magnetic Field

Let us start with the electrostatic potential 4. At a metallic contact 4 is
equal to the applied contact voltage. At a free boundary, the current density
should be tangential:
J - U , = oE'u, - ~/AHBE*u,
=0 (17)
or
v4.11, = ~ H B V $ . U , (18)
where u, = u, x u, is the unit tangential vector along the boundary (Fig. 1).
It should be noted that at a free boundary in a Hall plate the electric field can
show a nonvanishing normal component.
For a p-type semiconductor, the same calculations can be carried out. A
minus sign will then appear in the right-hand member of Eq. (18).
For Hall plates the current but not the potential through a contact is
usually given. It is then easier to introduce the current potential defined by
Eq. (16). At a metallic contact E should be perpendicular, or the tangential
FIG. 1. Hall-plate configuration for outlining the stream potential.
component E*u, must vanish. With Eq. (14) this gives rise to
~ E . u=, J . 4 + pH(J x B).u,= 0 (19)
or
V$.U, + ~ H B V $ * U=, 0 (20)
At a free boundary, J must be tangential, and due to Eq. (16) $ must be a
constant (Fig. 1).
At BB' the boundary value can be taken $ = 0. At the opposite side AA',
the value $o can be found from the known current 1 injected through the
con tact s :
I = - JAB' J . u , d l = JAB V$*u,dl = $(B') - $ ( A ' ) = -$o (21)
A similar treatment can be performed if more than two contacts are involved.
111. THEVAN DERPAUW METHOD
Van Der Pauw (1958)has presented an ingenious method for carrying out
resistivity and Hall-mobility measurements on thin layers with arbitrary
shape. In this article we shall restrict ourselves to Hall-mobility measure-
ments. It should be noted that Van Der Pauw's theory is only valid if the
following four conditions are fulfilled:
The layer must be perfectly flat
(1)
(2) The four contacts must be point shaped and placed along the
boundary
8 GILBERT DE MEY
(3) The layer must be homogeneous

(4) The geometry must be that of a singly connected domain
Only the second condition (point-shaped contacts) is difficult to meet. This
article is therefore mainly devoted to the influence of finite contacts on
Hall-mobility measurements.
In order to determine the mobility pH,a current I is fed through two
opposite contacts A and C (Fig. 2). The voltage across the two other contacts
B and D is then measured with and without the magnetic field B. The differ-
ence gives us the so-called Hall voltage. If the magnetic field B = 0, the
voltage drop V , between the contacts B and D turns out to be
When the magnetic field B is applied, one finds a voltage V2 between B and D
given by
V2 = = pl:J.dr + p p H I I ( Jx B).dr (23)
Because the contacts are assumed to be point shaped, the boundary con-
dition J-u, = 0 holds along the entire boundary. Since the basic equations
and boundary conditions for J are unaltered by the magnetic field, one
concludes that the current density field J remains unchanged. The J vector in
Eq. (22) is thus the same as in Eq. (23). The Hall voltage is then found to be
f-B
VH = V2 - V , = P,UH J (J x B).dr
D
with B = Bu,(u, directed perpendicular to the Hall plate):
FIG.2. Hall generator of arbitrary shape placed in a magnetic field B.

where d represents the thickness of the layer. Equation (25) enables us to

determine the mobility pH by measuring the Hall voltage provided that
p, B, I, and d are known.
OF THE GEOMETRY
1V. INFLUENCE ON
HALL-MOBILITY
MEASUREMENTS
Equation (25) of Van Der Pauw is only valid if the contacts are point
shaped. Actually, Hall generators always show finite contacts, and this will
alter the Hall voltage in a still unknown way. It is shown further on in this
article that the so-called geometry correction can be calculated provided
that the potential problem in a Hall plate is solved.
At this stage it is necessary to emphasize that there are two different kinds
of size effects in semiconductor components. The first (and best known)
effect is of a purely physical nature. Consider a semiconductor slice that is
very thin, in which the mean free path of the charge carriers can become
comparable to the thickness. One can easily understand that the mobility
will then depend upon the size of the sample (Ghosh, 1961). But in our case
the geometry (i.e., finite contacts) has no physical influence on the mobility
but will affect the measured Hall voltage. This second kind of size effect is of
a purely metrological nature.
We shall now review Section 111 for the case of finite contacts on a Hall
plate. Equations (22) and (23) are still valid; however, as J-u, = 0 no longer
holds along the entire boundary, the current field J will change in the presence
+
of a magnetic field. It is then necessary to replace J by J AJ in Eq. (23),
where AJ is the change in the current field caused by the magnetic field B.
The Hall voltage VH is then found to be
rB
VH = V , - V, = p ] AJ-dr - ppHB(l/d)
D
The absolute error ApH introduced by neglecting the influence of the finite
contacts is then
rs
ApH= (d/Bl)J AJ-dr
D
The problem is now to develop a geometry for which the correction (27) is as
small as possible in spite of the finiteness of the contacts. One has to find an
integration path from D to B in an area where AJ zz 0. This can be done if
contacts are placed at the ends of rectangular strips (Fig. 3). From field
10 GILBERT DE MEY
FIG.3. Hall generator with contacts placed at the ends of strips.
calculations presented in the next sections, it is shown that disturbances due

to the magnetic field (i.e., AJ # 0) only occur near the contacts. These areas
are shaded in Fig. 3, and an integration path is drawn which only goes
through the shaded areas at D and B. In Hall measurements, however, these
contacts are used to measure potentials, and hence the total current through
them is zero. Therefore J and, afortiori, AJ are very small in the neighbor-
hoods of D and B. The conclusion is that the correction (27) will be small for
the geometry of Fig. 3 where the contacts cannot be considered “point
shaped.”
Finite contacts with dimensions comparable to those of the Hall plate
offer several advantages. The resistance between two contacts is low. Hence
the noise induced in the measuring circuitry will be reduced. Small contacts
require precise mask positioning and hence cause technological problems
which can be avoided by using bigger contacts. The only disadvantage is the
introduction of an error ApH; however, in the following section we prove
that ApH can be calculated from the potential distribution in a Hall plate. It is
then rather convenient to call ApH a geometry correction instead of an error.
V. CONFORMAL TECHNIQUES
MAPPING
A. Basic Ideas
Conformal mapping is a very useful method for solving the Laplace
equation, provided that a geometry can be found for which the problem is
solved by inspection. For a Hall-plate problem such a geometry can be
found (Fig. 4). Indeed, for the parallelogram of Fig. 4, field lines are perpen-
dicular to the contacts AB and A’B‘. The current lines are parallel to the free
sides AA’ and BE’. The homogeneous E field obviously satisfies the Laplace
equation (6). One can easily verify that the boundary condition (18) is
A
FIG.4. Homogeneous field and current lines in a Hall plate having thz shape of a parallelogram.
exactly fulfilled if the angles at A, B, A', and B' have the values (n/2) - OH or
( 4 2 ) + as shown in Fig. 4. By mapping a given Hall-plate geometry on
the parallelogram of Fig. 4, the potential problem can be solved for a given
value of PHB = tan 8,.
Wick (1954) was the first to use conformal mapping techniques for the
study of the Hall effect. By using the Schwarz-Christoffel transformation
formula, the mapping function could be determined. For quite regular Hall
plates, however, calculations turned out to be very complicated, even for a
simple rectangular shape with two contacts (Lippman and Kuhrt, 1958a,b).
A considerable amount of work has been done by Haeusler (1966,1968,1971)
and Haeusler and Lippmann (1968), who calculated the Hall voltage for a
rectangular Hall plate provided with four finite contacts. The calculations
are so complicated that this method cannot be considered a general rule, i.e.,
applicable to all shapes of Hall plates. We present a semianalytical version
of the technique in which the Schwarz-Christoffel transformation formula
is still used, but the evaluation of various integrals and constants is done
numerically (De Mey, 1973a,b); we shall give an example of this later.
B. Approximate Analysis of the Cross-Shaped Geometry

A very useful geometry for Hall-mobility measurements is the cross-
shaped sample (De Mey, 1973b). It combines broad contacts with a low
geometry correction (27), which is now calculated. The cross-shaped form
(Fig. 5b) can be mapped onto a circle by the following equations (Haeusler
and Lippmann, 1968):
y = tan2 26 (28)
h/l = {2[(k/h) - l]}-'

12 GILBERT DE MEY
b c d e f g
H I J K L M N 0 A B C D D , E F G
FIG.5. Conformal mapping of a cross-shaped geometry (b) onto a parallelogram (a).
where F is the hypergeometric function. In the case that c sz 0, i.e., the

mapped contacts on the circle are small, one can introduce several approxi-
mations in Eqs. (28)-(30) so that one obtains finally (De Mey, 1973b)
2c = 2.58e-"hi' (31)
In his famous article, Van Der Pauw (1958) has given the following formula
for the geometry correction in the case of a Hall sample with circular
geometry and one finite contact:
ACIHICIH = 2ch2 (32)
A
-
20-
10-
P
\ -
5-
\s-
2 -
2-
I=
-
*0.5-
0 0.5 I I.5 2
h/l
FIG.6. Geometry correction as a function of h / / for a cross-shaped sample.
Note that Eq. (32) is only valid if 6 is sufficiently small. For four contacts, one
can simply multiply Eq. (32) by 4. For the cross-shaped sample, this yields
Ap(H/p(H = 4(2€/a2) = 1.0453e-"h/' (33)
This relationship is shown in Fig. 6. Note that for h = 1, the correction
ApH/pH becomes 4.5%, whereas for h = 21 a negligible value of 0.19% is
found. A correction of 5% means that experimental pH values have to be
increased by 5% in order to obtain the exact value.
C . Exact Analysis of the Cross-Shaped Sample

In order to calculate the potential distribution and hence the Hall voltage
in a cross-shaped geometry, the Hall plate has to be mapped onto the
parallelogram of Fig. 5a (De Mey, 1973b). For this simple geometry, the field
distribution is homogeneous. The current density is found to be
IJI = I/(IANIcos8(H) (34)
where I is the current supplied through the contacts AN and HG. I AN I
denotes the length of the contact A N . For unit values of I and p, the Hall
voltage is nothing other than the vertical distance between the Hall contacts
CDE and JKL (Fig. 5a), or
VH = IEI(IACI - ILNI)COS8H X (IACI - I L N I ) / ( A N ) (35)
14 GILBERT DE MEY
According to Van Der Pauw's theory, V , should be equal to pHB= tan OH,
and by comparing it with the value given by Eq. (35), the geometry correction
&/pH is found.
The conformal mapping of a circle into a semiinfinite plane is well known
and is described by the relation
x = tan(O/2) (36)
This half-plane is mapped onto the cross-shaped geometry according t o the
Schwarz-Christoffel transformation formula:
From Eq. (37), the values of h and I can be found:
I = I-+;
(u2 -dxx 2 ) 1 ' 2 ((c2
(b2 - x2)(f2 - x 2 )
- x2)(e2 - x2)(g2 - x2) >"'
h= job dx
[(b - x)(x - a)]'" (
(b x)(b2 - x 2 ) ( f 2 - x')
-
(x + u)(c2 - x2)(e2- x2)(g2- x 2 )
These are Gaussian-type integrals and can be easily evaluated numerically
(Abramowitz and Stegun, 1965).
The conformal mapping of the half-plane onto the parallelogram is
performed by
~ (C )z- ( P ~ / ~ ( Z - d ' ) ( z - e)-v2/n
dwldz = A ( z - u ) - ~ ~ / - ( z - g)-(Pl/n
x (z + U)-+'~/~(Z + C ) - ~ ~ / ~+( Zd " ) ( z + e)-'+'lln(z + g)-'+'2/n (39)

The unknown constants d' and d" in Eq. (39) are determined by saying that
the point E should coincide with C and J with L. From Eq. (39) one can
determine the distances I ACI - I L N I and I A N I needed to evaluate the Hall
voltage V,. These calculations also lead to Gaussian-type integrals, but
further details about the numerical procedure are omitted here. Figure 6
shows several numerically calculated values of ApH/pH.
We observe good agreement with the approximate equation (31). These
results also prove that conformal mapping can be adequately applied to
rather complicated geometries if one starts immediately with numerical
integrations once the Schwarz-Christoffel transformation formula has been
written.
D . Properties oJ the Cross-Shaped Hall Plute

The cross-shaped sample has contacts whose lengths are comparable to
the dimensions of the semiconductor layer. This means that the resistance
between two contacts is low, which offers many advantages when low Hall
voltages are to be measured. Technological problems caused by small
contacts no longer occur. Moreover, the geometry correction ApH/pH is
moderate. Even for h = 1, less than 5% is found. This also means that the
position of the contacts is not very critical. High-precision mask positioning
before evaporation of the contacts is no longer required.
VI. RELAXATION
METHODS
The relaxation method is the only purely numerical method that has been
used to calculate the potential distribution in a Hall-plate medium (Newsome,
1963; Grutzmann, 1966; Chwang et al., 1974; Mimizuka, 1971, 1978, 1979;
Mimizuka and Ito, 1972). If one is only concerned with the potential dis-
tribution for given voltages at the contacts, the relaxation method gives good
results; however, if the current through a contact has to be calculated, the
method is far from accurate. Impedance calculations carried out by Mimi-
zuka (1978) show relative errors of 10%. On the other hand, the relaxation
method allows us to include nonlinear effects such as the temperature effects
inside the Hall plate, a problem which cannot be treated with other techniques
(Mimizuka and Ito, 1972; Mimizuka, 1979).
We now use the relaxation method to calculate the geometry correction
ApH/pH for a cross-shaped plate. Because this involves knowledge of both
the Hall voltage and the supply current, we expect moderate accuracy. This
is mainly owing to the tangential derivative V 4 u, in the boundary condition
( 1 8), which is difficult to represent numerically.
In order to represent the Laplace equation numerically, the cross-shaped
form has to be divided into a mesh (Fig. 7). The Laplace equation is then
approximated by the well-known five-points formula :
44i.j - 4i+1 . j - A- 1.j - 4i.j+1 - 4i,j- 1 = 0 (40)
At a contact, the known potential values have to be inserted into Eq. (40).
The boundary condition (18) can be written as (Fig. 8)
1
4i,j - 4i- 1 . j = pHBT(4i,j+1 - $i,j- 1 (41)
-
At a free boundary, the Neumann boundary conditionV4 u, = 0 is written as
4i.j = 6(4i,j+1 + 4i.j-1 + W i - 1 . j ) (42)
which corresponds to putting 4’ = 4i-l . j (Fig. 8). This condition (42) gives
good results in numerical calculations. Note that Eq. (41) cannot be reduced
to Eq. (42) when p H B tends to zero. If pHB = 0, Eq. (41) reduces to 4i.j=
16 GILBERT DE MEY
0 0 0 0 0 0 0 0 0 0 0
FIG.7. Grid pattern for the finite-difference approximation in a cross-shaped Hall plate.
FIG. 8. Boundary conditions for the finitedifference approximation.
THEORETICALVALUE= L 51 90
: s : : : : : : : : : : : +
10 15
M f=NI
FIG.9. Geometry correction as a function of M :p H B= 0.1 ;h/l = 1.

h/I
FIG. 10. Geometry correction versus h/l calculated with finite difference approximation:
N = 7 ( 0 ) .10 ( W),
and 13 ( A ) .
4i- a boundary condition leading to a reduced accuracy and hence never

used. For Hall plates, only Eq. (41) can be applied. This fact can explain the
poor results.
Figure 9 represents the numerically calculated geometry correction
ApH/pH for h = 1 as a function of M (= N).Note that the results show errors
of more than SO%, and the convergence is extremely poor when the mesh
number M increases. Figure 10 represents the geometry correction A/.iH/pH
as a function of h / l . The analytic approximation (33) has also been drawn.
Only for low h / l values were acceptable results obtained. For higher h / l
values the calculations are meaningless.
We conclude that the relaxation method is not suitable for calculating
the geometry correction ApH/pH of a Hall plate. Several reasons can be
advanced. It turns out that contact current cannot be calculated with
sufficient accuracy because it involves numerical differentiations. Also, the
numerical representation (41) of the boundary condition (18) is not optimal
from a numerical point of view. It should be noted, however, that the high
relative errors found here are made on a geometry correction, i.e., the relative
difference between the numerical and theoretical Hall voltage, two numbers
that are normally close together. The accuracy on the geometry correction
will hence be much lower than the accuracy obtained on the Hall voltage.
18 GILBERT DE MEY
VII. THEBOUNDARY-ELEMENT METHODFOR POTENTIAL

CALCULATIONS IN HALLPLATES
A . Introduction
Several years ago, the boundary-element method (BEM) was a rather
unknown technique. Its popularity has since grown considerably, and the
method now competes with the finite-element method (FEM). In recent
years, some textbooks on the BEM have been published (Jaswon and Symm,
1978; Brebbia, 1978a,b).
The BEM replaces the given Laplace equation and the boundary
conditions by a single integral equation. This equation involves only the
boundary of the given geometry. For a numerical solution the boundary has
to be divided into a number of elements. This explains the term boundary-
element method.
Because only the boundary is involved, the complexity of the problem is
reduced. The method requires less storage allocation and computation time
if it is solved numerically. The BEM can also be programmed on small-size
computers (e.g., 60K memory) in contrast to the FEM, which requires a
large memory.
The integral equation technique was very well known to people working
in the area of electromagnetic fields. Presumably, the reason is that many
problems in electromagnetic theory can be directly formulated in terms of
integral equations (Edwards and Van Bladel, 1961; Mei and Van Bladel,
1963a,b). Integral equations have also been used in other fields such as
thermal diffusion problems (Shaw, 1974; De Mey, 1976a, 1977a), drift-
diffusion problems (De Mey, 1976b, 1977b), elastic problems (Brown and
Jaswon, 1971; Symm and Pitfield, 1974),calculation of eigenvalues (De Mey,
1976c, 1977c), semiconductor-component analysis using abrupt depletion-
layer approximation (De Visschere and De Mey, 1977; De Mey et al., 1977),
and potential distribution in Hall plates (De Mey, 1973c, 1974a, 1976d,
1977d).
B. Integral Equation f o r the Potential Distribution in a Hall Plate

We consider the rectangular Hall generator shown in Fig. 11. The
potential satisfies the Laplace equation (6) and the boundary conditions
4= V, at A'B', 4 = 0 at A B
(43)
V4.u" = p H BV ~ . U , at AA' and BB'
In order to construct an integral equation, the Green's function G(r I r') of
the Laplace equation is used:
G(r1r') = (27c)-' lnlr - r'I (44)
FIG. I I . Rectangular Hall plate.
One can easily verify that G(r I r’) satisfies the Laplace equation provided
that a delta function is put in the right-hand member:
V:G(rlr’) = d(r - r’) (45)
The solution of our problem 4(r) is now written as
where p(r) is called the source function and is only defined along the boundary
C of the Hall-plate medium. Owing to Eq. (49, the proposed solution (46)
automatically satisfies the Laplace equation for every point r inside C.
In order to determine the unknown source function p(r), one has to
impose the boundary conditions (43) on the proposed solution (46). This
gives rise to
V,, rEA’B’
&fcp(r’)lnlr - r’(dC’= (47)
0, rEAB
where r E AA’ and r E BB’. Equations (47) and (48) constitute an integral
equation in the unknown source function p. Once this integral equation has
been solved, the potential 4 (and hence all quantities which can be deduced
from 4) can be found at an arbitrary point r by evaluating the integral (46).
The first term - p ( r ) / 2 ~occurring in Eq. (48) is a so-called fundamental
discontinuity. It is caused by the discontinuity of the normal component
V$*un at the boundary. The gradient V+ can be calculated at an arbitrary
point r by
V4(r) = (271.)-’ p(r’)[(r - r’)/(lr - r’l’)] dc‘ (49)

$C
20 GILBERT DE MEY
FIG. 12. Coordinate system (t, q ) to explain the fundamental discontinuities around ro.
If r lies on the boundary, the integrand of Eq. (49) becomes infinite because r
can coincide with r'. For the normal component the integrand should be
treated carefully by taking r close to the boundary at the interior side (Fig. 12).
The boundary C can then be approximated by a straight line and p(r) can be
considered a constant p. Introducing a (5, q ) Cartesian coordinate system,
one has
5
2
By taking the limit 5 + 0 (i.e., r tends to the boundary C), the term -p(r)/2
remains. This problem does not occur for the tangential component
(r - r')-u, because this integrand is an odd function of q. There are no
fundamental discontinuities for tangential components.
C. Numerical Solution of the Integral Equations

In order to solve the integral equations (47) and (48) numerically, the
boundary C has to be divided into n elements ACi. In each of them, the
unknown source function p is replaced by an unknown constant p i . Denoting
ri as the center point of the element ACi, the expression (46)for the potential
can be rewritten as
n
(#l(r)= 1 p.G(r(ri)IACjl
j= 1
J (51)
where IACjI denotes the length of the element ACj. In a similar fashion, the
-
integral equations (47) and (48) can be discretized. However, if r coincides
with one of the ris, one obtains G(ri 1 ri) In 0 = co ! In order to avoid this
infinity, the self-potential, i.e., the potential at ri due to the source pi itself on
Aci, has to be calculated otherwise. Replacing pi by a uniform constant

function along Aci, the divergence no longer occurs and the self-potential
turns out to be
The integral equations are then replaced by

PiIAciIClnOlAciI) - 11
V,,, riE A'B'
+ C" -pjIn(ri
1
- rj(IAcjl =
0, ricAB
(53)
j=l27L
j#i
= 0, riE AA', BB'

Equations (53) and (54)constitute a linear algebraic set with n equations and n
unknowns pi, which can be easily solved numerically by the Gauss pivotal
elimination method, for example. The potential can then be calculated at
an arbitrary point r by using Eq. (51).
D. Application t o a Rectangular Hall Plate

The rectangular Hall plate shown in Fig. 11 has also been treated by
conformal mapping techniques (Lippmann and Kuhrt, 1958a,b; De Mey,
1973a). The Hall voltage V , induced between the opposite point contacts P
and Q has been calculated accurately. For a square-shaped Hall plate, one
has V , = 0.522654 if V , = 1 and 0, = 45" or c(HB = 1. By comparing this
value with the numerical results, the accuracy of the BEM can be checked.
Figure 13 represents the relative error on the Hall voltage V , as a function
of n, where n is the number of unknowns on each side (hence 4n unknowns
have to be determined for this problem). One observes that good accuracies
are obtained. Precision of better than 1% can be attained with a moderate
number of unknowns.
The linear behavior of the results shown in Fig. 13 is remarkable. When
solving a problem with the BEM the relative accuracy always varies ac-
cording to a l/n law. In contrast to most other numerical methods, this l / n
law does not change when other algorithms for representing the source
function are used. It should also be noted that a theoretical explanation for
this phenomenon has not yet been published.
22 GILBERT DE MEY
I I I 1 1 1 1 1 1 1 1 *
3 1 5 7 9 11 13 151719
n
FIG.13. Relative error on the Hall voltage versus n for a rectangular plate: p H B= I ; V, = 1.
E. Zeroth-Order Approximation
It is a general experience in numerical analysis that higher order approxi-
mations yield more accurate results for the same number of unknowns. For
example, we can cite that the fact that Simpson's rule for the numerical
evaluation of an integral will be more accurate than the trapezoidal rule. In
the preceding section a rather crude approximation for the source function p
was introduced. Instead of replacing p by a series of constants pi concen-
trated at the points ri, one can see p as a piecewise constant function. In each
element Aci, p equals a constant pi.The expression (51) for the potential now
has to be written as
The integral equations (53) and (54) now read
1" p j g1I A c J l n l r -i r ' I d c ' =

j= 1
V,,
0
riEA'B'
riEAB
where ri E AA', BB'. This method has also been applied to the square-shaped
Hall generator. Figure 14 represents the relative error of the numerically
calculated Hall voltage VH. The results obtained in the preceding section
4-
3-
2-
E
4
3
\
c 1-
0.5-
3 -4 5 7 9 11 13151719
n
FIG. 14. Comparison between the direct method and a first-order approximation.
have been redrawn in Fig. 14. One observes a slightly lower error due to the
zeroth-order approximation. It is also remarkable that the same l/n law
appears again. Similar results have also been found with higher order
approximations of the source function (Stevens and De Mey, 1978). These
results are in full agreement with the statements made at the end of the
foregoing section. A l/n law always occurs, and the relative errors d o not
decrease remarkably when higher order approximations are used for the
source function.
The integrals appearing in Eqs. (55)-(57) can easily be evaluated by
complex integrations. By drawing an ( x , y ) coordinate system with the x
axis directed tangentially and the origin at ri (Fig. 15), the integral appearing
t’ r z x Cx*yCy
*
Zzxejy
FIG. 15. Complex plane used to evaluate some integrals.

24 GILBERT DE MEY
sAc,
in Eq. (55) can be written as
In ) r - r‘ I dC’ = Re
s + IAcJ1/2
- lAcJl/2
Indeed, the real part of ln(z - z’) is nothing other than lnlr - r’I and
In(z - z’)dz’ (58)
dc’ = dz’ along the x axis (Fig. 15). The ( x , y ) plane is considered to be a
complex plane, and the points z and z‘ correspond to r and r’. The complex
integral appearing in Eq. (58) is easily evaluated analytically:
s + lAcj112
-IAcjI/Z
In(z - z’)dz’ = [-(z - z’)ln(z - z’) + (z - z ’ ) ] ~ : ~ ! l ~ ~ l \ ~
If one uses a computer with complex-number facilities, all integrals appearing

in Eqs. (55)-(57) can be easily calculated as the real part of an analytic
function.
F. Numerical Calculation of the Current through a Contact

Using the constitutive relation (14) between the current density J and
the electric field E, the total current through a contact AB (Fig. 16) is given by
I = C J - u n d C= 0 E.u,dC (59)
JAB
because the tangential component of E is zero at a metallic contact. From

Eq. (46) the electric field in an arbitrary point r is found to be
r
E(r) = -V$ = - p(r’) VG(r1r’)dC’
9 C
The current I is then given by
I = -Os~dCun.~~p(r’)VG(rIr.)dC. (61)
If one intends to calculate Eq. (61) numerically, the electric field has to be
found at a number of points on the contact AB, and then the summation
along AB should be carried out. This procedure will require a lot of com-
putation time, and the accuracy will be moderate because two numerical
FIG. 16. Use of the angle a to calculate current through AB

approximations will be carried out. It is more convenient to interchange the

sequence of the integrations in Eq. (61), which gives rise to
I = -a~~p(r’)dC~~~’dCu..VG(rlr.) (62)
The last integral appearing in Eq. (62) is nothing other than the flux of VG
through AB. As G(r I r’) depends only upon the distance I r - r‘ 1, the field VG
will show a radial symmetry around r’. Hence the flux of VG through AB is
given by the angle a normalized to 2n (Fig. 16). The integral (62) reduces to
I = -(a/2n)
h dC’p(r’)a(r’,A, B )
where a(r‘,A, B) denotes the algebraic value of the angle under which AB is
seen from the point r’. For the numerical calculations the evaluation of an
angle poses no particular problems, and a single summation is sufficient. If
r’ coincides with a point on AB, then a = n and the flux is then 4.This is
another view of the fundamental discontinuities discussed in Section VII,B.
The numerical calculation of contact currents is very important for the study
of Hall generators because these devices are normally operated under a
constant current through two opposite contacts generating a Hall voltage
at the other contacts.
G. Applicution to the Cross-Shaped Geometry

The BEM is now applied to the calculation of the geometry correction
of the cross-shaped sample. The boundary conditions are shown in Fig. 17.
v0.+IH 8V@.q V@ .rn=,+,so@ ..;
--
@ =-vH/2 qj=h/2--
IH IH
v@..‘,=p,,sv@.q ~#.u;l=~,sv@.u;
.
26 GILBERT DE MEY
FIG. 18. Hall voltage ( VH), supply current (I), and geometry correction (ApH/pH)as a func-
tion o f m : h / / = I ; p H B= 0.1.
Normally, the source current I is given and the Hall current is zero. It is
difficult to impose the total current through a contact as a boundary con-
dition because this often leads to an ill-conditioned algebraic set. Therefore
the potentials at the contacts are given. The current-supplying contacts have
potentials 4 = + 1 and - 1. The supply current I is then found from the
potential distribution, as outlined in the preceeding section. The calculations
are carried out twice, once with V , = 0 and once with V , = pHB.Each time
the Hall current I , is also calculated. As V , varies linearly with I,, the correct
Hall voltage is then found by putting I, = 0.
Figure 18 shows some numerical results obtained for a geometry with
h/l = 1. The Hall voltage V,, the supply current I, and the geometry correc-
tion ApH/pH have been drawn as a function of m where m is the total number
of unknowns used during the numerical procedure. The geometry correction
is far from accurate, as can be seen by comparing the numerical value of 12%
with the exact value 4.5%. From this consideration one may not conclude
that the BEM method is useless. One calculates the Hall voltage numerically
and compares it with the theoretical value (25) according to Van Der Pauw's
theory. In order to obtain the geometry correction these two values, which are
normally close to each other, have to be substracted. This will increase the
relative error so that the numerically calculated geometry correction cannot
be used. Nevertheless, the potential and field calculations in the Hall plate
are still accurate. If one wants to calculate ApH/pH with a 10% accuracy (for
example), the Hall voltage and the supply current have to be known with
much higher precision. In Section VII,H, a method is presented for finding
- 0.86
- 5
21 60 120
m
FIG. 19. Difference between calculated and exact geometry correction versus m. Note that
double logarithmic scale is used.
ApH/pH directly, without the intermediate calculation of the Hall voltage

VH or the supply current I .
The same calculations have also been carried out with other h/l values and
other numerical approximations for the source function. The results and
hence the conclusions remain identical.
If the difference between the calculated and exact geometry correction is
plotted on a double logarithmic scale as a function of m, one obtains linear
behavior with a slope -0.86 (Fig. 19). This agrees with the results shown in
Figs. 13 and 14. It also proves that the numerical values converge, although
very slowly, to the exact value.
H . Direct Calculation o j the Geometry Correction

The geometry correction can be calculated directly, and, as is proved
later, this can be easily performed if one introduces the stream potential II/
defined by Eq. (16). The technique is outlined first for a Hall plate with two
point-shaped Hall contacts, and a formula is then derived for an arbitrary
geometry.
Let us consider the rectangular Hall plate with two point-shaped Hall
contacts P and Q (Fig. 20). A current 1 is supplied. Figure 20 indicates the
boundary conditions if the stream potential t,b is to be used. From Eqs. (15)
and (16) one finds the electric field
E = PP,BVt,b + P(U, x V*) (64)
28 GILBERT DE MEY
FIG.20. Boundary conditions for the stream potential on a rectangular plate.
where uz is the normal unit vector perpendicular to the Hall plate. In order
to obtain the Hall voltage VH, the electric field (64) has to be integrated
between the Hall contacts P and Q. One obtains
VH = j:E-dr = pp.sj;V+-dr + psypC, x v+).dr
= PPHB[$(P) - (Q)]- P
J; v$.(U; X dr)
= pPHB(l/d) - P IQP
v$ ’(u: dr) (65)
By Eq. (65),the Hall voltage VHcan be found from the stream potential $. It is
remarkable that the first term in the right-hand member of Eq. (65)is nothing
other than the theoretical Hall voltage (25) given by Van Der Pauw’s theory.
Hence, the remaining term in Eq. (65)is the difference between the theoretical
and the actual Hall voltage, or the so-called geometry correction. We remark
FIG.21. Stream potential used on an arbitrary shape.

that the last integral in Eq. (65) is the flux of V+ through a line connecting P
and Q. It has also been assumed that the flux V + * (u, x dr) is zero if PHB = 0,
which is certainly the case for the symmetrical configuration of Fig. 20. The
conclusion is that the geometry correction can be calculated directly as the
flux of the V+ field. This can be calculated numerically by the method
outlined in Section VII,F.
We now consider an arbitrary Hall plate provided with four finite
contacts (Fig. 21). The electric field is still given by Eq. (64). On the Hall
contacts two points P and Q have been chosen. The integration of the electric
field gives us the Hall voltage:
where A$(P) and A+(Q) denote the differences between the II/ values at P
and Q, respectively, and the values at the ends of the contacts. For point-
shaped contacts these differences are evidently zero and Eq. (66) reduces to
Eq. (65).The first term in the right-hand member of Eq. (66) is the theoretical
Hall voltage according to Van Der Pauw. The geometry correction is now
given by a term containing the stream-potential differences A+(P) and
A+(Q) and a second term involving the flux of V+ through PQ. If necessary,
the ohmic potential drop occurring when pHB= 0 has to be subtracted from
Eq. (66); however, most Hall-plate configurations show sufficient symmetry
for this operation to be dropped.
I . Application to a Rectangular Hall Generator

The method outlined in Section VII,H is applied to the rectangular plate
shown in Fig. 20. Because the Hall voltage has also been calculated with
conformal mapping methods, the accuracy of the numerical values can
be controlled.
Figure 22 shows the relative error of the geometry correction for various
values of PHB obtained by evaluating the last term of Eq. (65). In contrast to
previous results, the accuracy is fairly high. It should be noted that a relative
accuracy of 10% on the geometry correction means that the Hall voltage is
known with a precision much better than 10%. We remark again that the
30 GILBERT DE MEY
20 -t
FIG.22. Relative error on the geometry correction as a function of IM for a square-shaped

Hall plate: p H B= 0.1 ( 0 ) , 0 . 3 ( + ) , 0 . 5 ( A ) , a n d I ( 0 ) . IMisthenumberofunknownsperside.
error decreases in inverse proportion to the number of unknowns used in the

numerical procedure. We may conclude that the accuracies are sufficient
for practical purposes.
J . Application to a Cross-Shaped Geometry

Equation (66) is now used to analyze the cross-shaped geometry. Two
points P and Q have been placed at the centers of the Hall contacts (Fig. 23).
FIG.23. Cross-shaped Hall plate analyzed with the stream potential.

Figure 24 shows the relative error on the numerically calculated geometry

correction for h/l = 0.5. For practical purposes these accuracies are more
than satisfactory. For high values of pHB(>0.5), the error curve seems to
diverge. This does not mean that the method should fail. The reason is that
the relative error was calculated taking Eq. (33) as the exact value. However,
there is a slight nonlinearity between the Hall voltage and the applied
magnetic field B, a secondary effect not included in Eq. (33).
Figure 25 shows the geometry correction as a function of h/l calculated
for PHB = 0.1. The theoretical relationship (33) has also been drawn. For
h/l < 1.5, numerical results are close to the theoretical ones, especially for
high values of ZM,which is the number of unknowns for one side (the total
number of unknowns is then 12ZM). For h/l 2 1.5, the deviations become
remarkable. Nevertheless, the numerical results still converge to the exact
values. In order to obtain insight into the practical consequences, let us take
h/l = 2. The exact geometry correction is 0.2%, whereas the numerical value
obtained with ZM = 9 turns out to be 0.5%. It makes no difference in practice
if an experimental value is corrected by 0.2% or OS%, because this correction
is much smaller than the accuracy of the Hall-mobility measurements. Only
I 1
+
20 - 20 -
-ae -10 10 -
$ 5- 5-
9'
z
0
Q 2- 2-
0
Q
b
Q
-
: I-
0.5-
1-
0.5 -
0.2-
+ 0.2 -
FIG.24. Relative error on the geometry correction versus IM for a cross-shaped Hall plate:
(a) h/l = 0.5; p H B= 0.3 (+).
32 GILBERT DE MEY
!i
+8 z -
E
1-
0.5-
az - \
1 1 I
1
I I
1.5
I I
2
*
0.5
h/l
FIG.25. Geometry correction as a function of h /lfor a cross-shaped Hall plate: p H B= 0.1 ;

I M = 3 (A),5 (OL7 (+), 9 ( 0 ) .
=-
when the geometry correction is greater (e.g., 1%) can it be calculated with
sufficient accuracy. The conclusion is that the method can be applied in
actual cases. Only when the geometry correction is negligible can it not be
calculated with high precision; but in these cases knowledge of it is of no
importance. It should be noted that the cross-shaped geometry has four
reentrant corners, which makes its use difficult for numerical treatments.
Some equipotential lines have been drawn in Fig. 26. Owing to Eq. (16)
these lines are also the current lines or streamlines. Only 50% of the Hall
plate is shown because the remaining part can be found by symmetry. From
Fig. 26 the hypothesis formulated in Section IV can be verified.
POTENTIAL CALCULATIONS I N HALL PLATES 33
FIG.26. Streamlines in a cross-shaped Hall plate.
K . Application to Some Other Geometries

We first consider a rectangular Hall generator provided with four finite
contacts (Fig. 27). This type of Hall plate is often used experimentally
because monocrystalline semiconductors are usually available in the form of
a rectangular bar. It is not always possible to cut a monocrystalline semi-
conductor slice in the form of a cross; a rectangular Hall plate is then used to
carry out Hall-mobility measurements. This geometry has been extensively
studied theoretically by Haeusler, using conformal mapping techniques
(Haeusler and Lippmann, 1968). Nevertheless, this shape is far from optimal.
Even for small contacts, the geometry correction turns out to be 30-40%. A
34 GILBERT DE MEY
FIG.27. Rectangular Hall plate provided with four finite contacts.
small shift in the position of the contacts causes a nonnegligible change of the
geometry correction and hence an error in the experiments. Figure 28 shows
the numerically calculated value of A,uH/PHfor h/b = 2 as a function of a/b.
The geometry correction increases rapidly with the contact length a and can
attain very high values. For this geometry Haeusler found the following
approximate formula valid for sufficiently small contacts (Haeusler and
Lippmann, 1968):
Equation (67) is accurate up to 4% for h/b > 1.5 and a/b < 0.18. The first
condition is met, but the second one is not fulfilled because the numerical
BEM cannot be used if a contact is much smaller than the other sides of the
Hall plate. Hence we expect that Eq. (67) will not coincide perfectly with the
numerical results. For h/b = 2 and 8, small, Eq. (67) reduces to
APH/PH = 1 - (1 - e-")[l - (2/a)(a/b)] = 0.0432 + 0.609(a/b) (68)
The linear relationship (68)has also been drawn in Fig. 28, and the agreement
is fairly good. One observes a better fit when the number of unknowns per
side IM increases.
A second particular geometry is the octagon provided with four equal
contacts (Fig. 29). Although this shape has not been used in experiments, it is
the first geometry that has been studied theoretically by Wick using con-
formal mapping techniques (Wick, 1954). The reason is that the conformal
mapping of the octagon into a circle leads to calculations which can be
< 0.L 0.5 0.6 0.7 0.8

O/b
FIG. 28. Geometry correction for the rectangular Hall plate: h/b = 2; pHB = 0.1 ; IM =
(m),
3 (+I, 5 (A), 7 ( x ), 9 (Oh I 1 13 (01, 15 ( 0 ) .
performed analytically. Figure 30 represents the numerically calculated

geometry correction as a function of l / h for various values of the number of
unknowns. One observes again that the geometry correction is high. Wick
did not calculate any geometry correction, but from several graphs in his
article it was possible to deduce approximately ApH/pH. These results are
t1
FIG.29. Octagonal Hall plate.

36 GILBERT DE MEY
I I I I *
0.2 0.3 0.4 0.5
i/h
FIG.30. Geometry correction for the octagonal Hall plate: p H B= 0.1 ; IM = 3 (+), 5 ( A ) ,
7(x).9(0). I1 (m),1 3 ( 0 ) , 1 5 ( 0 ) .
shown in Fig. 30 and referred to as “theoretical.” One sees good agreement
between these theoretical points and the numerical results. Chwang ef al.
(1974) analyzed the octagon using finite difference techniques. From the
results published in their article, the geometry correction could be derived
for small values of l/h. These results are not very accurate because they differ
a lot from the results obtained by Wick (conformal mapping) and the BEM.
It is also possible to obtain an approximate formula for ApH/pH for the
octogon. Van Der Pauw published geometry corrections for a circular shape
provided with finite contacts. An octagon can be approximated by a circle
with a diameter h and a contact length 1. The geometry correction is then
(Van Der Pauw, 1958)
AA+/PH = 4(2/nZHI/h)= (8/n)’(1/h) (69)
This relation is also drawn in Fig. 30, and it turns out to be a good
approximation for l / h % 0.5. For other l/h values the replacement of an
octagon by a circle seems to be less adequate.
FIG.31. Unsymmetrical cross-shaped form.
A final interesting geometry is the unsymmetrical cross-shaped form.

Thin-film semiconducting Hall plates are usually made by evaporating the
materials through a metallic mask. For the contacts on the Hall plate
another mask is required, and thus a shift between the mask positionings can
never be avoided. This gives rise to an unsymmetrical cross-shaped form. A
fortiori, when the contacts are made manually with a conducting ink, for
example, a perfect symmetry cannot be guaranteed.
We now investigate a cross-shaped form where one arm differs from the
others (Fig. 31). The numerical results are shown in Fig. 32, representing the
geometry correction ApH/pH as a function of u/l. The geometry correction
increases when the arm length shortens, which can easily be understood.
These results can also be found theoretically. In Section V,B, we made the
assumption that each arm gives its own contribution to the geometry
correction. This being the case, the following formula was found for four
equal arm lengths:
Applying Eq. (70) to three arms with length h and one arm with length u, one
obtains
* nze x p ( i l n 2
PH = 4L
- :)[:exp( -n:) + aexp( - 31
= 1.04,3[:exp( -n)> + aexp( - n 3 1
For h / l = 1, Eq. (71) reduces to
ApH/pH = 0.03387 + 0.2613e-""" (72)
This relation has also been shown in Fig. 32, and the agreement with the
numerical results is very good. This also proves the hypothesis that each
38 GILBERT DE MEY
13
+
12
A
11
10-
- 9
s
L
z
+ 8
I
x
d
7
I I I 1 I I I C
0.4 0.5 0.6 0.7 0.8 0.9 I
0 4
FIG.32. Geometry correction for the unsymmetrical cross-shaped Hall generator: p H B =

IM = 3 (+), 5 (A), 7 ( U ) . 9 (0).
0.1; h / l = I ;
arm generates its own contribution to the geometry correction ApH/pH

independent of the other ones.
VIII. IMPROVEMENT
OF THE BOUNDARY-ELEMENT
METHOD
A . Introduction
In this section we present a modified BEM. By investigating the method it
was found that the highest errors occurred at the corners of the geometry.
Therefore, a method is given here which uses analytic approximations at the
corners while the remaining part of the potential is calculated numerically.
This technique can be seen as a combination of conformal mapping and the
BEM. The conformal mappings are used to calculate analytic approximations

at the corners of the Hall plate. The remaining part of the potential distri-
bution is found by numerical solution of an integral equation. The main
advantage of this method is that singularities occurring at a corner can be
represented exactly by the analytic approximation (De Mey, 1980).
B. Integral Equation
We consider a Hall plate as having the shape of an arbitrary polygon
(Fig. 33). Along each side zi the following boundary condition holds:
a4 + BV4.U" + yV4.q = fo (73)
Forametalliccontactatpotential Vo,onehasa = 1,p = y = O,andf, = Vo.
For a free boundary a = 0, /?= 1, y = -pHB, and fo = 0. All types of
boundary conditions along a Hall-plate medium can be represented by
Eq. (73). It makes no difference in the subsequent analysis whether one is
interested in the electrostatic potential or the stream potential II/.
We assume now that in the neighborhood of each corner an analytic
approximation for the potential is known. At the ith corner one has
4(r) AiPi(r) (74)
where cpi(r) is a known function satisfying the Laplace equation; Ai is a still
unknown proportionality constant. The constants Ai will be determined
together with the solution of the integral equation.
The potential inside the Hall plate is now written as
4(r) =
i
Aicpi(r) + f C p(r')G(rlr') dC' (75)
The function cpi(r) includes all the singularities at the corners. Note that
cpi(r) is not only defined at the ith corner, but in the entire region. This means
that cpi(r) will not necessarely be zero at the other corners but that the
functions cpi(r) will be chosen in such a way that cpi(r) is a smooth potential
\
z,-i
FIG. 33. Boundary condition at one side of an arbitrary polygon: aiQ + fii V+.u. +
yIvQ.u,=jO.
40 GILBERT DE MEY
distribution at the other corners. The basic idea is that 1 Aiqi(r)is not a
first-order approximation but a particular solution involving all singularities,
especially those occurring in the gradient field. The remaining part will then
be a perfectly smooth surface and is represented by the last term in Eq. (75).
Applying the boundary condition (73) to the proposed solution (75) gives
rise to
C A i [ a q i + PVqi.un + ~ v q i . ~ ]
i
- P+[p(r)l + (jcP(rWNG(rIr7
+ P V G ( r l r ’ ) - u , + yVG(rIr’)*u,]dC’ = fb(r) (76)
Equation (76) constitutes an integral equation in the unknown function p. We
remark that the constants Ai are also unknown and have to be determined
simultaneously with the numerical calculation of p.
C. Culculation of the Functions qi(r)

The purpose of this section is to determine the q i , potential functions
being good approximations at corners, and to include all singularities:
especially those occurring in the gradient fields Vq,.The method is outlined
for the corner A of the Hall plate shown in Fig. 20. The corner can be replaced
by two infinite sides. The function qi is then a potential satisfying the
boundary conditions indicated in Fig. 20 along two infinite sides. This can be
easily done, as it is always possible to map one corner onto another at which
the potential is reduced to a constant and homogeneous field. This conformal
FIG.34. Approximation of a corner by two circular arcs.

mapping is done by the simple analytic function 2'. Some problems may arise,
however. Consider again Fig. 20 for the particular case p H B = 0. In each
corner one side has a constant potential and the other side requires that the
normal gradient should be zero. The field is then homogeneous because the
corners are rectangular. The potential pi then varies linearly with distance.
The same situation is valid for all corners, however, which means that the four
functions cpl, cpz, cp3, and cp4 will be linearly dependent. In this case it is not
possible to determine the constants A , , A z , A 3 , and A, numerically. Linearly
dependent functions should always be avoided.
In order to eliminate this problem, a corner will not be approximated by
two infinite sides but rather by two circular arcs, as shown in Fig. 34, repre-
senting a corner with angle cp. One must now solve the Laplace equation
V Z q i= 0 with the condition cpi = 0 at one circular arc and Vcpi-u,+
-
p H BVqi u, = 0 at the other. In this case it is almost impossible to obtain
linearly dependent solutions at two corners. Because the circles are tangent
at the sides of the corners, qi is still a good approximation for the potential.
The solution of this potential problem is carried out with conformal mapping
techniques.
The geometry bounded by two circular arcs can be mapped onto a
semiinfinite plane using (Fig. 39,
2' = A"z/(z - Zo)]n'e (77)
where A' is a complex constant. Let us consider now the particular case where
c2 is a free boundary and c1 a metallic contact held at constant potential. We
FIG. 35. Conformal mapping on a corner with angle ( 4 2 ) - OH

42 GILBERT DE MEY
then have to transform into a corner with an angle 4 2 - OH ( w plane in

Fig. 3 9 , because for this geometry the potential can be written immediately.
The transformation from the z to the w plane is then
w = A ’ ( 1 / 2 ) - ( @ H / 2[) z / ( z - zO)](n/2o)-(@H/9)
= A [ z / ( z - ~ ~ ) ] ( n / 2 1 ) - ( @ H / O ) (78)
For a point z located on c1 (Fig. 35), one has

z/(z - zo) = ( z / ( z - zo)l e-jqlz (79)
then,
w = A 1 z/(z - z o )1 ( ~ / W - ( @ H / V ) exp( - j i n + jl-0
2 H) (80)
A point z on c1 is mapped onto a point w situated on the real u axis. Hence
Eq. (80) has to be real, from which A can be determined:
A = exp(jin - j + O H ) (81)
Because the function cpi is multiplied by a constant Ai in the proposed solution
(75), the absolute value of A has no influence on our problem. From the
transformation formula (78) one obtains
dw/dz = w [ ( $ n - &)/~][-zO/z(z - ZO)] (82)
Equation (82) is used later to calculate the transformation of the electric fields.
In the w plane, the potential cp corresponds to a homogeneous field
directed perpendicular to the u axis (Fig. 35) because the angle was taken to
be $n - 0,. Hence pi can be written as
Vi(U,O) = IJ (83)
The potential function can be derived from the complex potential W :
W = -jw [cpi = Re( W ) ] (84)
with
d W / d w = -j
The potential in the z plane can be easily determined because the complex
potential W is invariant. The electric field components in the z plane are
determined by
- - - - E , + j E.
dW acpi
= - - j - = -acpi
-= dwdw -j-dw (86)
dz y ax ay dw dz dz
Inserting Eq. (82) into Eq. (86) directly gives us the field components in the
original z plane. It has been proved that the potential cpi in a corner can be
determined by a simple conformal mapping.
It should be noted that cpi has only been determined up to a proportion-

ality constant. Because the functions cpi are preceded by an unknown constant
Ai in Eqs. (74) and (75), this causes no problems. For other situations at the
corners, e.g., two adjacent free boundaries at a corner, the procedure to deter-
mine cpi remains identical; however, some formulas may be appropriately
changed.
D. Application to a Rectangular Hall Generator

Figure 36 shows a rectangular Hall plate with the circles for the approxi-
mation of the potential in the corners. First, the influence of the radius R was
investigated. Figure 37 indicates the relative error of the Hall voltage and the
current I as a function of the radius R for various values of I M , which is the
number of unknowns per side. One observes that the error is almost indepen-
dent of R. As for the current, it seems to be an optimum around R = 1;
however, this is owing to the fact that the numerical value of the current minus
the exact value changes its sign when R x 1, so one cannot consider it
to be optimal.
Figure 38 shows the relative error of the Hall voltage and the supply
current as a function of IM. The numerical results obtained without the
analytic approximation C Aicpi have also been drawn in Fig. 38. The accuracy
of the Hall voltage remains practically unchanged, but the accuracy on the
current calculation increases by almost two decades. Because both quantities
are needed for practical applications, an overall improvement has been
achieved.
The geometry correction has also been investigated. Figure 39 shows the
relative error on the geometry correction as a function of IM. The same
FIG.36. Square-shaped Hall plate with approximate arcs at the corners.

5-
-z 2-
6
E 1-
-.,
2 a5-
c
2
a2-
0.1-
M5-l t
a1
R
FIG. 37. Relative error as a function of radius R : p H B = 1 ; AVH/VH ( 0 . A); A I / l

(0.0.
A).
3 5 9 15 29
1M
FIG.38. Relative error on the Hall voltage and the supply current: R = 1 ; pHB = I , with
( + ) and without ( 0 )analytic approximation.
i 5 rb .?a 50
IM
FIG.39. Relative error on the geometry correction as a function of IM:p,J = 0.I (+, x );
0.3 (A, A); 0.5 ( 0 ,0 ) ; I (W, 0); R = I ; with analytic approximation (+, A, 0 , 0);
without analytic approximation ( x , A , 0, m).
results without the analytic approximation are also drawn in Fig. 39. Gener-
ally, the introduction of analytic approximations yields better results,
especially for low values of p H B ,when the Hall voltage is low and difficult
to detect.
E. Application to a Cross-Shaped Form

The cross-shaped form has also been analyzed using the same method. In
this case, there are 12 corners and hence 12 qifunctions have to be evaluated.
The geometry correction was calculated using the method outlined in
Section VII,H. The results are shown in Fig. 40.The data obtained in Section
VII,J are also drawn in Fig. 40. One now observes, contrary to the results
found in the preceeding section for a square-shaped form, that the introduc-
tion of analytic approximations at the corners yields no improvement in the
numerical results. Noting the fact that the calculation of the cpi functions
involves computation time, one must conclude that the method of Section
VII,H should now be preferred. These results indicate that the present
method is only useful for a limited number of corners. The cross-shaped
sample has also been analyzed using an analytic function at a reduced
number ( c12) of corners, but this did not lead to any improvement.
46 GlLBERT DE MEY
1
I 1 1 I I 1)
a5 1 1.5 2
h/ I
FIG.40. Geometry correction versus h / l for a cross-shaped sample; pHB = 0.1 ; with an-
alytic approximation, IM = 3 (A),
9 ( 0 )without
; analytic approximation, IM = 3 (A),9 ).(
F . Application to Some Other Geometries

Because the number of corners seems to influence the numerical accuracy,
two geometries with eight corners are now investigated. The rectangular and
octagonal Hall plates treated in Section VII,K are now analyzed.
It should be noted that the rectangular Hall plate provided with four
finite contacts should be viewed as having eight corners. Because the electric
field at both ends of each Hall contact shows a singular behavior, the
introduction of analytic functions may be useful.
Figure 41 represents the geometry correction as a function of a/b. Some
results of Figure 28 have been redrawn. One sees that more accurate results
are now found.
60 -A
2- M-
*5
z
LO-
30 -
I
I I I
0.4 0.5 d.6 0.7 0.6
o/b
FIG.41. Geometry correction for a rectangular Hall plate: hlb = 2; pHB= 0.1 ; without
analyticapproximation, IM = 3 ( A ) , 15 ( 0 ) withanalyticapproximation,
; IM = 3 ( A), 15 (0).
0.2 0.3 0.6 as

I/h
FIG. 42. Geometry correction for an octagonal Hall plate: pHB= 0.1, without analytic
approximation, IM = 3 ( A ) , 15 ( 0 ) ;with analytic approximation, IM = 3 ( A ) , 15 ( 0 ) .
48 GILBERT DE MEY
Similar data for the octagonal shape are shown in Fig. 42. The same
conclusion still holds because one observes that the numerical results with
analytic approximations are closer to the theoretical curve.
As a general conclusion of this section, one can state that the introduction
of analytic approximations that include the singularities at the corners is
only useful if the number of corners is limited to eight.
1X. CONCLUSION
In this article two objectives concerning Hall-effect devices have been

outlined. First, a review was given concerning several methods for calculating
the potential distribution in a Hall plate. Second, an attempt was made to
answer the question: What can be done with such a potential distribution? It
turns out that the Hall voltage depends upon the geometry of the plate, an
effect which can only be calculated if the potential problem is solved. Because
many applications of Hall effects involve measurement of the Hall voltage, it
is important to know all the influences on this voltage. When several methods
for potential calculations were tested, not only did we draw streamlines or
equipotential lines, but we emphasized the evaluation of the geometry
correction to the Hall voltage.
The first objective of this article was to check several methods for potential
calculations in Hall plates. The potential satisfies the Laplace equation, but
the boundary conditions are rather unusual, so that common techniques such
as separation of variables or eigenfunction expansion cannot be applied.
Three methods were found to give acceptable results: conformal mapping,
finite difference approximation, and the BEM, the last two being purely
numerical. Conformal mapping was the first method used to investigate the
Hall effect. Some geometries can be analyzed with a completely analytical
treatment; however, the calculations are very lengthly and complicated.
Conformal mapping has also been applied successfully in a semianalytical
approach : after the Schwarz-Christoffel transformation formula is written
down further steps are performed numerically. In this way the high accuracy
associated with conformal mapping techniques can be achieved with only
moderate computational effort. The finite difference method is the most
obvious numerical method, but complications arise owing to the special
boundary conditions. This method can be used for a potential distribution,
but if current and impedances are calculated the finite difference approach
gives unacceptable results. In the BEM the special boundary condition
presents no particular difficulties. Potential calculations can be done accu-
rately. With some modifications, even small geometry corrections may be
calculated. Finally, we can say that either theconformal mapping or the BEM
can be used. Conformal mapping yields high accuracies, but for every new
geometry, calculations have to be done again. The BEM can be applied to all
geometries but its accuracy is lower. One of the two methods will be appro-
priate for each individual case. In the last section a combination of these two
methods was presented. The potential was still calculated by the BEM, but at
the corners analytical approximations obtained with conformal mapping
were introduced. I t turns out that this combined method only constitutes an
improvement for geometries with a limited number of corners.
The second objective of this article was to show the influence of the
geometry on the Hall voltage, i.e., the geometry correction. Van Der Pauw’s
theory gives a formula for the Hall voltage, but owing to geometic effects
(i.e.,finitecontacts) a lower value will be measured in practice. Thiscorrection
has been calculated for several geometries, and the conclusion is that the
cross-shaped form requires the smallest correction. This may be useful
because some geometric parameters are not always known with sufficient
accuracy. The cross-shaped form is also fitted with contacts comparable to
the dimensions of the Hall plate, so that the resistance between two contacts
and hence the noise in the measuring circuitry will be minimal.
APPENDIX
1. THETHREE-DIMENSIONAL
HALLEFFECT
The Hall generators studied in this article are essentially two dimensional.
Van Der Pauw’s theory, as outlined in Section 111, cannot be extended to a
three-dimensional semiconducting volume having four contacts on its
surface and placed in a magnetic field. Even when all contacts are point
shaped, a general formula for the Hall voltage cannot be given. For every
geometry one has to solve the potential problem from which the Hall voltage
is found.
Neglecting terms of order &B2, it turns out that the potential still
satisfies the Laplace equation (De Mey, 1974b). At a metallic contact the
potential is given, but on a free surface one has
where u, is the unit vector perpendicular to the semiconductor surface and

u, is the unit vector in the direction of the magnetic field. Note that u, and u,
need to be perpendicular, as in the two-dimensional case. Because u, and u,
can include all possible angles, the right-hand side of Eq. (87) turns out to be a
complicated boundary condition. A perturbation method is therefore used to
solve the problem. Treating PHB as a small quantity, the potential 4 can be
50 GILBERT DE MEY
written as a zeroth-order approximation 4oand a first-order perturbation :
4 = 40 + P(HB4I (88)
The equations and boundary conditions for do and 4, are then
v24, = 0 (89)
v40 'U, = 0 on a free surface (90)
4o = applied potential on a metallic contact (91)
v24, = 0 (92)
V 4 , 'u, = (V40 x uz)*u, on a free surface (93)
41 =o on a metallic contact (94)
These equations are to be solved for a cylindrical semiconducting bar, as
shown in Fig. 43. If a potential difference V is applied, the zeroth-order
potential is easily found:
40 = (V/a)x (95)
The zeroth-order current I, through the contacts is given by
I0 = (V/p)(nR2/a) (96)
For the first-order perturbation, an eigenfunction expansion is used. In a
circular section, the Dirichlet eigenfunctions are
where N,, is the normalization constant, Jn is the Bessel function of the first
kind of order n and xnpis the pth root of the transcendental equation
The first-order potential 41 is then written
where
,
the other coefficients being zero. The normalization constant N, is given by
3
FIG.43. Cylindrical Hall medium placed in a magnetic field.
as
0 1 1 1
1 1
.-
5
FIG.44. Hall voltage for a cylindrical semiconductor.
52 GILBERT DE MEY
From Eqs. (96), (99), and (lOO), the Hall voltage V , between the points P and
Q can be found for a given current I , :
VH = -4/&BpI,(nR)-'c ( x t p - l)-'[l - cosh(x,,/R)(a/2)]-' (102)
P
Figure 44 shows the Hall voltage normalized to pHBpIo/R as a function of
u/R. If u/R + co, i.e., the semiconducting bar becomes infinitely long, the
normalized Hall voltage equals 0.6324. This means that the cylinder gener-
ates the same Hall voltage as a two-dimensional Hall plate with the same
width 2R and a thickness R/0.6324.
2. ON THE EXISTENCE
APPENDIX OF SOLUTIONS
OF
INTEGRAL EQUATIONS
If the Laplace equation for a given geometry with suitable boundary

conditions is replaced by an equivalent integral equation, generally a
Fredholm equation of the first kind will be obtained. On the other hand, the
existence theorems for integral equations deal only with equations of the
second kind (Hochstadt, 1973).For these equations it is possible to construct
an iteration procedure, which can be used to prove the existence theorems.
With regard to Fredholm integral equations of the first kind, the literature is
quite contradictory. Some authors claim that the existence has been proved
(Symm, 1963), whereas other authors allege the reverse (Tottenham, 1978).
A particular geometry is now treated in order to investigate the existence
of a solution. The conclusion cannot be extended to arbitrary geometries,
but it will provide us with a deeper insight into the problem. Also, the con-
tradictions occurring in the literature will be better understood. As integral
equations for potential problems are mainly used as numerical approxi-
mations, the treatment in this appendix is also done from a numerical
point of view.
According to Fredholm theory (Courant and Hilbert, 1968) an integral
equation can be regarded as the limit of a set of algebraic equations whose
number of unknowns goes to infinity. We shall proceed now in a similar way.
In order to investigate the integral equation for the circular geometry shown
in Fig. 45, the boundary is divided into n equal parts Acj. The integral
equation
fcp(r')G(rlrc)dC'= V(r), reC (103)
is then replaced by the finite algebraic set

I'
FIG.45. Circular geometry used to study the existence of a solution of the integral equation.
This set will have a unique solution if the determinant of the matrix [A,,]
is not zero:
det[A,,] # 0 (105)
Similarly, the integral equation will have a unique solution if the Fredholm
determinant does not equal zero. In our problem, the easiest way to check if
Eq. (105) is fulfilled is to calculate all the eigenvalues of [A,,]. When all the
eigenvalues are other than zero, the determinant will also be nonzero and
vice versa.
For the particular geometry of Fig. 45, one can easily verify that
= A(l-J)mOdn (106)
if the boundary is divided into equal parts AC,. Equation (106) means that
the matrix [A,] is circulant. It can be proved that the eigenvalues are then
given by
11
i., = C A,, (107)

k= 1
n- I
e - 2 n jmkln
'm = 1A l . k + l
k= 0
, m = l , 2 ,..., n - 1 (108)
If the matrix dimension increases, the matrix still remains circulant. In the
limit for n + 00, the expressions (107)and (108)are then replaced by integrals.
54 GILBERT DE MEY
The eigenvalues of the integral equation (103) are then found to be

r
i o= (2n)-'fClnlr - r,IdC
i,, = (2n)- ' In I r - r, Ie-JmedC
where r, can be chosen arbitrarily on the boundary. FOPa circular geometry,
1;
the integrals (109) and (1 10) can be easily evaluated:
j
.
, = (R/n) ln(2R sin +6)d6 = R In R (111)
i,, = (R/n) [i ln(2R sin +6)cos m6 d6 = - R/2m (1 12)
If R = 1, the eigenvalue I, turns out to be zero. The integral equation will

then have no solution at all because the Fredholm determinant is zero. This
fact has been investigated by Jaswon and Symm (1978) and Symm (1964).
By using an appropriate scaling factor, this problem can be easily eliminated.
It has been verified experimentally that eigenvalues of the matrix [ A i j ]
(with dimension n) show behavior similar to the first n values l o ,R,, . . .,i n -
given by Eqs. (111) and (112). Especially, it was found that the smallest
eigenvalue is inversely proportional to the matrix dimension n. This implies
that I , , - , tends to zero when n + 00. One therefore concludes that the
Fredholm determinant of the integral equation (103) will always be zero,
meaning that Eq. (103) will have no solution at all. O n the other hand, as
long as n remains finite, all the eigenvalues are other than zero and the
algebraic set has a unique solution. It can thus be stated that the integral
equation only has a solution if one considers it to be the limit solution of an
algebraic set. If the solution of the set converges for n + co,one can define it
as the solution of the integral equation.
Many properties of Fredholm integral equations of the first kind have
not been studied or explained because this matter turns out to be a difficult
area of functional analysis. This explains why some numerical results, such
as the l/n law for the relative error versus the number of unknowns, have
not been declared yet. A theoretical background will indicate to us in which
direction research should be continued in order to improve the BEM.
APPENDIX THEOREM
3. GREEN'S
The method outlined in Section VIII,B is rather unusual for constructing

an integral equation for the BEM. Normally, Green's theorem is used,
leading to an integral equation where the unknown source function p has a

physical meaning. If the potential is given on a part of the boundary, the
unknown function turns out to be the normal component of the gradient and
vice versa. However, this can only be done if the so-called natural boundary
conditions are given, i.e., the boundary conditions only involve 4 or V 4 sun.
The boundary condition (18) is not a natural one. An attempt has been
made to integrate the tangential component along the boundary in order to
obtain the potential 4 rather than V 4 mu,. The procedure is outlined later.
Consider the Hall plate shown in Fig. 11. The potential satisfies
v24(r)= 0 (113)
and the Green's function satisfies
V2G(r(r')= 6(r - r ' ) (1 14)
Multiplying Eq. (1 13) by G and Eq. (1 14) by 4 and substracting the resulting
equations from each other, one obtains after applying Green's theorem
fc [4(r)VG(rlr')*un- G(rlr')V4-un]dC = 4(r') (1 15)
Equation (1 15) holds for each point r' inside C and can be transformed into an
integral equation if r' is placed on the boundary. When 4 is given (e.g., on a
metallic contact), then V 4 * u, is the unknown function. However, along the
side BB', Eq. (18) is valid and V4.u" is not given, so 4 cannot be treated
directly as the unknown function. Some intermediate steps are required:
The last integral in Eq. (1 16) can be transformed by partial integration:
pHB jBB'
G V$ * U, dC
= pHB[G(r1r')4(r)]:r:~,- pHB jBB'4

VG -u, dC (1 17)
Equation (1 16) now reads
jBB'[4VG-un- GV4.un]dC
= pHBG(rB,lr')vo-I-
IBB'[4VG*U, - p~BvG'll,]dC (118)
The tangential component V 4 -u, has been eliminated, and only 4 appears as
an unknown function in the intergrand. Hence it is possible now to consider 4
56 GILBERT DE MEY
the unknown function along BB’. The same procedure can be carried out
along AA’, so that Eq. (1 15) can be rewritten as
n n
- pHBG(rB lr’)Vo + [ V o V G . u , - G V d * u , ] dC
s.,A.
d[VG.U, - p H B V G ’ U , ] dC + pHBG(rA,lr’)Vo, CE c (119)

+ JAA,
I f r’ E C , Eq. (1 19) is an integral equation that can be solved numerically;

however, in Eq. (119) terms of the form VoG(rB,lr’) occur. Since G is a
logarithmic function, these terms will be singular at the corner points A ’ and
B’. Generally, it can be stated that singularities will occur at the end points of
every contact staying at a nonzero potential. These difficulties d o not
happen with the technique presented in Section VII,B, because the Green’s
function then always appears under the integral sign.
The integral equation (119) with r’E C has been solved numerically. In
the term G(rA,lr’), r‘ was put equal to ri, which is the center point of the ith
boundary element. Singularities were hence eliminated. Figure 46 shows the
relative error on the Hall voltage. These results are compared with those
\ 5
APPROXIMATION
I
1 1 1 I I 1 I I I I +
3 5 7 9 I1 13 15 n 19
n
FIG.46. Relative error on the Hall voltage calculated by using Green’s theorem.
obtained in Section V11. One observes that the error is very high, making the
method unusable. For p H B = 0, accurate results were found because the
singular terms in Eq. (1 19) vanish.
These results explain why Green’s theorem is not used to construct an
integral equation for the field problem in Hall plates. The methods outlined
in Section VII should be preferred in this case.
APPENDIX
4. THEHALL-EFFECT
PHOTOVOLTAIC
CELL
The Hall effect can also be used to convert solar energy to electric power.
The basic principle of all photovoltaic cells is the generation of electrons and
holes in semiconductors owing to the absorbed light. If electrons and holes are
accelerated in opposite directions a net current will be delivered to an
attached load. In junction solar cells the separation of electrons and holes is
done by the junction field. The same effect can be achieved with the Hall
phenomenon because electrons and holes are deflected in different directions
by the Lorentz force.
A possible configuration of a Hall-effect photovoltaic cell is shown in
Fig. 47. The incident light generates electron-hole pairs in the layer. Owing
to the exponential decay of the light intensity, a concentration gradient
for the charge carriers is built up. Hence electrons and holes diffuse in the y
direction. Owing to the Lorentz force, charge carriers will be deviated along
the x axis in such a way as to give a net current through the load resistor R,.
A theoretical analysis has been published (De Mey, 1979).Assuming that
one type of charge carrier has a much higher mobility than the other (which
-
gs
58 GILBERT DE MEY
is the case for InSb, InAs, etc.), the maximum attainable efficiency was
found to be
where p is the mobility, B the magnetic induction, E , the band gap of the
semiconductor, L the diffusion length, and u the light absorption coefficient.
For E , = 1 eV, and giving GILits optimum value, Eq. (120) reduces to
= 0.00625(puB)2 (121)
For p B = 1, the efficiency turns out to be 0.6%, a low value compared with
that of junction solar cells, which have efficiencies better than 10%. Note
that p B = 1 is a rather high value because B = 1 Wb/m2 is difficult to attain,
and there are only two semiconductors having p > 1 : InSb ( p = 7) and
InAs ( p = 3). Because Eq. (120)was developed for monochromatic light, the
efficiencywill be reduced at least by 0.44 in order to take the solar spectrum
into account, and because both InSb and InAs have small band gaps, these
conductors are not matched to the solar spectrum, which results in a much
lower spectrum factor (< 0.44). The conclusion is that the Hall-effect solar
cell is not suitable for energy production because of its low conversion
efficiency.
APPENDIX
5. CONTRIBUTIONOF THE HALL-PLATE
CURRENT TO THE MAGNETIC FIELD
A current must be supplied through two of the contacts of a Hall plate,

which gives rise to a Hall voltage (in combination with an externally applied
magnetic field). However, the current in the Hall place also generates its
own magnetic field, which can also influence the Hall effect. The mathe-
matical analysis of this secondary influence is rather complicated because the
magnetic field caused by a current distribution in a plate turns out to be a
three-dimensional problem. However, a crude approximate analysis indicates
that the contribution of the currents in the Hall plate to the applied magnetic
field is negligible.
The situation changes if an alternating magnetic field is applied. Eddy
currents are then generated in the Hall plate, creating an additional magnetic
field and influencing the original Hall voltage. It is even possible to get a Hall
voltage for a zero supply current, the eddy currents only being responsible
for the Hall effect. If the frequency of the alternating magnetic field is high,
the reaction of the eddy currents on the magnetic field can be important. A
theoretical analysis carried out for a circular Hall plate in an ac magnetic
field indicates that the parameter Q = 2n f o p o d / R should be considered (f,
frequency; 0 , conductivity; po , permeability; d, thickness; and R, radius)

(De Mey, 1976d). If R < 0.1, the contribution of the eddy currents to the
magnetic field is negligible. Eddy currents are calculated directly from the
applied field, and the Hall voltage is found by integrating the electric field
between the Hall contacts. For high values of R, the calculation of the
eddy-current pattern constitutes a complicated field problem. For practical
purposes (e.g., magnetic field measurements in electrical machines), one is
interested in working under the condition R < 0.1. The quantity d/R can
then be replaced by the quotient of the thickness and a typical dimension of
the Hall plate if the shape is not circular.
APPENDIX
6. LITERATURE
Most books on the Hall effect describe the physical aspects of Hall
mobility. Putley’s well-known book (1960) describes the galvanomagnetic
properties of a large number of semiconductors. The book also provides
many references. An excellent work has also been published by Wieder (1979),
which describes both the physical nature and the measuring techniques
related to the study of galvanomagnetic properties.
ACKNOWLEDGMENTS
I wish to thank Professors M. Vanwormhoudt and H. Pauwels for their continuous interest
in this work. I am also grateful to my collegues B. Jacobs, K.Stevens, and S. De Wolf, who
collaborated on several topics treated in this article. I want also to thank Ms.H. Baele-Riems
for careful typing of the manuscript and Mr. J. Bekaert for drawing the figures.
REFERENCES
Abramowitz, M., and Stegun, 1. (1965). “Handbook of Mathematical Functions,” pp. 888-890.
Dover, New York.
Bonfig, K. W., and Karamalikis, A. (1972a). Grundlagen des Halleffektes. Teil 1. Arch. Tech.
Mess. 2, 115-118.
Bonfig, K. W., and Karamalikis, A. (1972b). Grundlagen des Hall-effektes. Teil 11. Arch.
Tech. 3, 137-140.
Brebbia, C. (1978a). “The Boundary Element Method for Engineers.” Pentech Press, London.
Brebbia, C., ed. (1978b). “Recent Advances in Boundary Element Methods.” Pentech Press,
London.
Brown, I. C., and Jaswon, E. (1971). “The Clamped Elliptic Plate under a Concentrated Trans-
verse Load,” Res. Memo. City University, London.
Bulman, W. E. (1966). Applications of the Hall effect. Solid-State Electron. 9,361-372.
Chwang, R., Smith, B., and Crowell, C. (1974). Contact size effects on the Van Der Pauw
method for resistivity and Hall coefficient measurements. Solid-Stare Electron. 17, 1217-
1227.
60 GILBERT D E MEY
Courant, R.,and Hilbert, D. (1968). “Methoden der Mathematischen Physik I,” pp. 121-124.
Springer-Verlag, Berlin and New York.
Davidson, R . S., and Gourlay, R. D. (1966). Applying the Hall effect to angular transducers.
Solid-State Electron. 9,47 1-484.
De Mey, G. (1973a). Field calculations in Hall samples. Solid-Slate Electron. 16,955-957.
De Mey, G. (1973b). Influence of sample geometry on Hall mobility measurements. Arch.
Elektron. Uebertragungsrech. 27, 309-3 13.
De Mey, G .(1973~).Integral equation for the potential distribution in a Hall generator. Electron.
Lelt. 9, 264-266.
De Mey, G . (1974a). Determination of the electric field in a Hall generator under influence of
an alternating magnetic field. Solid-State Electron. 17,977-979.
De Mey, G. (1974b). An expansion method for calculation of low frequency Hall etTect and
magnetoresistance. Radio Electron. Eng. 44,321-325.
De Mey, G. (1975). Carrier concentration in a Hall generator under influence of a varying
magnetic field. Phys. Status Solidi A 29, 175- 180.
De Mey, G. (1976a). An integral equation approach to A.C. diffusion. In:. J. Heat Mass Transjer
19,702-704.
De Mey, G . (1976b). An integral equation method for the numerical calculation of ion drift
and diffusion in evaporated dielectrics. Computing 17, 169- 176.
De Mey, G. (1976~).Calculation of eigenvalues of the Helmholtz equation by an integral
equation. In:. J. Numer. Methods Eng. 10, 59-66.
De Mey, G. (1976d). Eddy currents and Hall effect in a circular disc. Arch. Elektron. Ueber-
tragungstech. 30, 312-315.
De Mey, G . (1977a). A comment on an integral equation method for diffusion. In!. J. Heat
Mass Transjer 20, I8 1 - 182.
De Mey, G . (1977b). Numerical solution of a drift-diffusion problem with special boundary
conditions by integral equations. Comput. Phys. Commun. 13,81-88.
De Mey, G. (1977~).A simplified integral equation method for the calculation of the eigen-
values of Helmholtz equation. In:. J. Numer. Methods Eng. 11, 1340-1342.
De Mey, G. (1977d). Hall effect in a nonhomogeneous magnetic field. Solid-State Electron.
20, 139-142.
De Mey, G . (1979). Theoretical analysis of the Hall effect photovoltaic cell. IEE Trans. Solid
Slate Electron Decices 3,69-71.
De Mey, G . (1980). Improved boundary element method for solving the Laplace’ equation in
two dimensions. Proc. In/. Semin. Recent Ado. Bound. Elem. Methods, I980 pp. 90- 100.
De Mey, G., Jacobs, B., and Fransen, F. (1977). Influence of junction roughness on solar cell
characteristics. Electron. Lert. 13, 657-658.
De Visschere. P., and De Mey, G. (1977). Integral equation approach to the abrupt depletion
approximation in semiconductor components. Eleclron. Lelt. 13, 104- 106.
Edwards, T. W., and Van Bladel, J. (1961). Electrostatic dipole moment of a dielectric cube.
Appl. Sci. Res. 9, 151-155.
Ghosh. S. (1961). Variation of field effect mobility and Hall effect mobility with the thickness
of deposited films of tellurium. J. Phys. Chem. Solids 19,61-65.
Gray, R. M. (1971). “Toeplitz and Circulant Matrices: A Review,” Tech. Rep. No. 6502-1,
pp. 16- 19. Information Systems Laboratory, Stanford University, Stanford, California.
Grutzmann, S. (1966). The application of the relaxation method to the calculation of the poten-
tial distribution of Hall plates. Solid-State Electron. 9,401-416.
Haeusler, J. (1966). Exakte Losungen von Potentialproblemen beim Halleffekt durch konforme
Abbildung. Solid-Stare Eleclron. 9,417-441.
POTENTlAL CALCULATlONS 1N HALL PLATES 61
Haeusler, J. (1968). Zum Halleffekt Reaktanzkonverter mit vier Elektroden. Arch. Elektr.
Uebertragung 22,258-259.
Haeusler, J. ( I 971). Randpotentiale von Hallgeneratoren. Arch. Elekrrotech. (Berlin) 54,
77-81,
Haeusler, J., and Lippmann, H. (1968). Hall-generatoren mit kleinem Linearisierungsfehler.
Solid-Stare Electron. 11, 173- 182.
Hochstadt, H. (1973). “Integral Equations,” Chapters 2 and 6. Wiley, New York.
Jaswon, M. A., and Symm, G. T. (1978). “Integral Equation Methods in Potential Theory and
Electrostatics.” Academic Press, New York.
Kobus, A., and Quichaud, G. (1970). Etude d’un moteur a courant continu sans collecteur a
commutation par un generateur a effet Hall en anneau. RGE, Rev. Gen. Electr. 79,235-242.
Lippmann, H.,and Kuhrt, F.( l958a). Der Geometrieeinflussaufden transversalen magnetischen
Widerstandseffekt bei rechteckformigen Halbleiterplatten. 2.Naturforsch. 13,462-474.
Lippmann, H., and Kuhrt, F. (1958b). Der Geometrieeinfluss auf den Hall-effekt bei recht-
eckformigen Halbleiterplatten. Z. Naturjorsch. 13,474-483.
Madelung, 0. (1970). “Grundlagen der Halbleiterphysik,” Chapters 37-39. Springer-Verlag,
Berlin and New York.
Many, A., Goldstein, Y., and Grover, N. B. (1965). “Semiconductor Surfaces,” p. 138. North-
Holland Publ., Amsterdam.
Mei, K.,and Van Bladel, J. (1963a). Low frequency scattering by rectangular cylinders. IEEE
Trans. Antennas Propagation AP-I 1,52-56.
Mei, K., and Van Bladel, J. (1963b). Scattering by perfectly conducting rectangular cylinders.
IEEE Trans. Antennas Propag. AP-11, 185-192.
Mimizuka, T. (1971). Improvement of relaxation method for Hall plates. Solid-Stare Electron.
14, 107-110.
Mimizuka, T. (1978). The accuracy of the relaxation solution for the potential problem of a Hall
plate with finite Hall electrodes. Solid-State Electron. 21, 1195-1 197.
Mimizuka, T. (1979). Temperature and potential distribution determination method for Hall
plates considering the effect of temperature dependent conductivity and Hall coefficient.
Solid-State Electron. 22, 157- 161.
Mimizuka, T., and Ito, S. (1972). Determination of the temperature distribution of Hall plates
by a relaxation method. Solid-State Electron. IS, 1197-1208.
Nalecz, W., and Warsza, Z. L. (1966). Hall effect transducers for measurement of mechanical
displacements. Solid-State Electron. 9,485-495.
Newsome, J. P. (1963). Determination of the electrical characteristics of Hall plates. Proc.
Inst. Electr. Eng. 110, 653-659.
Putley, E. H. (1960). “The Hall Effect and Related Phenomena.” Butterworth, London.
Shaw, R. (1974). An integral equation approach to diffusion. In:. J . Hear Mass Transjer 17,
693-699.
Smith, A,, Janak, J., and Adler, R. (1967). “Electronic Conduction in Solids.” Chapters 7-9.
McGraw-Hill, New York.
Stevens, K., and De Mey, G . (1978). Higher order approximations for integral equations in
potential theory. In:. J. Elecrron. 45,443-446.
Symm, G. (1963). Integral equation methods in potential theory. Proc. R. Soc. London 275,
33-46.
Symm, G. (1964). “Integral Equations Methods in Elasticity and Potential Theory,” Res.
Rep. Natl. Phys. Lab., Mathematics Division, Teddington, U.K.
Symm, G., and Pitfield, R. A. (1974). “Solution of Laplace Equation in Two Dimensions,”
NPL Rep. NAC44. Natl. Phys. Lab., Teddington, U.K.
62 GILBERT DE MEY
Thompson, D. A., Romankiw, L. T., and Mayadas, A. F. (1975). Thin film magnetoresistors
in memory, storage and related applications. IEEE Trans. Magn. MAG-11,1039-1050.
Tottenham, H. (1978). "Finite Element Type Solutions of Boundary Integral Equations."
Winter School on Integral Equation Methods, City University, London.
Van Der Pauw, L. J. (1958). A method of measuring specific resistivity and Hall effect of discs
of arbitrary shape. Philips Res. Rep. 13, 1-9.
Wick, R. F. (1954). Solution of the field problem of the Germanium gyrator. J . Appl. Phys.
25,741-756.
Wieder, H. H. (1979). "Laboratory Notes on Electrical and Galvanomagnetic Measurements."
Elsevier. Amsterdam.
ADVANCES I N ELECTRONICS A N D ELECTRON PHYSICS, VOL. 61
Impurity and Defect Levels (Experimental)

in Gallium Arsenide
A. G. MILNES
Department of Electrical Engineering
Carnegie-Mellon University
Pittsburgh, Pennsylvania
1. Introduction ..................................
11. Possible Nati mplexes . . . . . .
Ill. Traps (and Nomenclature) from DLTS Studies
IV. Levels Produced by Irradiation .............
V. Semi-Insulating Gallium Arsenide with and without Chromium ............... 91
VI. Effects Produced by Transition Metals .........................
....................................... 116
IX. Group IV Elements as Dopants: C, Si, Ge, Sn, Pb ........................... I18
A. Carbon .... ...... 120
B. Silicon. ............................................................ 121
.............................................
.............................................
........................... 123
X. Oxygen in GaAs ...... ....
XI. Group VI Shallow Donors: e ...................................... 127
X11. Other Impurities (Mo, Ru, Pd, W,Pt, Tm, Nd). ............................. 129
A. Molybdenum ............................................
B. Ruthenium ......................................................... 129
C. Palladium . .
E. Platinum . ... .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ......................... 130
A. In LPE-Grown GaAs Layers ..... ........................... 133
XIV. Concluding Discussion

References ............................................................. 142
63
Copyright (* IPR3 by Academic Press. Inc.

All rights of reproduction in any form reserved.
ISBN 0-1m 4 ~ 1 - 4
64 A. G . MILNES
1. INTRODUCTION
The present understanding of defects and deep impurity levels in GaAs is

represented by a substantial literature but still leaves much to be desired after
two decades of study. Speculation on the physical nature of complexes and
defect levels is still very tentative. It is not even possible to say with any
assurance that simple gallium or arsenic vacancies (VGaand V,,, respectively)
have been related to particular energy levels.
Gallium arsenide specimens grown by various techniques' such as LEC,
Bridgman, VPE (chlorine process), MOVPE, LPE, and MBE are now
recognized as likely to have different properties in terms of energy levels
within the band gap due to differences in defect levels and trace impurities or
impurity complexes. Gallium arsenide grown by LPE from a gallium melt,
for instance, may be expected to be low in gallium vacancy defects and
possibly high in arsenic vacancies. Liquid-phase epitaxy GaAs usually
contains energy levels at E, + 0.41 and 0.71 eV (where E, is the energy of the
valence-band edge). Vapor-phase epitaxial layers are grown under arsenic-
rich conditions and almost always contain an important level at about
mid-band gap. This is still not understood. Also, depending on the construc-
tion of the system and the care with which it has been prepared and cleaned,
the VPE layer may contain another handful of energy levels. So a crystal
grower can no longer feel his obligation fulfilled by producing material
characterized only in terms of mobility and dislocation count. He is now
interested in deep levels because of the effect they have on ( 1 ) the carrier
concentration from shallow doping; (2) the effect on mobility because of
ionized impurity scattering; (3) trapping effects that can affect transport and
frequency and switching performances of devices; and (4) the recombination
diffusion lengths and generation in bipolar structures such as solar cells and
other optical sensing devices.
Section I1 begins by listing some of the many defects and complexes that
may be expected to exist in GaAs. Section 111 presents many of the energy
levels obtained by deep-level transient spectroscopy (DLTS) and related
methods for determining trap and recombination levels in GaAs. Energy
levels produced by high-energy irradiation are next considered. Section V is a
discussion of the effects of Cr in GaAs, since this may be used to produce
semi-insulating GaAs. The possible role of oxygen or other levels in this
respect is also considered. The effects of Group I-VI impurities are then
considered. Section XI11 discusses recombination and generation and
minority-carrier lifetime and diffusion-length effects.
'
LEC, liquid encapsulation Czochralski ; VPE, vapor-phase epitaxy ; LPE, liquid-phase
epitaxy ; MBE, molecular beam epitaxy ; MOVPE, metalorganic vapor-phase epitaxy.
IMPURITY LEVELS IN GALLIUM ARSENIDE 65
11. POSSIBLE
NATIVEDEFECTS
AND COMPLEXES
The number of possible native defects in a binary semiconductor such as

gallium arsenide is large, as may be seen from Table I. None of these has been
identified with any confidence. Experiments for studying such defects tend to
be too uncontrolled. There may be changes of electrical properties after heat-
ing under various conditions of Ga and As overpressure (and other conditions
are mentioned later). This is particularly true since it is difficult to obtain
gallium arsenide with trace-impurity levels below the mid- 10’ atoms cm-
and difficult to perform heating operations on specimens under conditions
of comparable cleanliness. A hundredth of a monolayer of metallic impurities
(say, I O I 3 atoms cm-’) entering a GaAs layer to a depth of, say, 10 pm has
the potential for producing contamination levels with concentrations as
high as 10I6 cm-j. Such effects probably invalidate many experimental
studies of the past, and the literature must be considered in this light.
Complexing of impurity atoms with native defects may also be expected.
Table I1 sets out some possibilities, more to stimulate thinking rather than to
be exhaustive or to represent the conditions that have been suspected.
Several nomenclatures exist for describing traps and their actual and
effective charge states. In this review, various forms are used to match the
TABLE I
POSSIBLE I N GALLIUM
DEFECTS ARSENID~
Gallium Arsenic Three-

vacancies vacancies Antisite Interstitials component
involved involved defects involved defects”
Asi
GaiAsi
a Subscript indicates lattice sites and i stands for interstitial.

Many others are possible besides the three listed as examples.
66 A. G . MILNES
TABLE I1
SOME HYPOTHETICAL
COMPLEXES OF NATIVE
DEFECTS
AND ELEMENTAL
IMPURITIES FOR GALLIUM
ARSENIDE'
Group I1 Group IV Group VI Transition

impuritiesb impurities' impuritiesd impurities'
various styles used in the literature. The nomenclature of Kroger (1977) is

used in some papers (although it is not in widespread use in the GaAs
literature) and so is explained as follows:
Defects can be represented by symbols, atoms being represented by their normal chemical
symbol, vacancies by V. Subscripts indicate the lattice site, characterized by the atom
normally occupying that site. Thus V, is a vacancy at an A site. Interstitial sites are indi-
cated by a subscript i, Ai being A at an interstitial site. Electrons are represented by e,
holes by h.
Defects, like normal atoms, can have charges. It is useful to distinguish actual charges
and effective charges. The former are the charges present inside the bounds of the defect.
Effective charges are the difference between the actual charge at a site and the charge nor-
mally present at that site if no defect is present. Coulombic interaction between charged
defects is determined mainly by the effective charges. For interstitials, actual and effective
charges are the same; for substitutional atoms or vacancies, the two types of charges are
different. Actual charges will be represented by superscripts + and -, effective charges
by dots (.) and primes ('). An effective charge zero is indicated by a superscript x. Thus
A; is an interstitial n atom with a single positive charge, V; is an A vacancy with a double
negative effective charge. Similar symbols can be used for normal crystal constituents,
A: representing a normal A at an A site with effective charge zero.
Use of these symbols occurs in fallowing Logan and Hurle's discussion

(1971) of the behavior of GaAs in contact with As vapor. Consider the
~~ ~~~
"Reproducedby permission of Her Britannic Majesty's

Stationery Office.
significant species to be Vga and (VGa);, which, if they accept an electron,

assume the forms Vba and (VGa);;V;, is considered to be a donor that is
capable of shedding an electron to become V;,. If we limit ourselves to these
defects, then the reactions of GaAs in contact with As vapor are those
shown in Table 111. The K values are estimated by Logan and Hurle, who
conclude that defect concentrations in the range 10'8-10'9 cm-3 might be
expected at a temperature of 1100°C (40°C below the melting point of GaAs)
and a pressure pAs2of about 0.3 atm. The equation 0 + V&, + V;, corre-
sponds to Schottky disorder, for which the heat of formation is believed to be
less than 5 eV (Van Vechten, 1980).The threshold energy for Ga Frenkel-pair
formation is believed to be 17 eV and that for As Frenkel-pair formation to
be even higher. However, Hurle (1979 ) has developed revised equilibria
equations in which Frenkel disorder on the arsenic sublattice is considered
dominant. The important reaction is then AS,, + Vi = Asi + V,,, and it is
supposed that V,, can become ionized to form donors V;,. An activation
energy for this of 0.25 eV has been suggested by Il'in and Masterov (1976).
The arsenic interstitial is considered to be either neutral or a singly ionized
donor. Hurle considers that there are several substantial experimental
studies, notably the lattice measurements of Bublik et al. (1973), that suggest
that Frenkel disorder is present to a very important extent.
Analyses based on simple vacancy models at high temperatures, however,
may not predict quenched-in effects. Grown-in defects as a sample cools
down might be expected to cluster to form neutral complexes such as
VGaAsGaVGa. As Van Vechten (1976) points out:
these should be effective nonradiative recombination centers because they quickly bind
both electrons and holes in deep states tightly localized in the same region of the crystal,
so that electronic matrix elements will be large, and because large lattice distortions occur
about these defects so that the configuration coordinate mechanism may proceed. Such
defects would appear as a cylinder one atom wide, 12 A long and oriented in the ( I 10)
direction. The recently developed direct lattice imaging, DLI, technique of electron
68 A. G. MILNES
microscopy allows the observation of such defects and they have in fact been found in the
size orientation and concentration (10” cm-3) predicted in GaAs (LEC melt-grown
material).
Other workers have looked for such structures with no success. Antisite
defects must also be considered, rather than limiting the modeling to
vacancies. Van Vechten remarks ( 1 976):
Consider the antisite defects, e.g., As: and Ga;:, and the bound pairs of these,
As,,Ga,,, which constitute a neutral defect. Obviously it takes no energy at all to inter-
change the two atoms in the unit cell of Ge. It can not take a large energy at all to inter-
change the two atoms in GaAs, but it will take more energy to interchange the two atoms
in ZnSe, and still more for CuBr. It is interesting to compare the QDT (quantum dielectric
theory) estimates to those of Pauling’s theory of electronegativities. The two estimates are
quite close in many cases, e.g., 0.70 and 0.72 eV for GaAs. No one has given an argument
that the energy of formation of the pair defect should be much larger. These energies for
111-V’s are much less than the energies of formation of vacancies or interstitials and imply
that antisite defects are present in compound semiconductors in substantial concentrations.
In addition to the DLI observation of antisite-vacancy complexes noted earlier, high con-
centrations of antisite defects have been identified by infrared absorption and are inferred
from EPR and other experiments. No description of a compound semiconductor can be
complete without them.
However, it would be good as a start to find a simple and certain way of

producing a vacancy such as V,, so that its properties may be explored in a
convenient temperature range. Electron or proton irradiation may be thought
of as such a way; but in studies of Si, the time required for a single vacancy
created by electron irradiation to migrate to and complex with an impurity in
the purest available material is of order lo3 sec at T = 100 K, so the time
required for a single vacancy to complex with an impurity at room tem-
perature is of order l o v 3sec or less. Therefore, in silicon and presumably also
in GaAs, after a growth process, no feasible quenching procedure can prevent
the single vacancies which were present in their equilibrium concentration at
Tho, from forming complexes during the quench and before Hall-effect
measurements can be performed (Van Vechten, 1976).
In our pursuit of the effects of vacancies, we are tempted to turn to
experiments in which As overpressures are involved. In one such experiment
in which GaAs was annealed under low pressures of As, donors were formed,
and under greater As pressure, acceptors were formed (Chiang and Pearson,
1975a,b). So it may be surmised that V,, (or V,, complexes) are donors and
that V, (or V,, complexes) are acceptors. For this to be accepted as true,
many repetitions of this experiment under conditions of highest purity in
many laboratories must consistently yield similar results (a state we have not
yet arrived at). The modeling of Fig. 4 (to be discussed later) assumes V,, to
be acceptors and V, donors.
IMPURITY LEVELS I N GALLIUM ARSENIDE 69
TABLE IV
LUMINESCENCE
SPECTRAL
PEAKSAND VACANCY
ASSOCIATIONS"
Luminescence peak Luminescence peak

(eV) Vacancy association (eV) Vacancy association
a Proposed by Chang er a / . (1971). Reproduced with permission of the American Physical
Society.
Created after Cu diffusions.
Enhanced after 0 implantation.
Chang et al. (1971), in studies of luminescence spectra for GaAs subjected

to various treatments with controlled As pressures, arrive at the tentative
assignments shown in Table IV. (The possibilities of interstitials of As being
involved were neglected.) Most of these assignments are now no longer
accepted.
In other work from studies of density, lattice constant, and internal
friction, the energy of formation of defects has been ranked as follows:
UGa, > UvAs> UvGa> U A ~ , (1)
so arsenic institials may have to be considered as possibly occurring in the
experiment [see also Van Vechten and Thurmond (1976)l.
Consider now some evidence for the formation of complexes. Hurle (1977)
considers the incorporation of donors, particularly Sn and Te, in the lattice,
and postulates that donors combine with gallium vacancies to produce
complexes such as Sn,,V&, and TeAsV&,. By rejecting the common idea
that Sn, is the dominant acceptor, it is possible to explain why in Sn-doped
meltgrown, VPE, or LPE GaAs, the ratio of N D / N A is usually of the order of 4,
(a result that would otherwise certainly not be expected for LPE material).
Hurle's model also explains why the free carrier concentration ceases to be
proportional to the Sn or Te doping at high concentrations. Two such curves
are shown in Figs. 1 and 2.
Some of the features of Hurle's model (without going into all the details)
are as follows. The gallium vacancy is assumed to have a double negative
charge in n-type material so that the complex acceptor Sn,,V,, is formed
by the reaction
snd, + vE; = Sn,,V,, (2)
70 A. G . MILNES
’ -’?8
- lou8 ’1 -
i
m
I Ol6
,/ , 1 I
MELT CONCENTRATION (Tell
FIG. 1. Carrier concentration versus Te concentration in an LPE melt; T = I123 K. Data

from Goodwin er 01. (1969). (See Hurle, 1977; reproduced by permission of her Britannic
Majesty’s Stationery Office.)
Hurle then shows that V& and ND/NA should be independent of PAs2and
that this explains the similar compensation results for VPE and LPE layers.
It is not argued that Sn, does not exist, but only that it is not the dominant
acceptor. Furthermore, for Si doping, it is suggested that SLaV& occurs but
that Si, can be dominant.
During the annealing of n-type GaAs(Te), the formation of compensating
acceptors is a complicated function of the initial doping, arsenic pressure, and
1
(100)
. (211)Go
- Colculoted
-
n
10”
-
z“
I
0
z
1015
I , , ,.,,..I . . ,..,..I I * 1-
1014 10‘6 10’8
N,+NA ( ~ r n - ~ )
FIG.2. Carrier concentration versus total ionized impurity concentration for Sn-doped
VPE layers. Dashed curve is expected behavior if electron-hole generation is dominant at the
growth temperature. Full curve is expected dependence. Data from Wolfe and Stillman (1973).
(See Hurle, 1977; reproduced by permission of her Britannic Majesty’s Stationery Office.)
ARSENIC PRESSURE ( T o r r )
FIG.3. Acceptor density in the heat-treated GaAs crystals at IO00"C and 900°C as a func-
tion of arsenic vapor pressure. Heat treatments were performed for 67 hr. The initial electron-
carrier concentrations were [published with permission from Nishizawa ef 01. (1974)j:
Symbol Sample Conc. (cm-') TH("C)
0 A 9 x 1016 1000
0 C 1.8 x 1017 900
A F 5.4 x 10i7 I000
0 I 9.5 x 101' I000
0 K 2.8 x 10" I000
temperature as shown in Fig. 3. Hurle explains the cusped minima as follows.

He supposes that upon annealing, As atoms exchange with the ambient
arsenic pressure but that Ga atoms diffuse very slowly and no exchange is
possible. Following Hurle's (1977) explanation,
Te:, + Vk; = Te,,V& (3)
and
upon raising the crystal to the annealing temperature arsenic Frenkel defects and some
gallium Frenkel defects form
AS,, + Vi = VA, + AS^ (4)
Ga,, + Vi = V,, + Gai (5)
where denotes a vacant interstitial site.
Whilst from mass action, the Frenkel product [V,,][Asi] is constant, the ratio of [V,.]
to [Asi] will depend on the ambient arsenic pressure since exchange with the ambient
phase is deemed to occur. With the gallium Frenkel defects, however, we have:
[VG.] = [Gail (6)
since no exchange is allowed
72 A. G. MlLNES
We suppose that the gallium vacancies are strongly “gettered” by the donors via
reaction (3) to form complexes. For an arsenic pressure for which [As,] [V,,], the follow-
ing may be expected to occur as the crystal is cooled. Firstly the arsenic Frenkel defects
can recombine leaving the net excess of As,. The latter can combine with the Ga, via the
reaction
Ga, + Asi = Ga,As, (7)
and the interstitial pairs can aggregate at a somewhat lower temperature to form the ob-
served interstitial loops. Since the gallium Frenkel reaction is supposed to occur to a sig-
nificant extent only when the V& are “gettered,” the absence of interstitial loops in
heat-treated undoped GaAs is explained. For low annealing pressures, corresponding to
[As,] < [ViJ, we can hypothesize that the unstable Ga, is removed from solution by the
formation of the antistructure defect:
Ga, + V,, = Ga,,
as proposed by Van Vechten (1975).

If now the arsenic pressure corresponds exactly to stoichiometry, namely [Asi] = [V,.],
then there are now no sinks for the Gai as the crystal cools save for the complex formed
by reaction (3). We may therefore suppose that under these conditions, the interstitial
annihilates the complex to restore the donor state:
Te,,V& + Ga, = Teis + 2e. (9)
In consequence there will be a sharp (cusped) minimum in the number of acceptors in the
vicinity of the pressure corresponding to the stoichiometric composition p:,, as observed
experimentally.
In a discussion of lattice dilation associated with VGaGaiTei(Dobson

et al., 1978),a cusped minimum is also found, and Hurle speculates that this is
caused by the Te,,V,, complex. Following his development of a point-defect
+
equilibria model based on Frenkel disorder, As, V,,, Hurle has further con-
sidered Te, Sn, and Ge doping effects (Hurle, 1979b,c,d; see also Logan, 1971).
One problem in the study of native defects is that of distinguishing
between vacancies and interstitials. For instance, treatment under Ga-rich
conditions may be expected to result in material with either As vacancies or
Ga interstitials, and there is need for an experiment that establishes which is
dominant. One method that has been attempted is a study of positron
lifetimes in GaAs. Because of the repulsive force between ion core and
positron, the positron, after thermalization in the crystal, can be trapped at
vacancy-type defects, but not at interstitials. The trapping may result in an
increase in positron lifetime because the electron density at vacancies is
lower than that in the bulk.
The results obtained by Cheng et al. (1979)for such an experiment did not
-
show evidence for the existence of As-vacancy defects in heat-treated
( 1000°C, plus quench) samples. This suggested three possibilities:
(1) The positron cannot be effectively trapped at As vacancies
(2) The arsenic deficiency exists only in a narrow region (x < 200 pm)
below the surface
(3) The arsenic deficiency is not mainly in vacancy form (e.g., Ga

interstitials are dominant).
It is possible to muster some arguments in favor of all of these possibilities,
and in the present state of knowledge it is difficult to choose between them.
However, the experiment did generate some evidence for the presence of Ga
vacancies or multiple vacancies in LEC-grown materials.
This discussion has touched on only a few aspects of model-building
problems in understanding defect interactions in GaAs, and it should serve
to indicate some idea of the complexities of defect studies.
Electron paramagnetic resonance (EPR) studies of GaAs do give evidence
of site locations and symmetries for some impurities. Unfortunately, the EPR
studies do not work as well for GaAs as for Si because of the hyperfine line-
broadening effects and high nuclear spins of Group 111 and V elements
(Eisen, 1971). Rutherford backscattering can also sometimes distinguish
between substitutional and interstitial location of impurities in GaAs;
however, it is generally difficult to infer anything about trap structure except
by very indirect methods (Kudo, 1979).
An example of such a method comes from the study of Schottky barrier
formation on n- and p-type GaAs as monolayers of metal (or oxygen) are
added to “clean” surfaces. The Fermi-level pinning is found to be at about
E, - 0.75 eV for n-type GaAs and at about E, + 0.5 eV for p-type GaAs and
is nearly independent of the metal used, as shown in Fig. 4a.
eV / / / / / / / / / / / C8M
1.2
0.0-
-
- A A
GoAs (110)
A
0
t
0.4
‘.4c73
M I S S I N G As
0.75 eV ACCEPTOR DUE
TO MISSING A s 1.4
GoAs Eg
0.5 eV DONOR DUE
0 TO M I S S I N G G o
0
DONOR
(b) (C 1
FIG.4. Schottky barrier pinning and the unified defect model of Spicer el a/. (1980): (a)
final pinning position of the Fermi level for n ( 0 )andp ( A ) GaAs; (b) postulated defect levels;
(c) interface states.
74 A. G. MILNES
Spicer et al. (1980) explain their unified defect model in terms of states
produced in the GaAs near the interface, as suggested in Fig. 4b and c.
They comment that
these probably represent the simplest stable defects since the adatoms were applied very
“gently” and the samples were not heated. Thus, the energy levels in Fig. 4(b) which are
considered accurate to 0.1 eV, represent the simplest defect levels stable at room tempera-
ture. The energy levels are most accurately known; the acceptor or donor nature of the
defect is next-best known; whereas, there is the least certainty about the identity of the
missing atom.
Based on the low coverage needed for pinning and independence of pinning energy
on extreme changes in the chemistry of the adatom, it is concluded that the pinning mecha-
nism must be indirect, i.e., that the pinning can not be due to levels directly introduced
by the adatom, i.e., states that depend on the electronic orbitals of the adatoms. Rather
the new states must be generated indirectly, i.e., the perturbations of adatoms must create
new energy states which do not depend on the electronic structure or other characteristics
of the adatom.
Most importantly, there is a close correspondence in the pinning positions found in
these experiments with submonolayer coverages of adatoms and those found in practical
MOS or Schottky barrier devices with many monolayers of adatoms.
Once the indirect nature of the formation of the new states becomes apparent, it becomes
very attractive to assume that these new states are due to lattice defects induced by the
adatom since this provides the simplest mechanism consistent with the large range avail-
able. For oxygen, both valence-band spectroscopy and LEED show that the surface be-
comes disordered at low coverage. lt is easy to conceive how, in such a situation, vacancies
or more complex lattice defects may be created. The fact that the clean surface is heavily
strained by the large surface rearrangement makes it easier to see how the surface may be
disordered by a relatively small oxygen adsorption. For thick oxides, the 111-V atoms
move through the oxide to the outer surface of the oxide before they react with the oxygen.
This is in contrast to Si-Si02 where the oxygen moves through the oxide to react with the
Si at the Si-oxide interface. The outward movement of the 111-V atoms can aid in the for-
mation of additional defect states as the Ga and As atoms near the interface are consumed
in the growth process. This is complicated by the nonuniform chemical spatial distribution
in thick oxides.
The situation with regard to Schottky barrier formation is also interesting. One of the
most surprising results in the work leading to the unified defect model was the discovery
that, even for metals where no appreciable chemical reaction was expected, the semi-
conductor material moves out into the deposited metal in surprisingly large quantities.
This is now well documented in the literature, and a mechanism has been suggested for it
based on the heat of condensation of the metal on the semiconductor.
There has been a mystery concerning Schottky barriers on Ill-IV’s which the unified
defect model clears up. On Si the Schottky barrier height varies strongly, depending on the
cleanliness of the surface on which it is deposited. In contrast, 111-V Schottky barriers
have been found to be surprisingly insensitive to the surface oxygen or oxidation. This is
just what would be expected on the basis of the [Spicer]defect model since the same defect
levels and thus the same pinning position would be produced by oxygen on the metal;
thus, the Schottky barrier height on Ill-V’s would not be affected (to first order) by addi-
tion of oxygen to the surface before the metal.
The defect model of Spicer et al. (1 980) therefore suggests consideration of

E, - 0.75 eV and E, + 0.5 eV ( f0.1 eV) as native defect levels in bulk GaAs.
They are less certain, however, about the identity of the defects, although
speculating that the midgap state is related to missing As. A review by

Brillson ( 1982) discusses the matter. From photo-EPR measurements,
Weber and Schneider (1982) find double-donor action in a center they
identify as AsGa,and the energy levels involved (10 K) are Do/D+at E, - 0.77
eV and D+/D2+at E, + 0.50 eV, which are close to the Fermi-level pinning
energies of Spicer et al. (1980) (see also Ikoma and Takikawa, 1981).
A case can be made that the E, - 0.75 eV level is associated with missing
Ga since a level between 0.75 and 0.83 eV, presumed to be the Spicer level, is
commonly seen when growths are made under As-rich conditions (as in VPE).
Growth under various ratios of AsH,/GaCI in the range l/3-3/1, by Miller
et al. (1977), shows that the mid-gap-level (E, - 0.82 eV) concentration
increasesas the ASH, is increased. Several other studies identify the E, - 0.75
to 0.82 eV level as Ga vacancy related (E, is the energy of the conduction-band
edge). An example is the study of Bhattacharya et al. (1980) with variations
of the As/Ga ratio in organometallic growth of GaAs.
Another interpretation of As-rich growth results would be that the level
is AsGarelated or even a complex such as AsGaVCa or AsGaVAs. Bhattacharya
et al. also find an electron trap at E, - 0.36 & 0.02 eV. If we exclude impurities
such as 0 as the cause of this level, the candidates for this level might by V,
or AsGa(or, again, some complex of these). Arsenic-rich growth conditions
might also be expected to produce arsenic interstitials. Little is known about
the energy levels expected for Asi. Arsenic interstitials have been suspected
by Driscoll et al. (1974; also Driscoll and Willoughby, 1973) as possibly
responsible for a shallow donor level at E, - 0.035 eV. Pons and Bourgoin
(1981a,b) conclude that levels at E, - 0.45 eV, -140 meV, and -0.31 eV
involve arsenic sublattice Frenkel pairs.
Yet another interpretation of the Miller results has been proposed by Zou
(I98la). He assumes the following relationship.
PAsH3IPGaCI OC [VGa]/[VAs] OC [VGa]2/[VGal[VAs] OC [VGa12 (lo)

So, for the mid-band-gap trap (0.75-0.83 eV), Zou assumes that its config-
uration must be related to (VGa)2 and/or AsGa. He further assumes that
these are related by the reaction
AS,, + (V,,); = As, + V,,V,, + e- (1 1)

Thus the concentrations of (VGa); and As Ga might be proportional to each
other. Ozeki et al. (1979) stated that the mid-band-gap level may, in fact, be
two levels quite close to each other in energy, and Zou tentatively suggests
these may be (VGa); and AsGa.
Let us leave for the moment consideration of the mid-band-gap level(s)
and consider possible As-vacancy-related states. In studies of resistivity of
material grown by the VPE-Ga/AsCI,/H, process, Saito and Hasegawa
(1971)identify a deep acceptor at approximately 0.5 eV from the valence-band
76 A. G. MILNES
edge, which they infer to be related to an arsenic vacancy. Spicer’s model

requires a Ga-vacancy-related donor at E, + 0.5 eV, and so acquires no
support from the Saito and Hasegawa observation.
Growth under Ga-rich conditions, as in LPE growth from a Ga melt,
normally results in two minority carrier (hole) traps in n-type GaAs layers.
These hole traps termed A and B are at E, + 0.41 eV and E, + 0.71 eV,
respectively. The electron trap at E, - 0.75 to 0.83 eV is not seen in the
material. The physical natures of the A and B levels have still to be deter-
mined, but one surmise is that they have to involve either an As vacancy or a
Ga atom on an As site. Perhaps the A level at E, + 0.41 eV may be a manifes-
tation of Spicer’s interface defect level at E, + 0.5 f 0.1 eV (assuming their
identification to be changed to missing As), but this assignment is very
hypothetical.
One attempt to make an assignment of the A and B level, is that of Zou
(1981a). He concludes that the hole trap A (E, + 0.41 ev) might be GaAsVGa,
and hole trap B (E, + 0.71 eV) might be AsGaVGa.The argument is involved
and is not presented here. To accept V, complexes in LPE material grown
under such Ga-rich conditions, it is necessary to assume that the V, diffuses
from the substrate into the LPE layer during growth. It should be noted
that AsGaVGais a possible candidate for the mid-band-gap electron trap
EL2 (E, - 0.75 to 0.83 eV) seen in material grown under As-rich conditions.
Another candidate that has been suggested for EL2 from VPE growth is
ASG,VC~VA~ (Zou, 1981b; Zou et al., 1982). Attempts to show that in LPE
material hole trap B is identical with electron trap EL2 of VPE material fail,
however, because the electron-trapping action is not observed in DLTS tests
of material containing trap B.
111. TRAPS(AND NOMENCLATURE)

FROM DLTS STUDIES
In 1970, Sah and co-workers showed the considerable merits of studying

traps by emission in junction depletion regions from the associated capac-
itance changes. The circuit-processing technique known as deep-level
transient spectroscopy (DLTS) added to the convenience of this type of
measurement (Lang, 1974; Lang and Logan, 1975; Wada et a!., 1977). The
group at the Laboratoires d’Electronique et de Physique Appliqee (LEP)
has led this technology in France and has published useful catalogs of
electron and hole traps in bulk, VPE, LPE, and MBE GaAs.
Deep-level transient spectroscopy studies yield the emission coefficients
en,pwhich are functions of temperature. For electron emission, the theoretical
expression from the detailed balance of emission and capture events is
el = ( 0 1 (v1 )NrJ,go /Sl) exp( - A E / W (12)
where a1is the minority-carrier capture cross section, (v l ) the mean thermal
velocity of minority carriers, N,, the effective density of states in the minority-
carrier band, g1 the degeneracy of the trap level, go the degeneracy of the
level not occupied by the electron (often equal to unity), and AE the energy
separation between the trap level and the minority-carrier band.
However, the cross section ol may be activated thermally with an energy
Eb, and the trap depth AE may vary as a function of temperature (often as a
nearly linear function of temperature with c1 the variation in eV K - I .
Furthermore, N,,, depends on TZ. Application of these corrections yields
e,,(T) = ~,,T2e,g,(expa/k)exp[-(El~+ Eb)/kT] ( 12a)
where a,,, is the extrapolated value of a,,for T = 00, E I o the extrapolated
value of E l for T = 0, g,, the degeneracy factor, and y,, a constant equal to
2.28 x 10’’ cm-’ sec-’ K-’ in GaAs. The plot of T’/e,, as a function of
1/T yields a,gn exp(a/k) (denoted by ofla), as well as E I o + Eb (denoted by
E m ) ;these two parameters, emand Ens, are actually the “signature” of a trap,
even if they do not have a direct physical meaning (Martin et al., 1977b).
The lines presented in Figs. 5 and 6 are plots of T’/e,, and T2/e, versus
1000/T for electron and hole traps in GaAs (Mircea et al., 1977; Pons, 1980).
The corrections for Eb and AE have not been applied in Figs. 5 and 6 or in
Tables V and VI that provide the keys to the data.
Table V shows the capture cross sections of the electron traps and the
class of material in which they were observed. Martin et al. (1980a,b) have
revised certain of these values. In particular, the lines for T2/e, of the EL1
and EB1 have been raised by over an order of magnitude. This also matches
better the ni (intrinsic carrier concentration) studies of Blakemore (1982).
The letters T, I, B, and L in the labels represent the names of various
laboratories. It seems probable that EL2 = ETl = ESl = EB2 = ECl, and
this is a commonly found electron trap at about E, - 0.83 eV. (In the past,
this was considered to be oxygen related, but is now thought to be a complex
that involves V, or AS,, or AsGaVGa). The dotted lines are some trap levels
seen in VPE GaAs (-mid-10” cm-3 n type) at Carnegie-Mellon University,
Pittsburgh; EC3 is probably the same as EI1; EL4 is similar to EB5; and
EL5, EB6, and EC5 are probably the same trap. Other likely candidates for
equivalence may be seen by inspection of the energy levels and the locations
on the plot (the location being dependent on both the energy and the
capture cross section).
Discussion of this catalog is not appropriate at this time, except to say
that the nomenclature is now in general use. It may be remarked that the
levels seen in the vicinity of E, - 0.17 eV are thought to be Cu related.
Figure 6 shows a collection of hole-trap data similarly plotted and Table VI
provides the key to the data (Mitonneau, 1976; Mitonneau e f al., 1977).
78 A . G . MILNES
FIG. 5 . Electron-emission coefficients plotted as T 2 / e , versus IOOOIT for electron traps

communicating with the conduction band in GaAs. (After Martin er a/.. 1977b; reproduced
with permission of IEE, London.)
2 4 6 8
1000/T (K-')
FIG.6. Plots of T 2 / e ,as function of IOOO/T for all hole traps: (T) University of Tokyo,
(S)University of Sheffield, (B) Bell Telephone Laboratories, (L) LEP work. The curves B giving
Lang's results are extrapolated from his DLTS spectra. (After Mitonneau et a / . , 1977; repro-
duced with permission of IEE, London.)
TABLE V
TRAPS
ELECTRON IN GALLIUM
ARSENIDE"
Activation Cross
Label in energy section
Fig. 5 E,, (eV) b p . (cm) Observation*
ET 1 0.85 6.5 x 1 0 - 1 3 BM
ET2 0.3 2.5 x 1 0 - l 5 BM
ES I 0.83 1.0 x 10-13 BM
EFI 0.72 7.7 x 1 0 - 1 5 Cr-doped BM
El 1 0.43 7.3 x 10-16 VPEM
El2 0.19 1 . 1 x 10-14 VPEM
El3 0.18 2.2 x 10-14 VPEM
EB I 0.86 3.5 x Cr-doped LPEM
EB2 0.83 2.2 x As-grown VPEM
EB3 0.90 3.0 x lo-" El M
EB4 0.71 8.3 x EIM
EB5 0.48 2.6 x As-grown MBEM
EB6 0.41 2.6 x 1 0 - 1 3 EIM
EB7 0.30 1.7 x 10-14 As-grown MBEM
EBS 0.19 1.5 x 10-14 As-grown MBEM
EB9 0.18 Imprecise EIM
EBlO 0.12 Imprecise EIM
ELI 0.78 1.0 x 1 0 - 1 4 Cr-doped BM
EL2 0.825 (0.8-1.7) x VPEM
EL3 0.575 (0.8-1.7) x 10-13 VPEM
EL4 0.51 1.0 x l o - " As-grown MBEM
EL5 0.42 (0.5-2.0) x VPEM
EL6 0.35 1.5 x 1 0 - l ~ BM
EL7 0.30 7.2 x As-grown MBEM
EL8 0.275 7.7 x l o - " VPEM
EL9 0.225 6.8 x VPEM
ELI0 0.17 1.8 x 1 0 - 1 5 As-grown MBEM
ELI I 0.17 3.0 x VPEM
EL12 0.78 4.9 x I o - I z VPEM
EL14 0.215 5.2 x BM
EL15 0.15 5.7 x 10-13 EIM
EL16 0.37 4.0 x l o - ' * VPEM
a After Martin er a/. (1977b). Reproduced with permission of the IEE,
London.
* BM, bulk material; VPEM, VPE material; LPEM, LPE material;
MBEM, MBE material; EIM, electron-irradiated material.
Full confidence cannot be placed in the values given for several reasons
that reflect limitations in the measurement techniques. The activation energy
(T2corrected) measured from a T2/e, vs. 1000/T chart gives an energy
+
E , , Eb.To convert this to the true defect energy level E I o ,it is necessary
to know Eb.If this correction is desired, one must determine Eb from the
80 A . G. MILNES
TABLE VI
HOLEPARAMETERS
F,,* A N D up, OF TRAPSAS CALCULATED
FOR FIG.6"
Activation Emission
Label in energy section Excitation Chemical
Fig. 6 E,, (eV) up, (m-') Type of sample modeb origin
HTI 0.44 1.2 x lo-" VPE

HS I 0.58 2.0 x lo-''' LPE
HS2 0.64 4.1 x LPE
HS3 0.44 4.8 x 1 0 - 1 ~ LPE
HBI 0.78 5.2 x lo-'(' Cr-doped LPE Cr
H B2 0.71 1.2 x As-grown LPE
H8 3 0.52 3.4 x Fe-doped LPE Fe
H B4 0.44 3.4 x lo-'* Cu-doped LPE cu
HB5 0.40 2.2 x lo-" As-grown LPE
H B6 0.29 2.0 x lo-'' Electron-irradiated LPE
HLI 0.94 3.7 x lo-'' Cr-doped VPE EO Cr
H L2 0.73 1.9 x As-grown LPE EO
H L3 0.59 3.0 x lo-'' Fe-diffused VPE E Fe
H L4 0.42 3.0 x lo-'' Cu-diffused VPE EO cu
HL5 0.41 9.0 x As-grown LPE EO
H L6 0.32 5.6 x 10- VPE with p ' layer E
H L7 0.35 6.4 x lo-'' As-grown MBE E
H L8 0.52 3.5 x lo-'' As-grown MBE E
H L9 0.69 1.1 x lo-" As-grown VPE 0
HLlO 0.83 1.7 x lo-'' As-grown VPE EO
HLI I 0.35 1.4 x Melt grown 0
HLl2 0.27 1.3 x lo-'' Zn-contaminated LPE E
" Reproduced with permission of the IEE, London (after Mitonneau ef al., 1977).
EO, electrical and optical; E, electrical; 0, optical.
variation of capture cross section with temperature. For electrons this is

approximated over some temperature range as
a,,= a, exp(-Eb/kT)
where a, is the limiting value of a,,at T + 00.
This can lead to quite a significant correction. For example, one electron
trap in VPE GaAs was measured as having a T2 corrected activation energy
of E, - 0.48 eV, but Eb was determined to be 0.09 eV; hence the corrected
defect level was inferred to be 0.39 eV. Sometimes, Eb is found to be negative
in value and one then is concerned that the model (or the experiment) is
not adequate [see also Majerfeld and Bhattacharya (1978) and Sakai and
Ikoma (1974)l.
Other sources of uncertainty in DLTS measurements are the effects of
high electric fields in depletion regions (Vincent et al., 1979), the effects of
trap gradients, and the possibility of traps closely spaced in energy-level
interacting in the output and producing a nonexponential capacitance

transient that is misinterpreted as a true exponential by the processing
electronics, which then yield an incorrect answer.
-
For the EL2 level, Mircea and Mitonneau (1979) have found that if the
electronic field is high ( 1.9 x lo5 V/cm), the electron emission is large when
the electronic field is (111)oriented, pointing from As to Ga, and is signif-
icantly lower for the opposite field orientation. For the (100) and (110)
directions, the emission rates take intermediate values. This suggests that the
defect structure must initially have nearly tetrahedral symmetry and that it
tends to favor a simple structure for EL2 such as an isolated vacancy rather
than a complex such as an impurity coupled to a vacancy.
In a further study of the EL2, however, Makram-Ebeid (1980a) has
shown that at low temperature and moderate fields, the EL2 kinetics are
dominated by a combination of hot-carrier capture and impact ionization.
At high electronic fields the kinetics are dominated by phonon-assisted
tunnel emission of electrons. Thus the anisotropy observed by Mircea and
Mitonneau (1979) may be explained by these effects. Noise studies have been
made by Roussel and Mircea (1973) and photocapacitance studies by
Vasudev et al. (1977).
Another important measurement technique that has been developed is
deep-level optical spectroscopy (DLOS). In this technique, photostimulated
capacitance transients are studied after electrical, thermal, or optical excita-
tion of a junction or Schottky barrier. This technique provides the spectral
distribution of both o;(hv) and o,O(hv), the optical cross sections for the
transitions between a deep level and the conduction and valence bands.
Besides its sensitivity, DLOS is selective in the double sense that o:(hv) and
o;(hv) are unambigously separated, and that the signals due to different
traps can be resolved from one another. As a result, the aO(hv)spectra are
measured from their threshold up to the energy gap of.the semiconductor,
over a generally large temperature range (Chantre and Bois, 1980; Chantre
et al., 1981). In GaAs, it has been applied to a study of the EL2 level, the “Cu”
level at E, + 0.40 eV, and the ELI, EL3, and EL6 levels. A comparison
between thermal and optical ionization energies allows the Franck-Condon
energy caused by lattice relaxation to be determined. For the EL6 electron
trap the thermal activation energy is 0.31 eV, but the optical ionization
energy for o: is much larger (0.85 eV), corresponding to a Franck-Condon
energy of about 0.6 eV.
IV. LEVELS
PRODUCED
BY IRRADIATION
Irradiation of GaAs and GaAs devices has been carried out primarily to
determine the effects on performance of FETs, Impatt and Gunn diodes,
LEDS, heterojunction lasers, and solar cells; however, in a few instances, it
82 A. G. MILNES
provides interesting clues to the physical nature of defects. The radiation may
be in the form of high-energy electrons ( > O S MeV), protons, neutrons, or
high-energy gamma or X rays. Such radiation may produce displacement
defect traps that alter device performance unless annealed out again.
Ion-implantation damage and subsequent annealing are not discussed here.
The effects of radiation on FET-device performance have been reviewed
by Zuleeg and Lehovec (1980).They report that GaAs and Si FETs operating
with the channel electrons in thermal equilibrium (the normal mode of FET
operation) degrade almost to the same extent under fast-neutron action. In
short-channel devices with hot-electron action, the GaAs FET has some
advantages over Si. The present technology for small-scale integrated GaAs
circuits appears to allow a radiation tolerance for neutrons of lo5 neutrons/
-
cmz and for ionizing radiation dose lo7 rad with a dose rate for logic upset
of 10'' rad/sec. This may be compared with a total dose hardness of lo6
rad in Si MOS integrated circuits (Berg and Lieberman, 1975; Kladis and
Euthymiou, 1972).The effect of radiation on GaAs diodes has been examined
by Wirth and Rogers (1964) and by Taylor and Morgan (1976).
In GaAs solar cells, the main effects causing performance deterioration
are the creation of deep centers that cause recombination-generation in the
depletion region and a decrease of the minority-carrier diffusion length.
Since GaAs is a direct gap material, photon absorption takes place in only a
few micrometers of depth. This makes the junction depth an important
parameter in the radiation-resistance behavior. In general, GaAs solar cells
[including (A1Ga)As-GaAs structures] are more resistant than Si cells to
radiation damage by both electron and proton irradiation. The efficiencies of
electron-damaged GaAs cells tend to recover, in part, with anneals at
temperatures as low as 200-300°C (Li et al., 1980a).
We consider now some of the indications of the nature of defects obtained
from irradiation studies. One of the principal effects of radiation is that
acceptor-type defects are produced that remove electrons from n-type GaAs.
Conversion to p-type has been studied by Farmer and Look (1979) and is
attributed to the development of a relatively shallow acceptor at -0.1 eV.
The carrier removal rates observed by many investigators for 1 MeV electrons
are shown in Fig. 7. There is a range of a factor of 10 in An/4 values but no
systematic dependence on donor type or concentrations. Large carrier
removal rates seen by some investigators seem to be related to larger electron
fluxes (so the rate of damage creation can be a factor), but Farmer and Look
(1980)find the removal rate to be dependent on the Fermi level but indepen-
dent of the irradiation flux. No great differences are seen for LPE, VPE, or
bulk material, from which it may be inferred that the created defects tend to
be intrinsic in nature and not affected directly by defects or impurities in the
starting material. Table VII shows a collection of data assembled by Lang
(1977) for energy levels seen by a number of investigators.
The introduction rate for several defect levels by 1 MeV electrons shows a
crystal-orientation dependence and suggests that these defects (notably El,
E2, and E3, at E, - 0.1 3, E, - 0.20, and E, - 0.3 1 to 0.38 eV) can according
to Lang et al. (1977) be related to Ga-site displacements, or according
to Pons and Bourgoin (1981a,b) to As-site displacements. Understanding
of the knock-on process dependence with energy now favors the As-site
displacement interpretation.
-----O
lO
*'
+ (1 MeV electrons/cm2 1
FIG.7. Carrier removal An against 1 MeV electron fluence Q for n-GaAs. Note that An/Q
ranges from 0.5 to 5.0 cm-', but with no systematic dependence on donor concentration or
species (after Lang, 1977; reproduced with permission of the Institute of Physics, UK):
n0
Symbol (cm - Reference
0 4 1016 tos 1017 Aukerman and Graft (1962)

7 x 1015 Thommen (1970) (Stage 3)
0 6 x lo'* (Si, Te) Kahan er a/. (1971)
A 9 x I015(Sn) Kalma and Berger (1972)
A 2 x loi7(Si) Kalma and Berger (1972)
0 2.5 x 1OI6 Pegler et al. (1972)
X 1.3 x I O ~ ~ ( V P E ) Dresner ( 1974)
V 1 x I0I6(LPE) Lang and Kimerling (1975)
84 A. G . MILNES
TABLE VII
ENERGY
LEVELS AND INTRODUCTIONRATESOF DEEPLEVELSINTRODUCED INTO GaAS
BY 1 MeV ELECTRONOR Cob' IRRADIATION AT ROOMTEMPERATUR~
Energy levelsh(eV)
Electron traps Hole traps

Measurement
technique El E2 €3 E4 E5 HO H1 References
Thermal emission activation energy:

As measured -0.08 0.190.45 0.76 0.96 -0.09 0.32 Lang (1974). Lang
With T' correction - 0.180.41 0.71 0.90 - 0.29 and Kimmer-
With a ( T )correction - - 0.31 - - - - ling (1975)
Hall coefficient 0.12 0.200.38 - - 0.10 - Vitovskii el a / .
activation energy ( 1964)
0.13 0.170.31 - - 0.10 - Brehm and
Pearson (1972)
Photoconductivity - - 0.38 0.52 0.72 - - Vitovskii er d.
threshold ( 1964)
- - - 0.54 - 0.10 - O'Brien and
Corelli (1973)
Best value 0.13 0.20 0.31 - - 0.10 0.29 -
Introduction rate N/qi 1.8 2.8 0.7 0.08 0.1 - 0.7 Lang and
(cm-'), I MeV electrons, Kimerling
300 K (1975)
From Lang (1977). Reproduced with permission of the Institute of Physics, UK.
a
Energies of electron(ho1e) traps are measured with respect to the conduction(va1ence)

bands.
The threshold energy for atomic displacement in GaAs appears to be of

the order 10 eV (which corresponds to an electron energy of about 0.25 MeV).
Hence, 1 MeV electrons have sufficient energy to displace more than one
atom, and so the possibility exists that divancancies such as V,,V, may be
produced in addition to (VGa+ Gai) and (VAs + Asi). Attempts to observe
divacancy production by varying the electron energy over the range 0.25- 1.7
MeV were, however, inconclusive (Pons et al., 1980). This was explained by
suggesting that if such centers are created, either the energy levels are too
close to the band edges to be readily observed or that the centers anneal at
below room temperature. Annealing stages after electron irradiation have
been found at 235, 280, and 500 K. Pons et a / . (1980) propose that
there is a recombination (due to Coulomb attraction) of a divacancy with a neighboring
interstitial (Asi or Ga,), resulting in single vacancies (V,, or V,,) and in associations of
+
antisite defects (one Ga atom on an As site or vice versa) with single vacancies: (Ga,, VGa)
and (As,, + V,J. These reactions could occur in two separate stages: one stage due to
the Asi jump and the other one due to the Ga, jump.
10
10-6 I 1 I
(015 4 0’6 101’ 10’6
no (cm-3)
FIG.8. Anneal rate (200°C) for the E2 [ A , Lang er a/. (1976)l and A2 [ O , Ackerman and
Graft (1962)] levels seen in GaAs after irradiation. The proportionality to n2’3 suggests that the
annealing is by cornplexing with donors. (After Lang, 1977; reproduced with permission of the
Institute of Physics, UK.)
The suggestion is, therefore, that the two low-temperature anneal stages
are in some way associated with such reactions. Pons and Bourgoin (1981b)
at present favor the view that the El, E2, and E3 levels at E, - 0.045, -0.140,
and 0.330 eV are manifestations of As vacancies or interstitial Frenkel pairs.
The annealing around 500 K involves the El and E2 levels and tends to
be proportional to the 2/3 power of the preirradiation donor concentration
(see Fig. 8). Lang remarks that
this functional dependence is exactly what one would expect if the donor atoms were acting
as sinks for the defect (or defects) responsible for the (El, E2) group. This can be seen from
the following simple argument. The average distance between donors is proportional to
the inverse cube root of the donor concentration. The average number of random jumps
needed to travel a given distance is proportional to the square of that distance. Hence the
average number of jumps necessary for a defect to travel to a donor should be proportional
to the 2/3 power of the donor concentration.
Since the (El, E2) production rate is independent of doped species and
material type (LPE, VPE, LEC), it suggests that these defects at E, - 0.13 eV
and E, - 0.20 eV are predominantly intrinsic in nature and annihilate by
complexing with donors (or acceptors related to donor concentration)
during the 500 K annealing stage.
Another interpretation is proposed by Pons and Bourgoin (1981b) since
their experiments show that the charge state of the trap influences the
86 A . G. MILNES
annealing rate and that the dopant concentration is not significant in the way
proposed by Lang. The model proposes that El and E2 are different charge
states of the same defect site and that when El is lying above the Fermi level
at low donor doping concentrations (i.e., is empty of electrons), its annealing
rate is less than when it is below the Fermi level (Pons, 1981).
Annealing of the E3, E5, and H 1 levels ( E , - 0.3 1 to 0.38 eV, E, - 0.9 eV,
and E, + 0.29 eV) seems independent of the donor concentration. There are
some indications that the E5 (E, - 0.9 eV) and HI ( E , + 0.29 eV) introduc-
tion rates vary from specimen to specimen, presumably depending on the
impurity content of the preirradiated material. Annealing has also been
studied by Walker and Conway (1979).
Electron paramagnetic resonance studies of semi-insulating GaAs:Cr
after 2 MeV electron irradiation show signals that are assignable to the
antisite defect AsGa(Strauss et a/., 1979; Kaufmann and Schneider, 1982).
The possibility that the signal comes from a complex such as AsGa-X,
however, cannot be entirely ruled out. Kennedy et al. (1981; Wilsey and
Kennedy, 1981) believe that the AsGa antisite defects may be formed by
room-temperature diffusion of As interstitials to Ga vacancies.
Fast-neutron-irradiated GaAs also develops the Ash: signal and this
anneals out between 450 and 500°C. No direct ESR evidence yet can be
offered for the formation of the complementary GaAsantisite defect. Elliott
et a/. (1982) have attributed an acceptor level at 78 MeV to the GaAsantisite
defect, since the defect is tetrahedrally coordinated and appears in LEC
material grown from Ga-rich melts. Ta et a/. (1982 ) agree that this level is
associated with excess Ga, but find some evidence that the level increases
with concentration if there is increased B in the crystal.
From simple consideration of a group V element on a group 111 site,
AsGamay be expected to be a double donor. However, the energy level(s)
associated with AsGahave not yet been established. The complex (As,,VA,)
is considered by Lagowski et al. (1982) to be a candidate for the EL2 trap, and
some evidence for double-donor action exists.
Fast-neutron-irradiated GaAs may be expected to have incurred con-
siderable damage. The primary-knock-on effect is likely to transfer as much
as 50 KeV to the atom that is displaced and this, because of secondary
collisions, may produce many further displacements (Worner et al., I98 I).
This seems to be borne out by the 400-600°C high annealing temperatures
required. However, Coates and Mitchell (1975) suggest that close interstitial
vacancy pairs may also be involved.
Turning again to consideration of electron-irradiated GaAs, Farmer and
Look (1980) reach the following conclusions:
(i) Room-temperature defect production is nearly always sublineal with h e n c e . This
can be explained by a simple model of stable (or nearly stable) Ga vacancies competing
with traps and sinks for the mobile Cia interstitials.
(ii) The wide variation of reported free-carrier removal rates cannot be accounted for
by a flux dependence. Some of the variation is probably due to the formation of defect-
impurity complexes, which may include E5 and H I . Another factor, quite important in
high purity samples, is the position of the Fermi level, which can change the proportionality
between the defect-production rate and the carrier-removal rate.
(iii) The 200°C annealing stage includes two first-order substages with annealing rates
close to those reported in the literature. Besides being first order, the first substage (1,)
is also nearly independent of sample growth conditions and doping levels and has a rela-
tively low prefactor. These attributes can be explained by a model in which the i , substage
involves Ga-vacancy related defects annealing by interactions with Cia interstitials which
are themselves emitted by interstitial traps. The I, substage is also best described by a dis-
sociation process with decreasing n o , although the exact relationship is not clear.
(iv) The defect model most consistent with all of the data includes E3 as a donor, HI
as an acceptor, and El and E2 as the two charge states of a double acceptor. The exact
identifications of these defects are, however, somewhat in doubt and must await further
experimentation.
Interesting observations have been made on the E3 level introduced into

LPE (loo), ( 1 lo), ( 1 1 1 ) and (111)material by 1 MeV electron irradiation. The
dependence of production rate of the E3 level on orientation is shown in
Fig. 9. The production rates are large for orientations that favor easy knock-
on of Ga atoms. This anisotropy seems also to be true for the El and E2
defects, which also anneal at 500 K and are therefore also believed to be
related to Ga-site displacements. The annealing actions seen at 235 and
280 K most probably are Asi and V, related. The high mobility of Asi
C
- <loo>
easy
C Go + As
0
8 0
>
<Ill>
<100>t110><111>
Go As t110> <111>
hard hard As
Ga+As Go Ga+As easy Go
1 MeV electron beam
direction (110)plane: Go As unit cell
I:> interstitial o As Go
site
FIG.9. Orientation dependence of the introduction rate of the E3 leve Four samples with
(100). ( 1 lo), ( I I1)Ga. and ( I I1)As surfaces exposed to the I MeV electron beam were irradiated
simultaneously to a total fluence of 4 = I x 10” cm-’. The right-hand side of the figure is the
(1 10) plane of the GaAs unit cell and illustrates three of the four sample orientations used. The
“easy” and “hard” notations are explained in the text. (After Lang er ol. 1977; reproduced with
permission of the American Physical Society.)
88 A. G. MILNES
at room temperature then suggests the possibility that the formation of VG,
is followed by creation of an antisite defect AsG,. Lang and his co-workers
(1977), as a speculation, comment that the AsGaantisite defect should be a
double donor and might be expected to have two levels near the conduction
band, such as El and E2. If El and E2 are donors and are present in quantity,
then some other level that is an acceptor must be produced in quantity to
provide the electron-removal action, such as the 0.1 eV, relatively shallow
acceptor seen by Farmer and Look (1979). Frenkel-pair defects are unlikely
to be involved for El and E2, according to Lang, since the need for 500 K
for annealing suggests long-range motion of the defects and not close-pair
recombination. However, this view does not concur with that of Coates and
Mitchell (1975) and has been challenged by Pons and Bourgoin (1981), who
regard the El, E2, and E3 levels as related to As-site Frenkel pairs. Weber
and Schneider (1982) also have discussed the AsGaantisite defect.
In the work of Lang et al. (1977) the El, E2, E3, and other levels have
been studied as a function of crystal composition in the range GaAs-
A1,,,Ga0~,As with the results shown in Fig. 10. The E3 level is seen to be the
p - -
-0
1.2
(, - n L
zl
a t.0
- -
W
z
W
-
0.6 - - -- FB - -
IP
0 -
C u a
- -
0.4,;
0.2 - -
0 1 1 I I I I I
only one that tracks the valence-band edge. Since vacancies, according to
Lang, tend to be created from valence-band-type wave functions, this suggests
that E3 is a vacancy, and the assignment V, is made because the E3 produc-
tion is greatest in situations that Lang considers to favor Ga knock-on. Both
views have been questioned (Wallis et al., 1981a). Displacement effects
produced by electrons have also been considered by Watkins (1976).
Consider now the traps produced by proton and neutron irradiations.
Protons of energy 50, 100, and 290 keV applied to n-type GaAs (LPE) layers
with fluences of lo'', lo", and lo'* protons/cm2, resulted in the trap levels
shown in Table VIII. The proton penetration depths are 0.5 pm for 50 keV,
1 pm for 100 keV, and 2.5 pm for 290 keV. The trap species produced depend
not only on energy but also on dose. Comparison of these results with those
of Table VII for the traps produced by 1 MeV electron irradiation and
Tables V and VI for electron and hole traps in as-grown material show a
substantial commonality. The following traps appear frequently in electron-
and proton-irradiated and as-grown material (with no specific deep impurity
dopant added) :
E, - 0.1 1 to 0.130 eV E, - 0.71 to 0.83 eV
E, - 0.18 to 0.20 eV E, + 0.71 to 0.73 eV
E, - 0.31 eV E, + 0.40 to 0.44 eV
E, - 0.52 eV E, + 0.31 eV
The uncertainty (range) values are as indicated or may be taken as about
k0.02 eV. Loualiche et al. (1982)find dominant levels acting as acceptor-like
TABLE VIll
DEFECT
LEVELS
OBSERVED
IN LOW-ENERGY
PROTON-IRRADIATED
GaAs"
Defect 100 keV 290 keV 100 keV 290 keV

level 10" 10" 10" 10"
Trap level (eV) protons/cm2 protons/cm2 protons/cm2 protons/cm2
Electron traps E, - 0.I 1 X
E, - 0.14 X
E, - 0.20 X X
E, - 0.31 X X
E, - 0.52 X
E, - 0.71 X X X
Hole traps E, + 0.059 X X

E, + 0.17 X X
E, + 0.44 X X
E, + 0.57 X X
E, + 0.71 X X
(I From Li el al. (1980b). Reproduced with permission of the Metallurgical Society of AIME.
90 A . G . MILNES
traps at E, - 0.22 eV and E, - 0.33 eV after proton irradiation of VPE

n-type GaAs.
Shallower hole traps around 0.17,0.l0, and 0.06 eV are also observed in
irradiated material but not commonly seen in as-grown GaAs. As-grown
LPE material tends to contain only hole traps at E, + 0.41 eV and E, +
0.71 eV. The Ga melt presumably suppresses traps related to V, and also
acts as a gettering sink for impurities that might exist on the specimen surface
or the inside of the specimen growth system. (Direct studies aimed at showing
the residual contamination of the Ga melt in LPE appear not to be reported,
but variations of growth purity have been reported for a sequence of growths
from the same melt. One LPE growth technology (Ewan et al., 1975) uses a
deep Ga melt system, and one of the advantages of this is that frequent
system “cleaning” and melt loading is not needed.)
Where clustering of defects occurs, as in high-energy neutron, proton, or
heavy-ion bombardment, one may expect many interacting energy levels to
result (Ludman and Nowak, 1976; McNichols and Berg, 1971; Stein, 1969).
Thus, a broadening of DLTS peaks occurs as may be seen from Fig. 11 for
O + and He+ ion bombardment.
The effects of ion implantation for doping of GaAs and the annealing of
the associated damage by bulk thermal or laser heating are not considered
I MeV (e‘)
X0.25
E2 €3 €4 E5
TEMPERATURE ( K )
FIG. 1 1 . DLTS spectra of four n-GaAs samples irradiated at room temperature with
1 MeV electrons, 600 keV protons, 1-8 MeV He’ ions, and 185 keV 0’ions. Note the general
trend toward a broader and deeper spectrum as the mass of the high-energy particle increases
(Lang, 1977; reproduced with permission of the Institute of Physics, UK).
here. Papers available include those of Davies (1976), Degen (1973), Eisen
(1971), Hemment (1976), Donnelly (1979, Donnelly E t a/. (1975), Elliott er al.
(1978), Favennec et al. (1978), Hunsperger and Marsh (1970), Hunsperger
et al. (1972), Moore et al. (1974), Ilic et al. (1973), Littlejohn et al. (1971),
Sansbury and Gibbons (1970), Sealy et al. (1976), Surridge and Sealy (1977),
Takai et al. (1975),Woodcock and Clark (1979, Yu (1977),and Zelevinskaya
(1973).
V. SEMI-INSULATING
GALLIUM
ARSENIDE
WITH
AND WITHOUT CHROMIUM
Chromium in gallium arsenide produces material of high resistivity,
typically in the range 0.3-2 x lo9 0-cm at 300 K. Calculations from known
effective masses suggest that perfectly pure (intrinsic) GaAs of band gap
1.42 eV should have n, = pi = 1.5 x lo6 cm-3 at 290 K. One may also
calculate n , ( T )from (n*p*)'/', where n* = (e,/c,) and p* = (eP/cp),for two
successive charge conditions of a deep level. For the Cr2+ P Cr3+ transition,
Blakemore (1982) obtains ni = 1.8 x lo6 cm-3 for the 300 K value in very
satisfactory agreement. The experimental evidence supports a value
1.7 f 0.4 x lo6 cm-3 (Look, 1977). The Fermi level should then be at
E, - 0.637 eV. Semi-insulating GaAs may be obtained with substantial
addition of Cr (the simple model suggests that Cr is a deep acceptor at about
E, f 0.73 eV). For high-resistance material, sufficient Cr must be added to
compensate for the residual (ND - NA) shallow doping. The electrically
active solubility of Cr in GaAs has been estimated to be in excess of 5 x IOI7
cm-3 and therefore can, in principle, compensate net shallow donors up to
the low 10I7cm-3. If an attempt is made to reduce the net shallow doping to a
very low level, for instance, by LEC growth in B 2 0 3 capped conditions in
BN crucibles, the material obtained is frequently semi-insulating even
though no Cr has been added (Thomas et al., 1981). The reason for this is not
certain but some investigators believe that it is related to EL2, a deep im-
purity level (Fig. 5 ) at about midgap that is almost always present in a
concentration about 1 O I 6 ~ m - As~ discussed
. in Section 11, it seems probable
that this level is gallium vacancy or AsGa related. If oxygen is not a component
of the level, possible candidates might be AsGaVGa,AsGaVAsrand VGaAsGaVGa.
Any study of semi-insulating GaAs with Cr present must recognize the
existence of the EL2 donor level and other possible levels (such as H10).
From ESR studies, Cr tends to be present on the Ga site. The outer
electronic shells of Ga are 3d1°4s24p' where the 4 . ~ ~ electrons
4~' participate
in the lattice bonding. The corresponding structure for Cr is 3d34s24p1and
therefore neutral Cr in a Ga site may be characterized by Cr3+(3d3).This
neutral center condition in another nomenclature system might be termed
Cro or NOT. If one electron is trapped from the conduction band, or picked up
from the valence band by optical pumping, the electron goes into the 3d shell
92 A. G. MILNES
and the level may henceforth be described as Cr2+(3d4)or N ; . The first

excited state of this level is believed to be a few tens of meV into the r con-
duction band (Eaves et al., 1981 ). On the possibility that chromium might
accept a second electron to become Cr' +(3d5), there is convincing evidence
that this level is 1 15 meV into the r conduction band at 300 K and therefore
can be seen only if the conduction-band edge is raised by the application of
pressure (Hennel et al., 1980, 1981a,b; Hennel and Martinez, 1982). Con-
ceivably, Cr3+(3d)could exhibit electron-donor action in GaAs that would
result in it becoming Cr4+(3d2),but there is no convincing evidence that
this happens.
Chromium in GaAs, when excited optically (say with an argon laser),
exhibits an luminescence band at about 0.8 eV with a zero phonon line at
about 0.839 eV. This is luminescence from inside the Cr atom involving
relaxation in the 5E-5T2 crystal field scheme. The 5T2 is the ground state of
the Cr that has trapped an electron and become Cr2+(3d4).The crystal-field-
level structure has the form shown in Fig. 12, according to Vallin et al. (1970),
Williams et al. (1981), and Clerjaud et al. (1980). See also Bishop (1981),
FIG. 12. Energy levels of Cr2+(3d4)as predicted by crystal-field theory for an undistorted,
substitutional cation site of Td symmetry. The ground state is the P, of 6T, (k is small, about
0.060 meV for GaAs). (After Vallin er al., 1970; reproduced with permission of the American
Physical Society.)
Grippius and Ushakov (1981),Glinchuk et al. (1977),Klein and Weiser (1981),

Krebs et al. (1981), and Krebs and Strauss (1977).
The luminescence at about 0.8 eV was thought at one time to involve a
Cr-conduction- or valence-band transition. Now that it has been shown to
be a Cr internal transition, it is of less importance with respect to carrier
trapping activities. The structure of the luminescence is now almost eluci-
dated (White, 1979; Picoli et al., 1980,1981; Eaves, 1980; Eaves et al., 1981b;
Williams et al., 1981; Deveaud and Martinez, 1981. The 0.838 eV Line is
strong in emission but weak on absorption and exhibits (111) axial sym-
metry. Eaves considers it to be a complex (Crii -X) where X is an unidentified
impurity (oxygen is not excluded). There is also a strong 0.82 eV absorption
line that is considered to represent the 5T2+5E absorption of simple
substitutional chromium, CrGa(Clerjaud et al., 1980).
For practical purposes, as opposed to detailed physical study, the amount
of Cr in GaAs can be estimated by gross comparison of the optical absorption
in the region between about 0.8 and 1.2 eV for various slices as shown in
Fig. 13. The absorption is associated with electron transfer from the valence
band to the Cr3+(3d3)state, causing it to become Cr2(3d4).Absorption with a
threshold around 0.67 eV may involve further pumping of the CrZ+state, and
the electron is handed on to the conduction band. The shape of the absorp-
tion band (either at 300 or 4.2 K) does not allow accurate determination of
the energy transition (at 4.2 K it is roughly 0.85 eV for the transition Cr3+ +
evB+ Cr2+)since the Lucovsky model does not fit, and the temperature
effects suggest that lattice relaxation or electron-phonon interactions may
be taking place. The segregation coefficient for Cr inferred from these
absorption measurements is 8.9 x
Energy (eV)
FIG. 13. Absorption coefficient at room temperature as a function of the energy of the
photons for four different samples: undoped (sample A) or doped with Cr (samples C, E, and 1).
(After Martin er al., 1979; reproduced with permission of the American Physical Society.)
94 A. G. MILNES
More detailed absorption curves for 300, 77, and 4.2 K are found in the
studies of Martinez et al. (1981), Hennel et al. (1981b), and Martin et al.
(1981b). The latter group also examine the variation of EL2 throughout
ingots and conclude that it must be a complex defect involving lattice defects
growing in the presence of stress. Others take the view that EL2 is the simple
antisite defect AsGa.Yet others consider it to be due to complexes involving
AsGa(Li et al., 1982) or AsGaVAs (Lagowski et al., 1982) formed by As move-
ment into Ga vacancies or divacancies. Martin and Makram-Ebeid ( 1 982)
consider the outward diffusion of EL2 level at temperatures above 600°C.
They consider that EL3 or HL9 may be the AsGadefect.
Photoconductivity in epitaxial GaAs: Cr shows evidence of optical
threshholds between 0.75 and 0.79 eV for low and high illumination levels. A
broad level of trap distribution (225 meV) may be involved, and an electron-
capture cross section of 2 x ~ m - which
~ , is unexpectedly large,
results from the modeling. The Lucovsky model is not a good fit. Better
success is had with a photoionization theory that assumes a Is)-ls) transition
involving a tight orbit with thermal broadening described by a single phonon
energy of 0.03 eV and a Huang-Rhys factor of 3, and taking into account
nonparabolicity in the band by k.p. (Kronig-Penney) theory. The enthalpy
of the 0.75 eV level is therefore 0.66 eV (Amato et al., 1980). Some departure
from the Lucovsky model is also found by Vasudev and Bube (1978). See also
Monch et al. (1981) and Vaitkus et al. (1981).
In other studies of photoconductivity and photo-Hall spectra, Look
(1977)gives the energy-band diagram of Fig. 14. In more recent studies, Look
shows that mixed conductivity effects that complicate the interpretation of
Hall and resistance measurements in very high resistance material are not
Cr-SE(rxcit)
t CB
FIG. 14. A proposed room-temperature energy diagram for GaAs: Cr (CB, conduction
band; VB, valence band). (After Look, 1977a; reproduced with permission of the American
Physical Society.)
important if the resistivity is below 5 x lo8 R-cm. From analyses of many

Cr-doped and undoped semi-insulating specimens, Look ( 1980) concludes
that E, - E,, is 0.64 eV (if go/gl = 4/5) and that another dominant level is
at E, - 0.59 eV for go/gl = 1/2 (which he terms E, without necessarily
implying that it is oxygen related). Look finds that for Cr-doped crystals, the
Fermi levels at room temperature are slightly lower than for undoped
(or 0 doped) semi-insulating crystals.
Other important studies have been made of the properties of semi-
insulating GaAs. We have, for instance, the models of Zucca (1977),Lindquist
(1977),and Mullin et a/. (1977). The Mullin group comments that transport
studies have until recently led to a consensus view of Cr-doped GaAs that
interprets the activation energy for conduction as due to a Cr level at
-0.70 eV from the conduction band. They declare that this is a misleading
interpretation and go on to show that their material exhibits two dominant
levels, termed Em and E,, at 0.40 and 0.98 eV from the conduction band,
respectively. The Cr concentrations (10" cm- 3 , exceed the measured values
of E, ( 1014-1015cm-3), although Cr was considered clearly responsible for
forming the species E,. They suggest that the E, level may be related to a
Cr-0 complex.
Martin et al. (1980b) have presented one of the most recent studies of
compensation mechanisms in GaAs. Hall measurements made for their
Cr-doped (6 x 10" - 4 x 10'' ~ m - and ~ ) undoped semi-insulating ma-
terials are given in Tables IX and X. The variations between different speci-
mens are quite large, the Hall-effect mobilities need interpretation, and the
TABLE IX
RESULTSFROM HALL-EFFECT
MEASUREMENTS A N D OPTICAL ABSORPTION
MEASUREMENT
ON SEMI-INSULATING
MATERIALS DOPEDWITH C r CONCENTRATION RANGING
BETWEEN l o i 6 A N D l o i 7 c m - j
Optical Hall-effect data

absorption
data, (yRH)-' Slope of
concn. of at 400 K p,, at 400 K In(R,T3'*)-' Concn. of Si
Material Cr (cm-j) (cm-j) (cm' V-l sec-I) (eV) (cm-
DI Bridgman 3.5 x lo'* 1.5 x lo1' I600 0.762 0

D2 LEC 3.5 x IOl* 7.5 x 10' I800 0.762 0
D3 Bridgman 5 x 10l6 1.2 x 10" 2800 0.700 A few x 10"
D4Bridgman 3 x 10l6 7.0 x 10' 3200 0.757 (1-2) x lo1*
D5 LEC 9 x lo'* 1.6 x 10" 760 0.716 0
D6 LEC 9 x 8 x 10' I200 0.776 0
" After Martin el (11. (1980a). Reproduced with permission of the American Physical Society.
96 A. G . MILNES
TABLE X
MEASUREMENTS
RESULTSON HALL-EFFECT CARRIED OUTON SEMI-INSULATING GaAs
NOTDOPED
WITH C r
A1 Bridgman 1.1 x 10'1 4500 0.757 3 x loi6 0.539

A2 Bridgman 1.7 x 10" 4500 0.729 3 x 10l6 0.604
A3 Bridgman 1.3 x 10" 5500 0.750 4 x 10l6 0.533
A4 Bridgman 2.5 x 1013 4300 0.350' 5 x 10l6 0.352
A5 Bridgman 1.5 x 1013 4300 0.422' 5 x 10l6 0.370
A6 Bridgman 2.8 x 1013 2500 0.392' 5 x 1017 0.348
A7 LEC 1.6 x 10" 2800 0.60W 4 x 101~ 0.526
A8 Bridgman 8 x IO" I700 0.164' 0.312
A9 LEC 4 x 109 I850 0.728' 0.653
A1 LEC 1.1 x 1010 2600 0.762d 0.619
After Martin er 01. (1980b). Reproduced with permission of the American Physical Society.
' Deduced from p,,.
Slopeof In(R,T3/4)-1 = f ( l / T ) .
Slope of ~ I ( R , T ~ ' ~=) -l(l/n.
'
Fermi levels are quite variable. From an analysis of the Cr-doped specimens,
two dominant deep levels are found, namely, the Cr acceptor at E, 0.73 eV +
and a deep donor (termed EL2) at E, - 0.67 eV (400 K). As seen in Fig. 15,
their deep donor lies below the deep acceptor, in reasonable agreement with
Fig. 14. The EL2 donor is considered to be dominant in creating high
resistance undoped (or very lightly Cr-doped) material. The shallow doping
concentration has some effect on the Fermi-level position as shown in Fig. 16.
If the electron and hole concentrations in GaAs at 400 K are to be equal, then
the Fermi level will be at E, - 0.637 eV.
For very high resistivity GaAs with both electrons and holes contributing
to mixed conduction, the apparent Hall mobility is strongly dependent on the
precise doping conditions. Figure 17 shows calculated values expected if the
level EL2 is supposed not to be present. On the other hand, Fig. 18 shows the
curves if EL2 is present in a concentration 10l6cm-3. The lines are seen to be
in general agreement with experimental results for some materials from
Tables IX and X.
In optical DLTS studies of Cr-doped GaAs, Jesper el a/. (1980)find a level
with an electron activation energy of E, - ET = 0.72 eV. The photon cross
section of this increases with temperature, but not in a way that conforms to
1.375 a\
0.73130 rv
FIG. 15. Scheme of the band gap with different important deep levels, at 400 K: A, con-
duction band; B, shallow donor ND(Si); C, deep acceptor NAA(Cr);D, deep donor N,,(EL);
E, valence band. (After Martin er a/., 1980; reproduced with permission of the American
Physical Society.)
=02 T= 400 K
bond band
FERMI-LEVEL POSITION ( e V )
FIG. 16. Shockley diagram corresponding to a semi-insulating GaAs with a given concen-
tration of the deep Cr acceptor and the deep donor EL2, with or without shallow donors. (After
Martin e r a / . , 1980; reproduced by permission of the American Physical Society.)
98 A. G . MILNES
5000
J
J
a
I
1000 -
CONCENTRATION OF CHROMIUM
FIG. 17. Expected variation of the observed Hall mobility in semi-insulating GaAs, as a
function of the Cr concentration, for different values of N, - N, (the concentrations of shallow
donors and acceptors) assuming that the concentration of the deep donor EL2 is zero; T =
400 K. (After Martin el al., 1980;reproduced with permission of the American Physical Society.)
multiphonon effects. The level may be detected as either an electron trap or a

hole trap, depending on its charge state and the nature of the experiment. The
electron-capture cross section is about 2 x cm2 and is not temperature
sensitive. The authors conclude that the DLTS signal in n-type semi-
insulating material is dominated by hole emission [i.e., by electron pickup
fromthevalence bandasCr3+(3d3)+Cr2+(3d4)]andopis -3 x lo-’’ cm2.
The values found in the work of the last decade for the acceptor level of Cr
tend to range from 0.7 to 0.9 eV. This scatter is a measure of the difficulty of
interpretation of results as a function of temperature when additional deep
levels such as EL2 are present. In theory, the Fermi function from Martin
et a / . (1980a) is
where
Y I \\
-NA
levels)
1016 1017
CONCENTRATION OF CHROMIUM
1018
-
FIG. 18. Expected variation (full lines) of observed Hall mobility in semi-insulating GaAs,
as a function of the concentration of Cr, for different values of N, - N, (the concentration
of shallow donors and acceptors), and for a given concentration of the deep donor EL2 equal
to 10l6 cm-’; T = 400 K. Points correspond to experimental results on different materials:
( 0 )Bridgman grown under 0, overpressure; ( 0 )LEC; ( A ) Bridgman, not intentionally
doped with 0,; (0) Bridgman, doped with Si. (After Martin et d.,1980; reproduced with
permission of the American Physical Society.)
is the free energy of ionization of the level, AH its enthalpy, AS its entropy;
go is the degeneracy of the level not occupied by an electron and g , the
degeneracy of the level occupied by an electron. The detailed electrical
characterization of a level allows the determination, as a function of
temperature, of the following terms:
E’ = ET - kT ln(g,/g,)
E, - (16)
E’ = AH - T AS - kT ln(g,/g,) (17)
The band gap for GaAs as a function of temperature is
EG = 1.519 - r(5.4 x 10-4)TZ/(T+ 204)] eV (18)
and from hole-emission and capture-rate data, Martin et al. (1980a) conclude
that for T > 50 K, the Cr level above E, is possibly represented by
E’(Cr) = 0.81 - [ ( 3 x 10-4)T2/(T+ 204) + kT In 0.93 eV
100 A . G. MILNES
At 300 K, this yields

E'(Cr) = ET - E, = 0.81 - 0.05357 + 0.0018 = 0.805eV (19)
and at 400 K, Eq. (19) yields 0.73 eV. The variation with temperature is seen
to be quite significant and the level cannot be regarded as pinned to either
band edge.
Detailed DLTS and ODLTS (optical deep level transient spectroscopy)
studies of the Cr3+ and Cr2+ levels in GaAs have been made by Martin et al.
1980a. The electron-capture cross section of the chromium site for the
process Cr3+ + e + Cr2+ is 4 x cm2. The hole-capture cross section
for the process Cr2+ + h + Cr3+ is much larger, perhaps 2.5 x 10- cm2.
From comparisons of trap densities and chemical analyses for Cr, it is
concluded that most of the Cr atoms (within a factor of 2) are electrically
active. Recombination and trapping has also been considered by Li and
Huang (1972).
Field-effect transistors may be made by the creation of n channels by
ion implantation into Cr-doped semi-insulating expitaxial layers and
subsequent annealing. The redistribution of Cr into the implanted region is
significant. In VPE and MOCVD (metalloorganic chemical vapor deposition)
layers, the out-diffusion of Cr into the implanted region has been repre-
sented by D = 4.3 x lo3 exp( -3.4/kT). At high Cr concentrations, the
behavior is more complicated. Semi-insulating GaAs can also be produced
by MBE (Morkoc and Cho, 1979).
PRODUCED BY TRANSITION
VI. EFFECTS METALS
Transition metals produce many interesting effects in GaAs. We have

just reviewed the role of Cr in producing semi-insulating material; Fe also
gives a fairly deep acceptor (E, + 0.5 eV) and so can contribute to producing
fairly high resistance material ( - lo5 Q-cm, 300 K).
TABLE XI
MULTIPLE
CHARGE
STATFSOF TRANSITION-METAL I N 111-V SEMICONDUCTORS'
ACCEPTORS
State 3d2 3d3 3d4 3dS 3d6 3d7 3d8 3d9
A' V3+ Cr3+ Mn3+ Fe3+ co3+ Ni3+ Cu3+

A- V2+ Cr2+ Mn2+ Fez+ Co2+ Ni2+ Cu2+
A?- Cr + Fe+ Ni+
S I 312 2 512 0, 1/2, 1 312 0 112
a Kaufmann and Schneider (1980).

The core charge states of transition metals on a gallium site are shown in
Table XI. Although A’- states are shown, it does not follow that they occur
or have been identified in GaAs. Following Kaufmann and Schneider (1 980),
we can say that:
Paramagnetic resonance studies of the Fe acceptor in GaAs have shown that it resides
on the Ga sublattice and has a deformed state of a 3d wave function for a free atom. Similar
observations have been made for V, Ni, Cr and Co. This is also the case for Mn in ingot
material, though in Mn doped LPE GaAs the ESR spectrum shows reduced symmetry
due to the formation of (what is probably) a Mnca-VAs complex. The E, + 0.15 eV Cu level
+
is believed to correspond to Cu on the Ga lattice, with another level at E, 0.45 eV corre-
sponding to Cuz-. However, recent examination of the 0.15 eV defect suggests that its
symmetry may be lower than a simple Cu,, assignment, possibly due to complexing with
another defect. Thus, it appears that the Fe group transition elements (Sc-Cu) in GaAs
generally act as deep acceptors, and occupy sites on the Ga sublattice.
A. Energy Levels Considered in Relation to Atomic Size and Strain

The nonionized core state of Fe is Fe3+ (3d5), and on accepting an electron
from the valence band (equivalent to emitting a hole to the valence band), it
becomes Fe2+(3d6).The situation is thus similar to that of an isoelectronic
impurity, except that the inner 3d shell is incomplete and a hole may be
thought of as localized in the atomic core.
For isoelectronic impurities it is known that the energy of binding a hole
is determined by the deformation of the lattice about this impurity (J. W.
Allen, 1968). Thus an impurity atom of radius R,, which replaces a lattice
atom whose radius is 6 R larger, forms a field of elastic stresses around the
impurity, and the activation energy of the hole is given by
6E = f i d R & 6 R / a 3 = KR& 6 R
Here, a is the distance between nearest neighbors in an undeformed crystal,
d is the deformation potential, and K is a proportionality constant given by
K =fid/a3
The anisotropy of the activation energy can be neglected to a good
approximation. It has been shown for substitutional isoelectronic impurities
that displacements of the nearest neighbors result in a redistribution of the
valence charge such that the activation energy of a hole is reduced to near
zero. The residual activation energy, of the order of 0.01 eV, is determined by
the Born-Mayer forces between the impurity atoms and the nearest
neighbors.
The covalent, tetrahedral radii of various elements is a function of their
position in the periodic table. The radii depend on the number of electrons in
the 3d shell, and data can be taken from compounds in which the transition
metal ion has a formal charge of +2. If these metal ions had filled atomic
102 A. G . MILNES
core shells like Ca2+ and Zn2+,they would presumably act as isoelectronic
traps with near-zero activation energy.
It is interesting (although rather crude in a fundamental sense) to follow
up this line of thought by postulating that the binding energy of iron-group
transition elements in GaAs may be given by Eq. (20) with 6R equal to the
difference between the metal atomic radius R, and the extrapolated radius of
the atom with a filled (3d'O) shell, R,. This was first done by Bazhenov and
Solov'ev (1972) and more recently by Partin et al. (1979e), with the results
shown in Fig. 19.
Several factors complicate the task of identification of the Ga-site single-
acceptor level. In most cases EPR data are not available to establish the
symmetry and absence of complexing, so the modeling must be taken as very
tentative. The level predicted for Mn (0.36 eV) deserves special consideration
since it differs significantly from the experimental value of 0.09 eV. This
difference can perhaps be understood from the fact that Mn has a d shell
which is exactly half full (3d5)and thus, like Zn (3d"), is stable with respect
to transfer of a 3d electron to a higher energy crystal bonding orbital. Thus,
Mn acts almost like a normal acceptor, except that its radius differs signif-
icantly more from that of a Ga atom than does the Zn radius. There has
+
been speculation that a hole-trap level at E, 0.33 or 0.37 eV observed in
VPE material may be related to Mn (Mitonneau et al., 1980); however, the
weight of the experimental evidence supports E, + 0.09 eV as the first Mn
Ec
ELEMENT
FIG.19. Calculated (0) versus experimental (+) energy levels in GaAs. (The Co experiment
value should be 0.16 eV.) (After Partin er al., 1979; reproduced with permission of Pergamon
Press, Ltd., UK.)
acceptor (Mn”). It may also be remarked that since the study was made, it
has become almost certain that the CoGalevel is at 0.16 eV and not the value
of 0.56 eV also shown.
Whether or not the undulations in the experimental values of Fig. 19 are
adequately explained by the 6R model, one thing that is apparent from the
diagram is the definite downward trend of ET - E, as the transition elements
are changed from Ti to Cu. The trends of energy levels associated with
transition elements in GaAs have been considered by Il’in and Masterov
(1977) using a Green’s function method within the framework of a semi-
empirical description of the impurity Hamiltonian. The downward trend of
energy as one proceeds from Ti to Cu results from the study. The problem
has also been considered by Jaros (1971, 1982) in a wave-function study in
which screened long-range Coulomb potentials and short-range core
potentials are taken into account. Understanding of chemical trends of deep
energy levels in semiconductors has been greatly increased in a semiquantita-
tive sense by the Hjalmarson theory (Hjalmarson et al., 1980), involving
notions of hyperdeep and deep levels and related to neglecting the long-range
potential of the impurity and simply inputting the band structure of the
host, the positions of the atoms, and a table of atomic energies (Sankey
et al., 1981).
B. More on the Transition Elements

A few further comments on the behavior of transition elements from Ti to
Ni in GaAs follow. Titanium in GaAs produces photoluminescence peaks at
0.525 and 1.3 eV (Kornilov et al., 1974). From Hall measurements Bek-
muratov and Murygin (1973)assign a value E , - 0.45 eV to Ti as an acceptor.
This is in general agreement with other work, such as Gutkin et al. (1972,
1974), from which the level is E, + 0.98 eV.
Photoluminescence and photoconducting studies of V in GaAs suggest
an energy of 0.8 eV for transfer of electrons from the valence band to the V3+
neutral state. There is also luminescence at 0.17 eV and absorption at 1.0-
1.2 eV (probably involving intracenter transitions) (Vasil’ev et al. (1976).
Martin (1980) identifies the EL19 level as the electron trap associated with
the presence of vanadium. The ESR spectrum of the neutral acceptor state
V3+(3d2)has been discussed by Kaufmann and Schneider (1980).
Manganese in GaAs exhibits a characteristic photoluminescence peak at
1.405 eV, arising from the process of electron capture from the conduction
band to a neutral Mn level, A’ + e - + A - + hv [or, in another nomen-
clature, Mn3+(3d4)+ eCB+ Mn2+(3d5)+ hv]. This photoluminescence is
not an intracenter transition within the 3d shell as exists for most other
transition elements in GaAs (and as exists for Mn in Gap). There is also an
104 A. G . MILNES
infrared absorption band starting at about 0.11 eV and peaking at about

twice the energy, that is the photoionization complement, namely,
Mn3+(3d4)+ eVB+ Mn2+(3d5),of the 1.405 eV luminescence. The curve
which can be fitted by Lucovsky’s model is shown in Fig. 20.
From such behavior, Mn has been found to be a common impurity in
VPE-grown GaAs. Possibly it comes from the stainless-steel components of
growth systems. The solubility of Mn in GaAs is large ( > 10’’ ~ m - ~and ),
the diffusion coefficient is hn= 6.5 x lo-’ exp(-2.49/kT) cm2/sec ac-
cording to Seltzer (1965).Manganese is sometimes implicated in the thermal
conversion to p type of GaAs surfaces (Klein et al., 1980).
Consider now Fe in GaAs. The solubility has a maximum active concen-
tration of 1.3 x 1OI8cm-3 at 1100°Cand is strongly retrograde, as shown in
Fig. 21a. The result of a typical diffusion at this temperature is shown in
Fig. 21b. Boltaks et al. (1975) conclude that the Fe atoms diffuse through
both interstices and the vacant sites; the interstitial atoms diffuse far more
rapidly than the substitutional atoms. The interstitial atoms are transferred
to the vacant sites, and vice versa; transition from interstices to lattice sites
remains predominant until the Fe-atom concentration in the sites and
interstices attains its equilibrium value at the given temperature.
The diffusion coefficients of the bulk sector were determined by
comparing the experimental curves with the theoretical curves of In N vs. x.
The temperature dependence of the diffusion coefficient of Fe in GaAs is
represented by the exponential function D = 4.2 x l o p 2 exp{(- 1.8 f
t -
--
I l l I I I
N
I
120 - 5
-5
I-
- 3 9
‘0
I
z
w
2
I
I.
W
0
u
z
0
I-
n.
E
sm
a
FIG. 20. Infrared absorption associated with the ionization of the neutral manganese
acceptor in GaAs. (After Chapman and Hutchinson, 1967; reproduced with permission of the
American Institute of Physics.)
O.l)/(kT)} as shown in Fig. 21c. Other studies show the diffusion coefficient
to be very dependent on the conditions of As overpressure (Uskov and
Sorvina, 1974).
The activation-energy level for the transition Fe3+(3d5) Fe2+(3d6) --+
that takes an electron from the valence band (i.e., emits a hole) is about
E, + 0.50 eV. The maximum resistivity that can be created by Fe in GaAs
tends to be about lo5 Q-cm at room temperature, so Fe cannot be considered
a useful dopant as an alternative to Cr in creating truly semi-insulating GaAs
(Ganapol’skii et al., 1974). However, Fe has been found by one laboratory to
be the predominant deep acceptor in LEC GaAs grown in either SiOz or
PBN crucibles, the concentration being 6 x 1015 cm-3 (Wilson, 1981).
Intracenter electron transitions in Fe give photoluminescence in the range
0.34-0.37 eV (Bykovskii et al., 1975, Fistal et al., 1974). Excited states of Fe
in GaAs have been studied by Demidov (1977).
- I03/T (K-’)
(a)
-?1ot
1 (“C)
1160 1100 1000 900 830
I
(00
1
200
X (pm)
(b)
300
-”
10-9
10-10
0.7 0.8 0.9
10 ’/ T ( K - ’ )
(C 1
FIG. 21. Behavior of Fe in GaAs: (a) solubility versus temperature; (b) diffusion profile
of Fe in GaAs: T = 1 loPC, z = 24 x lo4 sec; (c) diffusion coefficient versus temperature for the
bulk region. (After Boltaks et al., 1975.)
106 A. G . MILNES
In LPE-grown layers of GaAs doped with -10l6 Fe cm-3 (2.7 x lo3

R-cm), photoionization cross-section studies at 98 K (Kitahara et al., 1976)
show a photoionization threshold Eiof 0.5 eV and a reasonable fit to the
Lucovsky theory which gives the photoionization model as
o(hv) = C(Ei)”2(hV - Ei)3’2/(hV)3 (22)
A lower threshold at 0.37 eV and a higher one around 0.80 eV also were
present in this material. Lebedev et al. (1976) and other workers also have
seen levels at around E, + 0.38 for Fe-doped GaAs. Whether this level is
directly Fe related is uncertain. Other elements such as Cu and probably Ni
cause a level to appear at or near this position in the band gap, and there is
also, of course, the E, + 0.41 eV defect level seen in LPE material that is
still due to unknown causes.
The main energy level of Fe has been found to vary with temperature as
EFe - E, = 0.52 - (4.5T/104). (23)

This temperature dependence is somewhat similar to that of the band gap of
GaAs (Haisty, 1965; Omel’yanovskii et al., 1970).
There has been a report (Schlachetzki and Solov, 1975) that in LPE, Fe
may give rise to n-doped conductivity; however, another study (Hasegawa
et al., 1977)finds only the usual acceptor at E, +
0.52 eV with a hole-capture
cross-section of about cm2. Ikoma et al. (1981) find that the E, +
0.51 eV hole trap appears in LPE GaAs only after electron-beam irradiation
and annealing above 200°C in Fe-doped wafers. This suggests that Ga
vacancies are necessary for Fe to form a deep state.
For VPE-grown FETs it has been reported that the use of an Fe-doped
lo5 R-cm buffer layer on the Cr-doped substrate has a beneficial effect on
the active-channel-region mobility and therefore on the transconductance
gain gm (Nakai et al., 1977). Iron doping of bipolar npn GaAs transistors has
some beneficial effect apparently on the trapping actions in the base region
that improves the frequency response (Strack, 1966).
Cobalt in GaAs, according to Haisty and Cronin (1964), has an energy
level at E, + 0.155 eV. Photoionization studies by Brown and Blakemore
(1972)and Andrianov et al. (1977a) showed a steep rise in response at 0.16 eV
and so confirmed the previous work. On the other hand, tunnel-junction
spectroscopy studies of Fistul’ and Agaev (1966) showed some action at
0.54 eV.
+
The E, 0.16 eV level is supported by photoluminescence seen at
1.30 eV (Ennen et al., 1980), which is considered to result from electron
+
capture from the conduction band according to A0(3d6) ecB+ A-(3d7).
Ennen et al. also observed optical absorption at 0.501 eV (the zero-phonon
line) and assign this to the intra-d-shell transition 4A2 + 4T2(F) of the A -
negatively charged acceptor state Co2+(3d7).Kornilov et al. (1974) observed
photoluminescence at 1.29 and 0.49 eV.
For diffusion of Co at 1000°C at atmospheric pressure of As, the surface
concentration is about 10" cm-3 and the penetration depth is about 20 pm
after 24 hr. The effective diffusion coefficient for No = N, erfc[~/2(DT)"~]
is D 2 x lo-" cm2/sec for the above conditions. The effects of 6oCo
N
irradiation have been studied by Share (1975).

Consider now the energy level(s) created by nickel in gallium arsenide.
Nickel is likely to occur in VPE-grown GaAs as a result of contamination
from the growth system when HCl is used. Nickel is also a component of a
standard ohmic-contact recipe (Au-Ge, Ni) for n-type GaAs. Nickel is a
medium-fast diffusing species in materials with Ga vacancies, and so it may
enter the crystal during contact formation. This is of significance because Ni
can have an adverse effect on minority-carrier hole lifetime in II GaAs.
At various times, energy levels at 0.08, 0.15, 0.21, 0.35, 0.42, and 0.53 eV
have been attributed to Ni in GaAs (Damestani and Forbes, 1981). Neutral
nickel on a gallium site can be represented by Ni3+(3d7)and after accepting
an electron by Ni2+(3d8)(Andrianov et al., 1977a; Murygin and Rubin,
1970; Fistul' and Agaev 1966). Certain workers favor E, + 0.20 eV as the
energy level for the Ni(3d7) neutral acceptor (Bimberg et al., 1981). Other
groups believe that nickel substituting on a gallium site has an acceptor
energy E, + 0.42 eV. For instance, optical absorption in GaAs:Ni has been
studied by Bazhenev et al. (1974). The absorption spectra exhibited maxima
at 0.84 and 1.14 eV and the absorption coefficient c( in the 0.8 eV region was
proportional to the amount of Ni in the crystal (determined chemically).
This led Bazhenev et al. (1974) to attribute this to the acceptor photoionizing
from the valence band and, following the Lucovsky model, the value of
+
0.42 eV (0.84/2) is assigned for the threshold process, Ni3+ eVB+ Ni2+,
with an activation energy E, + 0.42 eV. This is in agreement with the Hall
measurements of Matveenko et al. (1969). On the other hand, from attempts
to dope LPE GaAs with Ni in the Ga melt, Kumar and Ledebo (1981)
conclude that the E, + 0.40 eV level observed is caused by trace Cu
contamination and that no level specific to Ni is seen.
Optical absorption attributed to the process A' + A- plus hole photo-
ionization, with a Hall activation energy of 0.35 eV has been observed by
Suchkova et al. (1975, 1977). Suchkova and co-workers find an absorption
band at 1.15 eV which they attribute to electron transition within the d8 shell
of the compensated Ni2+ ions.
Ennen et al. (1980) find a strong zero-phonon line at 0.572 eV in
GaAs: Te: Ni, which they tentatively assign to an intra-d-shell transition of
the two-electron-trap state A2- of Ni+(3d9).Kaufmann and Ennen (1981)
108 A. G . MILNES
also report sharply structural bands of absorption and emission in the 2-pm
region in nickel-diffused n-GaAs that they assign to intra-d-shell transitions
of Ni acceptor-shallow donor near-neighbor associates of Ni-S, Ni-Se,
Ni-Te, Ni-Si, Ni-Ge, and Ni-Sn.
Deep-level transient-spectroscopy studies of VPE GaAs :Ni (Partin
et al., 1979a) show an energy level acting as a hole trap at E, + 0.39 eV
produced by Ni diffusion. The uncorrected energy value was 0.48 eV, but
the temperature dependence of cross section leads to the E(T = 0) value of
0.39 eV. An electron trap was also found at E, + 0.39 eV after Ni diffusion,
and the concentration was about that of the E, + 0.39 eV level. The physio-
chemical origin is unknown. No evidence was found for a level at E, + 0.2 eV
either before or after the Ni diffusion. For the E, + 0.39 eV level, the hole-
capture cross section was about cmz at 207 K. The electron-capture
cross section, however, was small, about 5 x lo-" cm2, and so the center at
E, + 0.39 eV was not an important hole-recombination site. (These capture
cross sections agree with those that Ledebo attributes to Cu.) On the other
hand, the E, - 0.39 eV level appears to act as a strong minority- (hole)
carrier recombination center, as discussed in Section XIII.
VII. GROUPI IMPURITIES:

Li, Cu, Ag, Au
A. Lithium
Although studies of Li in GaAs have been informative in the past, there
appears to have been little work with Li during the last decade (Grimm, 1972;
Norris and Narayanan, 1977). The following summarizes briefly the under-
standing of Li in GaAs reached in earlier work. The solubility of lithium at
800°C is 1.6 x l O I 9 cm-3 in undoped GaAs and about 2.3 x 1019 cm-3 in
GaAs doped heavily with Te or Zn. Lithium diffuses rapidly into GaAs with
an interstitial diffusion coefficient D = 0.53 exp( - l.O/kT), where kT is in
electron volts (Fuller and Wolfstirn, 1962).
When nominally pure crystals of GaAs are saturated with lithium by
diffusion at temperatures greater than 500°C and cooled to room temper-
ature, they are iompensated to a high resistivity. The compensation
phenomenon is caused by the action of Li' which permeates the crystal and
tends to form donor acceptor complexes.
The presence of the Li' interstitials may be expected to increase the
gallium vacancy concentration by orders of magnitude. The possible reac-
tions that follow result in the species Li'V-, Li-, Liz-, Li'Li2-, and
(Li')zLi2-, where Li' is interstitial and Li- and Liz- are substituted ions.
On cooling to room temperature after a high-temperature lithium diffusion,
excess Li' readily precipitates because of its high mobility. This leaves
relatively immobile excess acceptor complexes which give rise to the ob-
served p-type compensation. The activation energy observed is 0.023 eV from
the valence-band edge. Lithium is found to compensate n-type GaAs that
has tellurium concentrations as high as 5 x lo'* ~ m - ~ .
Chemical interaction among defects can result in ion-pair formation in
semiconductors. The evidence for pairing is mostly indirect; however, when
ion pairs have an IR-active local vibrational mode, IR spectroscopy provides
a sensitive tool for studying the pairing reactions and pair structure. Pairing is
recognized by the reduction of site symmetry and the consequent lifting of
degeneracies of vibrational modes. Local mode frequencies of unpaired
impurities must be known in order to recognize site symmetry reduction. If
more than one stable isotope is available for one of the impurity species, the
change in the absorption-band frequency due to changing the isotope can be
used to study impurities which are heavier than host lattice atoms, provided
they are paired with a light atom.
When the behavior of electrically active impurities is studied in GaAs by
local-mode IR spectroscopy, the free-carrier concentration introduced by
the impurities must be reduced by compensation with another electrically
active defect. This reduces the nonlocalized modes of vibration contributions
to the absorption. The behavior of Li has made it a good choice for the com-
pensating impurity: it is a rapid diffuser, and two stable isotopes are readily
available. Usually, it is the local mode of the Li paired with the original
dopant which is experimentally observed. The bandwidth range of interest is
'.
typically 300-460 cm- In GaAs, the pair systems studied include Te-Li,
Mg-Li, Cd-Li, Zn-Li, Mn-Li, and Si-Li (Lorimor and Spitzer, 1966;
Allred et al., 1968; Spitzer and Allred, 1968; Leung et al., 1972). In addition to
the local modes observed in GaAs doped during growth, absorption bands
have also been reported in pure GaAs into which Li has been diffused; the
corresponding absorption centers are believed to be Li complexed with
native defects.
Other impurities in the 1A column such as Na and K have apparently
received no further study since the tentative report of Na as an acceptor
(Hilsum and Rose-Innes, 1961). Potassium however has been found to be
very adverse as a surface-eroding contaminant in early GaAs devices
(Leedy et al., 1972).
B. Copper
Copper is a fast interstitial diffuser in GaAs and converts to a substi-
tutional site CuGa that is capable of acting as a double acceptor. The
interstitial diffusion coefficient D' is 1.0 x cm2 sec-' at 500°C and the
110 A. G . MILNES
diffusion activation energy is 0.53 eV. The solubility of copper in GaAs

depends on the other impurities present, but may be in excess of lo'* cm-3.
Surface and bulk diffusion has been studied by Boltaks et a!. (1971) who find
surface diffusion coefficients of the A and B faces to be unequal and therefore
the bulk impurity profiles near these faces to be different. See also Hasegawa
(1974).
Deep-level determinations up to 1970 suggest that the main energy
levels associated with copper in GaAs are 0.14-0.15 eV above E,, pre-
sumably the first acceptor level (Cu-), and 0.44 eV, presumably the second
acceptor level Cuz-. A level seen at E, + 0.24 eV was thought to be related
to a pair of copper atoms. A level at E, + 0.19 eV was attributed to a Cu-Te
complex, and miscellaneous levels at other depths were thought to be related
in some way to Cu. The photoluminescent studies showed a broad band at
1.35 eV, which was tentatively ascribed to VAsCuGa pairs or DAsCuGa pairs
where VA, and DA, denote, respectively, an arsenic vacancy and a donor
impurity occupying an arsenic site.
Since 1970, there have been numerous investigations of the behavior of
Cu in GaAs that have added to previous knowledge. Norris (1979) finds that
in melt-grown GaAs: Cu, a broad cathodoluminescent band centered at
1.36 eV shows injection-level and temperature characteristics that are
consistent with conduction-band-acceptor transitions. Norris (1979)
comments that if we follow
the suggestion that the peak at 1.36 eV represents a nophonon transition, then thermal
quenching of the 1.36 eV band should proceed with an activation energy of approximately
0.14 eV since the band gap of GaAs is close to 1 S O eV at 80 K. Indeed, transport measure-
ments have shown that Cu in GaAs introduces an acceptor level with a depth of 0.14-
0.15 eV. More importantly, a previous study of the temperature dependence of the 1.36 eV
luminescence in GaAs:Cu showed quenching with an activation energy of 0.147 eV, thus
making it plausible that the 1.36 eV transition terminates on the same acceptor levels
identified in the transport measurements.
Norris states, however, that frequency-response measurements conflict

with this attribution and indicate that the kinetics of the 1.36eV luminescence
are distinctly nonexponential. In heavily Cu-contaminated material the
kinetics of the edge emission and the 1.36 eV band are remarkably similar.
Nevertheless, variation in injection level causes a strong variation in relative
prominence between the 1.36 eV band and the edge emission; this injection-
level dependence is not related to the kinetics of the two bands. Norris there-
fore suggests that the 1.36 eV transition terminates on the lower con-
figuration-coordinate branch of a compact Cu-related complex with strong
coupling to the lattice and that the upper branch of the same complex is
degenerate with the conduction-band continuum and in slow exchange
with the same.
Copper in LPE GaAs has been studied by Chiao et al. (1978). From
photoconductivity and other measurements they state that
if the Cu impurity is diffused into an n-type LPE GaAs layer after growth, it behaves as a
conventional acceptor and converts the layer to p-type conductivity; if the Cu impurity
is present in the melt during the growth of an LPE GaAs layer in a C-H,-SiO, system, an
increase in electron density by a factor of 10 is observed compared to an undoped layer
grown under the same conditions.
Hypothetical mechanisms have been suggested by which the Cu impurity

might increase the final electron density in n-type layers, but an unambiguous
description is not presently available. The Cu impurity may form a complex
with a lower oxidation state in such a way as to make more donor impurities
available: Cu might react with 0 to produce Cu,O,, thus removing 0 from
SiO, in the melt and freeing Si to act as a donor in the GaAs; Cu might enter
interstitial sites, acting as a shallow donor; or Cu and 0 may form a neutral
complex that serves, in turn, as a reaction center for the remaining excess 0
to form donors. Copper might also react with lattice defects, the Cu serving
as a reaction center for the formation of electrically active defects. Chiao
et al. (1978) continue as follows:
One particular impurity level is found in both Cu-diffused p-type GaAs layers and in n-type
GaAs layers grown with Cu present in the melt. This level is associated with an impurity
center that acts as a sensitizing center for n-type photoconductivity in the Cu-diffused
layers. Its energy level lies 0.43 eV above the valence band, its hole-capture cross section
is about cm’, and its electron-capture cross section is about lo-’’ cm’. Its presence
in the Cu-diffused layers is associated with the acceptor action of the Cu impurity that
dominates under these conditions; its presence in the layers grown with Cu in the melt
indicates that some of the Cu still occupies acceptor-like sites, but that the major effect
of the presence of Cu overwhelms this behavior. This 0.43 eV level associated with Cu
appears identical in every way to the corresponding level reported for Cu in bulk-grown
GaAs crystals.
The distribution coefficient of Cu in these LPE studies was from

emission spectroscopy but 10 - 6 from photocapacitance effects. These
numbers can be compared with the previously reported for melt-
grown GaAs where the temperature is much higher than the 700-750°C for
LPE growth. The considerably smaller values of distribution coefficient
given by capacitance technique indicate that the active Cu for these measure-
ments is only about 1% of the total. Possible reasons for this difference are (1)
photocapacitance gives only a lower limit of the Cu concentration, since it
assumes that all Cu centers have been emptied under the steady-state photo-
excitation; (2) Cu may diffuse away from the junction region in the electric
field after junction formation; (3) a major proportion of the Cu present may be
involved in nonelectrically active complexes.
112 A. G. MILNES
In these studies, levels were seen also at E, + 0.20,0.65, and 0.70 eV, but
these were not clearly copper related, and there was no sign of the E, + 0.14
eV level observed in other work (for instance, Ashirov et al., 1978).
Boborykina et al. (1978) in photocapacitance studies at 77 K use a
configuration-coordinate diagram approach and find the impurity energy
level to be E, + 0.37 f 0.02 eV with Stokes losses E, = 0.10 k 0.01 eV, and a
phonon energy of 2.5 k 3 meV. The luminescence band is a maximum at
1.035 eV. Kolesov et al. (1975) believe the 1.24-1.26 eV luminescence in
GaAs :Cu to be related to donor-acceptor pairs which they suggest are
Cui-Oi complexes.
Dzhafarov (1971) suggests that since
the vacancies in gallium arsenide are charged (V& and VLJ it follows that positive copper
ions migrating along the interstices should precipitate at negatively charged gallium
vacancies V,, . This is also favored by the fact that the radius of Cu (1.35 A) is closer to the
covalent radius of G a (1.58 A) than to the radius of As (0.96 A). Copper which is precipi-
tated at gallium vacancies exhibits acceptor properties. If a copper ion becomes lodged
alongside a positively charged impurity (or vacancy), it may form a complex with this
impurity or vacancy because of the Coulomb attraction. In the surface region of gallium
arsenide, where the nonequilibrium concentration of the arsenic vacancies is high, copper
may form complexes mainly with one or several arsenic vacancies. In the bulk of gallium
arsenide, where the equilibrium concentration of the vacancies is low, the formation of
complexes between the copper and the original impurity is more likely. However, copper-
vacancy complexes may also form in the bulk region. Since a 0.18 eV level is observed
only in the case of the simultaneous presence of Cu and Te, we can attribute this level to
Cu-Te complexes.
Dzhafarov (1971) continues by stating that

investigations of the spectra of bound excitons in samples of GaAs doped with copper
by diffusion have established the symmetry of impurity centers in the surface region.
These centers have been identified as complexes of copper and two arsenic vacancies
{VAeCuGaVAs}-. The acceptor level at 0.5 eV found in the surface region in our crystals,
may be attributed to the same {VAsCuGaVAS}- centers. A copper ion bound to two arsenic
vacancies in the surface region should have a low mobility and this is confirmed by our study
of electrotransport. The jump of a copper ion, bound into a complex, from one equilibrium
position to another requires the breakup of the bond between the copper ion and the arsenic
vacancies. This means that the activation energy should increase by an amount equal to
the binding energy of such complexes. In the bulk of a crystal, where the equilibrium con-
centration of vacancies is low and the concentration of copper (8 x 10” ~ m - is~ higher)
than the concentration of tellurium (- 1 x 10’’ ~ m - ~many) , copper atoms are not bound
and they may diffuse freely in gallium arsenide.
Thus, the experimentally observed rapid fall of the copper concentration near the surface
of gallium arsenide is the result of the formation of complexes in the vacancy-rich region
and the presence of these complexes reduces strongly the effective diffusion coefficient of
copper.
Zakharova et al. (1 972) have examined the effects of Cu diffused into n-type
GaAs: Si at 900°C. Acceptors are formed having ionization energies of about
0.1 or 0.14 eV. Slow cooling and diffusion at an elevated arsenic pressure
favored the formation of acceptors having an energy 0.1 eV, whereas
quenching and diffusion under an equilibrium arsenic pressure at the same
temperatures led to the appearance of the level 0.14 eV. The 0.1 eV level is
attributed to Cui and the 0.14 eV level to Cu;,. However, Willmann et al.
(1971) conclude from the symmetry of the IR excitation spectrum of the
E, + 0.15 eV level that this acceptor level cannot be a simple substitutional
Cu ion on a Ga site. The inference is that Cu induces the centers in con-
junction with lattice defects, but the actual physical nature is unknown
although the symmetry has been established (Willmann et al., 1973).
Kamar and Ledebo (198 1) have shown that Cu in LPE GaAs produces an
acceptor at E, + 0.4 eV. The behavior of the “Cu” level (HL4) at E, + 0.40 eV
has been examined with deep-level optical spectroscopy (DLOS) by Chantre
et al. (1981). This level has optical and thermal activation energies that are
identical, whereas many other traps typically show shifts of 0.1-0.2 eV
because of lattice relaxations (Bltte and Willmann, 1971).
The discussion of Cu in GaAs is concluded with a few further short
observations. One is that the Ga-As-Cu phase diagram has been constructed
by Panish (1967).The second is that Cu contamination enhances the degrada-
tion of GaAs electroluminescence diodes (Bahraman and Oldham, 1972).
Indeed, Cu can move 30 pm or more in a few hours in illuminated pn junctions
(Boltaks et al., 1972).
Copper in high concentrations precipitates in GaAs (Morgulis et al.,
1973) and may also be involved in the generation of interstitial dislocation
loops in GaAs:Te annealed above 380°C (Hutchinson and Dobson, 1975).
C . Silver
Solubility and bulk and surface diffusion studies of silver in GaAs have
been made. The diffusion studies suggest that the bulk solubility of Ag in the
temperature range 500-1 160°Cis 2-8 x 1017cm-3 (Boltaks and Shishiyanu,
1964). In other work, 3 x 10l6 cmP3 has been reported for the solubility at
800°C. The distribution coefficient is low, less than 4 x Distribution
coefficient measurements for impurities in GaAs have been summarized by
Willardson and Allred (1966). The surface solubility of the Ag is in the range
7 x 10’’ to 4 x 10’’ ~ m - The~ . diffusion is interstitial and the line marked
Ag in Fig. 22a roughly represents the behavior in bulk melt-grown (Bridgman)
GaAs. This illustration, taken from Kendall’s (1968) review of diffusion in
111-V compounds, is helpful for visualizing the overall diffusion pattern for
GaAs. See also Fig. 22b, from a review by Crawford and Slifkin (1975) that
contains some newer information. The original literature should be con-
sulted where possible because diffusion profiles tend to be complex in shape
114 A. G. MILNES
and dependent on the As overpressure. Self diffusion in GaAs has been

studied by Palfrey et al. (1982). Phosphorus diffusion has been studied by
Jain, Sadana, and Das (1976).
From photoluminescence studies, Blatte et al. (1970) report the acceptor
+
level of Ag to be E, 0.238 eV (4 K) for diffusion at 950°C into bulk n-type
+
-
GaAs to a concentration of 1017 cm-3 in a sealed ampule with As added.
This Ag acceptor at E, 0.235 & 004 eV has been confirmed by Hall
measurements. In addition, a shallower acceptor level at E, 0.107 eV is +
seen with a concentration about two orders of magnitude less than the Ag
level at 0.235 eV. The concentration NA of the 0.107 eV energy level is depen-
dent on the diffusion time; a longer diffusion time results in an increase of
the relative concentration of this level in the range 1-3 x l O I 5 ~ m - This
~ .
level has previously been reported after Cu treatments of GaAs so it is not
certain to be Ag related.
+
From DLTS studies the E, 0.23 eV level for Ag diffused into LEC-
grown GaAs has a hole-capture cross section of about cm2 (Yan and
Milnes, 1982).
The Ga-As-Ag and Ga-As-Au ternary-phase diagrams have been
obtained by Panish (1967). The equilibrium phase diagram for the binary
gold-gallium system has been established by Cooke and Hume-Rothery
(1966). In general, if Ag is a component of a GaAs ohmic contact technology
IOOO/T ( K ) IOOO/T(K)
0.70 0.60 0.90 0.70 0.60 0.90
I ' I I I I I I I I
1
2
c i -
loe 0
AQ
-10
10 Au
Be
z -12 Tm
0 10 Mn
v)
3
IA.
Sn
k I614 S
a
As
-161 I I I -16 I I
Se
10 10
1200 1100 1000 900 1200 1100 1000 900
FIG.22. Diffusion coefficients in GaAs at the low concentration limit: (a) after Kendall
(1968); (b) after Crawford and Slifkin (1975).
such as (Ag-Ge:Ni), the alloying temperature needed is more than 100°C

higher than for the corresponding (Au-Ge :Ni) technology. The Ag-Ge
eutectic temperature is 650"C, and the Au-Ge eutectic temperature is
about 350°C.
D. Gold
The solubility of Au in GaAs varies from 2.5 x 10l6 to 1.6 x 10" cm-3
for the temperature range 900-1140°C. A diffusion at 1000°C for 4 hr
produced an acceptor concentration of 4.5 x lOI7 cm-3 in bulk-grown
GaAs that was initially n-type 1.1 x loi5 ~ m - Hall
~ . measurements over a
range 150-300 K can be fitted by an energy level of E, + 0.405 f 0.002 eV. A
shallow acceptor state at E, + 0.05 eV also appeared with a concentration of
loi5 ~ m - Hiesinger
~ . (1976) comments that this is consistent with photo-
luminescence measurements of Au-diffused bulk GaAs, but the photolumi-
nescence of Au-doped layers grown from a saturated Ga solution did not
reveal the shallow level. Therefore, the shallow acceptor may be a center
formed by Au and Ga vacancies introduced during diffusion. In other
studies an acceptor at 0.04 eV detected by photoconductivity measurements
in n-type GaAs may be formed as a result of an interaction between gold and
defects associated with the presence of copper. Hiesinger found no evidence
of the earlier published level of 0.09 eV for Au-doped GaAs.
From DLTS studies the E, + 0.40 eV trap in Au-diffused LEC-grown
GaAs has a hole-capture cross section of about cm2 (Yan and Milnes,
1982). This is significantly smaller than the cm2 cross section for a
Cu-related hole trap seen at about E, + 0.40 eV when Cu is diffused into
GaAs (Kumar and Ledebo, 1981).
A complex of Au-Ge is found in degenerate germanium-doped n-type
layers of GaAs grown from a gold-rich melt and produces an energy level at
E, + 0.16 eV (Andrews and Holonyak, 1972). The diffusion of Au in GaAs
shows a higher concentration surface branch and a lower concentration deep
branch (Sokolov and Shilshiyanu, 1964).The deep branch can be represented
approximately by an erfc function. The deep diffusion coefficient varies from
6 x lo-' to 6 x cm2 sec-' in the temperature range 740-1025°C.
The alloying of Au on the surface of GaAs at 550°C causes a heavily
disordered near-surface region. Outward diffusion of As may occur at
temperatures as low as 350°C and at higher temperatures, there is evidence of
considerable inward diffusion of Au and outward diffusion of Ga and As.
Transmission electron microscope studies have shown the presence of
hexagonal AuGa (Magee and Peng, 1975). Other studies have shown the
formation of Au7Ga2.
Schottky junctions of Au on n-GaAs deteriorate in barrier height after
heating to 350°C for 30 min (Kim et al., 1974). Gettering of Au from GaAs
116 A. G . MILNES
can be achieved to some extent by back-surface damage followed by an

anneal at 800°C (Magee et al., 1979).
VIII. SHALLOW
ACCEPTORS:
Be, Mg, Zn, Cd
The commonly used acceptors in GaAs are Be, Mg, Zn, and Cd: Be and
Mg are used for implant purposes; Zn and Cd tend to diffuse readily (Kadhim
and Tuck, 1972); and in LPE, Be or even Ge (an amphoteric dopant) may be
preferred for the fabrication of multilayer structures. Beryllium is the acceptor
of choice in MBE because its sticking coefficient is much higher than that
of Zn or Cd.
Implantation of Be at 100 keV in GaAs followed by activation at 550°C
produces good p-type doping results (Anderson and Dunlap 1979). Other
studies of Be implants in GaAs have been made by Bishop et al. (1977), Kwun
et al. (1979), Lee (1980), McLevige et al. (1978), and Nojima and Kawasaki
(1978).
Magnesium and cadmium have been studied by Aoki et al. (1976), Yu
and Park (1977), and Williamson et al. (1979), who observed surface pileup
and outward-diffusion effects. Laser annealing of implanted acceptor
(Zn, Cd) and donor (Se, Sn) ions in GaAs has been studied by Badawi et al.
(1979) and other workers. For Cd solubility see Fujimoto (1970).
A systematic study has been carried out by Ashen et al. (1975) of the
incorporation of shallow-acceptor dopants in highly refined VPE and LPE
gallium arsenide. A detailed study of the bound exciton, free-to-bound, and
pair luminescence of doped layers resulted in an identification catalog for
the various elements. Many specimens were specially prepared for the study
and several growth methods were used. The layers intentionally doped with
Zn and Cd were grown by VPE; Be- and Mg-doped layers were obtained by
LPE. Magnesium has a fairly large segregation coefficient (-0.2) and is
relatively easy to use for lo'* cm-, doping, but in more lightly doped melts
the loss by oxidation presents a control problem. The Be distribution coeffi-
cient is greater than 1.O, but small additions of Be are affected by oxidation or
reaction with the C of the LPE growth boat. Attempts to add Ca and Sr in
LPE growths showed no new photoluminescence lines. Carbon doping in
VPE growths (AsCl, :H, :Ga) was achieved with CC1, addition. Ashen and
his co-workers (1975) comment that
Silicon-doped VPE layers were grown with the Si dopant situated in the reactor immedi-
ately downstream from the source. None of the layers which were deposited under the
standard growth conditions, showed any evidence of Si acceptors, only additional donors.
Silicon is a ubiquitous contaminant in both VPE and LPE layers. An appreciable portion
invariably occurs as an acceptor in LPE growth as it does with all deliberately Si-doped
LPE layers. In the exceptional VPE layer grown in the presence of considerable excess
of CCI,, Si acceptors were identified, together with C acceptors and residual Zn acceptors.
IMPURlTY LEVELS IN GALLIUM ARSENIDE 117
It is believed that in this case the excessive C deposit which built up on the walls of the
reactor during the run was responsible for the reduction of significant amounts of silica
and that these non-standard growth conditions were favourable for the introduction of the
acceptors.
The C acceptor was almost always found in LPE layers and Ge con-
tamination was frequently found in undoped LPE layers. High-purity VPE
layers show a line at 1.4889 eV that Ashen et al. (1975) tentatively assign to O, .
The results for specimens obtained from many sources are well illustrated
in Fig. 23. The values assigned to the binding energies of C , Si, Ge, Be, Mg,
Zn, and Cd are 26.0, 34.5, 40.4, 28.0, 28.4, 30.7, and 34.7 meV, respectively.
L L L L V V V V L L L L V V L L L V V V V V V V V L V V \
1.495
- -
C- -- -- -- - -
-
-
1.490
Zn-
eV
sil*
1.480
Ge -
1.475
FIG.23. Energies of photoluminescence bands in GaAs below 15 K produced from a variety

of sources. The spectra have been shifted by an energy which brings the shallow exciton lumin-
escence into coincidence. The arrows marking the positions of the elements on the energy axis
represent the free-to-bound transition for a conductive electron captured at a neutral acceptor
site. The dotted levels represent emission bands observed between 10 and 15 K-some tend to
become obscured at very low temperatures due to the growth of the pair luminescence. (After
Ashen et al., 1975; reproduced by permission of Her Britannic Majesty's Stationery Office.)
118 A . G. MILNES
In other work, there have been reports of deep levels apparently associated
+
with Zn and Cd doping. A deep level, E, 0.65 eV, possibly associated with
Zn doping, has been observed by Su et al. (1971) in LPE-grown GaAs. A deep
acceptor level at about 0.36 eV above the valence band has been observed
with Cd doping and is attributed to lattice defects or to Cd lattice-defect
complexes (Huth, 1970). However, systematic confirmation of such levels is
needed before general acceptance since trace contamination (for instance, of
Cu) must always be considered as being possibly present.
Kanz (according to Kendall, 1968) in a diffusion of Hg203 at 1000°C
found low surface concentration (5 x 10” cm-3) and low diffusion coeffi-
cient (5 x cm2 sec-I). Mercury, therefore, does not appear to be an
interesting dopant in GaAs.
IX. GROUPIV ELEMENTS

AS DOPANTS:
C, Si, Ge, Sn, Pb
For future device applications, large-area GaAs slices are needed, and as
discussed in Section V these must be semi-insulating yet relatively free of Cr.
Since the Cr is added to compensate Si donors, there is need to reduce the Si
content of the crystals. Acceptor impurities such as C are also undesirable
because of possible effects of ionized impurity scattering on electron mobility.
There therefore is a trend toward LEC growth (that is, GaAs pulled from a
melt encapsulated with boric oxide) from crucibles of BN rather than S i 0 2
or C. The result is that the GaAs contains boron. Laithwaite et al. (1977)
comment as follows:
After samples are rendered transparent, either by electron irradiation or by copper dif-
fusion, the optical absorption due to localized modes of vibration (LVM)indicates that
a large fraction of the boron atoms occupy gallium lattice sites; this is not surprising as
boron and gallium are both group 111 elements. However, other boron atoms are located
in another site which we have labelled B(2). Because of the existence of the two naturally
occurring isotopes loB and “B, it was possible to show that a B(2) centre involves only
one boron atom. These centres may be remote from other defects, but considerable pairing
with Si,,Se,,, or Te,, donors also occurs; similar complexes were observed in gallium
phosphide (Morrison et a/., 1974). It was deduced that an isolated B(2) centre had Td
symmetry and that the paired boron in the B(2) complexes was surrounded by four nearest-
neighbor gallium lattice atoms, apart maybe from one such atom in the [B(Z)-Si,,] pairs.
There would appear to be only two possibilities: (i) the boron atom occupies a substitu-
tional arsenic site; or (ii) the boron is in the tetrahedral interstitial site with gallium neigh-
bows. Because of the strong pairing with the known donor species it was argued that a
B(2) centre should be an acceptor.
Laithwaite et al. (1977) conclude that the B2 level is an acceptor but the
structure of the defect remains uncertain: they incline to the view that it is
possibly an interstitial boron atom rather than the antisite defect BAS.
Undoped GaAs prepared under the purest possible conditions tends to

have many residual impurities in concentrations of the order of 5 x l O I 3
cm-j. Wolfe et al. (1977) and Stillman et al. (1976) have shown that it is
possible, in favorable circumstances, to determine the chemical nature of
shallow donors in concentrations somewhat higher than this from high-
resolution far-IR photoconductivity measurements of transitions between
the ground state and the first excited state. The shallow donor levels are very
nearly at the energy level 5.737 meV predicted from the hydrogenic model,
but central cell effects can cause small chemical shifts (increases) of the
ground-state energies. The shifts and observed line shapes are magnetic
field dependent and this helps in the resolution.
Table XI1 shows the central cell shifts (above 5.737 meV) observed when
known dopants are added and some observations of levels that can then be
interpreted for residual donor levels seen in VPE and LPE specimens given
similar examination.
The ground-state values arrived at by Wolfe and co-workers for Se, Si, S,
and Ge are 5.854, 5.808, 5.890, and 5.908 meV, respectively. Some idea of the
precision, and difficulty, of the method may be obtained from the similar
experiments of Ozeki et al. (1979) that yield values 5.812, 5.799, 5.845, and
5.949 meV, respectively. Further work will no doubt give closer agreement,
but the use of the technique is limited to dopings less than 5 x 1014 cm-3
because of spectral broadening.
The work of Wolfe et al. (1977) in preparing the lightly doped specimens
led to estimates of segregation coefficients for their growth processes. Their
numbers and some other recent determinations are added to Table XIII,
which is otherwise the collection compiled by Willardson and Allred (1966).
TABLE XI1
CENTRALCELLCORRECTIONS OF GROUND-STATE
ENERGY
ABOVE
THEHYDROGENICLEVELOF 5.737 meV FOR KNOWNDONORLEVELS
AND TENTATIVE OF RESIDUAL
IDENTIFICATIONS DONORS"
Residuals observed Residuals observed

(meV) (meV)
Known donors Known donors
(mev) VPE LPE (meV) VPE LPE
Pb 0.036 - 0.041 S 0.153 - -

- - 0.064 - Te(?) 0.155 - -
Se 0.071 - - Ge 0.171 - -
Sn 0.080 - 0.079 C 0.200 0.200 -
Si 0.117 0.117 0.117
Reproduced with permission of the Institute of Physics, UK.

120 A. G. MILNES
TABLE XI11
DISTRIBUTION
COEFFICIENTS
OF IMPURITIES
IN GALLIUM
ARSENIDE~
Haisty and Willardson

Doping Cronin Edmond Weissberg Whelan and Allred Wolfe et al.
element” (1964) (1959) (1961a,b) et al. (1960) (1959) (1977)
Al 3 0.2
Ag 0.1 4 x 10-3
Bi 5 x 10-3
Be 3
Ca < 0.02 2 x 10-3
C 0.8
Cr 6.4 x 10-4 5.7 x 10-4
co 8.0 x 10-5 4 x 10-4
cu 2 x 10-3 < 2 x 10-3
Ge 0.03 0.018 0.01 1-2 x
In 0.1 7 x 10-3
Fe 2.0 x 10-3 3 x 10-3 1 x 10-3
Pb < 0.02 < I x 10-5
Mg 0.047 0.03 0.1
Mn 0.021 0.02
Ni 6.0 x 10-4 <0.02 4 x 10-5
P 2 3 3
Pb 1-3 x lo-’
Sb <0.02 0.016
Se 0.44-0.55 0.30 0.5-1.0
Si 0.11 0.1 0.14 0.14 1-2 x 10-4
S 0.17 0.3 0.5-1.0 0.30 1-3 x lo-’
Te 0.025 0.3 0.054-0.16 0.059 2-4 x lo-’
Sn 0.048 0.03 0.08 2-4 x 10-3
Zn 0.36 0.1 0.27-0.9 0.40
a See Romanenko and Kheifets (1975) for other data on Cr, Si, Sn, Te, Zn, and Stringfellow
(1974) for temperature dependence of the distribution coefficients of Se, Sn, and Te. Reproduced
with permission of the Institute of Physics, UK.
A. Carbon
Carbon concentrations up to lOI7 cm-3 can be obtained as shallow
acceptors by C doping of crystals during horizontal Bridgman growth.
Undoped Bridgman crystals show C concentrations of 5 x 10’’ cmP3 or
less. Measurements by Brozel et al. (1978)correlate the local vibration mode
IR spectroscopy line with acceptor action and conclude that carbon is
present, predominantly as C,, defects. In MBE layers, C at 1.6 K produces
photoluminescence at 1.493 eV by a free-electron transition to the CAs
acceptor (Kunzel and Ploog, 1980). Since the band gap is 1.519 eV a1 this
temperature, this agrees with the carbon acceptor found to be 0.027 eV above
the valence-band edge (Bimberg and Baldereschi, 1979; Bimberg, 1978;
Bimberg et al., 1982). Evidence for the existence of low concentrations of CGa
donors in epitaxal GaAs has been obtained by other methods (Wolfe et al.,
1977).It seems therefore that carbon is an amphoteric impurity, although the
ratio of the concentrations of the two species has yet to be determined for
any type of material.
Spark-source mass spectroscopy shows that the common etchant
bromine methanol tends to leave residual surface carbon that is not readily
removed. Etching with an H202:H,S0, solution leaves less carbon on the
surface. Implantation of C into GaAs has been studied by Shin (1976) and
Shin et al. (1978).
B. Silicon
Silicon is an amphoteric impurity in gallium arsenide, and 28SiGadonors
and 28SiAsacceptors give localized modes of vibration (LVM) absorption at
384 and 399 cm-l with linewidths of 1.4 and 2.0 cm-', respectively (Laithwaite
et al., 1977).Samples with relatively low carrier concentrations ( 5lo'* ~ m - ~ )
contain predominantly Sic, defects, low concentrations of Si,, , and show
negligible LVM absorption from other more complicated defects such as
(SiGa-SiA,)or (B2-SiG,) pairs in LEC material (Morrison et al., 1974).When
GaAs is grown from a melt of stoichiometric composition, silicon if present
is incorporated on gallium sites as a shallow donor. However, if GaAs is
solution grown from gallium, under conditions where there are few gallium
vacancies, Si is a shallow acceptor. The effect of the Fermi level on Si-site
distribution in GaAs:Si has been considered by Kung and Spitzer (1974).
In solution growth, by careful control of the temperature range and
cooling cycle, it is possible to grow np or pn amphoterically doped diodes
(Kressel et al., 1968a; Moriizumi and Takahashi, 1969).The GaAs is usually n
type if grown above 850°C and p type if grown below. At a temperature of
about 7OO0C, mol fraction of Si in the Ga melt produces a net hole
concentration of 1OI8 cm- '. In LED use, amphoterically doped junctions
exhibit a broad emission bandwidth, perhaps involving transitions to three
levels E, + 0.03, E, + 0.10, and E, + 0.22 eV. Kressel et al. (1969) suggest
that the 0.03 level is due to Si occupying random As sites and that the 0.10 eV
level is a complex involving Si. This might be (SiGa-SiA,)or (Sica-Vca)
(Spitzer and Allred, 1968). The 0.22 eV level is attributed to an arsenic-
vacancy complex that is related to the presence of silicon. These attributions
must be treated as hypothetical in the absence of direct supporting evidence.
Antell (1965) has examined the diffusion of Si in GaAs by junction-depth
studies. The diffusion rate is low and the profile is steep fronted and dependent
on the arsenic overpressure. Silicon-ion implantation in GaAs has been
studied by many workers (see, for instance, Nozaki, 1976; Tandon et al.,
1979).
122 A. G . MILNES
If GaAs is grown in the presence of silicon, Si02, and/or oxygen, some

workers believe that complexes of silicon and oxygen form. In particular,
Weiner and Jordan (1972) infer that silicon atoms on gallium sites pair with
interstitial oxygen atoms, forming a complex which behaves as an acceptor
with energy levels about 0.2 and 0.4 eV below the conduction band.
Wolfe et al. (1977) report that after adding G a 2 0 3to a Ga melt, the LPE
material grown contained a lower concentration of the shallow silicon donor.
This is in agreement with the observation of other workers that S i 0 2 disso-
ciation is suppressed (Huber et al., 1979; Martin et al., 1981a). No shallow
oxygen donor was seen and a search was not made for the deep-acceptor
levels reported by Weiner and Jordan (1972). Silicon is an effective dopant in
MBE (Stall, 1980).
C . Germanium
Germanium on a gallium site introduces a shallow donor at about 6 meV,
but on an As site it becomes a shallow acceptor at 40 meV. GaAs: Ge grown
from a Bridgman melt is n type or closely compensated p type. Addition of Ge
compounds during VPE gives n-type material. Hurle (1979d) has considered
the compensation of this doping. In LPE, the Ge appears mainly on the As
sites and so the layers are p type. Hole concentrations up to 10l9 cm-3 are
obtainable. In LPE, Ge is a valuable shallow acceptor because its vapor pres-
sure is much less than that of Zn or Cd and its diffusion coefficient in LPE
material is low (Neumann et al., 1977; Shih and Petit, 1974; Springthorpe
et al., 1975).
In MBE, Ge is a useful n-type dopant. Its sticking coefficient is good, and
under suitable growth conditions the tendency to give compensating acceptor
centers is no worse than that of Si. However, it does exhibit some residual
doping effects after the Ge effusion cell is shuttered, whereas Si does not. If
the substrate temperature is high and the arsenic flux low, Ge acts as a p-type
dopant (Ploog, 1981). The zero-phonon E, - GeAsphotoluminescence line
is at 1.4778 for 4 K. Inplantation of Ge into GaAs has been studied by
Kozeikin et al. (1977), Surridge et al. (1977), and Stoneham et al. (1980).
D. Tin
Tin is predominantly a shallow donor (5.8 meV) in GaAs. With the
electron concentration in the range 10'7-10'9 cmP3,VPE-grown GaAs-Sn
is found to have less than 8% of the Sn in an acceptor state (at E, + 0.167 eV
from photoluminescence measurement; Schairer et al., 1976;.Maier et al.,
1981). Hall mobilities for electron concentrations between 10'5-10'8 cm-3
tend to lie on a curve that is independent of the growth procedure (LPE, VPE,
or MBE) parameters such as the temperature or the As-Ga ratio in the gas
phase. This is a further indication that the films are not heavily compensated.
It is not known whether the acceptor that does form in limited quantities is a
complex deep level or is simply due to Sn on isolated As sites. Hurle (1 979c)
has a thermodynamic model that assumes the acceptor to be the complex
Sn,,V,, . Photoluminescence of the deep Sn related level has been studied by
Zemon et al. (1981).
In MBE, Sn has been widely used as an n-type dopant, and there seems no
evidence for acceptor formation, but it has been found to have a surface rate
limitation to the incorporation of Sn atoms, so until a certain surface popu-
lation, up to 0.1 monolayer, is formed the free-donor density increases
exponentially. There has been some tendency, therefore, to prefer Si as an
n-type dopant or to consider SnTe as a dopant source for MBE growth since
this seems to be better behaved (Wood and Joyce, 1978; Ploog, 1981).
Tin diffusion into GaAs from doped SiO films has been studied by
Yamazaki et al. (1975) and by Gibbon and Ketchow (1971). Kachurin et al.
(1975) conclude that Sn diffuses via gallium vacancies. In the diffusion of
tin into GaAs through a window in a phosphosilicate glass mask, care must be
taken to reduce interfacial stress between the mask and the substrate, other-
wise a lateral diffusion of tin may extend up to 500 times the junction depth
(Baliga and Ghandhi, 1974).
Ion implantation of Sn in GaAs has been studied by Finstad et al. (1974).
After a suitable anneal, about half of the Sn atoms (up to 1.7 x 10'' atoms
cmP3)are on substitutional sites. Effects of surface states have been examined
by Wolfe and Stillman (1973). Reflectivity measurements on Sn doped GaAs
have been made by Rheinlander et al. (1969).
E. Lead
Lead is also a shallow donor, but the low segregation coefficient (-
results in it being little used. The photoluminescence is broad band and
suggests a compensated acceptor level at about E, + 0.12 eV (Kressel et al.,
1968b).
IN GaAs
X. OXYGEN
Oxygen (0.J in GaAs was considered for many years to be responsible

for the deep donor EL2 level seen at about E, - 0.75 to 0.83 eV and present in
concentrations up to 10'7 ~ m - If~ net . shallow acceptors (NA- ND) are
present in the crystal (at a lesser concentration), the deep donor provides the
compensating electrons needed by the net acceptors. Hence the free carrier
concentration becomes very small and the GaAs is semi-insulating ( lo6 R-cm
or more). There is now considerable doubt about the role of oxygen in the
124 A. G . MILNES
formation of the EL2 level. One good reason for supposing that oxygen is
involved is the study of Wolford et al. (1981, 1982) that commences with the
oxygen substitutional on the phosphorus site in GaP with a donor level at
E, - 0.89 eV and tracks this energy level through the alloy range of
GaP,As,-, as x goes from 1 to 0.6. Extrapolation of this energy line to
x = 0, as shown in Fig. 24, gives an energy level that nearly coincides with
that of the EL2 trap; however, this does not allow us to presume that the EL2
level is substitutional oxygen on an arsenic site. Alternatively, one might
accept that OAsdoes exist with an energy close to that predicted by the
extrapolation in Fig. 24, but that its concentration is low and its effect
sometimes obscured by the presence of a non-oxygen-related complex, the
2.2 -
0 0.5
GaAs X
FIG.24. The 0 levels for GaP and GaP,As, - * fall on a line that may be projected to GaAs,
with a result in the vicinity of the EL2 level; T = 5 K. (After Wolford et al., 1981, 1982 and
S. Modesti, D. J. Wolford, B. G. Streetman, and P. Williams, private communication; repro-
duced with permission of the authors.)
EL2 electron trap, that just happens to have an almost identical energy level.
Photoluminescence at 0.64 eV has been attributed by Taniguchi and Ikoma
(1982)to an oxygen-related trap that is distinct from EL2. The possibility also
exists that oxygen-related complexes may also be present in the GaAs that
are quite different in energy than the EL2 position. For studies of GaAs, - x P x
see Sankey et al. (1980).
Huber et al. (1979) state that oxygen is not involved, either directly or as
part of a complex defect, in the origin of the main electron trap (EL2) at
E, - 0.75 eV. From secondary-ion mass spectrometry (SIMS) they deter-
mine the oxygen content of certain crystals to be in the low lOI5 ~ m - ~ ,
whereas the EL2 concentration is one order of magnitude higher. Attempts to
create the E, - 0.75 eV level in LPE GaAs or boat-grown GaAs by the
addition of G a 2 0 3to the Ga melt are unsuccessful, although the residual Si
content may be reduced in the crystal by such addition (Kaminska et al., 1981).
Consider other observations of the effects of oxygen on trap levels in
GaAs. Weiner and Jordan (1972) suggest that silicon atoms on gallium sites
pair with interstitial oxygen atoms, forming a complex which behaves as an
acceptor with energy levels near E, - 0.2 eV and E, - 0.4 eV. It is assumed
that this complex can dissociate upon annealing at 800°C by the reaction
= (si,,O,)O
2(siG,0,)- + Siba + 3e-
Gallium arsenide grown in quartz under a B20, layer converts from
semi-insulating to n type. This is accompanied by disappearance of the 0.2
and 0.4 levels. Annealing at much higher temperatures (1 100°C)reconverted
the material to semi-insulating type. Gallium arsenide grown in A1203 did
not show these conversion effects so, clearly, Si is involved. Weiner and
Jordan are uncertain whether the E, - 0.2 and E, - 0.4 eV levels are different
charge states of the Si,,Oi level or represent different configurations of the
complex (such as spacing of the Si and Oi atoms).
The Weiner and Jordan specimens were Bridgman-type material.
However, ion-beam implantation of oxygen in n-type GaAs also results in
levels at E, - 0.2 eV and E, - 0.4 eV from conductivity versus temperature
measurements and from photovoltage measurements (Abdalla et al., 1975).
Other studies of the effects of oxygen implantation in GaAs have been made
by Deveaud et al. (1975) and Favennec (1976).
Introduction of O2 into the gas stream during VPE growth of GaAs does
not affect the EL2 concentration or produce any other electron traps (Wallis
et al., 1981b). However, if Al,Ga,-,As is grown, even in trace quantities O 2
produces an acceptor level at E, - 0.41 eV for x c 5%, and this level seems
to deepen with increasing x. Hall studies by Akita et al. (1971) of oxygen-
doped boat-grown GaAs and VPE layers grown in the presence of oxygen
provide support for levels between E, - 0.176 eV and E, - 0.21 eV. Look
(1981) also has studied levels in this region of the band gap.
126 A. G . MILNES
Monemar and Blum (1977) in a study of O-implanted GaAs find a broad

photoluminescence that peaks at 1.23 eV, but they attribute this to a VGa-
donor transition. Bound exciton-donor luminescence has been studied by
White (1974).A survey of photoluminescence has been presented by Williams
and Bebb (1972).
Vasudev and Bube (1978) have studied photocapacitance effects in bulk
n-type GaAs grown by the horizontal Bridgman technique at 1200°C with
G a 2 0 3added. They see dominant deep levels at E, - 0.46 eV, E, - 0.79 eV,
and E, - 1.10 eV. The E, - 1.10 eV level is considered likely to be trace
contaminated by iron (the possibility of Cu or Ni contamination might be
considered). The E, - 0.46 eV level (present in lower concentration than the
0.79 eV level) is unidentified, but it is considered that this may be an oxygen-
associated level. The range of specimens available did not permit a systematic
study of the trap concentration versus the Ga,O, added or versus the actual
oxygen content of the material. Photocapacitance studies have also been
made by White (1976).
Aleksandrova et al. (1973) in studies of VPE layers (Ga-AsC1,-H,
system) grown in the presence of oxygen find deep centers at E, - 0.35 eV,
E, - 0.56 to 0.62 eV, E, - 0.8 and 1.01 eV. The acceptor level at 0.35 f
0.03 eV is considered the most clearly related to oxygen. The 0.56-0.62 eV
levels are stated to be donor-like in nature and to be related to lattice defects.
The 0.8 eV level is attributed to Cr acceptors.
Further evidence that there is a significant level at about E, - 0.4 eV in
n-type boat-grown GaAs: 0 occurs in the studies of Hatch and Ridley (1979).
The concentration of the center was about 7 x lo', cm-,. The photoioniza-
tion threshold was 0.41 eV, and the spectral dependence agreed with the
Lucovsky model (suggesting that the center involved was donor-like and
neutral in the n-type material). In other work, with similar n-type GaAs: 0,
levels at E, - 0.15 eV and E, - 0.23 eV have also been observed (Leach and
Ridley, 1978).
Some evidence that oxygen can be present in an interstitial position in
GaAs comes from the IR absorption band near 840 cm-' that is confirmed
to be oxygen related by the introduction of the O'* isotope (Akkerman
et al., 1976). The relatively high vibration frequency suggests interstitial
position for the oxygen atoms and covalent binding to the neighbors. The
vibrations of substitutional impurities (for instance, C") lie at longer
wavelengths (582 cm- ). '
Studies of 0,-doped n-type GaAs by Li and Huang (1973) conclude that
photoinjected excess carrier recombination at 20 K is dominated by band-to-
band radiative recombination and that any deep-level oxygen impurities
that may be present are neither acting as recombination centers nor as
trapping centers for the excess carriers.
The effective diffusion coefficient of oxygen in GaAs from mass spectro-

metric data has been stated to be DcR = 2 x exp( - 1.1 eV/kT) cmz
'
sec- between 700" and 900°C (Rachmann and Biermann, 1969); however,
this work reports oxygen contents of 10'9-1020 atoms cm-3 instead of the
10I6 atoms cm-3 typically measured when considerable precautions are
taken in the measurement (Blackmore et al., 1976). High solubility of oxygen
in GaAs ( > 10" atoms cm-3) has also been reported by Borisova et al. (1977)
in systematic doping studies. On the other hand, a study of oxygen by
charged-particle activation with the 160(3He,p)l'F reaction in which "F is
created by 3He bombardment suggests the oxygen concentration to be
4 x lo1' atoms cmP3in LEC material and 2 x 10I6 cm-3 in other material
(Emori et al., 1981).
In summary, then, the oxygen-related levels in GaAs are not well known.
There are a number of observations that suggest that levels at about E, -
0.2 eV and E, - 0.4 eV should be considered oxygen related, and there is
growing evidence that the E, - 0.75 to 0.83 eV level is not oxygen related.
Evidence that the "oxygen level" when filled with an electron may have a
metastable state (Bois and Pinard, 1974; Bois and Vincent, 1977; Vincent and
Bois, 1978; Bois et al., 1978) is accumulating (there is good evidence for sim-
ilar effects associated with 0 in Gap). See also Mitonneau and Mircea (1979).
XI. GROUPVI SHALLOW

DONORS:
S, Se, Te
Sulfur, selenium, and tellurium are shallow n-type dopants in GaAs. The
activation energies, as discussed in Section IX, are all quite small. At high
doping concentrations, the behavior of Se and Te in GaAs is disturbed by
impurity complexing or precipitation (Abrahams, 1974; Casey et al., 1971;
DiLorenzo, 1971; Greene, 1971; Hutchinson and Dobson, 1975; Morgulis
et al., 1974; Sealy, 1975; Shifrin et al., 1974; Takahashi, 1980).
A broad emission band centered at 1.2 eV is observed in GaAs doped with
group VI elements, whether melt or vapor grown, or grown by LPE with
gallium as the solvent. This photoluminescence increases in intensity,
relative to the band-gap radiation, with increasing dopant concentration in
the lo1' cmP3range. The luminescing recombination centers responsible for
this 1.2 eV band may be complexes such as V, + 3Se or V,, + 3Te. These
respresent the solid solution of Ga,Se, or Ga,Te, in GaAs; VGaTeAscom-
plexes also have been surmised. With increasing dopant content, the solubility
limit is eventually exceeded and precipitates of Ga,Se3 or Ga,Te, are formed.
With the dopant concentration increased beyond 2 to 3 x 10'' ~ m - the ~ ,
radiative efficiency falls off sharply. Kressel et al. (1968b) suggest that the
128 A . G. MILNES
falloff is partly due to nonradiative recombination via small precipitates,

each containing on the order of 1000 dopant atoms. (The radiative efficiency
of n-type silicon-doped GaAs prepared by LPE does not fall off up to
4.2 x lo1' cmP3)The formation of complex centers is to be expected since
the strain energy of the crystal is reduced when a vacancy relocates next to
an impurity atom that is somewhat large for its site position.
Nishizawa et al. (1974)have studied Te-doped GaAs crystals heated in the
presence of arsenic vapor. Complexes associated both with the donor
impurity (Te) and with the defects due to the deviation from the stiochio-
metric composition are formed. These complexes are responsible for both the
0.18 eV acceptor level and the 1.30 eV emission band at 77 K. Furthermore,
defects with acceptor levels of 0.12 and 0.15 eV are found. These defects are
unstable and disappear by reannealing at comparatively low temperatures
(300-400°C), leaving acceptor levels at 0.1 8 eV.
Precision lattice measurements on 4-8 x 10" cmP3 n-doped GaAs:Te
show superdilation effects (Dobson et al., 1978). A possible defect causing
superdilation could be an interstitial complex of Ga and Te on a split substi-
tutional site, VGaGaiTei,which could either be a donor or a neutrally charged
complex in equilibrium with the Te substitutional donor TeAs.Annealing at
880°C may remove this complex and form interstitial (GaiAsi)loops and the
complex TeAsVGamay form.
Tellurium is thought to interact with other impurities in GaAs such as
Zn (Blashku et al., 1971). (Te-Li complexes were mentioned earlier in
the discussion of lithium doping.) The diffusion of Te in GaAs has been
studied by Karelina et al. (1974), who conclude that arsenic vacancies play an
important role in the diffusion process. Ion implantation of Te in GaAs has
been studied by Pashley and Welch (1975) and Eisen and Welch (1977).
Selenium implantation has been studied by Gamo et al. (1977) and Lidow
et al. (1978).Tellurium is not a suitable dopant in MBE of GaAs because of a
pronounced surface segregation behavior (Cho and Arthur, 1975).
Sulfur is not usually used as a dopant in melt or LPE growth of GaAs
because of its high vapor pressure; however, it may be added to VPE-grown
layers because of the convenience of handling in the form of sulfur gas
compounds such as H,S, SCl, and SC1, (Savva, 1976). Diffusion of S into
GaAs is best accomplished by the use of a ternary diffusion source of Gas :As
(Matino, 1974).
Sulfur-ion implanted profiles in GaAs (200 keV) have been studied by
Comas and Plew (1976). A review of ion-implantation in compound semi-
conductors has been made by Degen (1973). Implantation is a technology
that is rapidly increasing in importance in GaAs processing.
XII. OTHERIMPURITIES:Mo, Ru, Pd, W, Pt, Tm, Nd
A. Molybdenum
Molybdenum has been studied as a possible barrier layer in ohmic
contacts to n-type GaAs to impede the tendency for the metals of the contact
structure such as gold to cause loss of arsenic and gallium and inward
diffusion of gold at temperatures of a few hundred degrees. In earlier work,
Mo did not seem to be a very effective barrier layer, but more recent studies
have suggested that barrier action is observed if the layer is sputtered in the
presence of nitrogen (Nowicki and Wang, 1978).
Studies of the diffusion of Mo into GaAs have not been found in the
literature. The photoluminescence peaks of GaAs: Mo have been reported as
at 1.34 and 1.43 eV (Kornilov et al., 1974), but no activation energy with
respect to a band edge has been reported.
B. Ruthenium
Chemisorbed Ru ions on the surface of n-type GaAs decrease the surface
recombination velocity of electrons and holes from 5 x 10' to 3.5 x lo4
cm sec-'. The ions in a one-third monolayer thickness are confined to the
surface and do not form a new junction by diffusing into the GaAs, even
after a 300°C treatment for 1 hr. This is the first known example of simple
chemisorption reducing the surface recombination of GaAs. The use of an
overlayer of Al,Ga, -xAs, however, on GaAs results in a recombination
velocity of 450 100 cm sec-'. (Nelson et al., 1980).
C . Palladium
Palladium is a part of certain recipes that make moderately satisfactory
contacts to n- and p-type GaAs. For contacts to n-type GaAs, Pd-Ge is used
and typical anneals are at 350°C or more for 10 min. The specific contact
resistances may vary from 5 x to 5 x R-cmZfor dopings between
1OI6 and 1 O I 8 electrons ~ m - respectively.
~ ,
For contacts to p-type GaAs, the metallization may be Au-Pd, Ag-Pd,
A1-Pd, and specific contact resistances are high, about R-cm2, even on
1 O I 8 cm-3 material. The surface metallurgy of the reaction of Pd with GaAs
has been studied by Olowolafe et al. (1979). The kinetics of Pd penetration
into GaAs was found to be controlled by diffusion with an activation energy
of 1.4 eV, corresponding to a characteristic (lle) distance of 160 A after 1 hr
130 A. G. MILNES
at 250°C and 730 8, after 10 min at 350°C. This considerable penetration may
be harmful in thin junction devices. The As and Ga compounds formed as a
result of contact reactions were identified to be PdAs, and PdGa at 250°C;
PdAs,, PdGa, and Pd,Ga at 350°C; and PdGa at 500°C. In a study of
Pd-Ge-n-type GaAs contacts, Grinolds and Robinson (1980) find that the
Pd-Ga and Pd-As interactions provide a situation favorable for greater
occupation of Ga sites by Ge than As sites and so the n-type doping of the
GaAs is increased in the sintering process. Their results confirm the specific
resistance of 3-4 x R-cm2 for 1OI6 cm-3 n-type GaAs, reported by
Sinha et al. (1975), with sintering temperatures of 500°C or more. There
appear to have been no studies of the energy levels of Pd in GaAs.
D. Tungsten
Tungsten (about 3000 A thick) offers an excellent barrier to diffusion of
Au and Pt atoms at temperatures up to 500°C. A structure such as Au-W-
PtGa-PtAs,-n-type GaAs exhibits a good Schottky barrier (& = 0.89 eV)
due to the formation of the PtAs, layer (Sinha, 1975). Tungsten directly on
the n-type GaAs has a barrier height of about 0.69 eV.
Tungsten or W: 10% Ti layers may also be used as barrier layers in ohmic
contact technology of the kind Au-Pt-W-Au: Ni: Ge/n-type GaAs. The W
serves as a diffusion barrier and also as a stress compensator (Rozgonyi et al.,
1974). The Pt layer (2000 A) helps improve the adhesion between the thick
outer Au layer and the W.
Tungsten-doped GaAs exhibits photoluminescence bands at 1.13 and
0.83 eV (Kornilov et al., 1974) and 0.707-0.651 eV (Ushakov and Gippius,
1980). Attempts have been made to dope GaAs with W by Czochralski
growth from a W-doped melt and by addition of a tungsten compound to a
chloride-gas transport system (750°C). Thermally stimulated current meas-
urements then show activation energies of 0.39-0.40 eV (Aleksandrova
et al., 1976).Since this level is similar to those seen in GaAs and attributed to
other causes (such as Cu or Ni), confirmation of a direct W relationship is
needed.
E. Platinum
The metallurgy of Pt layers (up to 9000 A thick) on GaAs when heat
treated in the range 250-500°C has been studied by Chang et al. (1975).When
Pt-GaAs was treated in vacuum, the Pt-GaAs interaction was initiated
with a rapid migration of Ga into Pt and simultaneous formation of an
As-rich layer at the Pt-GaAs interface. Gallium eventually traveled the entire
thickness of even the 9000-A Pt film, whereas As stopped abruptly at a
distance ~ 2 / of3 the way into the Pt. Little or no Au was detected (< 1 at. %)
in the Pt or GaAs after extensive Pt-GaAs reaction in Au-Pt-GaAs. The

reacted Pt film displayed a multilayered structure with each layer containing
different amounts of Ga, As, and 0.Pt-GaAs heated in air behaved similarly
with two added features: a surface layer containing mostly Ga and 0 formed
over the Pt, and 0 diffused through the (reacted) Pt to form an 0-rich layer
at the Pt-GaAs interface.
Attempts to dope GaAs with Pt have not been found in the literature and
the energy levels (if any) are unknown. Platinum is an effective recombination
center and lifetime controller in Si; however, a layer of Pt heated at 600°C
for 5 min on the surface of VPE n-type GaAs has been found to raise the
minority-carrier (hole) diffusion length (Partin et al., 1979d). This is presum-
ably because of gettering to the surface of a lifetime killing center rather than
the entry of Pt deep into the GaAs layer.
F . Rare-Earth Studies in GaAs:Tm and GaAs:Nd

Rare-earth elements are of interest because their incomplete 4f subshells
give spectral properties that are useful, such as stimulated emission when
incorporated in various crystal hosts. Relatively little work has been done on
the incorporation of rare earths in covalent semiconductors such as Si, Ge,
and GaAs since the solubilities are expected to be low.
The solubility and diffusion of thulium in GaAs have been studied by
Casey and Pearson (1964).Thulium was chosen because of its relatively small
atomic size (1.746 A) and the availability of a suitable radioactive isotope,
l7'Tm. The maximum solubility was found to be 4 x 10l7 cm-3 at 1150°C.
This low solubility was attributed to the fact that the Tm atom is 25% larger
than the Ga atom that it replaces substitutionally.
Richman (1964) has made initial studies of Nd doping of GaAs. The
pumping band of Nd closely matches the emission obtained from GaAs
luminescent diodes. Vapor-phase epitaxy was attempted and some inclusion
of Nd in the grown crystal was observed.
XIII. MINORITY-CARRIERRECOMBINATION, GENERATION,

LIFETIME,
A N D DIFFUSION LENGTH
Since GaAs is a direct band-gap material, non-phonon-assisted recom-

bination processes set an upper limit to the excess minority-carrier lifetime.
However, because of the high recombination center content in GaAs (bulk,
VPE, and LPE), the upper limiting lifetimes are hardly ever approached
(Ryan and Eberhardt, 1972; Acket and Scheer, 1971; Alferov et al., 1973;
Garbuzov et al., 1978).
132 A. G . MILNES
Mayburg (1961) has considered the direct recombination expected from

the recombination theory of Hall (1959). The expression obtained for the
probability of direct radiative recombination is
where n is the refractive index, m is the free electron mass, m, and m, are the
density of states effective masses of electrons and holes, and Eg is the band
gap. For GaAs at 300 K, with n = 3.6, m, = 0.068, m, = 0.5, and Eg = 1.43
eV, Bd, = 1.7 x lo-’’ cm3 sec-’.
The direct recombination lifetime therefore for an electron concentration
n cm-3 should be given by
td, N -
1
- 5.9 x lo-* sec for n
Bdrn - {5.9 x 10-6 sec for n
=
=
cm-3
101s cm-3 (25)
For n = 1 O I 7 and 10” cm-3, the 300 K hole mobilities tend to be about
300 and 400 cmz V-’ sec-’, respectively. Hence the hole diffusion lengths
from L, = ,/m’/’ and 0,= pkT/q may be expected to have upper
limits of 6.8 and 78 pm, respectively. Measurements of actual hole diffusion
lengths show considerable scatter from specimen to specimen (Sekela et al.,
1974). The largest values are often about one third of the L, values estimated
above. So recombination centers typically lower the minority carrier life-
times by an order of magnitude from the direct recombination lifetimes. The
scatter of diffusion-length values versus doping concentration indicates that
in the early 1970s, many laboratories were unable to control the recombina-
tion levels present, either because trace impurities were in every crystal or
because the processing used led to the formation of electrically important
native defect levels or defect-impurity complexes (Baliga, 1978; Bergmann
et al., 1976; Biswas, 1971; Bludau and Wagner, 1976; Chakravarti and
Parui, 1973; Kamm, 1976; Jastrzebski et al., 1980; Leheny et al., 1973;
Lender et al., 1979; Miki et al., 1979; Lomako et al., 1976; Matare and Wolff,
1978; Mathur and Rogers, 1977; Pilkuhn, 1976; Papaioannou and
Euthymiou, 1978; Peka et al., 1975; Seki et al., 1978; Kushiro et al., 1977;
Goodman, 1961; Kamejima et al., 1979; Chan and Sah, 1979; Mircea et al.,
1976; Hwang, 1971, 1972; Kravchenko et al., 1977).
Since the recombination processes are affected by the growth processes
used, effects in LPE, VPE, bulk (Bridgman or LEC) materials have to be
considered separately. The technique employed to determine GaAs diffusion
lengths is usually measurement of the variation of electron-beam-induced
current (EBIC) as a beam of electrons ( - 25 keV) in a scanning electron
microscope is moved across the surface of a pn junction or metal n Schottky-
barrier junction (Wittry and Kyser, 1965).At 25 keV beam energy, the beam
FIG.25. Carrier-capture cross sections versus inverse temperature for various deep levels
in GaAs. Subscript n denotes electron capture; subscript p denotes hole capture. Levels shown
for GaAs are due to Cu, Fe, Cr, 0, the E3 radiation-damage defect, and two unidentified, but
commonly occurring, levels A and B. Dashed lines show temperature dependence of cross
sections extrapolated to T = a.The value of urnpredicted by the theoretical model is indicated
on the ordinate. (After Henry and Lang, 1977.)
produces a carrier generation region roughly 0.5 pm in diameter and this

limits the accuracy of the diffusion-length measurement to about 0.2 pm
at best (Martinelli and Wang, 1973; Rao-Sahib and Wittry, 1969; Van
Opdorp et al., 1974; Vitale et al., 1979; Wu and Wittry, 1978).
Other factors that limit accuracy have been considered by Berz and
Kuiken (1976), Chi and Gatos (1977), Epifanov et al. (1976), Flat and Milnes
(1978), Hu (1978), Hu and Drowley (1978), Kuiken et al. (1976), Lastras-
Martinez (1979), Partin et al. (1979b), Rossin et al. (1979), von Roos (1978,
1979), Wight et al. (1981), and others.
A. In LPE-Grown GaAs Layers

Liquid-phase epitaxial-grown GaAs layers are relatively free of traps,
except for levels (A and B) at E, + 0.40 eV and E, + 0.71 eV, present in
concentrations of 10'4-1015 cmP3. The electron- and hole-capture cross
sections for these traps vary with temperature as shown in Fig. 25. From such
data and from other measurements, Partin et al. (1979a) arrived at the cross
sections shown in Table XIV for a specimen of an LPE GaAs (ND- NAabout
2.8 x lo'' ~ m - containing
~ ) the A and B levels in the concentrations shown.
The expression for the excess hole lifetime may be expected to be of the
form
134 A. G . MILNES
TABLE XIV
CONCENTRATIONS,
CAPTURE CROSSSECTIONS,AND EMISSION
RATES
AT 300 K"
FOR THE A AND B LEVELS
Level N , (cm- 3, 0" (cm') up (cm') ep (sec-')
A 2.4 x 1015 1.4 x 10-17 4.2 x 1 0 - 1 5 1.2 x 106

B 2.7 x l o i 5 4.2 x 2.1 x 1 0 - l ~ 1.7 x lo6
Reproduced with permission of the American Physical Society.
With hole and electron velocities up = 1.6 x lo7 cm sec-' and 0, =

4.5 x lo7 cm sec-' and the data from Table XIV, we have for level A:
u p o p p = lo5 sec-'; u,a,n = 1.8 x lo6 sec-'; e p = 1.2 x lo6 sec-'; and
vpopN,= 1.6 x lo8 sec-'. Thus, upopp is negligible, but e p is comparable to
v,a,n. Thus, z p is 1.0 x sec for level A at 300 K.
Level B is much deeper than level A, so ep is negligible. Also, for p =
2 x 10l2 cmV3,upapp = 7 x lo4 sec-', and n,a,n = 5 x lo3 sec-', and so
upopp dominates the numerator of Eq. (26). Therefore, Eq. (26) reduces to
zp = [p/(v,a,N,)] = 1.5 x sec. Thus, zp determined by level B is in-
jection-level dependent and is much larger than zp determined by level A. The
large value of z p for level B is caused by its very small electron-capture cross
section, which makes electron capture the rate-limiting recombination step.
Therefore, zp is determined only by level A and is independent of hole-
injection level. The value of Dp in p-type GaAs at NA - N,, = 3 x 10l5 is
9 cm2 sec- '. If we assume that Dp is the same in n-type GaAs with ND - NA =
3 x lo", then for L, = (Dpzp)'/2 a diffusion length of 3.0 pm is obtained. The
hole diffusion length measured by Partin et al. was 3.4 pm, so, they conclude
that the A level in LPE GaAs is the minority-carrier (hole) lifetime controller.
The A- and B-level concentrations, which tend to be closely similar,
depend on the details of the LPE process in some way that has not been well
characterized. To see these levels by DLTS, it is necessary to form a p n
junction unless optical injection of holes is used. The p region preferably
should be grown by LPE. If an LPE n region is grown and followed by a Zn
diffusion to create the p layer needed, the junction tends not to exhibit the A
and B levels. The Zn diffusion process in some way getters or destroys the
A and B levels. The physical natures of the A and B levels are still the subject
of speculation, as mentioned in Section 111 (Ashley and Beal, 1978; Casey
et al., 1973; Kressel, 1974).
Electron diffusion lengths in p-type GaAs grown from Ga solution with
Ge as the dopant are quite large. Values in the range 23-18 pm have been
observed for doping levels of 1.1 x 10l8 to 5 x 10l8 cm-3 by optical micro-
probing of beveled junctions (Ashley et al., 1973). These values correspond
TABLE XV
ELECTRON
DIFFUSION
LENGTHS
IN LPE p-TYPEGaAs :Ge"
Ga/Ge
atomic Estimated
ratio Pn D* S
in N A - N, (cm' V - ' (cm' (lo6 cm T,,,, Calc'
solution ( ~ m - ~ ) sec-') sec-l) Lc (pm) Ld (pm) sec-') (nsec) L (pm)
10 1.1 x 10'' 1500 40 5.5 4.6 - 11 6.6

100 2.0 x 10'' 3500 90 10.5 9.0 1.1 12.5 10.6
1000 6 x 10l6 6500 165 20.0 26.0 0.74 38 25
10 1.2 1019 - - 5.9 - - - -
From Ettenburg et al. (1973). Reproduced with permission of the American Physical
Society.
Calculated from estimate of electron mobility.
' Determined from laser beam scan measurement.
Determined from particle-scan measurement.
Calculated from luminescent decay time (T) measurement.
to electron lifetimes of about 5 x l o w 8sec. The nature of the recombination

process (band-band or band-recombination center) was not studied.
In similar studies of LPE-grown GaAs: Ge, the electron diffusion lengths
were examined by two direct methods: (1) He-Ne laser beam scan on beveled
pn junctions, and (2) an a-particle scan on cleaved samples; and by an in-
direct method, that of studing the electroluminescent decay time. The results
obtained for the three methods were in good agreement (see Table XV). The
diffusion length for 2 x l o r 8 cm3 doping is seen to be about 10 pm or about
half of that observed in the Ashley et al. study. The literature shows lower
diffusion lengths (3-6 pm) for p-type LPE material doped with Si, Zn, or
Cd; however, it is not certain that this must be generally true.
In p + n solar cells grown by LPE, the hole diffusion lengths for the
n-type GaAs (NDtypically 8 x 1 O I 6 cm-,) tend to be about 3 pm. After
irradiation with 10" protons (of energies 100-290 keV), the diffusion lengths
become less than 1 pm, and the recombination is attributed to deep-level
+
defects at E, 0.57 and 0.7 eV, and E, - 0.52 and 0.71 eV.
B. I n VPE-Grown GaAs Layers

In studies of the Ga-AsC1,-H, VPE growth process for GaAs, Humbert
+
et al. (1976) find levels at E, 0.4 eV and E, +
0.75 eV predominant. These
are present in about equal concentrations, and they consider these two levels
to correspond to two ionization states of a unique center involving a native
defect (perhaps a Ga vacancy), a very common impurity, or a complex of
136 A. G. MILNES
both, rather than believing that these two levels are the typical “Cu” and
EL2 levels. See also Iwasaki and Sugibuchi (1971).
One commonly used VPE growth process for GaAs involves the reaction
of ASH, and GaCl at about 700°C. The dominant deep impurity seen is then
at E, - 0.82 eV (which has been termed the EL2 level, see Section 111).In one
study, by Miller et al. (1977), the ratio of ASH, to GaCI was varied from 3/1
to 1/1 to 1/3 and the layers grown were n type and about 6 x 10’’ cm-, in
the absence of intentional n doping. The density of the deep level (Table XVI)
was found to correlate with variations of the ratio AsH,/GaCl, and the hole
lifetime varied in agreement. Since the smallest deep-level densities and the
largest lifetimes are obtained under Ga-rich conditions, the recombination
level is probably related to gallium vacancies or As,,-related defects.
Mitonneau et al. (1979) believe, however, that this is not related to the EL2
trap, which they show is not a significant hole recombination center since the
hole-capture cross section is only 2 x lo-’’ cm’. They find a hole trap
HLlO at E, + 0.83 eV to have cr = 5 x cm2 and to be the main
recombination level. See also Ettenberg et al. (1976) and Ozeki (1979).
A growth study has been performed for Zn-doped ( p = 10’’ cm-,)
VPE-grown GaAs, also with AsH,/GaCl ratios of 3/1, 1/1, and 1/3. For
electroluminescence decay studies this yielded electron lifetimes of 1.3, 3.2,
’,
and 3.4 nsec. For an electron mobility of 1500 cm2 V-’ sec- the diffusion
lengths are 2.3, 3.6, and 3.6 pm. Direct measurement of diffusion lengths by
laser probes gave 2.5, 4.8, and 4.8 pm. So, Ga-rich conditions during VPE
growth minimize recombination actions.
Vapor-phase GaAs may also be grown from the organometallic com-
pound trimethyl gallium and arsine. The dominant (electron) trap in n-type
MOVPE GaAs was again at E, - 0.82 eV, and the density of this trap varied
with As/Ga ratio, as shown in Fig. 26 (Bhattacharya et al., 1980). Hole traps
were also observed at E, + 0.36 eV and E, + 0.31 eV and ascribed to
TABLE XVI
THEEFFECTS
OF GAS-PHASE
STOICHIOMETRY
ON DEEP-LEVEL
CONCENTRATION
AND MINORITY-CARRIER
LIFETIME^
~~
Density of
Ratio of ASH, Minority-carrier dominant
to GaCl lifetime deep level
Sample in gas phase 5 (nsec) (cm-3)
1471 31 1 5 9 x 1013
1470 111 11 4 x 1013
1500 113 15 2 x 1013
From Miller et al. (1977). Reproduced with permission of the

American Physical Society.
2 4 6 8 10 12
As/Ga
FIG.26. Dependence of the concentration of the dominant 0.82 0.02 eV electron trap on
thP A o / G * mtin in thn pnit-vial I ~ v P ~ .The Ant- .
&
-
t
o ;nA;,.otnA h
., ,.lr\oeA ,.irrlnr ..,PWP Antar-
mined by Miller et al. (1977). The bars indicate the dispersion in the value of NT.(After Bhat-
tacharya er al., 1980; reproduced with permission of American Physical Society.)
unknown impurities. In other studies of M OVPE-grown GaAs, Samuelson
a concentration in the range 5 x 10I3 to 3 x 1014 cm-3 that varies in pro-

portion to P& . This result is compatible with the EL2 level being gallium
vacancy or AsGaVGa related. However, the results of Zhu et al. (1981) are
less decisive.
Considerable care is needed in processing GaAs. For instance, Ni has
been shown to degrade L, in VPE n-type GaAs when diffused inward from
the surface at moderate temperatures (600-700°C) (Fig. 27). Since epilayers
are usually only a few micrometers thick, it is conceivable that a tenth or
hundredth of a surface monolayer of Ni (1014-1013cm-2) could produce
very high concentrations of recombination centers if it all entered the layer.
The trap levels that are clearly Ni related are at E, 0.39 eV (believed to be +
Ni,,, although see the discussion in Section VI) and at E, - 0.39 eV. The
trap concentrations measured in one VPE specimen after a 7OO0C, 20-min
treatment are shown in Fig. 28 together with the estimated effective lifetimes
associated with the possible recombination paths. The dominant recombina-
tion action is seen to be through the E, - 0.39 eV level which has a capture
cross section op of about cm2. This level is present in low concentra-
tion (typically, rnid-10I3 crnp3) in as-grown VPE n-type GaAs because of
inadvertent contamination, but the concentration is greatly increased by
the Ni diffusion.
The as-grown VPE n-type GaAs (Te doped 5 x 10'' ~ m - also ~ ) con-
tained an energy level at E, - 0.75 eV (the EL2 level at E, - 0.82 eV of
other studies). This was in substantial quantity ( ~ 1 0 cmP3) ' ~ and was
considered to be a complex of a gallium vacancy with a presently unidentified
138 A. G. MILNES
I I I I I I
STRIPE 5 0 0 i N i ; H E A T E D 600°C,5 M i N .
V P E ( N , = ~ ~ I O ern-31
' ~
+ + +
t
+
0
0
0 0
0 0
0
0
0 0
0 0 0
0 24OC DATA
+ Il0"C DATA
0
0.3688
k N i 4
I I I I I I I
0
9 10 li 12 13
P O S l T i O N A L O N G T H E W A F E R (mrn)
E F F E C T O F N I C K E L I N DEGRADING Lp IN n V P E G o A s
FIG.27. Effect ofnickel in degrading L, in VPE GaAs. The dip in the diffusion length occurs
at the location of the applied Ni stripe. (After Partin e l al., 1979a; reproduced with permission
of the American Physical Society.)
impurity (or defect), namely (VGa+ X). When Ni was diffused into VPE
n-type GaAs, the concentration of the complex decreased and the E, - 0.39
eV-level concentration increased, and in equal quantity a level at E, +
0.39 eV appeared. This E, + 0.39 eV level is considered to be Ni substitu-
tional on a Ga site. (It has parameters quite different from those of the A level
in LPE material.) The model tentatively proposed for the interaction of Ni
with the E, - 0.75 eV level is that inward-diffusing interstitial Ni dissociates
the gallium complex (VGa+ X) to form Ni,, and X. These become the levels
+
E, 0.39 eV and E, - 0.39 eV, respectively. A Zn-diffused layer is found to
getter both of these levels and to improve the hole diffusion length.
An Ni layer heated to 600-700°C is not effective in altering the character-
istics of LPE layers, presumably because the absence of gallium defects
impedes its entry. It is found that surface treatments of VPE GaAs with many
other metals (for instance, Au, Cu, W, and Pt), followed by 600°C 5-min
heat treatments had the effect of increasing the hole diffusion length in the
GaAs (say, from 3 pm to 5 or 6 pm). Partin et al. (1979d) believe that these
metals form shallow surface regions and that the electric fields created have
the effect of gettering to the surface the recombination center that is con-
trolling the hole diffusion length in the n-type side of the junction. It is noted
that in these experiments Au had no effect on L, in GaAs, even when heated
at 700°C. Gold has acceptor levels in GaAs at E, + 0.405 eV and possibly at
0.05 eV. However, it is also known that an nf layer forms under a heated Au
Ec
I I I
----- -------- ------- T--------'--

-------I I
I EF"
0.39eV I
0.75 eV I I
I I
I I
I
I
II
+
I I
I I I
I I
I
I
I
1I I
I
I
I
1.43 eV
I I
I I I
0.63 eV
JI
I
I
I J III i-
0.39 eV i
a
21I 2;
b"l "b"' 'I
IP
b
1
FIG.28. Band diagram showing the concentration and value of zp for each defect level in
(Ni, Zn)-diffused VPE n GaAs; T = 700°C. The recombination rate-limiting defect parameter
= 1.4 pm; L, = 1.6 pm (SEM). (After Partin et al., 1979a;
is also indicated: L, = (Dp~p)l’*
reproduced with permission of the American Physical Society.)
contact on GaAs. This effect is related to outward diffusion of Ga and can be

suppressed by deposition of a Au-Ga alloy of eutectic composition. If it is
assumed that the hole lifetime controlling center at E, - 0.39 eV is in some
way related to Ni in the bulk of the crystals in a positively charged and some-
what mobile form, then the electric field of a pn junction would tend to getter
the centers into the p region; however, an n’n junction would cause repulsion
of Ni-related centers from the surface.
Further independent studies are needed to confirm these effects and to
explore other possible interpretations. If leakage currents in GaAs pn junc-
tions are involved in such studies, care must be taken to control or eliminate
surface-generated recombination currents (Henry et al., 1978).
C. DifSusion Lengths in Bridgman-Grown GaAs

The studies of Sekela et al. (1974) of hole diffusion lengths in Bridgman-
grown n-type GaAs resulted in the data points in Fig. 29 that are not prefixed
with a symbol V for VPE or L for LPE. Most of the specimens were in the
5 x 10’6-10’8 cm-3 doping range, and very few of these exhibited hole
diffusion lengths in excess of 2 pm. No correlation of diffusion lengths with
etch-pit counts (which tended to be in the 103-105 cm-* range) or with the
dopant type (Si, Sn, Te, Se) was observed. Deep-level transient spectroscopy
equipment was not available at the time of the studies and so the trap levels
140 A. G . MILNES
v7
L17
v 11
VlOV9 V8
L18
v4
v12
54 “g3 V2 52 L7 3532 4
25
Vl 49
ES E3V6
LlS
51 ;y 2&422
E6
L16
vs L14 L13
48 395$28~5
S634275
kZom
5780 58 65
El 66 4348
1 0 5 0 4:t:44,38 31
’8 L2
L12 ’413
17
67 62 42 I8 18
1615 63
11 64 2
E4 v5 59
29 61 Ll
65
0.2 1 1
10” 10“ 10 1 7 lOl8 10”
No- NA ( ~ r n - ~ )
FIG.29. Hole diffusion length as a function of net electron concentrationfor n-type gallium
arsenide samples. Sample numbers prefixed with “V” are VPE material, “L’ samples are
LPE, “ E ’ are unspecified epitaxial samples, and those with no prefix are from single-crystal
boules prepared by either the LEC or the horizontal Bridgman process. (After Sekela et al.,
1974; reproduced with permission of the Institute of Physics, UK.)
in the specimens were not studied. Electron diffusion lengths in Zn- and
Cd-doped ingot material were found to be larger than L,, mainly in excess of
5 pm, as might be expected from the fact that the electron mobility is much
larger than the hole mobility.
Experiments relating measured diffusion lengths to trap determinations
in bulk GaAs have not been found in quantity in the literature. Workers have
tended to concentrate instead on diffusion lengths and traps in epitaxial
layers since the device implications are clearer here. Also, suppliers of bulk
GaAs tend to provide the material only on two forms: doped in excess of
10” cm-3 or semi-insulating, and this limits the supply of material for trap
studies. However, as Tables V and VI show, electron traps at E, - 0.85 eV
and E, - 0.3 eV and a hole trap at E, + 0.35 eV have been observed. In
studies by Li (1978) and Li et al. (1980a), the following levels are reported in
their n-type GaAs (2 x 10l6 cm-3 doped):
(1) A donor-type defect center at E, - 0.38 eV with an electron-capture
cross section of about 3 x cm2 and no field dependence. It is possible
that this has association with oxygen. Okumura et al. (1976)have seen a level
in this vicinity in many n-type GaAs specimens.
(2) A donor-type defect level at E, - 0.65 eV with a thermal capture

cross section of about 3.4 x cm2 with no field dependence.
(3) A donor-type level at E, - 0.78 eV with electron-capture cross
section of 1.4 x cm2 that is not field dependent. This is presumably the
EL2 center of Table (V), although the capture cross section is smaller than in
some previous studies.
In recent work with Bridgman-grown material, the EL2 trap concentra-
tion has been found to be inhibited by shallow donors (Si or Te) and enhanced
by shallow-acceptor doping (Zn). This is compatible with EL2 being related
to the AsGa antisite defect and perhaps AsGaVAsor AsGaVGa.Addition of
oxygen to the melt during the growth decreases the concentration of Si
shallow donors and thus it increases the concentration of EL2. Oxygen itself,
however, is not considered to be a part of EL2 (Lagowski et al., 1982).
Systematic studies seem desirable of Bridgman- and LEC-grown n- and
p-type GaAs with dopings in the range 2 x 1016-10'8 cm-3 to establish the
nature of the dominant traps, their roles in minority carrier recombination,
and the extent to which they match up with traps seen in typical VPE, MBE,
and LPE expitaxial material. Studies of GaP have been made by Hamilton
et al. (1979) and Scheffler et al. (1981).
XIV. CONCLUDING
DISCUSSION
Experimental energy levels in the GaAs band gap have been reviewed. A
number of factors make it difficult to attribute particular crystal complexes to
the observed energy levels. Since GaAs is a compound semiconductor, the
number of defects possible is larger than for Si, where antisite defects do not
have to be considered. The energy conditions in the crystal (although only
partially known) seem to favor complexing of defects, and the formation of
antisite defects. Another problem is that the surface of GaAs reacts with
many metals, and this can be significant even though the metal is present in a
submonolayer quantity. Defects may propagate into the first few microns of
the crystal as a result of such reactions at temperatures of a few hundred
degrees. Alternatively, gettering to the surface of important defect levels may
occur in the course of an experiment. Since experiments involving depletion-
layer capacitance transients interrogate regions close to the surface, the
potential for error and misinterpretation of results is substantial. A further
difficulty is that electron paramagnetic spin resonance experiments can be
applied only to a limited set of impurities and defects in GaAs. Experiments
for deciphering successfully the physical or chemical nature of defect levels
in GaAs tend therefore to be hard to devise and execute successfully.
Some progress, however, has been made. Certain chemical species have
been identified with particular levels. This is particularly true for the transi-
tion elements. The mid-band-gap level (at E, - 0.75 to 0.83 eV) seen in most
142 A. G. MILNES
GaAs (except LPE layers) is still unidentified. Evidence exists that it is not
oxygen related but that it involves an antisite defect AsG, or a gallium
vacancy. Candidates for this level are the complexes AsGaVAs, AsGaVGa,or
AsGaVGaAsGa,but this identification is very tentative. The evidence that the
level at E , - 0.33 eV, produced by irradiation, is the gallium vacancy V, has
recently been challenged and the suggestion made that arsenic vacancies are
involved. Also recently an acceptor at E, + 78 meV has been ascribed to the
GaAsantisite defect. With respect to oxygen, an accumulation of measure-
ments seem to indicate that the 0.64 eV luminescence band is involved and
that the oxygen may also be associated with an energy level at about E, -
0.40 eV. In VPE studies an energy level at about this position, E, - 0.39 eV,
is a strong recombination center. These levels need further study, as do the
E, + 0.41 eV and E, + 0.71 eV A and B levels in LPE GaAs. The one
prediction that may be confidently made is that it will be many years still
before all of the defect energy levels in GaAs are identified and understood.
ACKNOWLEDGMENTS
The support of the Solid State Division, Westinghouse Research and Development Center,
Churchill, Pennsylvania, during the preparation of this article is gratefully acknowledged.
REFERENCES
Abdalla, M. I. el al. (1975). Transport properties of O f implanted in GaAs. Conf. Ser.-Inst.
Phys. 24, 341.
Abrahams, M. S . et al. (1974). Interdependence of strain, precipitation, and dislocation forma-
tion in epitaxial Se-doped GaAs. J . Appl. Phys. 45, 3277.
Acket, G. A,, and Scheer, J. J. (1971). Relaxation oscillations and recombination in epitaxial
n-type gallium arsenide. Solid-state Electron. 14, 167.
Akita, K. et al. (1971). Electrical properties of n-type gallium arsenide at high temperatures.
Jpn. J . Appl. Phys. 10, 392.
Akkerman, Z. L. et al. (1976). Infrared absorption spectra of oxygen-doped gallium arsenide.
Sou. Phys.-Semicond. (Engl. Transl.) 10(5), 590.
Aleksandrova, G. A. et al. (1973). Dominant contribution of oxygen to the compensation of
high resistivity GaAs films. Sou. Phys.-Semicond. (Engl. Transl.)6(7), 1170.
Aleksandrova, G. A. et al. (1976). Investigation of deep trapping centers in epitaxial GaAs:
Cr:W by the method of thermally stimulated discharge of a capacitor. Sou. Phys.-Semi-
cond. (Engl. Transl.) 10, 1162.
Alferov, Zh. I. et al. (1973). Radiative recombination in epitaxial compensated gallium arsenide.
Sou. Phys.-Semicond. (Engl. Transl.) 6, 1718.
Allen, G. A. (1968). The activation energies of chromium, iron and nickel in gallium arsenide.
Br. J . Appl. Phys. [2] 1, 593.
Allen, J. W. (1968). Energy levels of nitrogen-nitrogen pairs in gallium phosphide. J . Phys. C
1, 1136.
Allred, W. et al. (1968). Site distribution of silicon in silicon-doped gallium arsenide. Int. Symp.
Gallium Arsenide, 1968 [Conf. Ser.-Inst. Phys.].
Amato, M. A. et al. (1980). Photoconductivity of epitaxial GaAs: Cr. In “Semi-Insulating

111-V Materials” (G. J. Rees, ed.), p. 249. Shiva Publ. Ltd., Orpington Kent, U.K.
Anderson, C. L., and Dunlap, H. L. (1979). Low temperature annealing behavior of GaAs
implanted with Be. Appl. Phys. Lett. 35(2), 178.
Andrews, A. M., and Holonyak, N., Jr. (1972). Properties of n type Ge-doped epitaxial GaAs
layers grown from Au-rich melts. Solid-State Electron. 15, 601.
Andrianov, D. G. et al. (1977a). Magnetic and optical properties of Ni3+ and Coz+ ions of
3d7 configuration in gallium arsenide. Sou. Phys.-Semicond. (Engl. Transl.) 11,426.
Andrianov, D. G. et al. (1977b). Investigation of the neutral state of cobalt impurities in gallium
arsenide. Sou. Phys.-Semicond. (Engl. Transl.) 11, 858.
Antell, G. R. (1965). The diffusion of silicon in gallium arsenide. Solid-State Electron. 8,943.
Aoki, K. et al. (1976). Depth distribution of defects in Mg-ion and Cd-ion implanted GaAs.
Jpn. J. Appl. Phys. 15,405.
Ashen, D. J. et al. (1975). The incorporation and characterization of acceptors in epitaxial
GaAs. J. Phys. Chem. Solids 36, 1041.
Ashirov, T. K. et al. (1978). Photoionization spectrum of a deep impurity center in the presence
of a weak electron-phonon interaction in GaAs: Cu at low temperatures. Sou. Phys.-
Semicond. (Engl. Transl.) 12, 1036.
Ashley, K. L., and Beal, S. W. (1978). Diffusion lengths in amphoteric GaAs hereoface solar
cells. Appl. Phys. Lett. 32, 375.
Ashley, K. L. et al. (1973). Electron diffusion length in solution-grown GaAs:Ge. Appl. Phys.
Lett. 22, 23.
Bachelet, G. B. et al. (1981). Self-consistent calculations of the electronic structure for ideal
Ga and As vacancies in GaAs. Lund. In?. Con$ Deep-Leuel Impurities Semicond., 3rd,
1981 Abstracts, p. 31.
Badawi, M. H. et al. (1979). A comparison of laser annealed donor and acceptor ions implanted
into GaAs. Muter. Res. SOC.Meet., 1979.
Bahraman, A., and Oldham, W. G. (1972). Role of copper in the degradation of GaAs electro-
luminescent diodes. J. Appl. Phys. 43,2383.
Baliga, B. J. (1978). Recombination level selection criteria for lifetime reduction in integrated
circuits. Solid-State Electron. 21, 1033.
Baliga, B. J., and Ghandhi, S. K. (1974). Lateral diffusion of zinc and tin in gallium arsenide.
IEEE Trans. Electron Devices ED-21,410.
Bazhenov, V. K., and Solov’ev, N. N. (1972). Deep transition-metal centers in gallium arsenide.
Bazhenov, V. K. et al. (1974). Optical Absorption in GaAs:Ni. Sou. Phys.-Semicond. (Engl.
Transl.) 7 , 1067.
Bekmuratov, M. F., and Murygin, V. I. (1973). Investigation of some properties of titanium
and cobalt-compensated gallium arsenide crystals and diode structures. Sou. Phys. Semi-
cond. (Engl. Transl.) 7(1), 55.
Berg, N. J., and Lieberman, A. G . (1975). The effects of radiation-induced displacement damage
on impurity conduction in gallium arsenide. J. Appl. Phys. 46,3475.
Bergmann, Ya. V. et al. (1976). Diffusion length of holes in lightly doped gallium arsenide.
Sou. Phys.-Semicond. (Engl. Transl.) 10(8), 933.
Berz, F., and Kuiken, H. K. (1976). Theory of lifetime measurements with the scanning electron
microscope : Steady state. Solid-State Electron. 19,437.
Bhattacharya, P. K. et al. (1980). The trend of deep states in organometallic vapor-phase
epitaxial GaAs with varying As/Ga ratios. Appl. Phys. Lett. 36(4), 304.
Bimberg, D. (1978). Anomaly of the linear and quadratic Zeeman effect of an effective-mass
acceptor: C in GaAs. Phys. Rev. B 18, 1794.
144 A. G . MILNES
Bimberg, D., and Baldereschi, A. (1979). Non-hydrogenic magnetic behaviour of impurities

in semiconductors with degenerate bands: GaAs: C and GaAs: Sn. Conf. Ser.-Inst. Phys.
43,403.
Bimberg, D. et al. (1982). Defects in 111-V semiconductors. In “Landolt-Bornstein” (0.
Madelung, ed.), New Ser., Group 111, Vol. 17a. Springer-Verlag, Berlin and New York.
Bishop, S. G. (1981). Excitation mechanisms for transition-metal impurity luminescence in
11-VI and 111-V semiconductors. Lund. Int. Conf. Deep-level Impurities Semicond., 3rd,
1981 Abstracts, p. 62.
Bishop, S. G. et al. (1977). Effects of implantant depth distribution on photoluminescence
spectra in Be-implanted GaAs. Appl. Phys. Lett. 31,845.
Biswas, S. N. et al. (1971). Doping dependence of the internal quantum efficiency of spontaneous
electron-hole recombination in n-type samples of GaAs excited by a 30 kV electron beam
at 300°K. Phys. Status Solidi A 7,233.
Blackmore, G. W. et al. (1976). Concentrations of carbon and oxygen in indium phosphide and
gallium arsenide crystals grown by the LEC technique. J. Electron. Muter. 5(4), 401.
Blakemore, J. S.(1982a). Intrinsic density ni(T) for GaAs, deduced from bandgap and effective
mass parameters, and derived independently from Cr acceptor capture and emission
coefficients. J. Appl. Phys. 53(1), 520.
Blakemore, J. S. (1982b). Semiconducting and other major properties of gallium arsenide.
J. Appl. Phys. 53(10), R123.
Blashku, A. I. et al. (1971). Interactions between zinc and tellurium in gallium arsenide. Sou.
Phys.-Semicond. (Engl. Transl.) 5, 664.
Blatte, M., and Willmann, F. (1971). Photo-ionization of deep acceptors in gallium arsenide.
Opt. Commun. 4,178.
Blatte, M. et al. (1970). Photoluminescence of Ag-doped p-type GaAs. Solid State Commun.
8, 1265.
Bludau, W., and Wagner, E. (1976). Carrier lifetime controlled by capture into deep shallow
centers in GaAs. Appl. Phys. Lett. 29,204.
Boborykina, E. N. et al. (1978). Electron-phonon interaction in optical transitions between
a deep impurity level and allowed bands in copper-doped gallium arsenide. Sou. Phys.-
Semicond. (Engl. Transl.) 12(7), 768.
Bois, D., and Pinard, P. (1974). Optical absorption on localized levels in gallium arsenide.
Phys. Rev. B9,4171.
Bois, D., and Vincent, G. (1977). Mise en evidence d’un &tatmetastable du centre associe B
l’oxygtne dans GaAs. J. Phys. (Paris) 38, L-351.
Bois, D. et al. (1978). Optical properties of the “oxygen” centre in GaAs. Conf. Ser.-Inst.
Phys. 43,295.
Boltaks, B. I., and Shishyanu, F. S. (1964). Diffusion and solubility of silver in gallium arsenide.
Sou. Phys.-Solid State (Engl. Transl.) 5, 1680.
Boltaks, B. I. et al. (1971). Influence of the polarity of the surfaces of A”’BV compounds on the
diffusion of impurities. Sou. Phys.-Solid State (Engl. Transl.) 12, 2631.
Boltaks, B. I. et al. (1972). Effect of light on the movement of a p-n junction in GaAs. Sou.
Phys.-Solid State (Engl. Transl.) 13, 2877.
Boltaks, B. I. et al. (1975). Diffusion and solubility of Fe in GaAs. Inorg. Muter. 11,292.
Borisova, L. A. et al. (1977). Solubility of oxygen in GaAs. Neorg. Muter. 13,908.
Brehm, G . E., and Pearson, G. L. (1972). Gamma-radiation damage in epitaxial gallium
arsenide. J. Appl. Phys. 43, 568.
Brillson, L. J. (1982). The structure and properties of metal-semiconductor interfaces. Surf.

Sci. Rep. 3, 123.
Brown, W. J., and Blakemore, J. S. (1972). Transport and photoelectrical properties of gallium
arsenide containing deep acceptors. J . Appl. Phys. 43,2242.
Brozel, M. R. et al. (1978). Carbon, oxygen and silicon impurities in gallium arsenide. J. Phys.
D 11, 1331.
Bublik, V. T. et a[.( 1973). Nature of the point defects in GaAs single crystals as a function of the
composition of the melt during growth. Sou. Phys.-Crystallogr. 1% 218.
Bykovskii, V. A. et al. (1975). Luminescence of gallium arsenide due to intracenter electron
transitions in iron ions. Sou. Phys.-Semicond. (Engl. Transl.) 9, 1204.
Casey, H. C., Jr., and Pearson, G. L. (1964). Rare earths in covalent semiconductors: The
thulium-gallium arsenide system. J . Appl. Phys. 35, 3401.
Casey, H. C., Jr. et al. (1971). Influence of surface band bending on the incorporation of im-
purities in semiconductors: Te in GaAs. J . Phys. Chem. Solids 32, 571.
Casey, H. C., Jr. et al. (1973). Variation of minority-carrier diffusion length with carrier con-
centration in GaAs liquid-phase epitaxial layers. J . Appl. Phys. 44, 1281.
Chakravarti, A. N., and Parui, D. P. (1973). On the diffusivity-mobility ratio and the lifetime
ofthe minority carriers in degenerate specimens of GaAs. Indian J . Pure Appl. Phys. 11,584.
Chan, P. C. H., and Sah, C. T. (1979). Computer-aided study of steady-state carrier lifetimes
under arbitrary injection conditions. Solid-state Electron. 22,921.
Chang, C. C. et al. (1975). Interdiffusions in thin film Au on Pt on GaAs (100) studied with
Auger spectroscopy. J . Appl. Phys. 46,4237.
Chang, L. L. et al. (1971). Vacancy association of defects in annealed GaAs. Appl. Phys. Lett.
19, 143.
Chantre, A., and Bois, D. (1980). Deep level optical spectroscopy in GaAs. J . Phys. SOC.Jpn.
49, Suppl. A, 247.
Chantre, A. et al. (1981). Deep level optical spectroscopy in GaAs. Phys. Rev. B 23,5335.
Chapman, R. A., and Hutchinson, W. G. (1967). Photoexcitation and photoionization of
neutral manganese acceptors in gallium arsenide. Phys. Reu. Lett. 18,443.
Cheng, L. J. et al. (1979). Positron lifetimes in GaAs. J. Appl. Phys. 50(4), 2962.
Chi, J.-Y., and Gatos, H. C. (1977). Nondestructive determination of the depth of planar p-n
junctions by scanning electron microscopy. IEEE Trans. Electron Devices ED-24, 1366.
Chiang, S. Y., and Pearson, G. L. (1975). Properties of vacancy defects in GaAs single crystals.
J . Appl. Phys. 46,2986.
Chiang, S. Y., and Pearson, G. L. (1975b). Photoluminescence studies ofvacancies and vacancy-
impurity complexes in annealed GaAs. J . Lumin. 10, 313.
Chiao, S.-H. S. et al. (1978). Photo-electronic properties of LPE GaAs:Cu. J . Appl. Phys.
49, 261.
Cho, A. Y., and Arthur, J. R. (1975). Molecular beam epitaxy. Prog. Solid State Chem. 10,
157.
Clerjaud, B. et al. (1980). Cr2+(3d4)Absorption in GaAs. Solid State Commun. 33,983.
Coates, R., and Mitchell, E. W. J. (1975). The optical and electrical effects ofhigh concentrations
of defects in irradiated crystalline gallium arsenide. Adu. Phys. (GB) 24, 593.
Comas, J., and Plew, L. (1976). Beryllium and sulfur ion-implanted profiles in GaAs. J . Electron.
Muter. 5,209.
Cooke, C. J., and Hume-Rothery, W. (1966). The equilibrium diagram of the system gold-
gallium. J . Less-Common Metals 10,42.
146 A. G. MILNES
Crawford, J. H., Jr., and SliRin, L. M. (1975). “Point Defects in Solids,” Vol. 2. Plenum, New
York.
Damestani, A., and Forbes, L. (1981). Characterization of the iron and nickel impurities in
GaAs,,,P,,,. J . Electron. Muter. 10, 879.
Davies, D. E. (1976). Implantation in GaAs and other 111-V compounds. Proc. Conj. Sci. Ind.
Appl. Small Accelerators, 4th, 1976 p. 420.
Degen, P. L. (1973). Ion implantation doping of compound semiconductors. Phys. Status Solidi
A 16, 9.
Demidov, E. S. (1977). Impurity states of iron group ions in gallium arsenide and silicon. Sou.
Phys.-Solid State (Engl. Transl.) 19, 100.
Deveaud, B., and Martinez, G . (1981). The Luminescence of Cr2+(3d4)in GaAs induced by
hydrostatic pressure. Lund In?. Conj: Deep-Level Impurities Semicond., 3rd, 1981 Abstracts,
p. 80.
Deveaud, B. et al. (1975). Photovoltaic effects on O f implanted GaAs. Solid State Commun.
17, 1253.
DiLorenzo, J. V. et al. (1971). Analysis of impurity distribution in homoepitaxial n on n + films
of GaAs which contain high-resistivity regions. J . Appl. Phys. 42, 729.
Dobson, P. A. et al. (1978). Superdilation and defects in tellurium-doped gallium arsenide.
Conf. Ser.-Inst. Phys. 45, 163.
Donnelly, J. P. (1977). Ion Implantation in GaAs. Gallium arsenide and related compounds.
Conf. Ser.-Inst. Phys. 33b, 166.
Donnelly, J. P. et al. (1975). Silicon- and selenium-ion-implanted GaAs reproducibly annealed
at temperatures up to 950°C. Appl. Phys. Lett. 27,41.
Driscoll, C. M. H., and Willoughby, A. F. W. (1973). Study of point defects in gallium arsenide
using precision lattice parameter measurements. Proc. Inti. Conf. Radiat. Damage Defects
Semicond., 1972 p. 377.
Driscoll, C. M. H. et al. (1974). A photoluminescence study of defects in non-stoichiometric
gallium arsenide using concurrent electrical and structural characterization. J . Muter. Sci.
9, 1615.
Dzhafarov, T. D. (1971). Interaction between copper and imperfections in gallium arsenide.
Sou. Phys.-Solid State (Engl. Transl.) 12,2259.
Eaves, L. et al. (1980). Photoconductivity, photoluminescence and Zeeman spectroscopy of
Cr in GaAs, GaP and InP. In “Semi-Insulating 111-V Materials” (G.J. Rees, ed.) Shiva
Publ. Ltd., Orpington, Kent, U.K.
Eaves, L. et al. (1981a). High magnetic field Zeeman splitting and anisotropy of Cr-related
photoluminescence in semi-insulating GaAs and Gap. Springer Ser. Solid-State Sci. 24.
Eaves, L. et al. (1981b). The observation of a sharp peak in the deep-level photoconductivity
spectrum of GaAs (Cr) due to the Cr2+(5T2-5E)intracentre transition. J . Phys. C 14, L693.
Edmond, J. T. (1959). Behavior of some impurities in 111-V compounds. Phys. SOC.73(4), 622.
Eisen, F. H. (1971). 111-V compound review. Radiat. Efl. 9,235.
Eisen, F. H., and Welch, B. M. (1977). Tellurium implantation in GaAs. Solid-state Electron.
20, 219.
Elliott, C. R. et al. (1978). Ion Implantation damage in GaAs: A TEM study of the variation
with ion species and stoichiometry. Solid-State Electron. 21, 859.
Elliott, K. R. et al. (1982). Characterization of the 78 meV acceptor in GaAs. Electron. Muter.
Conf. Abstracts, p. 29.
Emori, H. et al. (1981). Determination of oxygen in GaAs by 3He activation analysis. Int.
Symp. Gallium Arsenide Relat. Compd., 1981. Con$ Ser.-Inst. Phys. 63.
Ennen, H. et al. (1980). Intra d-shell excitations of cobalt and nickel acceptors in gallium
arsenide. Solid State Commun. 34,603.
Epifanov, M. S. et al. (1976). Photon transfer of the non-equilibrium carriers excitation in

gallium arsenide. Phys. Semicond., Proc. Int. Con$, 13th, 1976 p. 275.
Ettenberg, M. et al. (1973). Minority carrier diffusion length and recombination lifetime in
GaAs:Ge prepared by liquid-phase epitaxy. J . Appl. Phys. 44, 827.
Ettenberg, M. et al. (1976). Effect of gas-phase stoichiometry on the minority-carrier diffusion
length in vapor-grown GaAs. Appl. Phys. Lett. 29(3), 141.
Ewan, J., Kamath, G . S., and Knechtli, R. C. (1975). Large area GaAlAs/GaAs solar cell devel-
opment. Conf. Rec. IEEE Photovoltaic Spec. Conf. 11,409.
Farmer, J. W., and Look, D. C. (1979). Type conversion in electron-irradiated GaAs. J . Appl.
Phys. 50(4), 2970.
Farmer, J. W., and Look, D. C. (1980). Electron-irradiation defects in n-type GaAs. Phys.
Rev. B 21,3389.
Favennec, P. N. (1976). Semi-insulating layers of GaAs by oxygen implantation. J . Appl.
Phys. 47,2532.
Favennec, P. N. et al. (1978). On the ion implantation of the group VI impurities into GaAs.
Solid-State Electron. 21, 705.
Finstad, T. G. et al. (1974). Annealing characteristics and lattice site location of 40 keV Sn
implanted GaAs. Phys. Status Solidi A 25, 515.
Fistul’, V. I. (1974). Properties of iron-doped semi-insulating gallium arsenide. Sou. Phys.-
Semicond. (Engl. Transl.) 8,311.
Fistul’, V. I., and Agaev, A. M. (1966). Determination of the deep-lying levels of Fe, Ni, and
Co in gallium arsenide. Sou. Phys.-Solid State (Engl. Transl.) 7, 2975.
Flat, A., and Milnes, A. G . (1978). Interpretation of scanning electron microscope measure-
ments of minority carrier diffusion lengths in semiconductors. Int. J . Electron. 44, 629.
Fujimoto, M. (1970). Physico-chemical properties of zinc and of cadmium in gallium arsenide.
Rev. Phys. Chem. Jpn. 40,
Fuller, C. S., and Wolfstirn, K. B. (1962). Diffusion, solubility, and electrical behavior of Li
in GaAs single crystals. J. Appl. Phys. 33, 2507.
Gamo, K. et al. (1977). Selenium implantation into GaAs. Solid-State Electron. 20,213.
Ganapol’skii, E. M. e f al. (1974). Mechanism of the compensation of iron-doped semi-insulating
gallium arsenide. Sou. Phys.-Semicond. (Engl. Transl.) 7, 1099.
Garbuzov, D. Z. et al. (1978). Temperature dependences of the intensities and times of radi-
ative transitions in a direct gap semiconductor of the GaAs type. Sou. Phys.-Semicond.
Engl. Transl.) 12(7), 809.
Gibbon, C. F., and Ketchow, D. R. (1971). Diffusion of tin into GaAs from doped SiO, film
sources. J . Electrochem. Soc. 118,915.
Gippius, A. A,, and Ushakov, V. V. (1981). Luminescence of transition elements implanted
in GaAs and Gap. Lund Int. Con$ Deep-Level Impurities Semicond., 3rd, 1981 Abstracts,
p. 66.
Glinchuk, K. D. et al. (1977). The scheme of electronic transitions with 0.94, 1.0, 1.2, and
1.3 eV radiative centers in n-GaAs. Phys. Status Solidi A 41, 659.
Goodman, A. M. (1961). A method for the measurement of short minority carrier diffusion
lengths in semiconductors. J . Appl. Phys. 32,2550.
Goodwin, A. R. et al. (1969). Liquid phase epitaxial growth of gallium arsenide. Conf. Ser.-
Inst. Phys. 7, 36.
Greene, P. D. (1971). Behavior of selenium and tellurium dopants in the solution growth of
gallium arsenide. Solid State Commun. 9, 1299.
Grimm, A. (1972). Infrared absorption by point defects in gallium arsenide. J. Phys. C 5,
1883.
148 A. G. MILNES
Guinolds, H. R., and Robinson, G. Y. (1980). Pd/Ge contacts to n-type GaAs. Solid-State
Electron. 23,973.
Gutkin, A. A. et al. (1972). Photoconductivity of semi-insulating chromium doped gallium
arsenide subjected to strong electric fields. Sov. Phys.-Semicond. (Engl. Transl.) 6,928.
Gutkin, A. A. et al. (1974). Electroabsorption of nickel-doped gallium arsenide. Sou. Phys.-
Haisty, R. W. (1965). On the temperature dependence of the iron acceptor level in GaAs. Appl.
Phys. Lett. 7, 208.
Haisty, R. W., and Cronin, G. R. (1964). A comparison of doping effects of transition elements
in gallium arsenide. Phys. Semicond., Proc. Int. Conf., 7th, 1964 p. 1161.
Hall, R. N. (1959). Recombination processes in semiconductors. Proc. Inst. Electr. Eng., Part B
106, 823.
Hamilton, B. et al. (1979). Deep-state controlled minority-carrier lifetime in n-type gallium
phosphide. J. Appl. Phys. 50(10), 6373.
Hasegawa, F. (1974). Different diffusion behavior of copper in epitaxial and in bulk GaAs.
J . Appl. Phys. 45, 1944.
Hasegawa, H. et al. (1977). Iron-doped liquid-phase-epitaxial GaAs layers. Jpn. J. Appl. Phys.
16, 1251.
Hatch, C. B., and Ridley, B. K. (1979). HF laser excited photoconductivity associated with the
0.41 eV level. Private communication.
Hemment, P.L. F. (1976). Ion implantation into compound semiconductors. Conf. Ser.-Inst.
Phys. 28,44.
Hennel, A. M., and Martinez, G. (1982). Charge transfer CrZ+(d4)-Cr”(d5) induced by
hydrostatic pressure in chromium doped GaAs. Phys. Rev. 13,24.
Hennel, A. M. et al. (1980). Activation of Cr’+(3d5) level in GaAs:Cd induced by hydrostatic
pressure. In “Semi-Insulating 111-V Materials” (G. J. Rees, ed.), p. 228. Shiva Publ. Ltd.,
Opington, Kent, U.K.
Hennel, A. M. et al. (1981a). The double acceptor state Cr” of Cr in GaAs. Lund. Int. Con5
Deep-level Impurities Semicond., 3rd, 1981 Abstracts, p. 79a.
Hennel, A. M. et al. (1981b). Investigation of the absorption of CrZ+(3d4)in GaAs. Phys.
Rev. B 23,3933.
Henry, C. H., and Lang, D. V. (1977). Non-radiative capture and recombination by multi-
phonon emission in GaAs and Gap. Phys. Rev. B 15,989.
Henry, C. H. et al. (1978). The effect of surface recombination on current in Al,Ga,-,As
heterojunctions. J. Appl. Phys. 49, 3530.
Hiesinger, P. (1976). Hall-effect levels in Ag- and Au-doped p-type GaAs. Phys. Status Solidi
A 33, K39.
Hilsum, C., and Rose-Innes, A. C. (1961). “Semiconducting 111-V Compounds.” Pergamon,
New York.
Hjalmarson, H. P. et al. (1980). Theory of substitutional deep traps in covalent semiconductors.
Phys. Rev. Lett. 44,810.
Hu, C. (1978). Determination of non-uniform diffusion length and electric fields in semi-
conductors. IEEE Trans. Electron Devices ED-25, 822.
Hu, C., and Drowley, C. (1978). Determination of diffusion length and surface recombination
velocity by light excitation. Solid-state Electron. 21,965.
Huber, A. M. et al. (1979). Direct evidence of the non-assignment of oxygen of the main electron
trap in GaAs. J. Appl. Phys. 50(6), 4022.
Humbert, A. et al. (1976). Influence of the growth conditions on the incorporation of deep levels
in VPE GaAs. J. Appl. Phys. 47,4137.
Hunsperger, R. G., and Marsh, 0. J. (1970). Electrical properties of Cd, Zn, and S ion-im-
planted layers in GaAs. Radiat. Efl. 6,263.
Hunsperger, R. G. et al. (1972). Mg and Be ion implanted GaAs. J. Appl. Phys. 43, 1318.
Hurle, D . T. J. (1977). Impurity-point defect complexes in GaAs. Conj: Ser.-Inst. Phys.

33a, 113.
Hurle, D. T. J. (1979a). Revised calculation’ of point defect equilibrium and non-stoichiometry
in gallium arsenide. J . Phys. Chem. Solids 40, 613.
Hurle, D. T. J. (1979b). Solubility and point defect-dopant interactions in GaAs-I. Tellurium
doping. J . Phys. Chem. Solids 40, 627.
Hurle, D. T. J . (1979~).Solubility and point defect-doping interactions in GaAs-11. Tin doping.
J . Phys. Chem. Solids 40, 639.
Hurle, D . T. J . (1979d). Solubility and point defect-dopant interactions in GaAs-111. Germanium
doping. J . Phys. Chem. Solids 40, 641.
Hutchinson, P. W., and Bastow, B. D. (1974). The nature of inclusions in heavily tellurium-
doped gallium arsenide. J . Muter. Sci. 9, 1483.
Hutchinson, P. W . , and Dobson, P. S. (1975). Interstitial condensation in n + GaAs. J . Muter.
Sci. 10, 1636.
Huth, F. (1970). Deep acceptor levels in Cd-doped GaAs. Phys. Status Solidi A 1, K37.
Hwang, C. J . (1971). Doping dependence of hole lifetime in n-type GaAs. J . Appl. Phys. 42,
4408.
Hwang, C. J . (1972). Quantum efficiency and radiative lifetime of the band-to-band recombina-
tion in heavily doped n-type GaAs. Phys. Rev. B 6, 1355.
+
Ikoma, T., and Takikawa, M. (1981). Origin of E, 0.51 eV level in GaAs. Lund. Int. Con/.
Deep-level Impurities Semicond., 3rd, 1981 Abstracts, p. 64.
Ikoma, T . et al. (1981). Change of deep levels in Fe-, 0- and non-doped liquid phase epitaxial
GaAs after electron-beam irradiation and annealing. Int. Symp. Gallium Arsenide Relat.
Comp., 1981 [Conf: Ser.-Inst. Phys. 631.
Ilic, G . e f al. (1973). Cadmium implanted gallium arsenide-A study of residual bombardment
disorder and atom location. Radiat. EJ: 18,47.
ll’in, N. P., and Masterov, V. F. (1976). Electronic-statesof neutral vacancies in gallium arsenide
and phosphide. Sou. Phys.-Semicond. (Engl. Transl.) 10,496.
Il’in, N. P., and Masterov, V. F. (1977). Electron structure of deep centers in gallium arsenide
doped with iron-group transition elements. Sou. Phys.-Semicond. (Engl. Transl.) 11, 864.
lwasaki, H., and Sugibuchi, K . (1971). Origin of high resistance at the epitaxial layer-substrate
interface of GaAs grown by vapor epitaxy. Appl. Phys. Lett. 18,420.
Jain, G. C., Sadana, D. K . , and Das, B. K. (1976). Phosphorus diffusion in gallium arsenide.
Solid State Electron. 19,731.
Jaros, M. (1971). Ground states and photoionization of iron group transition metal impurities
in semiconductors. J . Phys. C 4, 2979.
Jaros, M . (1982). “Deep Levels in Semiconductors.” Adam Hilger, Bristol, U.K.
Jastrzebski, L. et a / . (1975). Application of scanning electron microscopy to determination of
surface recombination velocity: GaAs. Appl. Phys. Leri. 27, 537.
Jesper, T. et al. (1980). Optical and thermal properties of Cr in GaAs. In “Semi-Insulating
111-V Materials” (G. J. Rees, ed.), p. 233. Shiva Publ. Ltd., Orpington, Kent, U.K.
Kachurin, G. A. et al. (1975). Enhancement and retardment of diffusion by defects in GaAs.
Conj. Ser.-Inst. Phys. 23, 461.
Kadhim, M. A. H., and Tuck, B. (1972). Isoconcentration diffusion of zinc in GaAs at IOOO’C.
J , Muter. Sci. 7, 68.
Kalma, A. H., and Berger, R. A. (1972). Electrical properties of electron-irradiated GaAs.
IEEE Trans. Nucl. Sci.NS-19, Part 2, 209.
Kamejima, T. et a/. (1979). Transmission electron microscope study of microdefects in disloca-
tion free GaAs and InP crystals. J . Appl. Phys. 50(5), 3312.
Kaminska, M. et al. (1981). Oxygen induced level in GaAs. Lund Int. Conf. Deep-levellmpuritirs
Semicond., 3rd, I981 Abstracts, p. 40.
Kamm, J . D. (1976). A method for investigation of fluctuations in doping concentrations and
150 A. G . MILNES
minority-carrier diffusion length in semiconductors by scanning electron microscope. Solid-

State Electron. 19, 921.
Karelina, T. A. et al. (1974). Diffusion of tellurium in GaAs. Inorg. Muter. 10, 194.
Kaufmann, U., and Ennen, H. (1981). Donor-acceptor pairs in GaF’ and GaAs involving the
deep nickel acceptor. Lund Int. Con$ Deep-level Impurities Semicond., 3rd, 1981 Abstracts,
p. 69.
Kaufmann, U., and Schneider, J. (1980). Optical and EST spectroscopy of deep defects in 111-V
semiconductors. Festkorperprohleme, 20, 87. [ “Festkorperprobleme XX Advances in
Solid State Physics” (Treusch, ed.). Vieweg, Braunschweig/Wiesbaden.]
Kaufmann, U., and Schneider, J. (1982). Point defects in Gap, GaAs and InP. Ado. Electron.
Electron Phys. 60, 81.
Kendall, D. L. (1968). Diffusions. In “Semiconductors and Semimetals” (R. K. Willardson
and A. C. Beer, eds.), Vol. 4, p. 163. Academic Press, New York.
Kennedy, T. A. et al. (1981). EPR of defects in electron-irradiated GaAs:Zn. Bull. Am. Phys.
Soc [2] 26,255.
Kim, H. B. et al. (1974). Analysis of metal-GaAs Schottky barrier diodes by secondary ion mass
spectrometry. Conf. Ser.-Inst. Phys. 24, 307.
Kitahara, K. et al. (1976). Photo-ionization cross-section measurements of deep levels in iron
doped GaAs. Jpn. J. Appl. Phys. 15(ii), 2275.
Kladis, D. L.,and Euthymiou, P. C. (1972). Carrier lietimes in low-resistivity GaAs upon
electron bombardment and annealing. Phys. Status Solidi A 10,479.
Klein, P. B., and Weiser, K. (1981). Time dependent photoluminescence studies of deep centers
in 111-V semiconductors. Lund. In?. ConJ Deep-level Impurities Semicond., 3rd, 1981
Abstracts, p. 71.
Klein, P. B. et al. (1980). Thermal Conversion of GaAs. J . Appl. Phys. 51(9).
Kolesov, B. A., et al. (1975). Nature of the 1.24-1.26 eV luminescence band of gallium arsenide
doped with copper and oxygen. Sou. Phys. Semicond. 9,163.
Kornilov, B. V. et al. (1974). Photoluminescence of gallium arsenide doped with deep-level
impurities. Sou. Phys.-Semicond. (Engl. Transl.) 8, 141.
Kozeikin, B. V. et al. (1977). Anomalous behavior of impurities in ion-implantation-doped
epitaxial n-type GaAs. Sov. Phys.-Semicond. (Engl. Transl.) 11, 577.
Kravchenko, A. F. et ul. (1977). Investigation of recombination of non-equilibrium carriers
in GaAs epitaxial layer. Phys. Stutus Solidi A 42, 647.
Krebs, J. J., and Stauss, G . H. (1977). EPR of Cr2+(3d4)in gallium arsenide: Jahn-Teller dis-
tortion and photo-induced charge conversion. Phys. Rev. B 16,971.
Krebs, J. J. et al. (1981). Chromium in GaAs: a review of electron paramagnetic resonance
results. Lund Int. Con$ Deep-level Impurities Semicond., 3rd, I981 Abstracts, p.
Kressel, H.(1974). “Gallium arsenide and (A1Ga)As devices prepared by liquid-phase epitaxy.
J. Electron. Muter. 3, 747.
Kressel, H. et al. (1968a). Luminescence in silicon-doped GaAs grown by liquid-phase epitaxy.
J. Appl. Phys. 39, 2006.
Kressel, H.et al. (1968b). Observations concerning radiative efficiency and deep-level lumi-
nescence in n-type GaAs prepared by liquid phase epitaxy. J . Appl. Phys. 39, 5139.
Kressel, H. et al. (1968~).New deep level luminescence in GaAs: Sn. J. Appl. Phys. 39, 5647.
Kressel, H. et al. (1969). Comment on “Local mode absorption and defects in compensated
silicon-doped gallium arsenide” by W. G . Spitzer and W.Allred. J . Appl. Phys. 4,3069.
Kroger, F. A. (1977). Defect chemistry in crystalline solids. Annu. Rev. Muter. Sci. 7,449.
Kudo, H. et al. (1979). Stoichiometry measurements of GaAs by means of 15 MeV H e + + back
scattering. J . Appl. Phys. 50(7), 5034.
Kuiken, H. K. e f ul. (1976). Theory of lifetime measurements with the scanning electron micro-
scope : transient analysis. Solid-state Electron. 19,447.
Kumar, V., and Ledebo, L.-A. (1981). Evidence for Cu-related deep states in GaAs, earlier
ascribed to Ni. J . Appl. Phys. 52(7), 4866.
Kung, J. K., and Spitzer, W. G. (1974). Determination of Fermi-level effect on Si-site distribu-
tion of GaAs: Si. J . Appl. Phys. 45,2254.
Kiinzel, H., and Ploog, K. (1980). The effect of As, and As, molecular beam species on photo-
luminescence of molecular beam epitaxially growth GaAs. Appl. Phys. Lett. 37(4), 416.
Kushiro, Y. et al. (1977). Improved properties of melt-grown GaAs by short time heat treat-
ment. J. Appl. Phys. 48, 1636.
Kwun, S. I. et al. (1979). Optical studies of Be-implanted GaAs. Appl. Phys. Lett. 50(u), 6873.
Lagowski, J. (1982). Deep levels related to native defects in GaAs. Electron. Muter. Con$,
I982 Abstracts, p. 28.
Lagowski, J. et al. (1982). Microscopic model of the EL2 level in GaAs. Conj: Ser.-Inst. Phys.
65.
Laithwaite, K. et al. (1977). Boron acceptor centers in Sidoped LEC gallium arsenide. Conj:
Ser.-Inst. Phys. 33a, 133.
Lang, D. V. (1974). Deep-level transient spectroscopy: A new method to characterize traps in
semiconductors. J . Appl. Phys. 45, 3023.
Lang, D. V. (1977). Review of radiation-induced defects in Ill-V compounds. Con$ Ser.-Inst.
Phys. 31,70.
Lang, D. V., and Kimerling, L. C. (1975). A new technique for defect spectroscopy in semicon-
ductors: Application to 1 MeV electron-irradiated n-GaAs. ConJ Ser.-Inst. Phys. 23,581.
Lang, D. V., and Logan, R. A. (1975). A study of deep levels in GaAs by capacitance spectros-
copy. J . Electron. Muter. 4, 1053.
Lang, D. V. et al. (1977). Identification of the defect state associated with a gallium vacancy in
GaAs and AI,Ga, -,As. Phys. Rev. B 15,4874.
Lastras-Martinez, A. (1979). Theoretical calculations of the minority-carrier diffusion length in
high-internal-quantum-efficiency GaAs. J . Appl. Phys. 50(6), 41 56.
Leach, M. F., and Ridley, B. K. (1978). On the origin of slow domains in GaAs:O. J . Phys. C
11,2249.
Lebedev, A. A. etul. (1976). Investigation of the photocapacitance of GaAs: Fe epitaxial diodes.
Sou. Phys.-Semicond. (Enel. Transl.) 10, 1028.
Lee, K. S. (1980). A comparison between atomic concentration profiles and defect density
profiles in GaAs annealed after implantation with beryllium. J. Electron. Muter. 9(1),
185.
Leedy, H. M. et al. (1972). A critical contaminant in GaAs devices and the related failure
phenomena. Con$ Ser.-Inst. Phys.
Leheny, R. F., Nahory, R. E., and Pollack, M. A. (1973). Millimeter wave determination of
photoinjected free-carrier concentrations in highly excited GaAs. Phys. Rev. B 8,620.
Lender, R. J. et al. (1979). Diffusion length measurements in Schottky barrier GaAs solar cells.
Solid-state Electron. 22, 2 13.
Leung, P. C. et al. (1972). Infrared absorption study of Li-diffused Mg-doped GaAs. J . Appl.
Phys. 43,4096.
Li, G. P., Ma, Y. R., and Wang, K. L. (1982). An ion implantation defect study in liquid phase
epitaxial GaAs. Electron. Muter. Con$ 1982 Metal SOC.AIME Abstracts, p. 31.
Li, S. S., and Huang, C. 1. (1972). Investigation of the recombination and trapping processes
of photoinjected carriers in semi-insulating Cr-doped GaAs using PME and PC methods.
J. Appl. Phys. 43, 1757.
152 A. G . MILNES
Li, S. S., and Huang, C. 1. (1973). Radiative recombination in 0,-doped n-type GaAs at low
temperatures. Phys. Status Solidi A 17, 659.
Li, S . S . e f al. (1980a). Deeplevel defects, recombination mechanisms and their correlation to
the performance of low-energy proton-irradiated AlGaAs-GaAs solar cells. IEEE Trans.
Electron Devices ED-27, 857.
Li, S. S. et al. (1980b). Deeplevel defects and diffusion length measurements in low energy,
proton-irradiated GaAs. J. Electron. Mater. 9,335.
Lidow, A. et al. (1978). Ion-implanted selenium profiles in GaAs as measured by secondary
ion mass spectrometry. Appf. Phys. Lett. 32, 15.
Lindquist, P. F. (1977). A model relating electrical properties and impurity concentrations in
semi-insulating GaAs. J. Appf. Phys. 48,1262.
Littlejohn, M.A. et al. (1971). The electrical properties of 60 keV zinc ions implanted into semi-
insulating gallium arsenide. Radial. Eff: 10, 185.
Logan, R. M. (1971). Analysis of heat treatment and formation of gallium-vacancy-tellurium
complexes in GaAs. J . Phys. Chem. Solids 32, 1755.
Logan, R. M., and Hurle, D. T. J. (1971). Calculations of point defect concentrations and non-
stoichiometry in GaAs. J. Phys. Chem. Solids 32, 1739.
Lomako, V. M. et al. (1975). Quantum efficiency and lifetime of non-equilibrium carriers in
gallium arsenide diodes. Sou. Phys.-Semicond. (Engl. Transl.) 10, 733.
Look, D. C. (1977). The electrical characterization of semi-insulating GaAs: A correlation with
mass-spectrographic analysis. J . Appl. Phys. 48(12), 5141.
Look, D. C. (1979). On the interpretation of photoconductivity and photo-hall spectra in semi-
insulating GaAs:Cr. Solid State Commun. 24,825.
Look, D. C. (1980). True mobilities in semi-insulating 0- and Cr-doped GaAs. In “Semi-
Insulating 111-V Materials” (G. J. Rees, ed.),p. 183. Shiva Publ. Ltd., Orpington, Kent,
U.K.
Look, D. C. (1981). New levels in GaAs at 0.13 < E, - E c 0.20. Bull. Am. Phys. Soc. [2] 26,
222.
Lorimor, 0.G., and Spitzer,W. G. (1966). Local mode absorption in compensatedsilicon-doped
gallium arsenide. J . Appi. Phys. 37,3687.
Loualiche, S . er al. (1982). Interpretation of profiles obtained by C(V)technique in presence of
deep traps : Application to proton irradiated GaAs samples. Solid-state Electron. 25, 577.
Ludman, J. E.,and Nowak, W. B. (1976). Traps in fast-neutron irradiated GaAs Schottky
diodes. Solid-State Electron. 19, 759.
McLevige, W. V. et al. (1978). Diffusion studies of Be-implanted GaAs by SIMS and electrical
profiling. Solid State Commun. 25, 1003.
McNichols, J. L., and Berg, N. (1971). Neutron-induced metallic spike zones in GaAs. IEEE
Trans. Nucl. Sci. NS-18,2 1.
Magee, T. J., and Peng, J. (1975). Electron microscopy studies of the alloying behavior of Au
on GaAs. Phys. Stntus Solidi A 32, 695.
Magee, T. J. et al. (1979). Back surface gettering of Au in GaAs. Phys. Status. Solidi A 55, 161.
Maier, M. et a f . (1981). Sn incorporation and electron concentration in vapor-phase epitaxially
grown GaAs. J . Appl. Phys. 52,342.
Majerfeld, A., and Bhattacharya, P. K. (1978). New technique for identification of deep level
emission to indirect conduction minima in GaAs. Appl. Phys. Lett 33(3), 259.
Makram-Ebeid, S. (198Oa). Effect of electronic field and current injection on the main electron
trap in bulk GaAs. Mater. Res. SOC.Meet., 1980.
Makram-Ebeid, S.(1980b). Effect of electronic field on deep-level transients in GaAs and Gap.
Appl. Phys. Lett 37(5) ,464.
Makram-Ebeid, S. era/. (1980). Phonon assisted tunnel emission of electrons from the Cr level
in GaAs. Conff:Ser.-Inst. Phys.
Martin, G. M. (1980). Key electrical parameters in semi-insulating materials and the methods to
determine them in GaAs. In “Semi-Insulating 111-V Materials” (G. J. Rees, ed.). Shiva
Publ. Ltd., Orpington, Kent, U.K.
Martin, G. M., and Makram-Ebeid, S. (1982). Manifestations of deep-level point defects in
GaAs. Int. Con$ Defects Semicond. 12th.
Martin, G. M. et al. (1977a). Study of experimental conditions in thermally stimulated current
measurements: Application to the characterization of semi-insulating chromium-doped
GaAs. J . Electrost. 3, 223.
Martin, G. M. et (11. (1977b). Electron traps in bulk and epitaxial GaAs crystals. Electron.
Lett. 13, 192.
Martin, G. M. et al. (1979). Measurement of the chromium concentration in semi-insulating
GaAs using optical absorption. J . Appl. Phys. 50(1), 467.
Martin, G. M. et al. (1980a). Detailed electrical characterization of the deep Cr acceptor in
GaAs. J . Phys. C. 13, 3855.
Martin, G. M. et ul. (1980b). Compensation mechanisms in GaAs. J. Appl. Phys. 51(5), 2840.
Martin, G. M. et al. (1981). Identification and analysis of near-infrared absorption bands in
undoped and Cr doped semi-conducting GaAs crystals. Conf: Ser,- Inst. Phys. 59, 000.
Martin, G . M. et al. (1982). Oxygen related gettering of silicon during growth of bulk GaAs
Bridgman crystals. J . Phys. C 15, 1841.
Martinelli, R. U . , and Wang, C . C. (1973). Electron beam penetration in GaAs. J . Appl. Phys.
44, 3350.
Martinez, G . et a/. (1981). Charge transfer Cr3+(3d3)-CrZ+(3d4)in chromium-doped GaAs.
Phys. Rev. B 23, 3920.
Matare, H. F., and Wolff, G. A. (1978). Concentration enhancement of current density and
diffusion length in 111-V ternary compound solar cells. Appl. Phys. 17, 335.
Mathur, V. K., and Rogers, S. (1977). Relaxation of photoexcited carriers in GaAs. Appl.
Phys. Lett. 31, 765.
Matino, H. (1974). Reproducible sulfur diffusion into GaAs. Solid-State Electron. 17,35.
Matveenko, Yu A. et al. (1969). Some properties of nickel doped gallium arsenide. Sou. Phys-
Mayburg, S. (1961). Direct recombination in GaAs and some consequences in transistor design.
Solid-State Electron. 2, 195.
Miki, H. et al. (197 1). Effects of impurities on electrical and optical properties of epitaxial GaAs
layers. Electrochem. Soc. Spring Meet., 1971 Abstract No. 83, p. 201.
Miller, M. D. et al. (1977). The effect of gas-phase stoichiometry on deep levels in vapor-grown
GaAs. Appl. Phys. Lett. 31, 538.
Mircea, A,, and Mitonneau, A. (1979). Strongly anisotropic field ionization of a common level
in GaAs. J.Phys., Lett. (Orsay, Fr.) 40, L-31.
Mircea, A. et al. (1976). Outdiffusion of deep electron traps in epitaxial GaAs. Appl. Phys. 11,
153.
Mircea, A. et ul. (1977). Temperature dependence of ionization energies of deep bound state
in semiconductor. J . Phys., Lett. (Orsay, Fr.) 38, L-41.
Mitonneau, A. (1976). Investigation of deep traps in GaAs by a capacitive method. Philips
Res. Rep. 31, 244.
Mitonneau, A., and Mircea, A. (1979). Auger de-excitation of a metastable state in GdAs.
Solid State Commun. 30, 157.
Mitonneau, A. et al. (1977). Hole traps in bulk and epitaxial GaAs crystals. Electron. Lett. 13,
666.
154 A. G . MILNES
Mitonneau, A. et al. (1979). Electron and hole capture cross-sections at deep centers in gallium
arsenide. Rev. Phys. Appl. 14, 853.
Mitonneau, A. et al. (1980). Defect characterization at the growth interface in GaAs epitaxy
by metalorganic and chloride depositions. J . Electron. Muter. 9,213.
Monch, W . et al. (1981). Surface photovoltage spectroscopy of chromium in GaAs. Lund. Znt.
Conj. Deep-level Impurities Semicond., 3rd, 1981 Abstracts, p. 85.
Monemar, B., and Blum, J. M. (1977). Optical characterization of deep 0 implants in GaAs.
J . Appl. Phys. 48, 1529.
Moore, J. A. et al. (1975). A comparison of the radiation damage produced in gallium arsenide
by monatomic and diatomic arsenic implants, Radiat. Eff.25,49.
Morgulis, L. M. e f al. (1973). Precipitation ofcopper in GaAs single crystals. Sou. Phys.-Solid
State (Engl. Transl.) 14, 2436.
Morgulis, L. M. et al. (1974). Electron-microscope investigation of tellurium-doped gallium
arsenide single crystals. Sou. Phys.-Solid State (Engl. Transl.) 16, 142.
Morkoc, H., and Cho, A. Y. (1979). High-purity GaAs and Cr-doped GaAs epitaxial layers by
MBE. J . Appl. Phys. !50(10), 6413.
Morrison, S. R. et al. (1974). The behavior of donor impurities in n type gallium arsenide and
gallium phosphide. J. Phys. C 7 , 633.
Morrizumi, T., and Takahashi, K. (1969). Si- and Ge-doped GaAs p-n junctions. Jpn. J . Appl.
Phys. 8,348.
Mullin, J. B. et al. (1977). Analysis of deep impurity levels in semi-insulating gallium arsenide.
Conf. Ser.-fnst. Phys. 33(s), 91.
Murygin, R. I., and Rubin, V. S. (1970). Some investigations of the electrical properties and
injection conductivity of high resistivity nickel doped gallium arsenide. Sou. Phys.-
Nakai, K. et al. (1977). Growth of iron-doped epitaxial layers for GaAs field effect transistors.
J . Electrochem. SOC.124, 1635.
Nelson, R. J. et al. (1980). Reduction of GaAs surface recombination velocity by chemical
treatment. Appl. Phys. Lett. 36(1), 76.
Neumann, H. et a/. (1977). Properties of liquid-phase epitaxial Ge-doped GaAs. Phys. Status
Solidi A 44,675.
Nishizawa, J. 1. et al. (1974). Nonstoichiometry of Te-doped GaAs. Jpn. J. Appl. Phys. 13,46.
Nojima, S . , and Kawasaki, Y. (1978). Annealing characteristics of Be ion implanted GaAs.
Jpn. J . Appl. Phys. 17(10), 1845.
Norris, B., and Narayanan, G. H. (1977). A study of the lattice imperfections in Li-diffused
GaAs by transmission electron microscopy. J . Appl. Phys. 48,2784.
Norris, C. B. (1979). Luminescence studies of the 1.36 eV band and the edge emission in melt-
grown GaAs:Cu. J . Appl. Phys. 50(5), 3658.
Nowicki, R. S., and Wang, I. (1978). Improvement of the diffusion barrier properties of rf
sputtered molybdenum. J . Vac. Sci. Technol. 15(2), 235.
Nozaki, T. (1976). Si ion implantation into GaAs. Jpn. J. Appl. Phys. 15, 1951.
O’Brien, J. K., and Corelli, J. C. (1973). Photo-conductivity of chrominum-compensated GaAs
after irradiation by 0.8 MeV electrons. J . Appl. Phys. 44, 1921.
Okumura, T. et af. (1976). Electron and hole traps in nGaAs crystals. Appl. Phys. 11, 187.
Olowolafe, J. 0.et al. (1979). Contact reactions in Pd/GaAs junctions. J. Appl. Phys. 50(2), 955.
Omel’yanovskii, E. M. e t a / . (1970). Deep levels in semi-insulating iron-doped gallium arsenide.
Ozeki, M . (1979). Effects of the growth conditions on the incorporation of deep levels in vapor-
grown GaAs. J . Appl. Phys. 50,4808.
Ozeki, M. et al. (1979). Self-compensation effect in VPE GaAs due to shallow and deep levels.
Con$ Ser.-Inst. Phys. 45, 220.
Palfrey, H. D., Brown, M., and Willoughby, A. F. W. (1982). Self-diffusion in gallium arsenide.
Electron. Muter. Conf. 1982 Abstracts, p. 30.
Panish, M. B. (1967). Ternary condensed phase systems of gallium and arsenic with group 1B
elements. J . Electrochem. SOC.114, 516.
Papaioannou, G. I., and Euthymiou, P. C. (1978). Profile of the carrier lifetimes in illuminated
GaAs as a function of depth. J . Appl. Phys. 49(4), 2578.
Partin, D. L. et al. (1979a). Deep-level transient spectroscopy studies of Ni and Zn-diffused
vapor-phase epitaxy n-GaAs. J. Appl. Phys. 50(1 I), 6845.
Partin, D. L. et al. (1979b). Surface effects in SEM measurements of hole diffusion length in
n-GaAs. J. Electron. Muter. 8,493.
Partin, D. L. et ul. (1979~).Effect of surface preparation and heat-treatment on hole diffusion
lengths in VPE GaAs and GaAs o,6Po,4.J . Electrochem. Soc. 126, 1581.
Partin, D. L. et al. (1979d). Hole diffusion lengths in VPE GaAs and GaAs,,,P,,, treated with
transition metals. J. Electrochem. Soc. 126, 1584.
Partin, D. L. et ul. (1979e). Transition metal deep centers in GaAs, GaP and Si. Solid-Stute
Electron. 22,455.
Pashley, R. D., and Welch, B. M. (1975). Tellurium-implanted n + layers in GaAs. Solid-Sture
Electron. 18, 977.
Peka, G. P. et ul. (1975). Recombination mechanisms in compensated GaAs. Phys. Stutus
Solidi A 29,47.
Picoli, G. et al. (1980). On the interpretation of the 0.839 eV optical transition in GaAs:Cr. In
“Semi-Insulating Ill-V Materials” (G. J. Rees, ed.), p. 254. Shiva Publ. Ltd., Orpington,
Kent, U.K.
Picoli, G . et al. (1981). 0.839 eV Luminescence in GaAs:Cr some complements to the model.
Lund In!. Con$ Deep-level Impurities Semicond., 3rd, I981 Abstracts.
Pilkuhn, M. H. (1976). Electron-hole recombination. Phys. Semicond. Proc. Int. Con$. 13th,
1976 p. 61.
Ploog, K. (1981). Molecular beam epitaxy of 111-V compounds: Technology and growth
processes. Annu. Rev. Muter. Sci. 11, 171.
Pons, D. (1980). Determination of the free energy level of deep centers, with application to
GaAs. Appl. Phys. Lett. 37(4), 413.
Pons, D. (1981). Charge state effects on the annealing of the electron irradiation induced defects
in GaAs. Con$ Ser.-Inst. Phys. 59,269,
Pons, D.! and Bourgoin, J. (1981a). Irradiation induced defects in n-GaAs. Lund. Int. Cant
Deep-level Impurities Semicond., 3rd, 1981 Abstracts, p. 26.
Pons, D., and Bourgoin, J. (1981b). Anisotropic defect introduction in GaAs by electron irradia-
tion. Phys. Rev. Lett. 47, 1293.
Pons, D. et al. (1980). Energy dependence of deep level introduction in electron irradiated GaAs.
J . Appl. Phys. 51(4), 2038.
Rachmann, J., and Biermann, R. (1969). Nachweis und Diffusion von Sauerstoff in GaAs.
Solid State Commun. 7, 1111.
Rao-Sahib, T. S., and Wittry, D. B. (1969). Measurement of diffusion lengths in p-type gallium
arsenide by electron beam excitation. J. Appl. Phys. 46,3745.
Rheinlander, B., Hadse, S., and Weinert, H. (1977). On the determination of the free carrier
relaxation time by transverse magneto-plasma-reflectivity measurements on Sn-doped
n-type GaAs. Phys. Status Solidi B 83, K83.
Richman, D. (1964). “Rare-earth Doping of GaAs, Synthesis and Characterization of Elec-
tronically Active Materials,” RCA Lab. Rep., Doc. AD 452 788. Natl. Tech. Inf. Serv.,
Springfield, Virginia.
Romanenko, V. N., and Kheifets, V. S. (1975). Partition coefficients of impurities in GaAs and
its solid solutions with AlAs, obtained by liquid epitaxy. Inorg. Muter. 11, 632.
156 A. G. MILNES
Rossin, V. V. er a / . (1979). Allowance for reemission in the problem of minority carrier diffusion
in direct gap semiconductors. Sou. Phys-Semicond. (Engl. Transl.) 13(6), 647.
Roussel, A., and Mircea, A. (1973). Identification of a bivalent flaw in n-GaAs from noise
measurements. Solid Stute Commun. 12, 39.
Rozgonyi, G . A. et al. (1974). Fabrication of GaAs IMPATT devices using electroplated heat
sinks. J . Electrochem. Sue. 121, 426.
Ryan, R. D., and Eberhardt, J . E. (1972). Hole diffusion length in high purity n-GaAs. Solid-
Sah, C. T. et al. (1970). Thermal and optical emission and capture rates and cross-sections of
electrons and holes at imperfection centers in semiconductors from photo and dark junc-
tion current and capacitance experiments. Solid-State Electron. 13,759.
Saito, T., and Hasegawa, F. (1971). Cause of the high resistance region at vapour epitaxial
GaAs layer-substrate interface. Jpn. J . Appl. Phys. 10, 197.
Sakai, K., and Ikoma, T. (1974). Deep levels in gallium arsenide by capacitance methods. Appl.
Phys. 5, 165.
Samuelson, L. et ul. (1981). Electrical and optical properties of deep levels in MOVPE grown
GaAs. J . Cryst. Grouith 55, 164.
Sankey, 0. F. et al. (1980). Substitutional defect pairs in GaAs, -xPx.Phys. Rev. Lett. 45, 1656.
Sankey, 0. F. et ul. (1981). A simple chemical theory of deep levels in semiconductors. Lund.
Int. ConJ Deep-level Impurities Semicond., Srd, 1981 Abstracts, p. 5.
Sansbury, J. D., and Gibbons, J. R. (1970). Properties of ion implanted silicon, sulfur, and
carbon in gallium arsenide. Radial. Eff: 6, 269.
Savva, M. A. (1976). Control of sulfur doping for gallium arsenide epitaxial layers. J . Electro-
chem. Soc. 123, 1496.
Schairer, W. et al. (1976). Piezospectroscopic and magneto-optical study of the Sn-acceptor in
GaAs. Phys. Rev. B 13, 3452.
Schemer, M. et al. (1981). Identification of deep centers in Gap. Lund Int. Conf. Deep-leuel
Impurities Semicond., 3rd, 1981 Abstracts, p. 27.
Schlachetzki, A., and Solov, H. (1975). High resistivity layers of GaAs grown by liquid phase
epitaxy. Appl. Phys. I , 195.
Sealy, B. J. (1975). An examination of tellurium ion-implanted GaAs by transmission electron
microscopy. J . Mater. Sci. 10,683.
Sealy, B. J . et a / . (1976). Donor activity in ion-implanted GaAs. Conf. Ser.-Inst. Phys. 28,69.
Sekela, A. et al. (1974). Diffusion length studies in n-gallium arsenide. Corzj. Ser.-Inst. Phys.
24,245.
Seki, Y. et al. (1978). lmpurity effect on grown-in dislocation density of InP and GaAs crystal.
Appl. Phys. 49, 822.
Seltzer, M. S. (1965). Diffusion of manganese into gallium arsenide. J . Phys. Chem. Solids 26,
243.
Share, S . et al. (1975). Properties of compensated GaAs light-emitting diodes following 6oCo
irradiation. Solid-State Electron. 18, 47 1.
Shifrin, S. S. er al. (1974). Nature of small etch pits in tellurium-alloyed GaAs single crystals.
Sou. Phys. -Crystallogr. (Engl. Transl.) 18, 8 18.
Shih, K. K., and Petit, G . D. (1974). Properties of Ge-doped GaAs and AI,Ga, -.As, Sn-doped
Al,Ga, -.As, and Si-Te-doped GaAs. J . Electron. Muter. 3, 391.
Shin, B. K. (1976). Carbon-ion implanted gallium arsenide. Appl. Phys. Lett. 29(7), 438.
Shin, B. K., and Park, Y. S. (1976). An abrupt dopant profile in GaAs produced by Te implanta-
tion. J . Electrochem. Soc. 123, 1588.
Shin, B. K. et al. (1978). Dual implantation of C + and Ga' ions into GaAs. J . Appl. Phys. 49,
2988.
Sinha, A. K. (1975). Metallization scheme for n-GaAs Schottky diodes incorporating sintered
contacts and a W diffusion barrier. Appl. Phys. Lett. 26(4), 171.
Sinha, A. K. et ul. (1975). Sintered ohmic contacts to n- and p-type GaAs. IEEE Truns. Electron
Devices ED-22,2 18.
Sokolov, V. I., and Shishyanu, F. S. (1964). Diffusion and solubility of gold in gallium arsenide.
Sou. Phys.-Solid Stare (Engl. Trans/.) 6,265.
Spicer, W. E. et ui. (1980). Unified defect model and beyond. J . Vuc. Sci. Techno/. 17(5), 1019.
Spitzer, W. G., and Allred, W. (1968). Local-mode absorption and defects in compensated
silicon-doped gallium arsenide. J . Appl. Phys. 39, 4999.
Springthorpe, A. J. et al. (1975). Te and Ge-doping studies in Ga,-xAI,As. J. Electron. Muter.
4, 101.
Stall, R.A. et al. (1980). Growth parameter dependence of deep levels in molecular beam epi-
taxial GaAs. Electron. Lett. 16, 171.
Straws, G. H. et al. (1979). EPR determination of the concentration of chromium charge states
in semi-insulating GaAs: Cr. J. Appl. Phys. 50(10), 6251.
Stein, H. J. (1969). Electrical studies of low-temperature neutron- and electron-irradiated
epitaxial n-type GaAs. J . Appl. Phys. 40,5300.
Stillman, G. E. et al. (1976). Donor impurities in GaAs: Evidence for a stoichiometric hydro-
genic shallow donor level. Proc. Int. Con$, I3th, I976 p. 623.
Stoneham, E. B. et ul. (1980). Formation of heavily doped n-type layers GaAs by multiple ion
implantation. J . Electron. Muter. 9(2), 37 1.
Strack, H. (1966). Iron-doped gallium arsenide transistors. Conf Ser.-Inst. Phys. 3, 206.
Stringfellow, G. B. (1974). Calculation of distribution coefficients of donors in Ill-V semi-
conductors. J. Phys. Chem. Solids 35,775.
Su, J. L. et al. (1971). A deep level in Zn-doped liquid phase epitaxial GaAs. Solid-Slate Electron.
14, 262.
Suchkova, N . I. et ul. (1975). Properties ofnickel-doped gallium arsenide. Sou. Phys.-Semicond.
(Engl. Trunsl.) 9, 469.
Suchkova, N . I. et ul. (1977). Behavior of the nickel impurity in the 3d8 state in gallium arsenide.
Sou. Php-Semicond. (Engl. Transl.) 11, 1022.
Surridge, R. K., and Sealy, B. J. (1977). A comparison of Sn-, Ge-, Se and Te-ion-implanted
GaAs. J. Phys. D 10, 91 I ,
Surridge, R. K. et al. (1977). Annealing kinetics of donor ions implanted into GaAs. Con{
Ser.-Inst. Phys. 33a, 161.
Ta, L. B. et ul. (1982a). Reproducibility and uniformity considerations in LEC growth of
undoped semi-insulating GaAs for large-area direct-implant technology. Conf. Ser.-Inst.
Phys. 65, 31.
Ta, L. B., Hobgood, H. M., and Thomas, R. N . (1982b). Evidence of the role of boron in un-
doped GaAs grown by liquid-phase Czochralski. Appl. Phys. Lett. 41, 1091.
Takahashi, K. (1980). Structural evaluation of GaAs crystals from etch pits. Jpn. J . Appl. Phys.
19, 1947.
Takai, M. et al. (1975). Lattice site location of cadmium and tellurium implanted in gallium
arsenide. Jpn. J. Appl. Phys. 14, 1935.
Tdndon, J. L. et al. (1979). Silicon implantation in GaAs. Appl. Phys. Lett. 34(2), 165.
Taniguchi, M., and Ikoma, T. (1982). Re-examination of the mid-gap electron trap (EL2) in
different GaAs wafers from photocapacitance spectra. COP$Srr.-Inst. Phys. 65.
Taylor, P. D., and Morgan, D. V. (1976). The effects of radiation damage on the properties of
Ni-n GaAs Schottky diodes. 11. Solid-State Electron. 19,481.
Thomas, R.N. et al. (1981). Growth and characterization of largediameter undoped semi-in-
sulating GaAs for direct ion implanted FET technology. Solid-Stute Electron. 24,387.
158 A. G . MILNES
Ushakov, V. V., and Gippius, A. A. (1980). Radiative transitions in Wz+ centers in GaAs. Sou.
Phys.-Semicond. (Engl. Transl.) 14(3), 333.
Uskov, V. A,, and Sorvina, F. P. (1974). Effect of vapor pressure of arsenic on diffusion and
solubility of Fe in GaAs. Inorg. Muter. 10, 895.
Vaitkus, J. et al. (1981). Impurity photoconductivity and scattering centers of carriers due to
deep levels of Cr and 0 in GaAs and due to dislocations in Si. Lund. lnt. Conf: Deep-level
Impurities Semicond., 3rd, 1981 Abstracts, p. 86.
Vallin, J. T. el a[. (1970). Infrared absorption in some 11-VI compounds doped with Cr. Phys.
Reu. B 2, 4313.
Van Opdorp, C., Peters, R. C., and Klerk, M. (1974). Use of Schottky-diode collectors for SEM
determination of bulk diffusion lengths. Appl. Phys. Lett. 24, 125.
Van Vechten, J. A. (1975). Nonstoichiometry and nonradiative recombination in Gap. J .
Electron. Muter. 4, 1159.
Van Vechten, J . A. (1976). Simple estimates of the thermo-chemical properties of impurities
and other defects in semiconductors. Phys. Semicond. Proc. Int. Conf., I3th, 1976 p. 577.
Van Vechten, J. A. (1980). Point defects and deep traps in 111-V compounds. Czech. J . Phys.
B30, 388.
Van Vechten, J. A., and Thurmond, C. D. (1976). Entropy of ionization and temperature
variation of ionization levels of defects in semiconductors. Phys. Reo. B 14, 3539.
Vasil’ev, A. V. ef al. (1976). Energy spectrum of V impurities in GaAs. Sou. Phys.-Semicond.
(Engl. Transl.) 10(3), 341.
Vasudev, P. K.. and Bube, R. H. (1978). Photocapacitance of deep levels in GaAs:Cr and
GaAs:O. Solid-State Electron. 21,1095.
Vasudev, P. K.el al. (1977). Deep levels in high-purity LPE GaAs by photocapacitance and
related techniques. Conf. Ser.-lnst. Phys. 33b, 154.
Vincent. G., and Bois, D. (1978). Photocapacitance quenching effect for “Oxygen” in GaAs.
Solid State Commun. 21,431.
Vincent, G. rt al. (1979). Electric field effect on the thermal emission of traps in semiconductor
junctions. J . Appl. Phys. 50(8), 5484.
Vitale, G. et al. (1979). Measurement of minority-carrier lifetime in GaAs using the transient
response of MOS capacitors. Appl. Phys. Lett. 34, 105.
von Roos, 0. (1978). Analysis of the interaction of an electron beam with a solar cell. Solid-
von Roos, 0. (1979). Recombination lifetimes and surface recombination velocities of minority
carriers in n-p junctions. A new method for their determination by means of a stationary
amplitude-modulated electron beam. J. Appl. Phys. 50(5), 3738.
Wada, 0. et al. (1977). Determination of deep electron traps in GaAs by time-resolved capaci-
tance measurement. Appl. Phys. 13, 5 .
Walker, G. H., and Conway, E. J. (1979). Annealing kinetics of electron-irradiated GaAs hetero-
face solar cells in range 179-200°C. Appl. Phys. Lett. 35(6), 459.
Wallis, R. H. et al. (1981a). Pressure dependence of the energy levels of irradiation-induced
defects in GaAs. Appl. Phys. Lett. 38(a), 698.
Wallis, R. H. et al. (1981b). Effect of oxygen injection during VPE growth of GaAs and A1,-
Ga,-.As. Conf. Ser.-Inst Phys. 56, 73.
Watkins, G. D. (1976). Displacement defects in semiconductors produced by electrons. Proc.
Conf. Sci. Ind. Appl. Small Accelerators, 4th, 1976 p. 588.
Weber, E. R., and Schneider, J. (1982). As,, antisite defects in GaAs. lnt. Conf. Defects Semi-
cond., 12ih, 1982, p. 398.
Weiner, M. E., and Jordan, A. S. (1972). Analysis of doping anomalies in GaAs by means of a
silicon-oxygen complex model. J . Appi. Phys. 43, 1767.
White, A. M. (1974). Bound excitation luminescence associated with the deep oxygen donor in
gallium arsenide. Phys. Semicond., Proc. Int. Conj:, 12th, 1974 p. 381.
White, A. M. (1979). The origin of sharp near infrared transitions in chromium doped 11 I-V
semiconductors. Solid State Commun. 32, 205.
White, A. M. et al. (1976). Photocapacitance effects of deep traps in epitaxial GaAs. J. Appl.
Phys. 47,3230.
Wight, D. R. et al. (1981). Diffusion lengths in p-type MOCVD GaAs. J. Cryst. Growth 55,183.
Willardson, R. K., and Allred, W. P. (1966). Distribution coefficients in gallium arsenide.
Gailium Arsenide and Related Compounh. Inst. Phys. p. 35.
Williams, E. W., and Bebb, H. B. (1972). Photoluminescence 11: Gallium arsenide. In “Semi-
conductors and Semi-metals” (R. K. Willardson and A. C. Beer, eds.), Vol. 8. Academic
Press, New York.
Williams, P. J. et al. (1982). High resolution optical absorption spectroscopy on Cr related
defects in GaAs and Gap. J. Phys. C 15, C1337.
Williamson, K. R. et al. (1979). Glow discharge optical spectroscopy of B-, Ge-, Mg-implanted
GaAs. J. Appl. Phys 50(12), 8019.
Willmann, F. et al. (1971). Complex nature of the copper acceptor in gallium arsenide. Solid
State Commun. 9, 2281.
Willmann, F. et nl. (1973). Optical properties of excitons bound to copper-complex centers in
gallium arsenide. Phys. Rev. B 7, 2473.
Wilsey, N. D., and Kennedy, T. A. (1981). EPR of radiation-induced defects in 111-V semi-
conductors. Lund Znt. Con$ Deep-level Impurities Semicond., 3rd, I981 Abstracts.
Wilson, R. G. (1981). Impurity densities and the redistribution of Fe and Cr in LEC GaAs.
Int. Symp. Gallium Arsenide Relat. Compd., 1981 [Conf. Ser.-Inst. Phys. 631.
Wirth, J. L., and Rogers, S. C. (1964). The transient response of transistors and diodes to ioniz-
ing radiation. IEEE Trans. Nucl. Sci. NS-11,24.
Wittry, D. B., and Kyser, D. F. (1965). Cathodoluminescence at p-n junctions in GaAs. J.
Appl. Phys. 36, 1387.
Wolfe, C. M., and Stillman, G. E. (1973). Effect of surface states on the amphoteric behavior
of Sn in vapor epitaxial GaAs. Solid State Commun. 12, 283.
Wolfe, C. M. et al. (1977). Residual donors in high-purity GaAs. ConJ Ser.-lnst. Phys. 33b,
120.
Wolford, D. J . et al. (1981). Deep traps in 111-V semiconductor alloys. Lund. Int. Con$ Deep-
level Impurities Semicond., 3rd, I981 Abstracts.
Wolford, D. J. et al. (1982). Photoluminescence of oxygen-induced states in GaAs,-,P,. Conf:
Ser.-Inst. Phys. 65.
Wood, C. E. C., and Joyce, B. A. (1978). Tin-doping effects in GaAs films grown by molecular
beam epitaxy. J. Appl. Phys. 49(9), 4854.
Woodcock, J. M., and Clark, D. J. (1975). The ion implantation of donors for n+-pjunctions
in GaAs. Conj: Ser.-Inst. Phys. 24, 331.
Worner, R. et al. (1981). ESR of As,, antisite defects in fast neutron-irradiated GaAs. Private
communication.
Wu, C. J., and Wittry, D. B. (1978). Investigation of minoritycarrier diffusion lengths by
electron bombardment of Schottky barriers. J. Appl. Phys. 49,2827.
Yamazaki, H. et al. (1975). Tin diffusion into GaAs from doped SiO, films. Jpn. J. Appl. Phys.
14, 717.
Yan, Z.-x., and Milnes, A. G. (1982). Deep level transient spectroscopy of silver and gold levels
in LEC grown gallium arsenide. J. Electrochem. SOC.129, 1353.
Yu, P. W. (1977). Excitation-dependent emission in Mg-, Be-, Cd-, and Zn-implanted GaAs.
J. Appl. Phys. 48, 5043.
160 A. G . MILNES
Yu, P. W., and Park, Y. S. (1977). Photoluminescence from Mg-implanted GaAs. Appl. Phys.
Lett. 30, 14.
Zakharova, G. A. et al. (1972). Conditions and mechanism of the formation of certain impurity
centers in copper-doped gallium arsenide. Sou. Phys.-Semicond. (Enyl. Trunsl.) 6 , 171.
Zelevinskaya, V. M. et af. (1973). Impurity-defect interaction in gallium arsenide doped by
implantation of tellurium ions. Sou. Phys. -Semicond. (Engl. Trawl.) 6, 1209.
Zemon, S. er al. (1981). Photoluminescent studies of a deep acceptor level in GaAs:Sn and
(AIGa)As: Sn. Lund In?. Conj: Deep-levels Semicond., 3rd, I981 Abstracts, p. 36.
Zhu, H.-z. et al. (1981). Deep levels in MOCVD GaAs grown under different Ga/As mol
fractions. J. Cryst. Growth 55, 154.
Zou, Y-x. (1981a). A defect model of GaAs and its possible applications. Private communica-
tion. Substantially similar material presented at the September 1980 Symposium on GaAs
and Related Compounds. Con$ Ser.-Znst. Phys. 56,213.
Zou, Y-x. (1981b). Some aspects of the recent work on 111-V compounds in China. Con$ Ser.-
Znst. Phys. 56, 213.
Zou, Y-x.et al. (1982). Interactions among certain deep-level traps in GaAs. Conf. Ser.-Inst.
Phys. 65.
Zucca, R. (1977). Electrical compensation in semi-insulating GaAs. J . Appl. Phys. 48(5), 1987.
Zuleeg, R., and Lehovec, K. (1980). Radiation effects in GaAs junction field effect transistors.
ZEEE Trans. Nucl. Sci.NS-27, 1343.
ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS, VOL. 61
Quantitative Auger Electron Spectroscopy

M. CAILLER
Laboratoire de Physique du Metal
Ecole Mationale Supkrieure de MPcanique
Mantes, France
J. P. GANACHAUD
Laboratoire de Physique du Solide
Institut de Physique de I' Universitk de Nantes
Nantes, France
D. ROPTIN
Laboratoire de Physique du Mkzal
Ecole Nationale SupPrieure de MPcanique
Nantes, Frunce
I. Introduction . . . . . ....... ....................

A. Purpose ...............................................
B. Basic Experiment.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
... . . . . . . . . . . . . . . . . . . . 164
.................... . . . . . . . 167
A. Cross Sections ...................................................... 167
B. Mean Free Path, Attenuation Length, and Escape Depth of the Electrons. . . . 168
Ill. Dielectric Theory of Inelastic Collisions of Electrons in a Solid . . .
A. General Equati
B. Normal Metals
D. Semiconductors and Insulators ........................................ 181

E. Depth Dependence and Anisotropy of the MFP ...
IV. Elastic Collisions

V. Auger Transitions
A. Introduction .................... ................... 187
B. Classification of Auger Transitions . ................... 188
VI. Quantitative Description of Auger Emission

A. General Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 I3
B. Theoretical Analysis of Auger Emission.. ............................... 222
V11. Auger Quantitative Analysis ...... ....... 244
A. Introduction . . . . ................................. 244
B. Operating Modes ................................. 245
161
Copyright rC' 1983 by Acadcmic Press, Inc.
All rights of reproduction in any form rcseerved.
ISBN n - i m 4 6 6 1 - 4
162 M . CAILLER, J . P . GANACHAUD, AND D. ROPTIN
C. Quantification of Auger Analysis ...............

D. Sample Preparation ...........................
E. Effects of Sputtering on the Surface Composition
of Multicomponent Materia . 273
F. Electron-Beam-InducedEffects in AES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
VIII. Conclusion . . ...... ............. 289
References.. . ...... . 289
I. INTRODUCTION
A . Purpose
This article reviews Auger electron emission from a target subjected to
electron bombardment. The mechanisms involved in generation of Auger
electrons by ion or photon bombardment are therefore not considered here.
Despite this restriction, we shall use the term Auger electron spectroscopy
(AES) without reference to the excitation mode throughout in what follows
to denote the analysis technique developed for electron-induced Auger
electron emission.
An Auger transition can occur when a vacancy is produced in an inner
subshell of an atom. This atom, being in an unstable state, will try to revert to
a lower energy state by filling up its vacancy with an outer shell electron.
There are two ways for the atom to get rid of its energy excess: (1) by emitting
a photon (X-ray fluorescence) or (2) by emitting a second outer shell electron
(Auger transition).
During the past decade AES has become one of the most widely used
techniques for analysis. Its two major features are the following:
(1) The energy of the Auger electron characterizes the nature of the
emitter atom and eventually its chemical environment.
(2) The electrons emitted from the target with this characteristic energy
can only emerge from the outermost atom layers at the surface.
Accordingly, AES is an analysis technique for elements a t surfaces
(eventually, a chemical analysis technique). However, by combining ion
sputtering with AES it is possible to probe the bulk of a sample and to plot
depth profiles of the element concentrations in it.
Because of its practical importance a number ofbooks and general papers
have been devoted to the AES technique. Sevier’s book (1972), although
having a wider scope, devotes an important part to AES, and this book is very
useful to experimentalists working on Auger transitions. Chattarji (1976) has
described the theory of Auger transitions, following the earlier book by
Burhop (1952).Although these books have different objectives, together they
QUANTITATIVE AUGER ELECTRON SPECTROSCOPY 163
constitute a rather complete reference on the physics of the Auger effect con-
sidered as an elementary mechanism. In the case of a complex sample, how-
ever, there are no direct means of anticipating the results of a quantitative
analysis in terms of first principles, although some obstacles have been
removed.
Most of the general papers on AES (Chang, 1971; Palmberg, 1973;
Bouwman et al., 1973a,b,c; Chang, 1974; Miiller, 1975; Defrance, 1976;
Le Gressus, 1978;Holloway, 1978,1980; Powell, 1978; Van Oostrom, 1979)
are intended for researchers confronted with analysis problems of multi-
component targets which are frequently of a technological nature. They
review more or less extensively the Auger analysis method and stress the
physical fundamentals, possibilities, limits, and applications of the method.
In these papers the quantitative description of the creation and emission
mechanisms of Auger electrons is phenomenological.
During the pa st decade some papers appeared that were devoted to
establishing the relation between the phenomenological description and the
more fundamental description of the elementary mechanisms. Usually this
relation is shown via a simulation model on a computer, and Shimizu and
Ichimura (1981) have published a very important contribution on this
question. The present review article describes, as far as possible, all the steps
from first principles to the real sample analysis, but, as mentioned above, the
path is far from complete.
B. Basic Experiment
A very short general description of a typical experiment in AES is given
first, to introduce some basic vocabulary, after which we outline the content
of this article. In this experiment, an electron beam emitted by an electron gun
collides with the sample to be analyzed. The electrons of this beam are called
primary electrons, They are characterized by their energies and their angles
of incidence on the target (Fig. 1). Quite frequently, the incident beam can be
considered monoenergetic and monodirectional. Some of the electrons in the
primary beam are reflected by the outermost atom layers without appreciable
electron
baam
\ =by the target
Target
FIG.1. Basic diagram of secondary-electron emission.

164 M. CAILLER, J . P. GANACHAUD, AND D. ROPTIN
energy loss. They are called elastically reflected electrons. The others pene-
trate to different depths into the sample, interact with it, and thus are deflected
and undergo energy loss.
In classical language the primary electrons undergo elastic and inelastic
collisions in the solid. In an inelastic process the energy and momentum are
transferred to the target and contribute directly or indirectly to the excitation
of electrons in the solid. Before excitation these electrons have an energy
lower than the Fermi level. After excitation they join the incident beam, and
can therefore undergo elastic and inelastic collisions themselves. This is
known as the cascade process.
At some point in their paths these electrons may reach the free surface of
the target and escape into the vacuum. If the escaping electrons belonged
initially to the primary beam they are called inelastically backscattered
electrons (or, simply, backscattered electrons). If they were dislodged from
the solid they are called secondary electrons. Since electrons are indistin-
guishable particles, this twofold classification is rather arbitrary. How-
ever, it is possible to a certain extent to separate these two contributions
experimentally (Gerlach et al., 1970).
The fundamental datum that we desire from the experiment is the
angular energy distribution of the electrons emitted from the target; how-
ever, experimental results about it are very scarce. The most frequently used
result is, undoubtedly, the energy distribution or its derivative curves. With
grid analyzers the electrons are captured in a large solid angle which can be
of the order of 2.n sr. When other spectrometers such as the cylindrical mirror
analyzer (CMA) are used, the electrons are collected in a rather small
acceptance angle.
In the energy spectrum of the electrons emitted by the target, several fine
structures emerge from the background. The elementary mechanisms that
create them are very diverse in nature (characteristic energy losses caused by
creation of elementary excitations in the solid, interband transitions, plasmon
decays, Auger transitions, etc). Because they represent specific responses of
the solid to an external excitation, they are fundamental clues to the nature of
the sample. In this article, only the information provided by Auger transitions
are considered, although some of the other elementary mechanisms are
included to give a general description of the secondary-electron emission
process (especially in computer-simulation models).
C . Contents
This article is divided into seven major sections in addition to the
introduction (Section I). In Section 11, some general definitions, such as
cross sections, mean free paths (MFPs), attenuation lengths, and escape
depths, are given. Stress is placed on what constitutes, according to the

authors, the links and the differences between these different quantities. Then
we present quantitative theoretical evaluations of the inelastic (Section 111)
and elastic (Section IV) MFP for a randiun-jellium model of the solid. In
this model the ions (randium) are randomly distributed in a free electron
gas (jellium).
The jellium contribution to the inelastic MFP can be calculated by means
of the dielectric function scheme, and the randium contribution can be
obtained using the classical binary-encounter theory of Gryzinski (1965a,b,c)
or with the help of the generalized oscillator strengths.
The elastic collisions that scatter the excited electrons result from their
interaction with the ions, which may be regarded as scattering centers.
Calculation of the elastic MFP by expanding the electron wave function in
partial waves is discussed only very briefly.
The various cross sections and MFPs defined in Section I1 are used as
fundamental data in computer-simulation models (Section VI) of the
secondary- and Auger electron emissions. For the authors of the present
article, that justifies the emphasis on discussion of methods of calculation
of these quantities.
Section V is devoted to a consideration of Auger transitions regarded as
elementary mechanisms. Classification of these transitions is considered in
Section V,B and evaluation of their energies in Section V,C. Special attention
is given to line-shape analyses of Auger transitions involving the valence band
because of the interest generated by this topic, especially in view of
quantitative measurements (Gaaren st room, 1 98 1 ).
Section VI gives a quantitative but approximate description of the
creation and emission of Auger electrons. In Section VI,A the presentation
is made on a phenomenological basis. This description is quite usual, and the
equations which are obtained constitute the fundamental expressions for a
further development of quantitative AES (cf. Section VI1,C). One of the most
important factors introduced in such a development is the backscattering
factor, and it is defined in Section VI,A.
Section VI,B is devoted to a theoretical quantitative analysis of Auger
emission by using computer-simulation models. In the first part, several
original results obtained on a pure A1 target are shown in order to test the
validity of the principal assumptions made in the phenomenological ap-
proach. Moreover, evaluations of the backscattering factor are given and
discussed. In the second part of Section VI,B we extend to the case of more
complex samples the evaluations of the backscattering factor. The major
part of this work was done by Shimizu and his co-workers.
Here it is important to stress the fact that results obtained by a simulation
method are only as good as the model. Therefore, the use of simple models,
166 M. CAILLER, J. P. GANACHAUD, AND D . ROPTIN
necessary to reduce the time and cost of a simulation, could lead to signhcant
errors: Thus such models must be checked.
In Section VII the different problems experienced in experimental quanti-
tative AES are reviewed. The adjustment of all the measurement devices and
the correct use of the apparatus must be carefully checked. The results of the
round robin realized by the ASTM 42 Commission show (Section VI1,A) that
in the absence of such checks, very important differences can be observed
between the experimental values.
Several operating procedures have been developed in AES. For instance,
it is usual in AES for the secondary-electron signal to be measured as a
function of the secondary-electron energy. Then, by derivation it is possible
to make more apparent the fine structure of this secondary-electron current.
Other operating modes have been used, such as the measurement, as func-
tions of the primary-electron energy of either the total secondary-electron
current [appearance potential spectroscopy (APS)] or the elastically reflected
primary-electron current [disappearance potential spectroscopy (DAPS)] .
These diverse operating modes are briefly sketched in Section VI1,B.
The various ways of quantifying experimental Auger analysis are pre-
sented in Section VI1,C. There have been few attempts to correlate the
experimental and theoretical values on an absolute basis. In most cases, the
investigators use a relative calibration technique. In the internal calibration
technique, the bulk composition of the sample is known. Assuming from the
start that the surface elemental concentrations are the same as in the bulk, it
is possible to define a sensitivity factor for each element of the sample. For a
given element, this sensitivity factor is equal to the ratio of the intensity of one
of its Auger peaks to the concentration of the element. Assuming then that a
variation of the surface concentrations has occured, the new Auger peak
intensity will give the new surface concentration of the element.
In the external calibration method, the sensitivity factors are defined for
the pure elements. In such a case, application of the elemental sensitivity
factor, without correction for the so-called matrix effects, to the quantitative
elemental surface analysis of a multicomponent sample will lead to errors. A
procedure for correction of the matrix effects was developed by Hall and
Morabito (1979) on the basis of the phenomenological description of
Section V1,A.
'
Before analysis, the samples have to be prepared. In any case the mode of
preparation raises some questions, at least about its influence on the results
of the analysis. These questions are reviewed in Section VI1,D. Because of its
practical importance, preparation by sputtering is considered separately
in Section VI1,E.
During analysis, the sample is submitted to the action of an electron
beam, which can induce a modification of the composition of the target

surface. The principal disturbances, associated with stimulation by electron
beam, are the following processes (Section VI1,F):
(1) adsorption or desorption of species at the surface;
(2) dissociation of species at the surface;
(3) dissociation of the solid sample itself; and
(4) diffusion along the surface or within the bulk.
In some extreme cases, therefore, AES could possibly destroy the system it
hopes to analyze and thus would not work. More usually, there is a risk that
contamination of the sample can occur.
Although this article is rather lengthy, it omits such very important
questions as theoretical evaluation of Auger line intensities or crystallog-
raphic angular effects in Auger emission. There were no special reasons for
these omissions; it was simply a matter of choice. If the reader is interested in
further information on theoretical evaluation of Auger line intensities, he will
find much to interest him in the original McGuire papers on this question.
11. GENERAL
DEFINITIONS
A . Cross Sections
Let us consider the interaction between a parallel beam of No mono-
energetic particles per unit area and a scattering center (e.g., a charged
particle). If dn is the number of particles scattered without energy loss per
unit time into the solid angle d o = sin 8 d 0 d 4 around the direction (0,4)
(Fig. 2), the elastic differential cross section a(0,4)is defined by
dn = N0a(8, 4) d o (1)
Similarly, it is possible to associate a particular differential cross section
with each type of reaction between the parallel beam and the target particle.
If dn is the number of incident particles that reacted with the target in a
particular way characterized by initial parameters ( a , ,pi, . . .) and final
parameters (af,&, . . .), the differential cross section or(cli,pi, . . .; orf, fit, . ..)
for this reaction is defined by
dn = Noa,.(ai, pi, . . .;u,.,pf, . . .) dor, dp, (2)
where the subscript r denotes the type of reaction (or channel). The energy of
the incident particles can be considered to be one of the parameters of
the reaction.
168 M . CAILLER, J. P. GANACHAUD, AND D. ROPTIN
FIG.2. Geometry of particle scattering.
For N , scattering centers, if we neglect the coherence effects between the

waves scattered by each center, the above differential cross sections have to
be multiplied by N , . For the reaction considered, the total cross section or is
deduced from the differential one by integrating over all final parameters
or(Mi, Pi, * * *) =
s br(Ni, P i , . * *; ut > P i > .. dP, (3)
B . Mean Free Puth, Attenuation Length, and Escape

Depth of the Electrons
There is some difficulty about properly defining these different quanti-
ties. We shall thus consider the mean free path (MFP) to be essentially a
theoretical notion which can form the subject of calculations, whereas
attenuation lengths and escape depths are considered only a s experimental
quantities. In some idealized cases, the MFP and the attenuation length could
be identified, but more generally, the determination of an MFP from atten-
uation-length data is possible only in an approximate way, by introducing a
physical model.
1. Mean Free Path

An MFP is generally associated with a given reaction by considering a
Poisson process. For that, one can consider a flow of N particles having
identical dynamic parameters (energy, direction of propagation) impinging
normally on an elementary layer of thickness dz a t a depth z in the target. In
this layer, these particles can undergo a reaction characterized by the total
cross section or(i),where i denotes the set of initial parameters of Eq. (3).
Along the infinitesimal path d z , according to Poisson's law, an incident

particle can undergo only one or no reaction with the sample. So the number
of particles which have undergone this reaction by crossing the layer is
dN' = NN,o,(i)dz = -dN (4)
where N, is the number of scattering centers per unit volume of the target.
Equation (4) can be written
d N = - N dz/A,.(i) (5)
where
&(z) = [N,o,(i)]-
is the M F P of the identical incident particles for the channel labeled by
subscript r.
Now, one can consider a flow of No particles, characterized by the initial
parameters i, impinging normally on a target of thickness t , in which they
can undergo only reactions of type r. The residual amplitude of the beam
which is transmitted through the target but which conserves its initial
dynamic parameters is
N = No exp[ - t/A,(i)]
Between two collisions, the mean distance traveled by an incident
particle is
( z ) = N,' J: z dN = A,(i) (8)
Usually, different channels are open to the incident particles, and between
two collisions of type r , collisions of another type may occur and cause the
trajectories to be different from straight lines. Equation (7) can be extended
to this case if the direction of propagation is no longer included in the set of
conserved dynamic parameters. For instance, if this set is restricted to the
initial energy E , only, the probability that a particle will run path 1 in the
target without undergoing the reaction r is
It is important to note that the definitions leading to Eqs. (7) or (9) apply
to the case of a homogeneous target, whereas Eq. (5) could be used even in
the case of a heterogeneous sample or for the surface layer of a solid. In these
cases N , may vary with the depth z and & may become z dependent. To the
contrary, attenuation lengths and escape depths are considered to be experi-
mental data where surface and volume effects are not distinguished but can
be defined and measured for any target.
170 M. CAILLER, J. P. GANACHAUD, AND D. ROPTIN
Some authors prefer using the concept of inverse mean free paths (IMFPs)
A;'(i) = N p , ( i ) owing to their additive character. However, in what follows
we promulgate the MFP notion, which is, in our opinion, a more direct one.
2. Total M F P
If we consider the set of possible reactions which can take place in the
target, a total MFP AT can be associated with an incident particle of energy
E , through the relation
G W O ) = Cn;'Wo) (10)
r
From Eq. (4) and owing to the additivity of the IMFPs, it is possible to
evaluate the probability
pT(EO) = exP[ -l/AT(EO)] (1 1)
of an incident particle traveling a path 1 in the target without undergoing
any collision.
The total MFP includes both elastic (noenergy loss) and inelastic effects.
However, the additivity of the IMFPs can be used to separate a total inelastic
MFP effect from the elastic one.
3. Attenuation Length
Quite frequently, the attenuation length is assimilated into the total
inelastic MFP. In fact, this identification is always approximative. Thus,
differences in the interpretation of attenuation lengths for hot electrons have
led to some controversy (Oxley and Thurstans, 1975a,b; Collins and Gould,
1975).
Reviews on measurement techniques and data have been given by Powell
(1974) and by Seah and Dench (1979). The principal experimental methods
are (1) the transmission technique; (2) the overlayer technique; and (3) the
semi-infinite-target technique, although low-energy electron diffraction
(LEED) and extended X-ray absorption-edge fine structure (EXAFS)
experiments can also give some informations on MFPs.
The first method can be considered the basic technique for measuring
attenuation lengths. In these experiments (Kanter, 1970; Stein, 1976), the
number of electrons transmitted with no energy loss through a film (self-
supported or not) is related to the number of incident electrons by
NT = NO exP[ - t/&] (12)
where A, is the attenuation length of the target. Consequently, the quantity A,
is obtained by measuring the dependence of the transmission coefficient
N T / N o on the thickness of the target. If the aperture of the analyzer is ideally
narrow, the measured attenuation length can be assimilated into the total
(elastic and inelastic) MFP. However, this condition is never realized. The
so-obtained apparent attenuation length can be interpreted as the MFP
between collisions in which the electrons are scattered either elastically at
large angles (i.e., out of the aperture) or inelastically.
Interpretation of the measurements also raises some questions related to
the variation of the thickness and of the morphology in thin films. According
to Stein (1976), it appears that the shortest apparent attenuation lengths are
the most representative of the ideal situation and can be accounted for by
inelastic scattering alone.
Some special information can be given about partial MFPs, e.g., by
varying the temperature of the target. Thus by separating temperature-
dependent effects from the others, Kanter (1970) obtained an evaluation of
the electron-phonon MFP.
The overlayer technique resembles the transmission method to the extent
that it consists of measuring the attenuation in a thin overlayer of an electron
beam of given energy created in the substrate by an excitation mechanism
(UV or X rays in general), either directly (photoemission) or indirectly (Auger
emission). Basically, the electrons created in a direction 8 are supposed to
have a straight-line path, traveling over a distance t/cos 8 in the overlayer,
according to an attenuation law given by
NT(@= No(0) exp[ - t/(A, cos O)] (1 3)
As for the transmission method, the dimension of the aperture of the
analyzer plays an important role in the measurement. This question con-
cerning energy analyzers has been investigated by Shelton (1974) and by
Norman and Woodruff (1978).
A rather complete bibliography on the overlayer technique can be found
in papers by Powell (1974) and Seah and Dench (1979). The principal types
of measurements concern
(1) X-ray photoemission (Battye et al., 1974, 1976a,b; Szajman et al.,

1978; Evans et al., 1977; Hattori, 1977)
(2) UV photoemission (Peisner et al., 1976)
( 3 ) Synchroton radiation photoemission (Norman and Woodruff, 1977,
1978)
(4) Auger electron emission (Palmberg and Rhodin, 1968; Seah, 1972,
1973; Gallon, 1969; Jackson et al., 1973). Using this method, Auger electron
emission of both the overlayer and the substrate is measured. In the latter two
papers, an analysis of the dependence of Auger intensities on the thickness of
the overlayer is proposed that assumes a model of layer-by-layer growth of
the deposit on the substrate.
The use of spin-polarization measurements to distinguish the electrons

emitted by the substrate from those emitted by the overlayer must also be
cited (Pierce and Siegmann, 1974).
The semi-infinite evaporated-target technique has been used by Powell
et al. (1977) to measure electron attenuation lengths in beryllium and
aluminum. The principle of this method is the creation of electrons in the
target by an excitation source and the measurement of the emission yield. In
this paper, the excitation source was a proton beam and the induced emission
that of Auger electrons.
It is still supposed that until they escape into the vacuum, the electrons
follow a straight-line path. Thus, the current collected by the analyzer is
proportional to the attenuation length. Of course, the estimation of these
attenuation lengths requires a knowledge of the instrumental characteristics
and of some other physical parameters such as the density of atoms in the
solid, the ionization cross section for a given proton energy, and the
fluorescence yield.
The method used by Koval et al. (1978) is intermediary to the semi-infinite
and the overlayer techniques. It consists in measuring the intensity of the
peak of the elastically reflected electrons from semi-infinite metals, say 1 and
2, and from thin layers of metal 1 deposited on a substrate 2. The depth
dependence of the reflection coefficient yields information on the attenuation
length if we suppose that the electrons suffer a very localized reflection either
in the overlayer or in the substrate but otherwise have a straight-line tra-
jectory. Along their paths the electrons can undergo inelastic collisions, and
the beam is attenuated according to an exponential law versus the effective
path traveled in the overlayer. Some measurements of the electron-plasmon
interaction MFPs can also be realized by determining the relative intensities
of the characteristic loss peaks and of the elastic one.
It is possible to evaluate inelastic effects from LEED experiments by
studying either the widths of the Bragg peaks or the LEED intensities. In
layer-by-layer techniques of LEED theories the attenuation effects are
usually incorporated as an imaginary part for the component of the electron
wave vector perpendicular to the surface via a complex optical potential
(Stern and Sinharoy, 1972; Payling and Ramsay, 1977) or via a complex
electron self-energy (Duke and Tucker, 1969).
Inelastic effects have also been introduced by a spherical attenuation
coefficient in the interpretation of EXAFS experiments (Ashley and Doniach,
1975; Lee and Beni, 1977). This wave-mechanical aspect makes somewhat
difficult the interpretation of these attenuation coefficients in terms of MFP
or of attenuation lengths. This is the type of difficulty pointed out by
Feibelman (1974) in his comments on theories of photoemission.
4. Mean Escape Depths

Let us now consider an inner distribution of monoenergetic electron sources
in a solid (e.g., Auger electron sources). Some of these electrons can pass
through the surface potential barrier and escape into the vacuum if their
energy and normal momentum are sufficient. Their mean escape depth he is
defined as the depth for which the escape probability, without energy loss,
is lle.
An evaluation of the escape depth necessarily makes use of a model.
Usually, the monoenergetic electron sources (of energy E) are supposed to be
uniformly distributed up to a maximum depth z, and to create n electrons
per unit volume and per unit time. Moreover, electrons are supposed to
follow straight-line paths up to the surface and to be attenuated according to
an exponential law. These are precisely the assumptions introduced to
define the attenuation length. Thus, the number of electrons escaping into
Jr
the vacuum is, per unit area and per unit time,
N = (n/47c)m
J
: Joz^ exp[ - ~ / ( A ~ cos
( E 0)]
) sin 0 d0 d+ dz (14)
where 0, is the half-angle of the emergence cone

cos 0, = (W/E)"' (15)
and W is the work function of the solid. After angular integration, one
obtains (Fitting et al., 1978)
N = fn J
0
W J Z / A ~ E,
, w)dz
where we is the escape probability.
If z , goes to infinity and if W/E goes to zero, the depth dependence of w e
is well described by the relation
we = exp( - 2z/Aa) (17)
In this case, the mean escape depth is approximately equal to half the
attenuation length.
THEORYOF
111. DIELECTRIC INELASTIC COLLISIONS OF
ELECTRONS
IN A SOLID
For the electron energies used in Auger spectroscopy, a rigorous

description of the inelastic interactions would be rather sophistocated. In
fact, it is a many-body problem that includes all the electrons of the incident
beam and of the target. An essential simplification of this problem is to
consider that the solid reacts as a whole to an external probe, which is here
the incident-electron beam.
With this in mind, the dielectric theory of the response of the solid can be
considered as having a sufficiently general frame. However, from a practical
point of view, it is advisable to break the dielectric response function into
several separate contributions, which makes description easier. Afterward,
we shall consider principally the case of metals and very briefly that of
semiconductors and insulators. Normal and “d-electron” metals are dis-
tinguished. By normal metals we mean those for which the energy states are
well separated into a valence (or conduction) band and core levels which can
be considered as nearly atomic states.
The d-electron metals include the transition and noble metals, for which a
separation of the electrons, as in the case of the normal metals, is difficult if
not impossible. Indeed, the electrons coming from the d atomic states,
called d-electrons, are rather localized on the ionic cores but have energies
in the range of the valence band.
A . General Equations
Let @&) be the initial wave-packet state for a particle incident on a
many-particle system in its ground state. Here K represents the four-vector
[K, iE(K)] and x is (r, it).
If the single-particle state predominantly characterized by the four-
momentum K is well defined, its lifetime z is high and can be evaluated from
the relation (Engelsberg, 1961)
T-’ = (2/h)Im M ( K ) (18)
In this relation M ( K ) is the four-dimensional transform of the self-energy
operator M ( x , x’)of the incident particle.
For a periodic system, @&) and E(K) are the wave function and the
energy of the band-structure theory (Adler, 1963, 1966), respectively. For a
translationally invariant system, OK(x)is a simple plane wave
E ( K ) = E,(K) = h2K2/2m (19)
M ( K ) = iA s4 W(q)G(K- q ) r ( K ,q) exp( - iw6) (20)
where q represents (a; io).In Equation (20)6 is a positive infinitesimal; W(q )

and G ( K - q ) are the Fourier transforms of the effective interaction operator
and of the electron Green’s function; and T(K, q) is the vertex function which
can be considered a sum of the functionals of G and W (Minnhagen, 1974,
1975). At the lowest order of development, T ( K ,q) = 1.
For a screened Coulombic interaction,
W(q)= (4ne2/q216- ’(q, 0) (21)
where t-l(q, o)is the dielectric response function of Martin and Schwinger
(1959; cf. also Falk, 1960).
If the incident particle is described in the free-electron approximation,
G(K - q) = [E,(K) - ho - Eo(K - q) + id’sgn(lK - q1 - k , ) ] - ’ (22)
Taking T(K, q) = 1 and letting 6 and 6’ approach 0, the imaginary part of
the self-energy becomes
Im M[K, Eo(K)]= -
a19J0 4
do Im[ -t-’(q, o)]
EOB) - Eo(K - 9)
x a[, - h
In this case, the inverse MFP can be written
2-1 =-
m 2 Im M[K, Eo(K)] -
--me 2 Im V[K, Eo(K)]
h IKI h hlKl h (24)
where V is the optical potential. Eqs. (23) and (24) correspond to the first
Born approximation and were developed by Quinn (1958, 1962) for the
free-electron gas.
The response of the solid is then entirely incorporated into the dielectric
function, a very general expression of which is (Kittel, 1967) the following:
4xe2
e-1(q7m) = -~ ~ ~ ( @ b ~ ) q I @ a ) ) 2 ( -k
hq2 b
Oba
( m + %)-’
-(o- m b a + iYb)-’] (25)
where 3, is the density fluctuation operator for the particles, and yb is a
small positive quantity. In Eq. (25), hwba= (Eb - E,) represents the differ-
ence of energies between excited states I @ b ) and the ground state 1 Oa). The
frequencies w b a depend on q and tend to be grouped in more or less extended
bands, every one giving a contribution to Im M (Ritchie et al., 1975). In the
case where 6 - depends only on the modulus Iq 1 of q, which is later written q
except where otherwise stated, the angular integration of Eq. (23) is direct.
176 M . CAILLER, J . P . GANACHAUD, AND D . ROPTIN
For simple (or normal) metals we can consider two independent con-
tributions to the imaginary part of the inverse dielectric function. The first
comes from the valence electrons and is well described by a free-electron gas
model (cf. Section 111,B). The second comes from the core electrons and can
be described in terms of the Bethe’s oscillator strengths [Bethe, 1930; lnokuti,
1971; Inokuti et al., 1978; (cf. Section lII,B)]. It can also be accounted for by
a classical description such as that developed by Grysinski (1 965a,b,c).
For transition or noble metals, the d electrons add their contribution to
that of the s-p electrons of the conduction band. Thus the free-electron gas
model is no longer suitable. Moreover, local field effects must be considered
in the calculation of the d-electron contribution (Wiser, 1963; Nagel and
Witten, Jr., 1975). Possible ways to approximate the sum of these two con-
tributions are given in Section II1,C. The ionization of the inner shells can
still be described as in the case of normal metals.
B. Normal Metals
As indicated above, the response of the conduction electrons is well
described by a free-electron gas model and is given, for instance, by the
Lindhard (1954) dielectric function EL(q, w). Some more refined dielectric
functions taking into account exchange and correlation effects in the electron
gas or finite lifetime of the elementary excitations have been developed (Hub-
bard, 1955; Singwi et al., 1968, 1970; Vashista and Singwi, 1972; Mermin,
1970).Results on the M F P obtained by using these dielectric functions can be
found in papers by Ashley and Ritchie (1 974), Ganachaud ( 1 977), Tung and
Ritchie (1977),and Ashley et al. (1979).The corrections brought to the results
obtained with the Lindhard dielectric function are rather small (nearly 10%).
A description of the many-body effects in an interacting electron gas has
been given by Hedin and Lundqvist (1 969). Numerical results using their
model can be found in several papers (Lundqvist, 1967, 1968, 1969a,b; Bauer,
1972; Lezuo, 1972; Shelton, 1974; Penn, 1976a).
1. Separation of the Contributions of the Plasmon Mode and

of the Valence-Electron Gas Individual Transistions
A remarkable feature of normal metals is that they show a well-charac-
terized collective oscillation (or plasmon) mode. It is thus possible to consider
for the valence electrons two different contributions to the M F P : the
plasmon mode and the individual transitions. This is now discussed in terms
of Eqs. (23) and (24) and Lindhard’s dielectric function.
Near the plasmon-dispeflion curve defined by the condition el [q, w,(q)]
= 0, t , can be expanded in powers of the difference [ w - w,(q)], and c2 is
FIG.3. Energy- and momentum-transfer domain in the Lindhard approximation. Domain

of individual excitations in the solid: I and 11.
1 --f hw = (h2/2m)(qZ + 2qk,), 3 - hw = (fiZ/2m)(2Kq- 42)

2 hw = (h2/2m)(q2- 2gk,), 4 + h o = (h2/2m)(2gk,- 42)
an infinitesimal number. Then, to a first approximation of the expansion,

the electron-plasmon inverse MFP is written
A:;, =
(26)
Here we have written K for I K 1, and fl(x) = 1 for x > 0 and 0 for x < 0.
Except near the plasmon threshold, qmax is equal to the cut-off wave
vector qc for which the plasmon line enters the individual transition domain
(Fig. 3). The individual transitions of the valence-electron gas are restricted in
Lindhard's approximation to the domain of the (0,q) plane limited by the
parabolas o = (h/2rn)(q2+ 2qkF) and o = (A/2rn)(q2 - 2qkF) and by the
restrictions 0) > 0 and o < h-'[E,(K) - E F ] . The electron-electron in-
verse MFP can be obtained from Eqs. (23) and (24) by integrating
Im[ - E ; (4, a)] on the individual transition domain. Figure 4 shows the
results obtained for both contributions to the MFP in aluminum.
A more detailed analysis of the inelastic collisions can be made; for
instance, the energy-loss function associated with the individual processes is
given by
4(4= (27)
178 M. CAILLER, J . P. GANACHAUD, AND D . ROPTIN
I
.
0
.
100
.
.
200
.
E (eV)
FIG. 4. Mean free paths in aluminum: (1) between two individual excitations Ae-e; (2)
between two collective excitations 2e-p,. Reprinted with permission from North-Holland
Publishing Co., Amsterdam.
In the assumption of an infinite lifetime of the elementary excitation, 4(0)

presents a singularity when the plasmon line enters the region of the in-
dividual collisions. In a more realistic description this singularity is softened
by plasmon damping (Fig. 5).
For a given value of O, several momentum transfers can occur. The
probability for a primary electron to transfer a momentum q to an electron
of the solid is
4(q, 0) = 4 - Im( - [& 4- 1 (28)
2. Collisions on the Inner Shell Electrons

If in Eq. (25) the values of w are restricted to positive values and if y b goes
to zero, one obtains
Equation (29) has to be integrated over q and w as indicated in Eq. (23). For
low values of q, Eq. (29) does not depend on the direction of the wave vector.
Because the inner subshell binding energies are well separated, it is possible
to associate with each subshell an oscillator strength I,&) according to
Eq. (30):
(30)
where Z,, and En,are, respectively, the occupation number and the binding
energy of the band (nl)supposed flat; ER is the Rydberg energy; and a. is the
radius of the first Bohr orbit.
hw (ev)
FIG. 5. Energy-loss function 4(0) for an individual collision in Al; dielectric function:
1, Lindhard (1954); 2, Singwi et al. (1968): 3, Mermin (1970) (y = 0.1).
By neglecting the q dependence of the oscillator strengths, the simple

ionization inverse M F P is written
where qminand qmaxare the roots of Eq. (32):

q2 - 2Kq + (2mEnI/h2)= 0 (32)
Expression (31) could be compared with the usual Bethe total cross
section (Bethe, 1930)
one obtains
where cn,is in cm2 and En, in electron volts. In the validity domain of the
Born approximation the agreement with experimental values is obtained by
fitting b,, and C,, (Burhop, 1940; Arthurs and Moiseiwitch, 1958).
Some modifications to the Bethe formulation have also been proposed,

but they are still in the general line of the Born approximation (Worthington
and Tomlin, 1956; Rudge and Schwartz, 1966; Kolbenstvedt, 1967; McGuire
1971a,b). Note: By using in Eq. (31) the boundaries given by Eq. (32), the
result for a primary electron of high energy will be exactly twice that of Eq.
(33) if B,, is taken as equal to En[. This factor could be representative of the
order of magnitude of the uncertainty in this cross section.
The difficulty of the quantum mechanical calculation and the uncertainty
mentioned above led frequently to the use of the classical expression of
Gryzinski (1965a,b,c)for the total ionization cross section, that is,
onl= 6.51 x 10-’4(Z,l/E~,)g(U,,) [cm2] (35)
with
g(U) = U - ’ [ ( U - 1)/(V + 1)]3’2{1 +$[1 - ( 2 U ) - ’ ] ln[e + ( U - 1)1’2]} (36)
where e cz 2.718 is the base of the natural logarithm.
This approximation considers that the collision between the incident
electron and an electron of the p21 subshell is binary. Equation (36) is interest-
ing because it gives very simply and directly a reasonable evaluation of cross
sections. In form, expression (35) can be compared with Eq. (33’).
Some semiempirical expressions have been proposed to account for the
energy dependence of the total ionization cross section. In a general way they
are empirical modifications of the expressions obtained in the Born approxi-
mation or in the binary collision approximation. Their aim is a better
agreement with experimental results (Drawin, 1961, 1963; Lotz, 1967, 1968,
1969, 1970; and Pessa and Newell, 1971).
C. &Electron Metals
A particular requirement in the study of the response function of rather
highly localized states, related to the polarization of these states, is to take
into account the existence of a local field differing from the mean macro-
scopic field. The effects of this local field on the dielectric function have been
considered by Adler (1963)and Wiser (1963)and more recently by Nagel and
Witten (1975) who have concentrated on the local field effects in the de-
pendence of the dielectric function on q. By assuming that the localized
electrons behave as polarizable charges that are strongly concentrated on
the lattice sites, Nagel and Witten have deduced a dipole approximation
expression for c(q, w).
Ritchie and Howie (1977) have analyzed the requirements coming from
the sum rules for the extension of the “optical” dielectric function to nonzero
values of q. Assuming that the experimental function Im{ -[t(O, o ) ] - ~ }
obeys the sum rules, then using a procedure similar to that of Ritchie and
Howie, Cailler et al. (1981) have approximated the loss function by
Im { - [&, o)]- } = M," Im { - [@,M ~ O )-] ] (37)
with
~f = (1 + hq)-' (38)
where b is a constant to be determined (in copper, h N 0.05 - 0.1 A). The
shape of the curve is the same for each value of b and is in agreement with the
behavior generally assumed for the MFP curve.
Two theoretical evaluations of the contribution of the d-like states to the
self-energy of the energetic electron have been made (Ing and Pendry, 1975;
Beni et al., 1976). Between the two estimations there is a factor of 2, which
can result from differences in the treatment of the screening of the electron
interaction. These results are not directly comparable with those deduced
from Eqs. (23) and (37) because the experimental dielectric function includes
the responses of the delocalized electron and of the 3d electrons. But it is
possible to combine them with a free-electron gas model MFP to obtain a
total inelastic MFP for delocalized and d electrons. With these conditions a
general agreement is obtained for copper by taking b N 0.05 - 0.1 A.
D. Semiconductors and Znsulators

In this section we shall give only a few indications of some calculations
made for semiconductors and insulators. Emerson et al. (1973) studied the
electron slowing-down spectrum in silicon and used the dielectric function of
Callaway (1959) and Tosatti and Parravicini (1971) to describe the valence-
band contribution.
Tung et al. (1977) also studied the electron slowing-down spectrum in
Al,O, by using, to describe the response of valence electrons, a model
similar to that used by Fry (1969) in which the valence electrons are repre-
sented in the tight-binding approximation in their ground state and by
orthogonalized plane waves when they are in excited states. Each atom of
aluminum and of oxygen gives, respectively, three and six electrons to the
valence bands. So, per molecule, there are 24 electrons shared out between
two bands, one occupied by 15 electrons (mean binding energy ho, N 9 eV)
and the other by nine electrons (ho,N 29 eV). The contributions of the
inner shell electrons were described by generalized oscillator strengths.
Penn (1976b) has suggested describing the 24 valence electrons of Al,O,
as if they were free and proposed for them the MFP
2, = E,(K)/{a,[ln Eo(K)+ b,]} (39)
where E,(K) is the electron energy expressed in electron volts and a, and b,
are two numerical constants.
E. Depth Dependence and Anisotropy of the M F P

Because a large proportion of the Auger electrons come from the first
atom layers of the surface, the question arises whether there is a depth
dependence of the MFP in this surface region. This aspect is usually neglected
in quantitative theories of Auger emission.
Gersten (1970) and Gersten and Tzoar (1973) introduced the limit
assumption of a S-type optical potential strictly localized at the surface. They
concluded that its effect was only noticeable in the range between surface-
and bulk-plasmon threshold energies for photoemission calculations. In
calculations for aluminum, Feibelman (1973)has shown that near the surface,
the increase of the total IMFP due to surface plasmon excitations is com-
pensated by the reduction of the bulk-plasmon creation probability. So, at
least within a simple model of surface effects, the assumption of a spatially
nonvarying IMFP is not in serious error, but is was suggested that this
constant IMFP had to persist some distance beyond the solid surface in an
improved theoretical description.
Rasolt and Davis (1980) evaluated the anisotropic contributions to the
imaginary part of the optical potential caused by the broken symmetry due
to the presence of surface. They concluded that these anisotropic effects
could be of the same order of magnitude as the isotropic effects. However, as
shown by these authors, these modifications do not bring major changes in,
for instance, LEED I-V theoretical profiles.
The importance of these effects has not yet been stated for the Auger
domain. However, one may hope that they do not represent the major
cause of uncertainty in a quantitative description of the Auger emission.
Additional work is requested to confirm this assumption. In what follows,
they are ignored.
F . Some Comparisons of Theoretical Evaluations

and Experimental Data
All of the inelastic mechanisms are important for a quantitative inter-
pretation of the Auger emission. In the previous sections, the stress has been
essentially on the MFP concept; however, energy-dissipation effects also
play an important role. For instance, the collisions with the inner shells
contribute an important part to the stopping power of the primary beam and
to the spatial variation of the Auger source function.
Ab initio MFP calculations must represent some standard references.

This is the reason for the previous, rather extensive, theoretical discussion.
However, most of the inelastic MFP evaluations have been made within a
jellium model, and more empirical estimations are obviously very useful
whenever theoretical values are lacking.
A substantial amount of MFP data have been published, for instance, by
Powell (1974), Penn (1975), and Seah and Dench (1979). Ionization cross
sections have been reviewed by Powell (1974, 1976) for K an L subshells.
The theoretical evaluations obtained by Gryzinski's expressions show
good agreement with other estimations using different formulations (Drawin,
1961) or Lotz (1967, 1968, 1969, 1970), i.e., within about 10%. These esti-
mations also compare rather well with the experimental determinations (cf.
Hink and Ziegler, 1969, for the K shell of Al, for instance).
The collisions with the inner shell electrons can essentially play a role in
the higher energy domain of the MFP curve. At low energies individual and
collective excitations of the delocalized electrons determine its shape. There
is apparently a somewhat universal behavior for the inelastic MFP in quite
different materials.
For a metal, the arguments are easy to understand. Near the Fermi level,
the Pauli exclusion principle models this shape. The decrease of the MFP is
approximately given by E-' (the zero-energy reference being the Fermi level).
The oscillator strengths for the excitation mechanisms are then gradually
exhausted as the energy grows. Thus, after a flat minimum for E values
ranging in the 50-100-eV region, the MFP increases according to an E/ln E
approximate law. In fact, most of the collision mechanisms can be described
by partial cross sections involving an E-'(AE)- ln(4EIAE) general term
whenever a mean excitation energy AE can be defined for the corresponding
process (Powell 1974).
For Penn (1976b) the general behavior of the total inelastic MFP can be
described by
I, = E/[a(ln E + b)] (40)
where a and b depend on the solid investigated. Two contributions can be
separated, one coming from the valence electrons and the other from the
core inner shells:
A,' = A;' + A,' (41)
where A,, for a free-electron-likematerial, can be written in a form very similar
to that of AT but with suitable parameters a, and b,, which have been eval-
uated by Penn for various electron concentrations; A, also has the same form,
and the corresponding parameters a, and b, have been given by Powell.
184 M . CAILLER, J . P . GANACHAUD, AND D. ROPTIN
TABLE I
OF MFP ESTIMATIONS
COMPARISON FOR ALUMINUM
Penn Seahand Seah and

E - E, Cailler ez al. (1976b) Dench (1979) E - E , Cailler et al. Penn Dench
(ev) (4 (‘4 (4 (eV) (4 (4 (4
25 4.1 - 4.8 500 12.0 12.1 11.8
50 3.4 - 4.3 150 16.2 16.5 14.5
100 4.4 - 5.4 1000 20.3 20.6 16.7
250 1.4 7.3 8.4
Another approach, which is a rule of thumb, has been given by Seah and
Dench. From an important compilation of all the published data on the
inelastic MFP, these authors propose a general relation
L = ( A / E 2 ) + BE”2 (42)
Different values for the parameters A and B pertaining to different groups of
materials (elements, inorganic compounds, organic compounds, adsorbed
gases) have been proposed.
It appears that the most reliable way to present a universal law is to
express the MFP in monolayers. This quantity A,,, is related to the usual
MFP 1 by
L = aL, (43)
where a is the monolayer thickness for the solid considered. We shall only
present some comparisons for elements for which the universal law proposed
by Seah and Dench is
A,,, = 538E-’ + 0.41(aE)”2 (44)
where the energies are expressed in electron volts and the distances in
nanometers.
We shall consider two examples: A1 and Cu. In the first case, several
theoretical evaluations have been proposed. Let us just mention, for in-
stance, since the work by Quinn (1962), calculations made by Kleinman
(1971), Bauer (1972), Shelton (1974), Penn (1976), Ganachaud (1977), Tung
and Ritchie (1977), and Ashley et al. (1979). This list is far from being
exhaustive.
Starting from Lindhard’s dielectric function, several ameliorations have
been proposed. These aspects have already been discussed in Section II,B.
The general conclusion is that all of these theories agree (within ~ 1 0 % )
for MFP-value predictions.
Table I gives a comparison between our own estimations, Penn’s calcu-
lations expressed by the general law [Eq. (40)], and Seah and Dench’s
TABLE 11
COMPARISON FOR COPPER
OF MFP ESTIMATIONS
Cailler et nl. Penn Seah and Cailler et (I/. Seah and

E - EF. (b = 0.05 (1976b) Dench (1979) E - EF ( b = 0.005 Penn Dench
W) A) (4 (A) W A) (A) (A)
25 2.8 - 4.2 500 6.9 7.0 10.0
50 2.3 ~ 3.7 750 10.2 9.3 12.2
100 2.6 ~ 4.6 1000 12.8 11.5 14.1
250 4. I 4.6 7.1
universal curve for aluminum for typical values of the energy. These figures
include the inner shell contributions.
We can conclude that there is in this case rather good agreement between
the theoretical values and the experimental results, except perhaps at low
energies, but this is not in fact the important domain for Auger transitions.
However, the universal law, based on an E1j2variation for E > 150 eV has
not the same shape as that given by the Elln E law generally predicted by
the electron gas theory.
For Cu, some elements of comparison have already been given in a
previous paper by Cailler et al. (1981), in which theoretical estimations and
experimental values were reviewed. We shall only present here some brief
comments. Table I1 shows the inelastic MFPs that we 'have obtained by
adopting an intermediate value of 0.05 A for the parameter b in Eq. (38).
These figures are compared once more with Penn's values and those of Seah
and Dench.
Our theoretical predictions are not far from Penn's values but differ
appreciably from the Seah and Dench universal law. Tung et al. (1979) have
also calculated the inelastic MFP for several solids including copper by
using an electron gas statistical model. Their values are substantially higher
a
than ours, especially at high energies ( 15 for E = 1 keV).
N
Although a universal law is certainly a very useful tool for immediate

investigations, it appears that some additional theoretical and experimental
work is still needed before a complete conclusion can be reached. For
instance, one could be tempted to reject experimental MFP values which
depart too strongly from the experimental law although they are, in fact,
correct.
COLLISIONS
IV. ELASTIC
The very weak energy losses (<0.2 ev) caused by electron-phonon

interactions can contribute significantly to the attenuation length of very slow
186 M . CAILLER, J. P. GANACHAUD, AND D . ROPTIN
electrons; however, in the energy range of Auger spectrometry these quasi-

elastic effects will play a minor role. We shall now consider the purely
elastic interaction of the electrons with the real part of the potential
surrounding the ionic cores in a solid.
In quantitative Auger analysis calculations (for instance, by Monte
Carlo methods), the ionic cores are assumed to be randomly distributed in a
highly delocalized electron medium. This randium model (Bauer, 1970) is
evidently a crude approximation which cannot account for coherent diffrac-
tion effects. The fine structures appearing in the angular distribution of the
Auger electrons have to be studied by multiple-scattering approaches, that is,
by LEED theories. However, the randium model is apparently sufficient for
describing more integrated properties of Auger emission such as the global
energy distribution of the electrons or the emission yield.
Bauer (1972) discussed the problem of the proper choice of the ionic core
potential qr). The charge redistribution coming from the delocalization of
the outer shell electrons when an atom is brought in a solid-state environ-
ment causes the failure of the free-atom approximation, especially at low
energies (c100 eV). Self-consistent potentials are then needed.
The most commonly used approximation is the muffin-tin picture. In this
model the spherically symmetrical potential V ( r ) varies inside spheres of
radius R centered on each ionic site. In the surrounding medium the true
potential is replaced by a mean term with a constant value V,.
Thus, the differential elastic cross section is given by a partial wave
analysis:
h K / m is the velocity of the incident particle; P,(cos 0) is the l-order Legendre

polynomial; and dl is the phase shift suffered by the lth partial wave.
The total elastic cross section is
m
o(E) = (4.1r/K2) C (21 + I) sin’ 6, (46)
l=O
and the elastic MFP A,(E) can be written
Figures 6 and 7 show results obtained for aluminum by Ganachaud (1977)

with a muffin-tin potential evaluated by Smrcka (1970), in complete
agreement with Baudoing’s calculations ( 1971).
It should be mentioned that the values of the phase shifts are very sensitive
to the choice of the potential. This explains the differences in 6, evaluations
by various workers [see, for instance, Pendry’s values (1974) obtained with a
Hartree-Fock exchange potential]. A simple Slater exchange term can, in
FIG.6 . Principal phase shifts in Al. Reprinted with permission from North-Holland
Publishing Co., Amsterdam.
50 100 150
E - V&V)
FIG.7. Elastic MFP in Al. Reprinted with permission from North-Holland Publishing Co.,
Amsterdam.
fact, incorporate some part of the correlation effects which are omitted from
the Hartree-Fock exchange operator and can give better MFP values. In
any case, the order of magnitude of the MFP for elastic collisions clearly
indicates that this is a strong mechanism which can play an important role in
the backscattering of the primary electrons and, consequently, in the source
function of the Auger electrons as well as in their escape process.
V. AUGER
TRANSITIONS
IN A SOLID
A. Introduction
An Auger transition is part of a global process involving the excitation
and relaxation of a solid. The excitation step leads to the ionization of a core
electron. The vacancy created in an inner shell is then filled by another

electron which can come from (I) an outer shell of the same ionic core or (2)
from the valence band. The excess energy is brought to another electron of
the solid, which can itself originate from these various states.
It is usual to assume that excitation and relaxation steps are independent.
So, the probability of an Auger emission is just the product of the vacancy-
creation probability and that of the relaxation process. This assumption has
been recently questioned (Abraham-Ibrahim et al., 1978, 1979). In fact, some
coherence effects may occur in the creation of the deep holes. This question is
certainly of basic importance. However, we shall adopt hereafter the simpler
decoupling scheme, and we shall consider the relaxation step of the Auger
transition in particular.
Section V is devoted to classification of Auger transitions (Section V,B);
theoretical evaluation of the Auger electron energies (Section V,C); and
study of the line shapes, when at least one of the electrons involved in the
Auger transition comes from a valence-band state (Section V,D).
B. Classification of Auger Transitions

Classification of purely intra atomic Auger transitions makes reference to
the atomic multiplet structure, and this is considered first. Transitions in
which at least one of the jumping electrons is in the valence band are
considered in Section V,B,2.
1. Purely lntraatomic Auger Transitions

The initial and the final state of an intraatomic Auger transition present
one and two holes in otherwise filled subshells. However, an equivalent
classification of these one- or two-hole states can be obtained by considering,
instead of holes, one or two electrons in an otherwise empty subshell. This
makes classification easier.
The atom, in a neutral or in an ionized state, can be described by a many-
particle wave function +, which is an eigenstate of the Hamiltonian
operator I9
H $ = E+ (48)
Different A's are associated with the different states of ionization of the
atom. A rather complete description of the Hamiltonian has been given by
Larkins (1975), but is not reproduced here. Let us just consider the
nonrelativistic form
-
Ze2
-
ri
+ -21 1
'' e2
j=l
-
rij
+ [(ri)li.si
j#i
For each of the Z' electrons (Z' = Z for a neutral atom), four contribu-
tions have been separated. They represent, respectively, the kinetic energy
operator, the Coulomb attraction of the nucleus, the Coulomb interaction
with the other electrons, and the spin-orbit coupling.
Determination of Auger electron energy entails the resolution of Eq. (48)
for both the initial and final states of the atom. This is, in fact, a very intricate
problem, and it is necessary to proceed by steps.
A basic simplification consists in neglecting the spin-orbit coupling and
in replacing the mutually antagonistic second and third terms of Eq. (49)by a
potential having a spherical symmetry around the nucleus. This central field
approximation is now briefly discussed.
a. Central Field Approximation. In this approximation, I? is replaced
by a sum of one-particle decoupled contributions
where U ( r i )is a spherically symmetric potential. For closed subshells, this

assumption is strictly true; for open subshells, this is only a very convenient
approximation.
In this scheme, the functions II/ can be represented by antisymmetrized
products of the eigenfunctions of (Slater determinants), each one-electron
state being characterized by the usual five quantum numbers n, 1, m,, s, and
m,.However, the total energy of the atom, described by the Hamiltonian of
Eq. (50),does not depend on the two numbers m, and m,. Thus, several Slater
determinants correspond to the same eigenvalue of 8'. In the general case,
the order of the degeneracy is given by
D = n
1
c,+z (51)
where z, is the occupation number of the subshell of orbital quantum

number 1.
Only incomplete subshells contribute to the degeneracy. For instance,
for an ( n p 2 )configuration, two electrons occupy the subshell (n, 1 = 1) and
all the other subshells, including (n, I = 0) are complete. Thus, the degeneracy
is C; = 15.
b. Coupling Modes. The approximate Hamiltonian I?' of Eq. (50)
differs from the original I? by two contributions:
’
where <(Ti) = (h2/2m2c2)r; [dU(ri)/dri] (if in this corrective term the
potential is evaluated in the central field approximation). The perturbation
theory is usually brought into play to include the effects of the corrections
H’ and H ” .
In extreme coupling schemes, only one corrective term is taken into
account, either only 8’(L-S or Russell-Saunders coupling) or only 8”
( j - j coupling), whereas, in the intermediate coupling mode, the corrections
coming from both 8’and fir’ are introduced. Intermediate coupling is
apparently the most reliable, but it is well known that for light atoms
Russell-Saunders coupling is sufficient, whereas j - j coupling works well
for heavy atoms.
s,
c. Russell-Saunders Coupling. Let 2, 3 be, respectively, the orbital,
spin, and total angular momentum operators for the atom (with, for instance,
2= ziJi). The Hamiltonian operators fil and fitseparately commute with
z2, s,.
E,, S2, and Thus, the eigenstates of this set of four operators can be
used to construct the eigenfunctions of 8’+ B‘,which are characterized by
the four corresponding quantum numbers L, M , , S, M , . They are represented
by the notation 2s+1X.
The multiplicity is 2S + 1 and the symbolic letter X is one of the sequence:
S, P, D, F, G, H, I,. . ., according to the value of L: 0,1,2,3,4,5,6,. ... The
degeneracy associated with the incomplete Hamiltonian fil is partially
raised in this new basis.
Let us return to our example of an ( n p 2 )configuration. It is easy to prove
that the closed subshells make a zero contribution to L, M , , S , and M s .
Letting I , = 1, = 1 and s1 = s2 = 1/2 be the quantum numbers for the two
electrons of the open subshell, L, M , , S, Ms can vary between the following
limits (by a unit increment):
11, - 121 < L < 1, + 12, -L < M, <L
(53)
IS1 - s2( < S < SI + ~ 2 , -S < Ms < S
So, the 15 degenerate states associated with 8’are split into
5 degenerate states ‘D ( S = 0, L = 2)
9 degenerate states 3P ( S = 1, L = 1)
1 state 1
s (S = 0, L = 0)
the degeneracy being given by (2L + 1)(2S + 1).

d. j - j Coupling. In this coupling scheme, the Hamiltonian is
i i
This operator remains separable, and the corresponding eigenfunctions

can be represented by Slater determinants involving the one-electron
functions which are now the eigenstates of &. This latter Hamiltonian
commutes with j^z and TZi
[with corresponding eigenvalues h2ji(ji + 1)
and hrnji].
For an (np2)configuration, the electrons of the closed subshells still play
no role in the classification,The two remaining electrons are characterized by
the two quantum numbers j , and j , , which can vary (by unit increment)
between the limits
11, - slJ < j , < 1, + sl, 11, - s 2 J < j 2 < 1, + s2 (55)
The 15 degenerate states of the original Hamiltonian 8' can now be
split into
1 state with j , = j , = 112
8 states with j , # j , , j , , j 2 = 112 or 312
6 states with j , = j , = 312
Each one-electron state is denoted by naj where a is a symbolic letter asso-
ciated with the 1 value. For instance, np,,, represents a state of the n shell
with 1 = 1 and j = 112.
The corresponding subshell is denoted by X , where the label z is 1,2,3,4,
5,6, 7,. .., for the states slI2,pIl2,k l 2d3/2,
, d5,,, f 5 / 2 , f7,,, ..., respectively.
For example, the states 2sIl,, 2plI2, 2p3,,, 3d3,,, 3d,,, belong, respectively,
to the L,, L,, L,, M,, M, subshells.
e. Intermediate Coupling. By applying both correction 8'and 8'' to A',
the degeneracy can be more completely raised. When 8''is applied after H',
2 and $ no longer commute with 8. On the other hand, application of 8'
after a'' sets up an electrostatic coupling between the electrons of the atom,
which can no longer be described as individual particles with a given total
angular momentum.
However, in this domain, corresponding to the intermediate coupling,
the total angular momentum for the atom f still commutes with A. These
two operators have the same eigenstates, which are denoted by using a double
notation coming from the two extreme coupling modes, the L, S designation
being completed by a right-hand subscript label equal to the value of J .
The term J varies by unit increment from L - S ] to J L + S J ,and for a
I
2p2 configuration, the 15 states are now broken into
1 state L2L, ( , S o )
3 states L,L3 ('PI)
5 states L2L3 (ID,)
1 state L3L3 (3P0)
5 states L3L3 (3P2)
where the multiplicity is now 25 + 1.
192 M. CAILLER, J . P . GANACHAUD, AND D . ROPTIN
f. Designation of Auger Transitions. This designation concerns the initial

and final states of the atom. We have already mentioned that from the point
of view of classification, the states involving holes in otherwise closed sub-
shells could be replaced by states with electrons in otherwise empty subshells.
This is not true for energy calculations. The corresponding modifications
have been indicated by Condon and Shortley (1970).
Following Larkins's notation, an electron configuration such as (2s2)(2p4)
corresponds to a [2p2] hole configuration. Accordingly, an Auger transition
in which a vacancy in the L, subshell is replaced by a double hole in the final
state M,M, ['Po] is denoted by L,M3M3 ['Po] in the intermediate coupling
scheme.
However, more imprecise designations often refer to the extreme modes
of coupling and, for example, the notation L,M,M, ['PI or L2M3M3or
even L23M23M23 can be used. In the latter case, all transitions from L, or L,
states to M,M,, M2M3, and M3M, states are supposed to be included.
We should mention that the L, S designation is sometimes completed by
indicating the parity of the atomic state, which is the parity of the sum of the
Ii values. An odd state is denoted by the symbol ''0." For instance, a
multiplet arising from a (2s, 2p) configuration has an odd parity. This is
indicated by the notation 3P0(Slater, 1960).In an Auger process, the parity of
the atomic state does not change. Thus, some transitions can be forbidden
[e.g., the L,L3( ,PI) line of the KLL spectrum] by this parity rule.
On the other hand, collective excitations can take place, for instance,
plasmon effects in a metal. They can lead to an extra satellite structure which
will obscure the purely atomic spectrum. Good examples are aluminum
(Dufour et al., 1976) and sodium and magnesium (Steiner et al., 1978).
Another possibility is that of satellite peaks due to a double ionization of the
atom, the Auger transition taking place in an atom with two holes initially
present. This kind of transition has been presented in a paper by Melles et al.
(1974), for instance, for the L2,, Auger emission bands of Al, Si, and P.
It is a bit difficult for a noninitiate to the multiplet-structure theory to
understand the rather complex Auger spectrum notation. Table I11 indicates
the possible Auger transitions for given final configurations with their
complete designations. For example, the ninth column corresponds to
[np,n'p] final configurations ([pp'] for the sake of brevity); that is, to
[ X , , Xi], [ X , , Xi], and [ X , , X i ] states where X and X ' are symbolic
letters (L, M, . . .) representing the subshells. When n = n', the two holes of
the final state are said to be equivalent, and the configuration is denoted by
[p' ] (eighth column). The rows correspond to the intermediate coupling
notation, and the numerals in the table give the degeneracy of the final states.
To illustrate how Table I11 can be used, let us mention, for instance, that
the [L2, L2], [L2, L3], [L3, L3], [M,, M,], [M,, M3], [M3, M31r "2,
TABLE 111
DESIGNATION
AND DEGENERACY
OF AUGER-TRANSITION
FINALSTATES
12 14 16 18 23 23
l1 l1 13 15 17 19 33 33
s2 ss' sp sd sf sg p2 pp' pd pf pg d2 dd' df f2

~ ~~ ~ ~ ~ ~~
IS, 1 1 1 1 1 1 1
3s, 3 3 3
'PI 3 3 3 3 3
3p0 1 1 1 1 I 1 1 1
3p, 3 3 3 3 3 3 3 3
3p* 5 5 5 5 5 5 5 5
'D2 5 5 5 5 5 5 5 5
3D1 3 3 3 3 3
3D2 5 5 5 5 5
'D3 7 7 1 7 7
IF3 7 1 1 1 7
3F2 5 5 5 5 5 5
3F3 7 I 7 7 7 7
3F4 9 9 9 9 9 9
'G4 9 9 9 9 9
3G3 7 7 7 7
3G4 9 9 9 9
3G, 11 11 11 11
'H, 11
3H4 9
3H5 11
3H6 13
'16 13
N2], . . . final states correspond to the eighth column and to the 'So, 3P,,
3P,, 3P,, ID,, . . . lines. A good example of what appears in the true Auger
spectrum is given by Fig. 8a for the [L3M23M23]lines of copper according to
Mariot (1978). Some additional examples are also given (Fig. 8b-d) for
[MI, M z , ~ ] , M,,,], [N,,,, N,,,] final configurations.
Determination of Auger line energies and calculation of these Auger
transition probabilities have been the object of numerous papers. It is beyond
the scope of the present article to give a full account of such a vast theoretical
field. The interested reader will find detailed information in the books by
Chattarji (1976) or by Sevier (1972) and also in the original papers (Asaad
and Burhop, 1958; Asaad, 1963a,b, 1965a,b; Rao and Crasemann, 1965;
McGuire, 1970, 1971a,b, 1974, 1977, 1978; Larkins, 1971a,b, 1973a,b, 1974,
1976a,b, 1977a,b,c, 1978a,b; Crotty and Larkins, 1976,1977; Bambynek et al.,
t
:L-. a
0. .Experiment
L 780
FIG. 8. Auger transition line shapes. (a) Cu (L3-Mz.3M2,3); (b) Cu (L3-M1M2,3);(c)

Cu (L3-Mz.3M4,5);(d) Ag (M4,5-N4,5N4,5). Reprinted from Mariot (1978) by courtesy of the
author.
+Ms-Nes N+ +M
4 +%-
-N
(d)
FIG. 8. (Continued)
1972; Price and Rao, 1972; Shirley, 1973; Yin et al., 1972, 1974; Kowalczyk
et al., 1973a,b, 1974; Ley et al., 1973; Kim et al., 1976a,b;Mariot and Dufour,
1977a,b; Aksela et al., 1980; Vayrynen et al., 1980; Chen et al., 1980a,b,c).
However, some general considerations about Auger transition energies are
given in the next section.
g . Conjiguration Interaction. Let us return to the second step of the cal-
culation leading from the Russell-Saunders coupling to the intermediate
coupling. The corrections introduced to account for the A’’ perturbation
s2,
were evaluated by using eigenfunctions of the four operators E2, t,, and
s,. However, as indicated above, the full Hamiltonian r? still commutes
with J2 and Jz.
In fact, in the former discussion the analysis was made configuration by
configuration. This is an unnecessary supplementary restriction. For
instance, a given configuration was associated with a given J value. Other
configurations corresponding to the same value of J have to be introduced to
describe the many-particle wave function of the atom. This interaction of
configurations a priori brings into play many Slater determinants. In practice,
only functions corresponding to nearly the same energy are mixed.
Let us consider the example of the KLL spectrum, or more precisely, the
function corresponding to J = 0. In practice, the associated subspace was
separated into two subspaces related to either a [2s2] or a [2p2] configura-
tion. The first subspace was spanned by the [L,L,, ‘ S ] Russell-Saunders
eigenstate, whereas the [L2L2,‘S] and [L,L,, ,PP] states gave a basis for the
196 M . CAILLER, J . P. GANACHAUD, AND D . ROPTIN
second subspace. (In the intermediate coupling these three terms became
(L,L,,'So]; [L2L2, 'So]; and [L3L3, 3Po], respectively.) When the dif-
ferent configurations are treated separately, two distinct secular equations
have to be solved, the secular determinants being of rank 1 in the [2s2] case
and of rank 2 for the [2p2] configuration.
Taking account of the configuration interaction (CI) consists now in
introducing a three-dimepsional subspace spanned by the three basis wave
functions of the Russell-Saunders scheme. So this leads to consideration of a
secular determinant of rank 3. Returning to Table 111, we may remark that
the [2s2p] configuration yields a 3P0term. A priori, this term would also
mix with the above basis functions. However, this 3P0function does not have
the same parity as the three other eigenstates and cannot be brought into
play in the above CI calculation.
2. Lines Involving the Valence Band

A designation with respect to the multiplet structure is not suitable for
this case. When the two holes of the final state are in the valence band, the
transition is simply denoted by CVV, where C indicates, in the j-j coupling
mode, the subshell in which the initial hole has been created and V represents
valence (or conduction). When only one hole is present in the valence band
in the final state, the other hole appearing in a C inner subshell, the line will
be designated CC'V (KL,V or KL,,,V, for example).
Owing to the information it can give about the chemical environment of
atoms on solid surfaces, the shape of these lines has been the object of several
studies. These are reviewed in Section V,D. Although in the present study we
are mainly concerned with metals, we should first like to examine the special
case of ionic crystals.
The ionic crystals can be described by their band structure (see Cheli-
kowski and Schliiter, 1977, for SO,), but they have also been depicted in
terms of molecular orbitals (Fisher et al., 1977).
These molecular states are designated by the irreducible representations
of the symmetry group of the molecule. However, some of these molecular
states are very close to atomic states of a given component of the compound.
They are rather localized, and their energy level is nearly equal to atomic
levels of this component. They can be called quasi-atomic molecular states.
The CVV Auger lines of the compounds have been often denoted by
referring to the molecular orbital (MO) designation (Bernett et al., 1977;
Ramaker et al., 1979b). Some were also considered to be interatomic or
crossover transitions (Citrin et al., 1976)if the initial vacancy was in a core of
a given component and the final vacancies were in the quasi-atomic molecu-
lar states of another component.
It could therefore be conjectured that there are different physical

processes leading to two holes in the solid valence band (CVV Auger tran-
sitions) or in molecular valence states or yet in quasi-atomic states (inter-
atomic transitions). In fact, for solids they are simply different pictures of the
same mechanism. From this point of view, a comparison between the densities
of states calculated by Chelikowski and Schliiter (1977) and the XPS
valence-band assignation in terms of M 0 s is enlightening.
The case of Si02 is used to illustrate the terminology of the ionic
compound CVV Auger transition. First, it is necessary to indicate the nature
of the atom and of the subshell in which the initial vacancy occurred. Con-
sidering the case of S O z ,it is, for example, the subshell L2,3of a Si atom. The
central atom is bound to four oxygen atoms in a tetrahedral configuration.
The corresponding symmetry group is denoted Td. This symmetry group
has 5 irreducible representations :
2 one-dimensional representations, A and A ,
1 two-dimensional representation, E
2 three-dimensional representations, T , and T2
The MOs centered on the Si atom have the symmetries of these five
representations; for instance, the various orbitals of the representation Al
are denoted la,, 2a,, 3a1, . . . in the order of the increasing energies. Similar
notation is used for the MOs of other representations.
In the particular case of SiO,, the MOs la,, 2a1, . .., It,, 2tz, . . . are
nearly identical to the Si(ls), O(ls), Si(2s), Si(2p) atomic orbitals. In the
absence of s, p hybridization for the oxygen atoms (which should be weak
according to Yip and Fowler, 1974),the MOs of the valence band correspond
to the Si and 0 atomic orbitals, as indicated in the accompanying tabulation,
Number of
orbitals
4a, +-+ O(2s) and Si(3s) 1

3t, t* O(2s) and Si(3p) 3
-
5a1 tt 0 ( 2 p f ) and Si(3s)
4t1 0(2p+) and Si(3p)
le, 5t,, I t
1
3
8
where le, 5t2,and It are antibonding orbitals corresponding to a 2p oxygen

orbital, with some mixing with the Si orbitals. The ( 2 ~ ' )orbitals of the 0
atoms are those pointing towards the central Si atom. The other 2p orbitals
give the eight orbitals of the le, 5t2, and Itl configurations.
On the whole, ther? are 14 MOs in the valence band: that is, the number
of the atomic orbitals from which they are issued. To be more rigorous, there
198 M. CAILLER, J. P. GANACHAUD, AND D. ROPTJN
is in fact some hybridization of the s, p orbital of oxygen, and each of the 4al,
3t, states is in fact a linear combination of 0(2s), Si(3s),and Si(3p).
Auger transitions involving the valence band are thus designated by
Si(Ll)Si(L,,,)4a, or Si(L,,,)4a13t,. Each oxygen atom is bound with two
silicon atoms. Its environment has the symmetry of the CZygroup; A,, A,,
B,, B , are the irreducible representations of this group. The KVV lines
starting from a vacancy in the 0 atom are denoted by K(0)6a,6al or
K(0)6a15b,(S), where S indicates that it is a singlet state.
As far as the energy level of at least one of the MOs playing a role in the
Auger transition is essentially determined by that of an atomic orbital (AO),
one can speak of interatomic or crossover transitions and designate them by
A(C)A(C')B(C") or A(C)B(C')B(C)(cf. Citrin et al., 1976).
C. Kinetic Energy of Auger Electrons: Purely Atomic Transitions

To evaluate this kinetic energy we shall assume that, the initial hole being
created, the Auger transition follows two successive and independent steps :
(1) annihilation of the initial hole in the C subshell of atom A; and
(2) creation of the two final-state holes.
As indicated earlier, these two holes can appear in C' and C" subshells of
the same atom A or in the valence band.
To make the formulation more explicit, we shall adopt notation
corresponding to the first case. An extension to a more general situation
would not be too difficult. Moreover, we shall assume that a classification of
the states adapted to the intermediate coupling scheme in the central field
approximation can be used.
The first step of the Auger transition in an isolated atom can be described
by the equations:
A(Z - 1;ATC) + e- + A(2) (56)
AEl = E [ A ( Z ) ] - E[A(Z - 1; A,C)] = -Ek(Z+Z - 1; ATC) (57)
Equation (56) indicates that the A atom has initially one hole in the C
subshell (A,C)and that by collecting an electron, it returns to its ground
state; Z is the atomic number of the neutral atom. Equation (57) relates to
the energies and indicates that the energy difference AE, between the initial
and the final states described by Eq. (56) is the opposite of the ionization
energy for an electron in the C subshell (the arrows indicate the dynamical
nature of the transition). An energy equal to this binding energy is thus
liberated, a fraction of which is carried by the Auger electron. Here we shall
consider that the electron by which the hole is annihilated was initially at the
Fermi level or, in other words, that the binding energy is referred to that level.
The second step of the Auger transition obeys the following equations:
A(Z)+A(Z - 2 ; A,C'; A,C", x) + 2e- (58)
which specify that the two holes of the state have taken place in the atom A,
one in the C' subshell, the other in the C" subshell. These two holes interact
and, assuming that the central field approximation still holds for this two-
hole situation, one can make reference to the multiplet structure in the
intermediate coupling scheme to designate the final state of the atom; here
x represents the more complete notation "+lX J ' -
If the holes were created in the valence band, A,C' and A,C" would be
replaced by P. Similarly, for interatomic transitions, -and would
become B,C'and B,C".
The kinetic energy of an Auger electron, measured from the Fermi
level, is given by
E(A; CC'C"; X) = - AEI - AE, (60)
Asaad and Burhop have used the formalism leading to Eq. (57), (59),and (60)
to evaluate the kinetic energy of the Auger electrons of the KLL atomic lines.
Within the frozen orbital (or sudden) approximation, they obtained
the result
E"(A,KLL';x)=E;(Z+Z- l;K)-E$(Z+Z- 1;L')
- Ei(Z*Z - 1 ; p)- F(A,L'L";x) (61)
In Eq. (61) the quantities E; represent the binding energies in the atom
measured from the vacuum level, whereas the last term F represents the
interaction energy of the two holes in the final state. This interaction energy
has been evaluated by Asaad and Burhop in terms of Slater integrals and the
spin-orbit coupling constant. The binding energies were set equal to the
first ionization energies, measured for the neutral atom. We should note that
these binding energies differ from the electronic orbital energies, the creation
of a hole (or its annihilation in the reverse transition) causing a perturbation
in the electronic cloud. The most important effect is that for the subshell in
which the hole is created and for the more external subshells. This pertur-
bation has been called dynamic relaxation by Shirley (1 973).
The relation between the binding energy and the orbital energy for the C
subshell of atom A is written
E,A(Z * Z - 1 ;ATC) = ~ ~ A,C)
( 2 -; Edr(Z --t Z - 1; C C ) (62)
where dr is the dynamic relaxation (Fig. 9). For a metallic environment this
relation must be modified to account for the screening of the holes by the
electrons of the valence band.
Kowalczyk et al. (1973a,b) and Ley et a/. (1973) supposed that for this
environment (denoted by M in what follows), the screening of a hole was due
to the occupation of an additional quasi-atomic state appearing around the
vacancy in the atom, according to Friedel’s model (1952). Equation (57) has
to be replaced (Hoogewijs et al., 1977; Larkins, 1977a,b) by
AEl = Ef(Z-tZ - 1 + n ; A,C, AC,)
= E[M(Z - 1 + n ;A T ; AC,)] - E[M(Z)] (63)
In Eq. (63), C, is the atomic orbital of the atom, ionized in the C subshell in
which the screening electron takes place.
Larkins (1 977a,b) considered that this screening did not involve a unit
electronic charge but a fractional charge n set equal to 0.6 per hole, whereas
Hoogewijs et al. chose n = 1. In both cases, the screening charge for the
two holes was set equal to twice the screening charge of one hole. In these
conditions, Eq. (59) is replaced by
AE, = E[M(Z - 2 + 2n; A,C’; A,C”; X ; AC,, AC,)] - E[M(Z)] (64)
and the kinetic energy of the Auger electron can be, at least in principle,
calculated from Eqs. (62)-(64), provided that the energies of the right-hand
members of Eqs. (63) and (64) are evaluated in the adiabatic approximation,
for instance, by a self-consistent field method.
However, it is possible to follow a different approach initially proposed by
Shirley (1973) for an atom. He introduced the concept of a static relaxation
energy which was adapted to the case of metals by Kowalczyk et al. (1973a,b)
and Ley et al. (1973)and was then ameliorated by Kim et al. (1976a,b)to take
Reference level
T
--L- ------
75,(=-1,A7)
-c (Z,A,C)
FIG. 9. Relation between the binding and orbital energies: for an atom, the reference level
is the vacuum level; in a metal, it is the Fermi level.
QUANTITATIVE AUGER ELECTRON SPECTROSCOPY 20 1
into account the dynamic relaxation corrections for the energies. More
recently, Hoogewijs et al. (1977) and Larkins (1977a,b) returned to this ap-
proach with different calculation procedures but equivalent points of view,
as shown by Larkins (1978a,b).
By adopting the main features of the Shirley and Kim et ul. scheme, but
with slight adaptations, one may consider the second step of the Auger
process to have three stages.
(1) Creation of a first hole in the C' subshell, with screening in the C,
subshell, according to the energy relation
6EI = E [ M ( Z - 1 + n ; A,C'; AC,)] - E [ M ( Z ) ]
= E,M(Z+Z - 1 + n ; p, AC,) (65)
(2) Creation of a second hole in the C" subshell, with screening but with
no repulsion between the two holes:
6E2 = E [ M ( Z - 2 + 2n; A T ; c; ACI; AC:')]
- E [ M ( Z - 1 + n; A T ; AC,]
(3) Intrgduction of the repulsion between the two holes:

6E, = E[M(Z -2 + 2n; A,C' ; A,C" ; x;ACI ; AC;)]
- E[M(Z - 2 + 2n; A T ; A,C"; ACI; AC:')]
= F(mA,C";
; x) (67)
This last energy term is assumed to be independent of the screening. Con-
versely, one assumes that the screening does not depend on the multiplet
structure. Adding Eqs. (65)-(67) evidently gives back Eq. (64).
As first noted by Shirley, stage (2) does not exactly correspond to the
ionization of the C" subshell, which itself obeys the relation
6E; = E[M(Z - 1 + n; A T ; AC,)] - E[M(Z)]
=Ez(Z+Z -1 + n ; A T ; AC,) (68)
where E f is related to the orbital energy by
EF(Z+Z - 1 + n ; A T , AC,) = - e M ( Z ; AC")
- Ed,(Z+Z - 1 + n ; A T ; AC,) (69)
whereas
6El = -P(Z - 1 + n ; A T , AC"; ACI)
- &,(Z - 1 + n - 2 - 2 + 212;n, A,C"; AC;; AC:) (70)
So, from Eqs. (66) and (68) we obtain

6E2 = E F ( Z + Z - 1 + n ; A T , AC,)
+ E [ M ( Z - 2 + 2n; A T , A T , ACA; AC:)]
- E [ M ( Z 1 + n ; A X AC,)]
-
- E[M(Z - 1 + n ; A,C"; AC,)] + E[M(Z)] (71)

which, from Eqs. (69) and (70), can also be written
6E2 = E F ( Z + Z - 1 + n ; F,AC,) - EZ - AEE (72)
with
E,"r = c'(Z - 1 + n ; A,C'; AC"; AC,) - tM(Z;A,C") (73)
AE:, = Ed,@ - 1 + n + Z - 2 + 2n; A T , A T , AC,'; AC:)
-Ed,(Z+Z- 1 +n;AF,AC,) (74)
Equation (73) defines the static relaxation energy, according to Shirley, and
Eq. (74) defines the dynamic relaxation-energy correction.
In the atomic case, the sum E;r + AE& has been called the cross-relax-
ation energy R; by Hoogewijs et al. It is identical with the adiabatic
relaxation correction K(C'C") of Larkins (1976a,b).
From Eqs. (60),(63), (65)-(67), and (72) we obtain
E'(A,CCC"; X) = Ef(Z --t Z - 1 + n ; ATC; AC,)
- E!(Z + + n ; A,C'; AC,)
Z-1
- E ; ( Z + Z - 1 + n ; A,C", AC,)
+ F(A,C', A,C"; x) + EZ + AEYr (75)
The sum of the last three terms is identical with the term UM(A,C', A T ,
x) from Larkins. This author sets Ci = Cf = C, and, for that case,
UM(A,C';A,C"; X) = E[M(Z - 2 + 2n; AC'; AC"; X; AC,, AC,)]
+ E[M(Z)] - E[M(Z - 1 + n ; AC',AC,)]
- E[M(Z - 1 + n ; Ac",AC,)] (76)
For an isolated atom this relation would become
U a ( A T ; AC";X) = E[A(Z - 2; AC'; AC"; x)] + E[A(Z)]
- E[A(Z - 1 ; AC')]- E[A(Z - 1; AC")] (77)
and could be evaluated by energy self-consistent field (SCF) calculations. It
could also take the form:
U a ( A T A T , X) = F a ( E , AC",X) - Ka(ACT;A T ) (78)
and U'(AC', A T , x) can be calculated by referring to U " ( E ,AC", x)

according to
UM(AC', AC",X) = U"(AT, C,
x) + GU(AC'; A T , X) (79)
where the 6U corrective term is defined by substracting Eq. (77)from Eq. (76).
Various estimations of this corrective term (assumed to depend on the
nature of the element but not on the particular transition considered)
have been given by Larkins. Thus, for instance, in the equivalent core
approximation,
6U = 2n(CC,)Z+' - n y C , C , ) Z + ' (80)
where the terms between angle brackets represent pair interactions for the
electrons of the C and C, or C, and C, subshells in an atom having 2 + 1 as
its atomic number.
From the above theory, it is possible to reproduce with great precision the
experimental results for KLL, LMM, MNN, N O 0 lines of a large number of
elements with an approximation of 1-2 eV when the binding energies are
known.
Larkins supposed that the (CC, ) pair-interaction term was relatively
independent of the subshell C and proposed to take for it the subshell of the
penultimate occupied shell in the atom. The choice of n and C, raises some
questions. For Larkins a good approximation is obtained by giving n the
value 0.6 by taking for the screening orbital the first available orbital in an
atom having atomic nuqber 2 + 1 and by using results obtained by the
nonrelativistic Hartree-Fock method for the pair-interaction values. For
example, the screening orbital he chooses for zinc is 4p. Hoogewijs et al.
take the same orbital for zinc but use n = 1. On the other hand, Kim et al.
(1976)suggest that the initial configuration of solid nickel has more 3d charge
than the free atom in its ground configuration. Then the screening of a first
hole comes from the mixed occupation of 3d and 4s orbitals, whereas the
screening of the second hole takes place nearly exclusively through the
occupation of the 4s orbital.
1. Computational Aspects
From Eq. (75) the kinetic energy of the Auger electrons emitted by an
atom labeled A is
EM(A; CC'C; x)
= ~ f ( z _ + z - i +~;A,C;AC,)-E,M(Z+Z-~ +~;A,C';AC,)
~ _ _ _
- E?(Z + Z - 1 + n ; A , C ; AC,) + UM(A,C';A , C ; X) (81)
where ~- --
U'(A,C'; A,C"; X) = F(A,C'; A,C";x) + EZ + AEZ (82)
For an isolated atom, Eq. (82) is written

E"(A; CC'C"; x)
= E ~ ( z - 2- 1; ATC) - pB(z+z
- 1; A,C')
- F~(z-2- 1; A,C") + u a ( mA T 7x ) (83)

with
U"(A7, A T , X) = F ( A 7 , A,C"; x ) + g, + AE& (84)
Let us consider this case first and neglect the relaxation corrective terms
E:, +
AGr.It is the Asaad and Burhop approximation.
2. Asaad and Burhop Approximation

Basically these calculations necessitate making an evaluation of the
energies of the initial and final states involved in Auger transitions and
subtracting them. Let us consider the case of the KLL lines. For example,
using the notations given by Eqs. (57) and (59), we obtain
E ( A ;KL,L,; 'Po) = E[A(Z - 1; AK)]
- E[A(Z - 2; c;
6;3Po)] (85)
which we shall write in the simplified form
E(KL,L,; 3p0)= E[R] - E[L,L,; 3p0l (86)
--
Neglecting spin-orbit coupling, the energy of the final state is E(L,, L, ;
3P)and can be expressed via the average energy of the [2sp] hole configura-
tion associated with it, plus some additional terms. For the three [2s2],
[2sp], and [2p2] configurations, the average energies are the following
(Slater, Vol. 11, Appendix 21):
Ea,(2S2)= 2ELl + (2s, 2s), E,,(2SP) = E L 1 + EL2,3 +(& 2P)
(87)
E&P2) = 2EL,*, + (2P7 2P)
where ELIand EL2,3 are the binding energies for the L, and L2,3subshells,
respectively. In our notation,
EL, E i ( Z + Z - 1 ; AL, ) (88)
The (2s, 2s), (2s, 2p), and (2p, 2p) terms represent the interaction energies
for two 2s, one 2s and one 2p, and two 2p electrons, respectively.
The interaction energies for pairs have been tabulated by Slater for
equivalent and for nonequivalent electrons. For example,

(2s, 2s) = F0(2s, 2s)
(2P, 2P) = F0(2P, 2P) - M 2 ( 2 P ,2P)
( 2 ~ 2p)
, = F0(2s, 2p) - &G1(2s,2p)
where F and G are Slater integrals defined by
Rk(ab,cd) = e2 J; J: Rala(Tl )Rflbl&21

r(
pi 1
Rnclc(r*
x Rnal&Z k3.;d r , dr2 (90)

where Fk(i,j ) = Rk(&ij) and Gk(i,j ) = Rk(ij,j i ) . The additional terms have
also been listed by Slater. Thus, for a [2sp] configuration one obtains a
triplet state for which the corrective term is - $G1(2s, 2p) and a singlet state
associated with a $G1(2s, 2p) term.
The multiplet energies, without spin-orbit coupling corrections, are
E(LILl; 'S) = 2EL, + F0(2s, 2s)
E(L,L2,3; 'P) = EL1 + + F0(2s, 2p) - $G'(~s, 2p)
EL2,a
E(L,L2,3; 'P) = ELI + EL2,3+ F0(2s, 2p) + $G1(2s, 2p)

(91)
E(L2,3L2,3 ; "1 + F0(2p, 2p) - iF2(2p, 2p)
= 2EL2,,
E(L2,3L2,3 ; ID) = 2EL2,, + F0(2p, 2p) + &F2(2p, 2p)

E(L,,,LZ,, ; '9= 2EL2,, + F0(2P, 2P) + 5F2(2P,2P)
The spin-orbit interaction corrections can be obtained from the above results
and from the spin-orbit interaction matrices given by Condon and Shortley
(1970).For example, their sl spin-orbit interaction matrix gives the following
result for the [2sp] configuration:
In matrix (92)the rows have been ordered according to the sequence 'P2, 3P1,
'PI, 3P0,and the signs have been changed from the original Condon and
Shortley formulation because we are concerned here with holes, not electrons.
The coupling parameter izl= $(EL,- EL3)and can be obtained from
experimental measurements of the binding energies.
206 M . CAILLER, J. P. GANACHAUD, A N D D . ROPTIN
The matrix obtained by adding to the diagonal terms of matrix (92) the
energies of the corresponding multiplets (without spin-orbit coupling), now
has to be diagonalized. For instance, for the 3P1 and 'P, multiplets the
matrix to be diagonalized is
[ E(L2,3L2,3;
-121ld5
3p)f (121/2)
E(L1,
-[21/&
L2,3;
] (92')
We obtain
E(L1L3; + F0(2s, 2 ~ +) +C2]
3p1)= ELI+ EL2,3
- { [fG1(2s, 2 ~ -) $1211 + i1$1])1'2 (934
E(L1L2; lP1) = E L , + E L 2 . 3 + F0(2s, 2P) + $ 1 2 ,
+ {[+G1(2s,2p) - $[,,I2 + $1$1}li2 (93b)
The final expressions for the energies of the (KL,L, ; 3P1)and (KLlL3;
'PI) Auger lines can be obtained by introducing in Eqs. (93a) and (93b),
respectively, the relations
= EL2,3 - $ 1 2 1 (944
EL2 = EL2,3 + 121 (94W
and by subtracting from E(K) the obtained results. For more complete
tabulation of the KLL spectrum the reader is referred to Chattarji (1976).
Larkins (1976a, 1977c) has tabulated the Slater integrals [F0(2s, 2s);
F0(2s, 2p); . . .; G1(2s, 2p); . . .] for all the elements with atomic numbers
between 10 and 110; For that he has used a semiempirical method. These
tables also give adiabatic relaxation cohections [K(2, 2); K(3, 3); . . .] asso-
ciated with the different atomic shells (L, M, . . .) and also solid-state correc-
tions as calculated from relation (SO). The agreement between calculated and
experimental values is, in most cases, of the order of 1 or 2 eV and never
exceeds 5 eV.
D. Auger Line-Shape Analysis

Auger transitions arc part of a global process where the atom (or the solid)
suffers an ionization-relaxation mechanism via an intermediary unstable
state. If we separate the relaxation mechanism but retain for it the name of
Auger transition, we can associate with this step the following characteristics :
(1) It is a many-body dynamical process (interaction between the two
final-state holes, dynamic relaxation-energy corrections).
(2) It is a local phenomenon. This means that, except perhaps for ionic
compounds, the two final-state holes are created near the ionic core which
contains the initial hole. Thus it might be important to take into account the
local modifications of the charge density caused by the screening of the holes.
(3) It obeys parity-selection rules and kinetic momentum conservation
(Feibelman et a!., 1977).
Considerable progress has been obtained in line-shape analysis. Interest in
such analysis stems from the possibility of using Auger spectroscopy as a
local microprobe for the chemical environment of the atomic species present
in a traget. It does not seem that a complete theory that accounts for all the
above-mentioned aspects yet exists in a form comparable with what has been
done, for instance, by Nozieres and De Dominicis (1969) for X-ray absorption
and emission.
For simple metals, Von Barth and Grossmann (1979) have shown that,
except for the edge singularities, most of the results of the many-body
theory could be recovered by a static one-body analogy, provided that the
final state was conveniently described. When evaluating matrix elements,
local densities of states are quite different in initial and in final states. For
instance, compared to the ground state, the presence of one hole gives rise to a
strong screening charge with an-s symmetry in the low-energy part of the
valence band. This screening charge represents the onset of a bound state
which could separate from the low-energy tail of the valence band if the
potential due to the hole were sufficiently strong. No effect coming from this
screening charge appears in the experimental emission XL,,, spectra for
sodium. However these spectra are in good agreement, at least for their
shape, with thedensities of states of the atom in its ground state. This obser-
vation is the basic confirmation for the final-state rule given by Von Barth
and Grossmann (1979).
The same rule would lead, for Auger emission, to the search for a corre-
spondence between the line shapes and the two-hole local density of states
of the final configuration. In the general case, for CVV lines, this local density
does not result from a simple self-convolution of the one-hole valence-band
local density of states due to the correlations between the two holes and to
the matrix-element effects. The many-body aspect of the correlations between
the two holes has been very well analyzed by Cini (1976, 1977, 1978a,b) and
by Sawatzky (1977) and Sawatzky and Lenselink (1980).
To present this analysis, let us start from the energy relation
E [ M ( C , C’,C ” ;x)] = €[M(C)] - E[M(C‘, C”; x)] (95)
which can be written
E [ M ( C , C’,C”; x)] = &(C) - &(CT - EB(C”) + U ( C’, C ” ;X) (96)
I u i
band states bound states
Energy (arbitrary units)

(a ) (b)
FIG.10. (a) Two-hole density of states; (b) local two-hole density of states. From Sawdtzky
and Lenselink ( I 980); reprinted with permission of the American Institute of Physics.
using simplified notations with respect to Section V,C. In the last expression,
U represents the effective interaction energy of the two holes in the final state
of the x configuration.
The importance of the correlation effects can be checked by comparing
the effective interaction energy U to the width W of the x line in the valence
band. The narrower the bands (small W ) ,the stronger is this effect. On the
other hand, it is weak for crystals with strong covalent bonds such as silicon,
for which Coulombic correlations would not be so important, the screening
charges being rather delocalized.
For U >> W , the two-hole density of states departs considerably from a
one-hole density self-convolution. The band is split into two parts separated
by a gap of width U . The low-energy part contains N,,(N,, - 1) states (Nalis
the number of sites). Its width is approximately 2 W . It essentially corresponds
to states having two holes at different sites.
The high-energy band is very narrow, its width being b W 2 / U ,and
resembles an atomic band. It contains N,, states, most of them corresponding
-
to cases where the two holes are at the same site (although they are moving
in the solid).
However, as indicated above, Auger emission is a local mechanism. Thus,
the two-hole densities of states that we have,to consider to describe the
transition are principally local densities where the two holes are at the
same site.
Under these conditions, the Auger CVV spectrum presents a narrow
quasi-atomic peak and a large weak band on its low-energy side. The ratio
R of the integrated intensities for the peak and for the other part is
approximately
R = ( U / W )- 1 (97)
It is thus possible to obtain an estimation of the correlation energy from the
Auger spectrum:
U rr b ( R + 1) (98)
Sc Ti V Cr Mn Fe Co Ni Cu Zn Sc Ti V Cr Mn Fe Co Ni Cu Zn
(a1 (b)
FIG.11. (a) Two-hole effective interaction energy in the final state of the Auger transitions
L3-M4,5M4,5(U), M2,3-M4,5M4,5( O ) ,M1-M4,5M4,5(*) of the first series of the transition
metals. (b) Comparison between the two-hole effective interaction energy in the final state of
the Auger transition L3-M4,5M4,5,the width of this Auger line, and the width of the 3d band for
the first series of the transition metals: ( 0 )2r (M4,5);(*) r (L3-M4,5M4,5);(U) Ci(M4,5M4,5).
From Jardin (1981). courtesy of the author.
For copper (Madden et al., 1978), the L,,,VV lines and their satellite
structures bear a rather good resemblance to those calculated with a purely
atomic approximation (McGuire, 1978) but with the contribution of Coster-
Kronig L, L3V and L,L,V transitions, which reinforce the L,-hole popu-
lation (in a Coster-Kronig transition, one of the final-state holes is in the
same shell as the initial-state hole). M,VV and M,VV lines have a narrow
structure, in good agreement with the line-shape calculations made with an
atomic model by McGuire (1 977).
The theoretical approaches of Cini and of Sawatsky have been generalized
by Treglia et al. (1981) to the case of unfilled d-band metals. Starting from the
two-hole Green's function, correlation effects between the hole in the final
state of the Auger transition and the holes of the ground state are accounted
for by replacing the usual one-electron density of states n(E)by a one-particle
spectrum C(E) defined in terms of a local self-energy Z(E), evaluated for an
arbitrary number of holes in the band. The two-particle spectra are, in fact,
strongly sensitive to the hole population of the d band.
The values of .the effective interaction energies between the two holes
have been measured by Jardin (1981) for the first series of transition metals
and for L3cM2,?.M4,5 L3cM4,5M4,5
3 M1-M4,5M4,5
9 , and M2,3cM4,5M4,5
Auger transitions. The results for the three transitions involving the M4,5M4,5
final configuration are shown in Fig. 1 la. The effective interaction energy for
the L3-M2,3M4,5Auger transition has a slightly smaller value but presents a
quite similar variation with 2.
The effective interaction energy U of the L3-M4,5M4,5 Auger line is
compared with 2r (where r is the M4,5d-band width) in Fig. 1 Ib. As can be
seen from this figure, U exceeds 2r as soon as Z > 28. Consequently, the
Auger line must be of the bandlike type and have a width 2r for Z < 28. For
higher Z values, it must be of the quasi-atomic type. This conclusion is
confirmed by analysis of the shapes and widths of the LVV and MVV
Auger lines (Jardin, 1981).
The M,VV spectrum has, on its high-energy side, a broad shoulder which
does not appear in L,VV and M,,VV lines. This shoulder has been initially
explained by Madden et al. (1978) as being the broad structure predicted by
Sawatzky (1977). Particularly because theoretical predictions lead to an in-
tensity too weak for this structure and because it does not appear in L,,VV
and M,,VV lines, which show the same final states as the M,VV line, this
interpretation has been questioned by Jennison (1978~)and described by a
convolution of the partial s and d densities of states in the valence band. The
ratio of the intensities coming from the atomic d-d part (sharp structure in
the M,VV spectrum) and from the s-d part (large shoulder on the high-
energy side) does not match MacGuire's theoretical predictions within the
assumption of a 3s- '(d'Os) initial configuration where the initial hole is not
screened. On the contrary, Jennison has shown that the agreement could be
largely improved by considering an initial 3s-l(d1 Os) configuration where
the initial hole is screened by an s electron.
Thus, we are led to consider now the case of a weak interaction between
the two final-state holes ( U << W ) .The shape of the lines is a result of the
self-convolution of the one-hole local densities of states (including screening
effects) and of the Auger matrix-element effect.
For normal metals the screening charge has essentially an s symmetry. It
is the same for noble metals. However, for transition metals, the screening
could come from an s and a d charge mixing.
This s-screening charge reinforces the s partial local density of states in
comparison with densities having other symmetries. This can lead to an
enhanced contribution from the final s-s, s-p, and s-d configurations.
However, the contribution of the s-s configuration remains negligible.
According to Jennison, the high-energy-side shoulder of the M,VV line of
copper could result from the s-d enhanced contribution, the interaction
between the two holes being weak and giving rise to a bandlike line. We
should note that Jennison describes the screening of an initial state [here, the
3s-'(d'Os2) configuration], whereas Von Barth postulates a final-state rule.
The initial-state screening effect has been reexamined by Jennison et al.
(1 980) for beryllium. Simplifying their model somewhat, the screening charge
effect was introduced via a scaling factor for the partial densities of states
(PDOS). The best fit with experiment was obtained by taking a screening
density of a charge of pure s symmetry, which reinforces the s-s and s-p
contributions compared to the unscreened case. For lithium, the screening
effect might be important; however, it has little chance to be observed because

s-p and p-p contributions have similar shapes and also because they are
positioned at the same energy.
By considering the final-state rule, Lasser and Fuggle (1980) were led to
study the KLV Auger spectra of Na, Mg, Al, and Si. Indeed, the KL,,,V
Auger transition of sodium led to the same final state as L2,3V X-ray absorp-
tion. In considering a final-state rule, one should find in both lines the same
manifestation of screening. For the KL,,,V line it explains the presence of a
peak in the low-energy part of the Auger spectrum. This effect is even more
pronounced for the KL,V line. The difference in the importance of these
peaks can be understood from theoretical considerations involving the
angular part of the Auger matrix elements. These considerations indicate
that, for the Auger KL,V transition, the final hole in the valence band has an
equal probability of having either an s symmetry or a p symmetry. On the
contrary, for a KL,,V line, the probability of occurrence of an s-symmetry
hole in the valence band is only one-third of that of a p-symmetry hole. Thus,
enhancement by screening of the s-symmetry charge in the valence band
leads to a stronger effect for the KL,V line than for the KL,,,V line.
The case of silicon seems slightly different. Indeed, if for Na, Mg, and Al,
the s-symmetry screening charge reinforces the density of states at the bottom
of the valence band, in Si it is due to the creation of a bound state below the
valence band (at about 5 eV under the bottom of this valence band according
to Lasser and Fuggle).
It also appears from all the measurements made by Gsser and Fuggle
that the concept of a screening resulting from the occupation of one atomic
orbital, which was used in Section V,C to evaluate the kinetic energy of the
Auger electrons, could be a relatively rough approximation. In fact, the
screening effect leads to important modifications of the local densities of
states in the entire valence band.
Before the Esser and Fuggle observations it was admitted that, for
covalent crystals, the screening was rather delocalized and that this effect
might not be very important. However, in this case, the shape of the Auger
line still does not result from a simple self-convolution of the one-hole local
density of states because the Auger matrix element depends on the kinetic
momentum, especially on its angular component. Feibelman et al. (1977a,b)
advanced this important effect and presented calculations of the L2,VV
and L,L,,3V line shapes in silicon by using a Fermi golden rule. The two
wave functions of the valence band involved by the calculation of the matrix
elements were described by a LCAO approximation, with a basis composed
of (3s) and (3p) atomic orbitals centered on the various atomic sites. Owing to
the strong spatial localization of the initial hole, it was admitted that the main
contributions to the Auger matrix element was due to the very neighborhood
of the hole that is composed of (3s) and (3p) orbitals centered on the initially
ionized core (valence local density of states). A study of the selection rules for
parity and for angular mementum conservation shows that for the damping
of an L2,3hole, 46 channels are open, leading to a final p-p configuration,
whereas there are 24 channels for an s-p configuration and only 3 for the s-s
configuration. The angular parts of the various matrix elements are propor-
tional to these numbers of channels. Thus, the damping mode leading to the
creation of two (3p) holes in the valence band is strongly favored with respect
to the others in a L,,,VV transition.
However, the “angular momentum” character of the Si valence band
varies all along this band. It has an s type near the bottom of the band and a
p type near its top. So, the dependence of the Auger matrix elements on the
angular momentum acts as an energy-dependent weighting factor on the
(ss, sp, pp) convolutions of the s and p local densities of states. Consequently,
this causes the presence of an intense peak on the high-energy side of the lines.
In the calculations by Feibelman et al. (1977), the ppp nature of the
vacancy final states in the Auger transition was favored, which led to a much
better agreement between the calculated and the experimental lines. However,
the s-p character of the L,,,VV line was still overestimated. In fact, experi-
ments seemed to indicate that the LVV line of silicon in practice arises only
from a self-convolution of the p component of the optical density of states.
The same observation can be made for lithium and aluminum.
Jennison (1978a,b) pointed out some inconsistencies in the theory of
Feibelman et al., who used atomic orbitals to evaluate matrix elements,
whereas Wannier functions were introduced to calculate the partial local
densities of states. Consequently, the reduction of the local atomic charge due
to the contribution of part of the charges to the covalent bond was not
accounted for. This is due to the fact that the Wannier functions have no
binding charge. By reevaluating the partial local density of states for the
initial state with an atomic orbital basis, Jennison could show that the (3s)
electrons contribute more than the (3p) electrons to the binding charge. The
direct consequence was that the contribution of the (3s) states to the local
atomic density (which acts, together with matrix elements, to determine the
Auger line shape) was more reduced than that of (3p) states. Additional
calculations using the Von Barth and Grossmann final-state rule, the
screening effects, and the Auger matrix elements effects need to be developed.
However, it seems that the problem is now well outlined.
Analysis of the lines shapes for X0:- ions (where X represents the
elements of the third row of the periodic table) shows that the LVV Auger
lines of the X central atom in a tetrahedral environment present a character-
istic doublet separated by about 14 eV (Bernett et al., 1977).The origin of this
doublet is the splitting of the (3p)states of the central atom into two molecular
orbitals 4t2 and 3t,. The energy levels of these orbitals are essentially deter-
mined by the (2s) and (2p) levels of oxygen. The two lines of the doublet are
associated with (2p4t24t2) and (2p4t23t2) transitions. The (2p3t,3t2)
transition is also possible, but it has a low intensity.
These results have been discussed by Losch (1979) from experimental
Auger spectra of sulfur obtained from Cu,FeS, surfaces by segregating the
sulfur at the surface of Cu-Au alloys. A method for the quantitative analysis
of the surface composition from LVV spectra of sulfur in SO:-, SO2- 3 9 S2-
anions has been proposed by Turner et al. (1980). They compared the lines
measured for compounds of these anions with a combination of the lines
calculated for pure materials. The precision obtained was 5% for SO;-, S2-
compounds, but only 30% for SO:-, SO;- compounds.
The lines shapes of silicon and oxygen in silica has also been studied in
detail (Ramaker and Murday, 1979; Ramaker et al., 1979a,b; Ramaker, 1980)
from the experimental and theoretical points of view. Charge effects and some
dissociation due to the incident beam take place in these kinds of targets
(Carriere and Lang, 1977; Ichimura and Shimizu, 1979).
Grant and Hooker (1976)proposed to analyze the line shape of carbon to
distinguish its different chemical states at a surface (graphite overlayers on
metallic surfaces and carbides). Gaarenstroom (198 1) adapted the principal-
component analysis technique to take advantage of the modifications of the
line shapes in compounds (when the spectra of the pure components overlap).
For such cases, our own opinion is that this method could be very efficient
and quite comparable to self-deconvolution techniques (Tagle et al., 1978),
although their mathematical bases have not been completely and clearly
detailed. We shall return to this aspect in Section VI1,C.
VI. QUANTITATIVE EMISSION

OF AUGER
DE~CRIPTION
A. General Description
Let us consider a flow of N p electrons per second impinging upon the
surface of the target. A detailed analysis of the incident beam would indicate
that just before reaching the surface of the target it is spatially distributed and
shows some dispersion in energy and angle. If no(0) = no(xo,y o , Eo, no) is
the differential distribution of the primary electrons, one has
214 M . CAILLER, J . P. GANACHAUD, AND D. ROPTIN
In these relations, CEO = (cos do, 4,) where 8, is the incidence angle and c$o
the azimuthal angle measured around the normal to the surface.
By crossing the surface region, these primary electrons are slightly
accelerated and deflected because of the surface barrier of potential. Some of
them will be elastically reflected in the very first atom layers, but most of
them will penetrate the solid. Owing to the collisions in the solid, the electron
beam is scattered and slowed down. Moreover, its intensity varies because
of the cascade process.
At a mean depth z,, the distribution of the energetic electrons is given by
NflzI(l)where n,,(l) is related to n,(O) by
r
As a rule, the h( 1,O) function is quite complicated. It must account for the
slowing down and scattering of the primary beam. It must also account for
the existence of a backscattered electron beam and for the creation of the
secondary electrons in the cascade process. Finally, it must account for the
roughness of the surface. However, such a description is valuable only if the
height of the roughness is small in comparison with the electron MFPs.
Clearly, and analytical evaluation of nzl(1) is quite impossible. Only simula-
tion or numerical methods, carried out on a computer, can bring some
information about this function. However, a rough approximation of h( 1 , O )
and similar functions, which are introduced in this section in a general way,
will be given later.
Focusing attention on the ionization process giving rise to the Auger
effect, the number of electrons scattered by ionization of atoms A at the
depth zl, is, per unit path length,
with
VJl) = N A ( Z 1 )[n,,(l)/cos 61 1 4 C ( E , ) (103)
In Eq. (103),oX,,(E,)is the total ionization cross section for the subshell C
of an atom A. This cross section depends on the energy of the ionizing
electron. The d z , cos-' 8, term accounts for the effective distance covered
in a layer of thickness dz, by the electrons propagating in the direction
n,(cos el, 4, ). The N , term is the numerical density of atoms A. For the
sake of simplicity, it has been tacitly assumed that NAonly depends on the
depth z1 (the sample is laterally homogeneous).
Now, it is assumed that after ionization, the atoms A can create Auger
electrons via a transition denoted as ( A , x). At a depth z , , the creation rate of
these Auger electrons is, per unit length,
dNt;"/dz, = Np J d(2)vtiX(2)
where vti"(2) can be deduced from v,, (1) by using a general equation similar
to Eq. (101) but extended in this case to five variables:
I- I-
Considering the whole generality, the function h ( z , , 2 I z l , 1) is certainly

very complex, but approximations of it are given later.
Among the N,vt;"(2) dz, Auger electrons created in a layer of thickness
dz, at depth z 2 , only some of them reach the free surface and escape into the
vacuum. Thus, their distribution just after they enter the vacuum would
be given by
rw r
where the subscript zero recalls that this distribution is obtained near the
plane z = 0.
It should be noted that in the distribution n$9"(3) most of the Auger
electrons created in the solid have lost their characteristic energy in inelastic
collisions. In fact, there are three domains in the energy distributions of
these Auger electrons. The first domain corresponds to the characteristic
Auger energy and is occupied by the true Auger electrons. The second domain
ranges several electron volts under the characteristic energy. It corresponds
to the energy-loss spectrum of the Auger electrons. Finally, the third domain
is very far below the characteristic Auger energy. The energy distribution of
the Auger electrons in this domain is flat and is hardly separated from the
background of the secondary electrons.
Depending on the analyzer, all or only a part of these electrons enter the
collector, and at the spectrometer output they have a distribution nAv"(4)
given by
nA3"(4)=
s d(3)h(4, 3)n$*"(3) (107)
The function h(4, 3) accounts for the capture of the electrons by the collector,
but also accounts for the transmission function of the spectrometer and for
the broadening due to the voltage modulation (in the derivative mode). At
the spectrometer output the energy distribution of Auger electrons is
NA3"(E4) =
s dx4 dy4 da4nA*"(4) (108)
For practical analyses, one has to simplify the preceding general descrip-
tion. These principal simplifications are now reviewed. Throughout the
presentation, all unnecessary subscripts are dropped; they are retrieved in
the final result.
(1) Let us suppose that all the primary electrons which make up the
incident beam have the same energy Epand the same direction of propagation
so that
(2) For simplicity, we neglect the acceleration and the deviation caused
by the potential barrier. For the case where the primary electrons are of high
energy, this approximation is good. The elastic reflection can be taken into
account with the help of a transmission coefficient T < 1 which depends on
the surface topography and on the parameters Epand $2, of the primary beam.
(3) The effect of the roughness of the surface is important at both the
entrance and the exit and is usually taken into account by a multiplying
factor on the intensity of the Auger peak. This effect is not well understood
and deserves much further research. Here we consider it to be included in the
factor of transmission T , at least concerning its effect on the primary beam.
Therefore, 8, is just an average incident angle.
(4) Inside the target, it can be assumed that there are two electron beams:
the primary beam and the backscattered electron beam, whose contributions
to the creation of Auger electrons can be added. In the first step, the contribu-
tion of the backscattered electrons is omitted. Then, at the end of the
calculation it is introduced in the form of a correction factor to the result. If
we neglect the effects of coherent diffraction, of incoherent scattering, and of
slowing down of the primary beam in the target, h(1, 0) becomes a simple
product of Dirac functions, thus,
with
P
Nevertheless, by introducing Eq. (111) into Eq. (103) and then introducing
the result into Eq. (102), we have
dNz,/dzl = NA(zl)(c/cOs d,,)N,,T (1 13)
( 5 ) After ionization in the subshell C, the probability of the reorganiza-
tion of atom A taking place by Auger transition, denoted x, is yA,c,x. Gen-
erally, yA,c,x is not the complement to unity of the fluorescence yield. The
difference is due to the contribution of the Coster-Kronig transitions and of
the possible cascading and filling up of the holes in the ionic core. If several
simultaneously possible channels exist, then their contributions to the cre-
ation of the Auger electron can be summed up to give Ec c ~ ~ & , ) y , . , ~ , ~ .
However, for simplicity we drop the summation symbol in what follows.
Usually, we consider that the Auger transition is local; in other words, it
takes place where the ionization has occurred. For transitions involving the
valence band, this assumption can be questioned.
We can assume that an overall compensation occurs in the bulk of the
solid, giving validity to the hypothesis of locality. However, at the surface,
because of the loss of symmetry, this assumption is more doubtful. However,
this reservation is disregarded. Thus, using the hypothesis of locality and
also separating the spatial and the angle-energy parts, we can write
w, 1) = Y*,C,xw.2 - rl )C(E,, a,)
with
Consequently,
s dE2 da2C(E,, 0,)= 1
= YA,C,XC(~~
n2)NA(z2)[Tnz,(x2 Y2)/c0s
3 5 ep]aC
dN$;"ldz2 = N, j d(W;;"(2) eP)%

=Y A , C , X ~ A ( ~ Z ) ( ~ ~ ~ / ~ O ~
(6) By integrating Eq. (116) over x2 and y,, one obtains

V 3 E 27 a,) = $;"C(E, a,) 9
with
= ~ A , C , X N A ( ~ep)aC
~)(~/~~~
In the usual quantitative description of Auger electron emission, it is

tacitly assumed that the distinction between the true Auger electrons (i.e.,
those that have kept their characteristic creation energy upon reaching the
sample surface) and the inelastically scattered Auger electrons can be made
unambiguously. Then, assuming that these true Auger electrons emitted in
the direction of the solid surface reach it after following a straight-line path
along which they are exponentially attenuated, their angle-energy distribu-
tion just before the surface is
In this expression, no,= refers to an internal distribution (denoted by the

subscript i) in the neighborhood of z = 0 and considered for one primary
electron.
(7) If one takes into account the effect of the potential barrier, Eq. (120)
represents the internal distribution of the true Auger electrons. From this,
one can derive the external distribution with the help of
nA,x
o,T (Ee, a,1 =
f dEi dQh(e, i)n2,(Ei, Q ) (121)
where the subscripts e and i represent external and internal, respectively. For a
rough surface the function h(e,i) is, a priori, very complex. For a smooth
surface, for simplification we have
h(e, i) = - $i)b{cos 8, - [ ( E cos’
~ Oi - w ) / E , ] ~ ’ ’ }
x 6(Ei - E, - W ) (122)
We usually neglect this correction and assume that Eq. (120) also represents
the external distribution [denoted by the subscript 3 in Eq. (106)l of Auger
electrons.
Nevertheless, this omission is less justified for Auger electrons of low
energy than for primary electrons. For a rough surface the result is usually
corrected by a simple multiplying factor as indicated above. Logic dictates
that the correction factor must be a function of the surface roughness
and of the Auger electron energies and their angular distributions: We
denote it as T ’.
(8) Frequently, the MFPs involved in the energy range of Auger elec-
trons are very small so that the exponential term in Eq. (120)decreases from
1 (for z2 = 0) to negligible values within a thickness so narrow that the
variation of the concentration NA(zZ)can be neglected in this domain.
Calling NA the density of atoms A near the surface, one obtains

Npn>”(E3,03) = y(NpT/cos 8,)aT‘C(E3, n,)NAn(E3) cos 83 (123)
Guglielmacci and Gillet (1980) evaluated the corrections to the atomic
density NAfor the case of a binary alloy A-B when the concentration gradient
in z cannot be neglected. If N g is the mean density of the atom A in the
escape zone of the Auger electrons, and N,(z) and C,(Z) are the actual
density and concentration, respectively, then N , and N,(z) are related by
NZ - NA(z) = (NgLA,” + Ng&,y)(dCA/dz) ( 124)
where jlA,”, J.B,y are the escape depths associated with the Auger peaks (x for A
atoms and y for B atoms). In the case of a linear variation of the concentration
with z and if LA = A,, the correction would be simply written as
NA(Z) = N T ( z - A) (125)
(9) The total primary-electron contribution to the true Auger electrons
emitted into the vacuum is obtained by integrating Eq. (123)over R, and E, .
It is recalled here that the total number of Auger electrons leaving the target
is obtained from Eq. (123) after correction by a backscattering factor.
In principle, Eq. (123) would give the angle-energy distribution of the
Auger electrons at the surface of the solid. However, it is necessary to keep
in mind that this equation was derived from the assumption of an exponential
attenuation of the Auger electrons along their straight-line escape paths. In
other words, it is assumed that there is no broadening of the natural Auger
line in the sample. Under these conditions, the natural line shape of the
Auger transition is obtained by integrating Eq. (123) over R,.
Clearly, the omission of the sample broadening is not realistic in studies
of line shapes, and Eq. (123) must be corrected to take account of this broad-
ening. Deconvolution procedures have been proposed to find the original
natural line shapes from the experimental shapes. They rest on estimations of
the so-called sample transmission function and of the spectrometer function
(Ramaker et al., 1979) and are reviewed in what follows.
(10) Another point that is disregarded in Eq. (123) is the spatial
distribution of the Auger electrons. The effect of such a spatial distribution
could be important in an absolute measurement of the Auger peak intensity
in relation to the capture of the Auger electrons by the collector. For relative
measurements, the problem is certainly less important. However, Holloway
and Holloway (1977) have shown that for a target of pure gold the ratio of
the intensities of the low- and high-energy Auger peaks depends on the size of
the primary beam. For a scanning analysis the effect was found to be as large
as 30%. They also mentioned that a misalignment of the primary beam spot
and the analyzer focal point affects the peak intensity ratio.
It is well known that compensation of the electromagnetic fields is
indispensable for measuring well the low-energy electron distributions.
Holloway and Holloway have suggested that the variations of the peak in-
tensity ratios that they observed were due to the electromagnetic field effects.
The question of the influence of the axial position of the sample on Auger
measurements was considered by Sickafus and Holloway (1 975). They
reported that the apparent energy and the intensity of an Auger peak are
sensitive functions of the sample position. They also showed that this depen-
dence is basically the result of the finite size of the exit aperture of the CMA.
Despite these instrumental difficulties, we suppose that the angle-energy
distribution of the true Auger electrons is given by Eq. (123).
(1 1) Moreover, it is assumed that this distribution is not perturbed along
the electron path from the target to the collector. Therefore, the number of
Auger electrons which are collected by the analyzer is
s
N;,” = Np dE3 dQ3n$7”(E3,Q,)
where R, is the acceptance angle of the collector.

Incidently, it can be seen that the anisotropy effects can considerably
affect the measurements. In the present simplified description, these effects
are included in the function C(E,, a3). Then, they arise only from the
creation process. However, it is clear that anisotropy effects can originate
from the transport process of the Auger electrons from their place of creation
to the surface. These effects are important for monocrystalline targets.
(12) For polycrystalline samples, the anisotropy effects can be neglected
so that
w 39 Q 3 1 = (474- C(E31 (127)
If, moreover, the Auger line is sufficiently narrow to allow any variation
of the MFP within this energy domain to be neglected, practically,
F
where EA,” is considered the characteristic energy of the infinitely narrow

Auger line.
Then, Eq. (126) is written
(13) For a four-grid analyzer, integration over $2, concerns a cone with
an aperture 0, and
N?'x = (4)-'yA,C,xNP(~'/cosOp)$A,,(E,)NAn(EA,xX1 - cosz e C ) (130)
For a CMA the integration now concerns a cone of revolution with an
aperture AO, for 0, and the integral in Eq. (129) becomes
s cos 8, d$2, = 2a
s
Bc + W c / 2 )
8c - (ABcIZ)
For a narrow aperture A@-, Eq. (131) becomes

cos 0, d(cos 0,) (131)
s cos 8, d$2, = 7c AO, sin 20,
Dejinition of the Backscattering Factor

In the very simplified description developed above, the number of Auger
electrons N;," of the (A, x) line collected by the detector is given by Eq. (129).
An actual measurement of this number gives a very different result that can be
written N;,"(exp). In fact, this difference comes from several origins. However,
it is quite generally admitted that the main reason lies in the fact that we
have neglected the contribution of the backscattered electrons to th? creation
of the Auger electrons.
To make this description more realistic, after Eq. (131) has been evaluated,
one multiplies the right-hand member of this relation by a multiplicative
factor [l + I~,,~,(E~,,, &)I, called the backscattering factor. To estimate
this backscattering factor, let us return to the general formalism.
The combined action of the penetrating primary beam and of the back-
scattered electrons (among which we shall include the secondary electrons of
the cascade) results in the substitution, for the various functions n and v used
in the above calculations, of new functions in which the backscattering
effects are included. We denote them by a prime symbol.
Thus, for instance, Eq. (1 11) becomes
nk1(1) = nz,(x1,Yl)Wl- Ep)6($2,- a,)+ n:,(l) (133)
where n,: (1) is the distribution (for one primary electron) of the backscattered
electrons.
222 M. CAILLER, J. P. GANACHAUD, A N D D. ROPTIN
A calculation very similar to that developed above leads to the replace-

ment in Eq. (126) of TT'olcos 0, by
where
P
is the integral of the second term of Eq. (1 33) evaluated near the free surface
of the solid. Practically, nB(E,,0,) is the angle and energy distribution of the
emitted backscattered electrons (for one primary electron) and can either
be experimentally determined or estimated, for instance, by a simulation
method.
The backscattering factor is defined by the relation
1 +r =N~A9"/N~~" ( 136)
and has the value
B. Theoretical Analysis of'Auger Emission
1. Introduction
The purpose of this section is to link together the theoretical aspects
developed in the previous sections and some more practical considerations
necessary for a quantitative analysis of the Auger emission of a solid target. A
quantitative Auger analysis is obviously based on interpretation of experi-
mental results via a manageable model for the creation of Auger electrons in
the target and for their transport toward the free surface. Accordingly, one
needs a theory that allows
(1) an estimation of the parameters which appear in the usual formula-
tions of the quantitative analysis (MFP, backscattering factor, etc.);
(2) a possibility of gaining some insight into the validity of some
assumptions currently admitted for the "hidden" variables of Auger emission
(depth dependence of the Auger source functions, escape process of the
Auger electrons, etc.).
This is the type of problem more generally encountered in secondary-
electron emission. When one considers the emission of a pure metal, in some
special instances, the elementary processes are sufficiently well known to

allow either an analytical treatment or a Monte Carlo simulation method. By
analytical treatment we mean the resolution of a Boltzmann-type equation,
including (in principle) all the collision effects. This kind of approach has been
used by several authors, and has given very interesting results (see, for
instance, Chung and Everhart, 1977; Bindi et al., 1980; Schou, 1980; Rosler
and Brauer, 1981a,b for the most recent accounts for this approach and all
the references to previous work on the same subject countained therein).
However, it seems that analytical treatments do not represent the best
way to obtain all the important information one must expect from a suffi-
ciently precise model of Auger emission if one wants to have reasonable
computation times. In our opinion, the complexity of the calculations re-
quired by practical problems (such as the emission of alloys or of compound
materials) is such that the Monte Carlo method represents the most efficient
technique. For instance, in one calculation it allows detailed statistics to be
obtained about the individual events which occur to cause Auger emission,
by suitably marking their separate interventions in a computer program. This
kind of flexibility is difficult to obtain with an analytical method.
2. Some General Features of’ a Monte Carlo Simulution Method

A primary beam impinging in a solid target suffers elastic and inelastic
collisions with the components of the solid. The effect of the inelastic col-
lisions is to bring the electrons of the solid to the upper levels so that they can
themselves take part in the transport process. This is the cascade effect.
In a study of Auger emission, one must describe more precisely the
ionizing collisions that are caused by the primary electrons which penetrate
the target or those that are elastically turned back towards the free surface. In
addition, we must follow those which are inelastically backscattered and,
more generally, all the electrons of the cascade which maintain an energy
above the ionization threshold of the atomic subshell in which an Auger
hole can be created. All these processes can be incorporated a priori into a
Monte Carlo simulation, and the main aspects of this method are now
briefly recalled.
The Monte Carlo method is based upon the statistical concept of a
trajectory for the particles which take part in the transport process in the
solid. A trajectory is a sequence of free paths si separated by interaction
“points.” During a free path, the particle propagates along a straight line,
keeping its classical parameters (energy and velocity) unchanged. This
behavior is modified by collisions with other components of the solid which
restrict the amplitude of this free path. In practical calculations, one admits
that the effects of these interactions can be located at some points where the
angle and energy characteristics of the particle suffer their accidents. Most
frequently, these are the end points of the free paths, and the accidents
determine completely the random concatenation of the rectilinear portions
of the full trajectory.
With the assumption of a linear response of the solid to the excitation pro-
voked by a primary electron, each primary particle is followed individually,
as well as the electrons of the cascade which it generates.
In principle, calculations with a Monte Carlo simulation method require
theoretical knowledge of all the cross sections associated with the various
types of collisions. Whenever these requirements are fullfilled, a direct
simulation is possible. However, for practical problems, the complexity is
such that one has to accept less elaborate but more efficient descriptions.
Except in a close neighborhood of the surface, the solid is considered to
be a homogeneous medium described by a randium-jellium model. The
delocalized electrons of the jellium are assimilated to an electron gas. The
other component of the solid is thus composed of the ionic cores. For the
sake of simplicity, one assumes that they are randomly distributed, according
to a uniform background model (randium). This, of course, precludes any
direct treatment of the diffraction effects.
Under these conditions, the propagation of an electron in the solid
becomes a stochastic process, and one generally admits that the collisions it
suffers can be described by a Poisson distribution law. Except near the
surface, the total MFP 12, depends only on the energy E of the electron but
not on its position or its direction in the solid. This MFP &(E) is deduced
from the total cross section (oT(E)= (Nat,IT(E))-' obtained by summing over
the elementary cross sections associated with the elastic and the inelastic
processes a, = oel+ Gin.
The probability of an electron having a free path in the [s, s + ds] domain
between two collisions is given by
P(s)ds = 12;le-sfATds (138)
This, in principle, allows the determination of the extent of the different
straight-line portions si which constitute the zig-zag path of an electron in
the target.
As already indicated, ,IT is most frequently imprecisely known. However,
the universal curve described in Section JII,F indicates that in the energy
domain concerned by Auger analysis, AT is small, of the order of a few
angstroms. This means that a primary electron can suffer an important
number of collisions before its energy falls below the ionization threshold
of many Auger lines of practical interest. Because the description of an
individual collision is itself a rather complex problem, the computational
times may quickly become prohibitive.
Another approach is to describe some condensed histories for the

particles and to make use of the so-called continuous slowing-down approxi-
mation. It has been extensively used in electron microprobe analysis (see, for
instance, Bishop, 1976, or Shimizu et al., 1972, for reviews on this subject).
The energy lost by an electron per unit path length in a pure element
solid target is most frequently represented by the Bethe expression:
dE/ds = -2ae4(N,,Z/A)E-' ln[1.166(E/J)] (1 39)
where N,, is the atomic density of the target, 2 its atomic number, A its
atomic weight, E the energy of the electron at a given stage of its history, and
J is a mean ionization potential for which several analytical expressions have
been proposed (for example, Shimizu et al., 1972).
To describe a condensed history, a simple solution could be to choose for
each si an arbitrary constant value, not too small, however, in order to
reduce the computation times. During si, the electron thus generally
undergoes several collisions, the mean number of collisions being
Pi = (pNat/A)dE)si (140)
The individual atom cross section has often been taken to be of the
Rutherford type and it may include, through simple approximate expressions,
both elastic and inelastic collisions (Reimer, 1968). Under such conditions,
the angular deviation aspect must be treated by a Lewis-type multiple-
scattering theory, that is, by the relation
where a, is itself a Legendre coefficient of the individual cross section.

Additionally, one may assume that the energy transfers, deduced from
Eqs. (139) and (141),and the angular deviations are in a practical simulation
simply located at the end point of each si segment. Some improvements have
been proposed (for instance, Shimizu et al. 1972,1975)either by choosing si in
order to keep pi approximately constant for each segment, or by locating the
angular deviation at an intermediary point of si.
For several reasons, which will become clearer after the analysis of the
Auger emission of aluminum presented in Section IV,B,3, the condensed-
history approach is more suited to the analysis of the penetration of the
primary beam than to the intricate problem of Auger emission. Thus, the end
of the present section is devoted to the presentation of the direct simulation
method. This method can be used mainly for standard reference calculations.
Beyond its ability to give much data about the emission of a few solids, its
secondary role is to allow a judgment as to the validity of other methods, less
precise but much more efficient for complex targets, by applying both
methods to the standard cases. This is discussed in Section IV,B,4.
In a direct simulation method, the free path s(E)followed by an electron
between two successive collisions can be calculated from Eq. (138) by
where ys is a random number uniformly distributed between 0 and 1.

An elastic collision represents a global interaction process with the
potential field surrounding the ionic cores. An inelastic collision can result
from various effects (individual collisions with the electrons of the jellium,
collective excitations of the electron gas, ionization of the inner shells), for
which the energy losses and the angular deflections follow too dissimilar laws
to remain, a priori, undifferentiated. The cross sections o,,,,cre,pl, and
correspond to these various inelastic effects.
By drawing a random number ys, the free path is determined according to
Eq. (142). Knowing the initial position and direction of a simulated electron,
this fixes the location of the collision. To know which type of collision occurs,
one must use a new set of random numbers. For instance, if the number y1
selected is such that
Oel/OT < Y1 < (143)
the collision will be an elastic one.
The next step is to analyze in more detail the angular deflection 8 suffered
by the electron and thus to fix the new direction of propagation of this particle
after the collision. This is done by drawing a new random number y2 ; 8 is
determined from the differential elastic scattering cross section o,,(E, 0) by
using the relation
and by evaluating for which angle 0 this relation is satisfied.

For an inelastic collision which can put into play several elementary
mechanisms, the occurrence of a particular process is determined by a method
analogous to that of Eq. (143), that is to say, in proportion to the probabilities
associated with the various mechanisms. By a method similar to that de-
scribed via Eq. (144), new sets of random numbers allow us to define the
energy ho transferred in the collision and the corresponding momentum
transfer kq by using the differential cross sections o(E,q, 0).
In inelastic collisions, electrons of the solid reach excited states and have
to be followed themselves by the same simulation method. For a study of
Auger emission, all electrons which have an energy lower than the ionization
threshold of the Auger line investigated can be forgotten. This allows us to
set aside many important aspects in a general secondary-emission study

(plasmon damping, for instance) and to gain a substantial amount of
computation time.
The next section is devoted to the application of the direct simulation
method to the LzJVVemission of aluminum. This is done for the purpose of
checking some assumptions and of evaluating some hidden parameters.
More precisely, we shall try to give some information about the source
function of Auger electrons by delimiting the respective contributions of the
primary and the backscattered electrons. The depth dependences of these
contributions are studied. Auger-electron escape is also investigated. This
should clarify the participation of the backscattered electrons in the Auger
yield (backscattering factor and variations of this quantity with the depth).
In order to keep computation times reasonable, the Auger process itself has
been extremely simplified. In our model, Auger electrons are created at one
energy (that of the main Auger line, ignoring its linewidth), and the creation
has been assumed to be isotropic. The same kind of simplification has been
retained for the angular distribution of the electrons ejected in the initial
stage of the ionization process.
Because Auger emission is, in fact, located in the near subsurface of the
solid, we shall briefly recall the assumptions that we have retained for the
surface region. As discussed in Section 111,a simple assumption is to consider
that in a thin layer of the solid (with an extension a, of typically 1 A), the bulk
collective excitations are replaced by surface excitations, all the other types
of collisions remaining unchanged. This surface layer has also been extended
outside the solid (its thickness being a, 5 a i )to account for the possibility of
surface plasmon excitations outside the solid. In this spatial domain it was
assumed that this type of excitation was the only possibility for a particle to
suffer energy losses.
3. Results Obtained f o r Aluminum by the Simulation Method

a. Elastically ReJected Primary Electrons. As indicated above, diffrac-
tion effects have been ignored in the present model. Thus, the elastically
reflected primary electrons are those which have suffered one or more
individual elastic collisions but no energy loss. The total angular deflection
due to these processes must be sufficient to turn them back towards the
free surface.
The number of these particles per primary electron is defined as the
elastic reflection coefficient. In the energy range considered in the present
section (0.1-1 keV), this coefficient is of the order of a few percent. It decreases
as the primary energy E, increases. This tendency is all the more pronounced
as the primary electrons reach a normal incidence. The overall behavior of
-w
n
F
0.025
L
500 1000
Ep(aV)
FIG.12. Elastically reflected primary beam, influence of the primary energy and incidence
angle on the reflection coefficient: tIp = 60" (V),45"( a),0" (m).
this reflection coefficient can be approximated by a (cos tl,)-a law, where 0, is

the incidence angle and ci is a positive exponent which depends on E,. It
increases with E, and is nearly zero at low primary energies. These results
are illustrated in Fig. 12.
The inelastic collisions are frequent for all the energies considered, so the
FIG.13. (a) Differential contribution n(E,, z) of the different atom layers to the reflection
coefficient ( E , = 700 eV); (b) integrated contribution dz n ( E , , z ) ( E , = 700 eV).
primary electrons which are elastically reflected originate at the near sub-
surface of the solid. This maximum-penetration depth results from a balance
between two processes which are strongly energy dependent. As indicated in
Fig. 13, for a normal incidence, 50% of the elastically reflected primary
electrons originate from the first three angstroms for Ep = 700 eV, whereas
for Ep = 150 eV, it is 80% (Fig. 14).
Let us also remark that for Ep = 700 eV, 40% of the elastically reflected
electrons have suffered one collision, 25% two collisions,207; three collisions,
and 15% more than three collisions; whereas at Ep = 150 eV, these
percentages become 60, 25, 7 and S%, respectively.
b. Creation of Auger Electrons in a Solid. In the present calculations,
only electrons coming from the L,,,VV line of aluminum have been con-
sidered. These Auger electrons have been created by primary incident
electrons and by those which have been backscattered in the solid. In this
latter class we include all the electrons with an energy between (E, - 1) eV
and the ionization threshold of the L2,3level. The separate contributions of
these two groups of electrons to the dependence of the Auger source function
on the depth z have been presented in Fig. 15a for Ep = 700 eV and a normal
primary incidence.
Primary electrons clearly do not represent the main contribution to the
Auger current. The corresponding source function decreases rather strongly
0.0 :
OD 2
0.5
0.0 1
FIG.14. (a) Differential contribution n(E,,, z ) for E,, = 150 eV; (b) integrated contribution
for E, = 150 eV.
0.0 2 1 (Ep> z
,
0 50 100
0 50 100
ztii,
(b)
FIG.15. Creation of Auger electrons in Al: (P) contribution of the primary electrons; (B)
contribution of the backscattered electrons; (T)total; (a) E,, = 700 eV, 8, = 0";(b) E,, = 700 eV,
ep = 45";(c) E,, = 150 ev, e,, = 0".
FIG.15. (Continued)
with the depth z due to the inelastic collisions and it is in fact rather con-
centrated near the surface, with a quasi-exponential decrease. Assuming a
true exponential law of the form S,(z) = Sp(0)e-'/L;L is found to be about
15 A for this set of primary parameters.
For an incidence angle 0, = 45" (see Fig. 15b), the general behavior of
the Auger source function is the same. The contribution of the primary
electrons can still be roughly approximated by an exponential law with an
exponential parameter of about 10 A. This is quite consistent with an overall
exp( - z / L cos 0,) behavior.
Creation of Auger electrons by the backscattered electrons takes place for
the most part in the bulk. For Ep = 700 eV, the high-energy electrons can
suffer several elastic and inelastic collisions (including ionizations) before
they fall under the Auger ionization threshold. The corresponding source
function presents a maximum which for 0, = 0" is located at about 30 8, from
the surface. For 0, = 45", this value becomes 20 A. This seems to indicate that
the position of the maximum still varies roughly like (cos 0,)-'. Thus, at least
in the domain of sufficiently high values of E,, the backscattered-electron
contribution is representative of a rather general law for the stopping power
in the bulk of the solid.
For Ep = 700 eV, this contribution of the backscattered electrons is by
far the strongest (90% of the total Auger electron creation) and dominates
the shape of the Auger source function. However, the Auger electrons (which
are those of the L,,,VV transition) have a low initial energy, and the major
part of them can not be emitted unless they originate at moderate depths.
232 M. CAILLER, J . P . GANACHAUD, AND D . ROPTIN
Consequently, in Auger emission, the primary-electron contribution is

approximately equal to that of the backscattered electrons.
At lower energies (E, = 150 eV for Fig. 15c), the contribution of the
inelastically backscattered electrons is strongly reduced. That of the primary
electrons also decreases but relatively less. Thus, at low energies, there is an
equilibrium between these two contributions. So, the total source function is
strongly decreasing near the surface. This shape is principally due to the
primary-electron contribution. For the emission, both groups of electrons,
coming from similar depths, will still have rather comparable contributions.
c. Emission of Auger Electrons. In this section, we shall focus our
attention upon Auger electrons emitted at the free surface without having
lost too large a fraction of their initial energy along their path in the solid
(less than 5 eV, for instance). They can still contribute to the shape of the
L,,,VV line located at about 67 eV above the vacuum level. The other
electrons of the Auger source function which are emitted in the “background”
have not been considered.
iiOOli
FIG. 16. Contribution nA(z) of the different atom layers to the emission of the Auger elec-
trons in A1 ( E , = 700 eV, 0, = 0”): (a) total emission n,(z); (b) contribution of the primary and
backscattered electrons; (c) relative integrated contribution [nA = S$dzn,,(z)]; P, B, T as in
Fig. 15.
Because the absorption law is nearly exponential, only Auger electrons

created near the surface can reach it with no energy loss. In our model, we
have neglected the initial Auger linewidth as well as the electron-phonon
collision mechanisms, so that only a very small fraction of the internal Auger
electrons can reach the surface with an energy loss lower than 5 eV. Thus the
width of the corresponding external Auger line obtained by a simulation
method is in fact very small. This approximation is not very severe if we focus
our interest on the intensity of an Auger line and not on its precise shape.
Figure 16 shows that for a primary energy E, = 700 eV and at normal
incidence, 90% of the emitted Auger electrons come from the first five ang-
stroms. The primary- and backscattered-electron contributions have nearly
the same values. The total Auger yield is 0.75%. It becomes 1% for the same
primary energy but for an incidence angle of 45" because the creation mecha-
nism has then been reinforced near the free surface of the solid. As shown in
Fig. 17, for a primary energy Ep = 150 eV, the Auger yield is nearly 0.56%,
but in this case, 70%of the emitted Auger electrons have been created by the
primary beam and only 30% by the backscattered electrons.
The dependence of the total Auger yield on the primary energy is repre-
sented in Fig. 18 where reduced variables have been used. The reduced-
energy variable is (E, + 6)fj L2,3 where Epis the kinetic energy of the primary
0.00' 1
c z(X) 45
40005
(b)
FIG. 17. Contribution nA(z) of the different atom layers to the emission of the Auger elec-
trons in Al (E, = 150 eV,8, = 0"): (a) total emission nA(z);(b) contribution of the primary and
backscattered electrons; (c) relative integrated contributions [nA = J;dzn,(z)]; P, B, T as in
Fig. 15.
electrons in the vacuum, Cp = W - EF is the work function of the target, and

EL,,3is the ionization energy of the L2,3 level (73 eV). This Auger yield is
compared with the ionization cross-section curve of Gryzinski plotted versus
the same reduced-energy variable. These two curves have been fitted at a
common reduced energy ( ~ 4 . 8 5 )corresponding to Ep = 350 eV, the
Gryzinski curve itself being normalized in such a way that its maximum
amplitude becomes unity.
One can remark that the reduced Auger yield estimated by the simulation
method for several primary energies shows rather good agreement with the
Gryzinski curve. There are no a priori evident arguments to predict this;
however, some reasons can be found from the fact that the creation rate of
the Auger electrons is not strongly z dependent near the free surface of the
solid and that the true backscattering factor varies rather moderately with
Ep,at least in the energy range considered in our study.
By a true backscattering factor we mean the ratio of the true number of
Auger electrons escaping from the solid and the number of Auger electrons
emitted after they have been excited by a primary electron. This definition
differs from the usual definition of backscattering factor, which one can de-
note as the effective backscattering factor. This latter concept comes from
several assumptions (for instance, a uniform spatial distribution for the Auger
electrons created by the primary beam and also by the backscattered elec-
0 5 10 15
Epi 1 En1
FIG.18. Dependence of the normalized total Auger yield nA/nA,msx

on the primary energy:
( A ) primary electron contribution; ( 0 )total.
trons, at least in the region of the solid which can contribute to the Auger
emission). If instead of the total Auger yield we consider the partial yield due
to the primary electrons only, and use the same normalization conditions as
indicated above, then agreement with the Gryzinski curve is expected to be
better still. Thisis precisely the case at low primary energies.For higher values
of E , ,some slight discrepanciesappear, but the overall agreement is still good.
This discussion seems to indicate that an experimental study of the variation
of Auger yield with E, might make it possible to measure the ionization cross
sections of a target and to compare them with their theoretical estimations.
The variation of the Auger yield with the incidence angle of the primary
electrons has been represented in Fig. 19a for several values of E,. For
relatively high energies, this yield varies as (cos O J l . However, for lower Ep
values, its variation is not so rapid. This distortion from the (cos O,)-' law
has its origin in the contribution of the primary electrons (See Fig. 19b),
whereas that of the backscattered electrons seems to fit this ideal law rather
well whatever the primary energy (Fig. 19c).
The reason for the distortion from the ideal law which appears in the
primary-electron contributions has to be found in the reduced penetration of
these electrons. In fact, a more detailed analysis of the influence of the in-
cidence angle on Auger electron creation rate, made layer by layer, reveals a
more subtle behavior. For instance, at E, = 150 eV, in the first two or three
angstroms, the creation rate of the Auger electrons by the primary electrons
increases with e,, and the ideal (cos OP)-' law is well followed. On the con-
trary, beyond a depth of 5 or 6 A, this creation rate becomes weaker at a 45"
incidence angle than at normal incidence. Because this spatial domain is still
within the escape depth of the Auger electrons, the variation of the Auger
yield with 0, is not so rapid as that for high energies.
These results, obtained by the simulation method can be compared, at
least qualitatively, with the experimental observations made by Matsudaira
and Onchi (1978) for nickel. In order to make such a comparison, one has to
think in terms of reduced energies. This can approximately be done by
comparing results corresponding to identical values of the ratio EA,../E,,
where EA,x is the Auger electron energy and E, the primary-electron energy.
Matsudaira and Onchi (1978) presented results for EA,.. = 61 and 850 eV
and for a primary energy Ep = 2000 eV. This leads to the values 0.0305 and
0.425 for EA,../E,. For the L,,,W line of aluminum, EA,.. = 67.5 eV, and for
the primary energies, Ep = lo00 and 150 eV, the ratio is EA,/E, = 0.0675
and 0.45.
The results presented in Fig. 19d correspond to the curve n(d,)/n(O),
obtained for E,,,/E, = 0.45. This curve, plotted versus O,, increases much
less rapidly than (cos O P ) - l , For instance, at 0, = 60", the value 1.52is reached,
0.5 1 0.3 1
cos e, cos ep
(0) (b)
0
I 3
0
' S'O" 9'0"
cos 8, 8,
(C) (d1
FIG. 19. Dependence of the Auger yield on the incidence angle: (a) total emission; (b)
contribution of the primary electrons; (c) contribution of the backscattered electrons; (d) curve
n(O,)/n(Oo)for the total emission; E, = (A)150 eV, (v)350 eV, ( 0 )700 eV, (*) 1000 eV.
'
whereas (cos gP)- = 2. The same behavior appears for the experimental
results on nickel for EA,,/Ep = 0.425, whereas for 9, = 60°, the function
takes approximately the value 1.6. On the other hand, for low EA,,/Ep
values, the (cos OP)- ' law is followed well in both cases up to 0, = 60". Some
disparities appear in the case of nickel, but they correspond to grazing
incidences which have not been simulated for aluminum.
The fact that in the first few angstroms the ideal law is satisfactorily
followed, even when at low primary energies important distortions appear
for greater depths, suggests that the same behavior could occur for an
adsorbed layer. Bui Minh Duc (198la) envisioned this hypothesis to explain
the results of Matsudaira and Onchi (1978). These authors have experi-
mentally observed that a monolayer of sulfur deposited on nickel reinforces
'
Auger emission according to a law which is very close to (cos 6; ), 0, being
the incidence angle.
An Auger transition characterizing the sulfur monolayer can be induced
by electrons backscattered in this layer or in the bulk of the nickel substrate,
but for these groups of electrons, the ideal law is correctly followed. The
primary-electron signal, itself confined to moderate depths, presents in this
case quite identical variations. Thus the simulation method seems to confirm
the set of experimental results obtained by Matsudaira and Onchi, at least
qualitatively, because the material that we have simulated is not the same as
those they have experimentally studied.
d. Buckscattering Factor. One of the most important parameters in the
quantitative analysis of Auger emission is the backscattering factor or, more
precisely, what we have called the effective backscattering factor. This
quantity is frequently expressed as a function of the ratio U-' = En,/Ep,
where En, is the binding energy of the nl subshell. The results do not differ
significantly from one author to the other (Bishop, 1967; Jablonski, 1978;
our own results are presented in Fig. 20). Additional results obtained by the
present authors (but which have not been reported here), indicate that the
backscattering factor is nearly independent of the incidence angle, at least at
high energies. However, as can be seen from Fig. 20, the backscattering factor
',
is not simply a function of U - but it also depends separately on the value of
the primary energy E p ,although the differences between the curves obtained
for different Ep values never exceed 10% in the energy range studied.
We should also remark that this backscattering factor differs more
strongly from the true backscattering factor defined as the ratio of the total
Auger yield to the partial yield due to the primary electrons only. This factor
cannot be rigorously defined, for instance, by an experiment. However, a
simulation method allows one to estimate it, provided one can distinguish
I
-+ l.5 ’
U-’
-
FIG.20. Dependence of the backscattering factor on the primary energy and on the ratio
U-’= E,,/E,, (Epi= Ep + W - E F ) . ‘B’ refers to Bishop’s (1967) results, and ‘J’ refers to
Jablonski’s (1978) results.
between the contribution of the primary electrons and that of the back-
scattered electrons. Such an estimation is presented in Fig. 20. It is compared
with various values obtained for the effective backscattering factor.
One notices that, on the average, the true backscattering factor exceeds
the effective one by 20-30”/:, depending on which estimation and which
value for U are considered.
In Fig. 2 1, we have indicated how the backscattered electrons of different
energies contribute to the effective backscattering factor (1 I ) . This +
evaluation has been made by varying the lower limit of integration Einf
Eh-EI
FIG.21. Contribution of the backscattering electrons of different energies to the effective

backscattering factor (1 + r ) : E,, = Ep - 1 ; El is variable.
in the expression
where the value of Einfranges from the ionization threshold to the value Esup
[taken as ( E , - 1) eV]. These results are presented for various values of E,.
e. Angular Distribution of Emitted Auger Electrons. It is generally
admitted that Auger electrons are absorbed according to an exponential law
along their path towards the free surface of the solid. If one assumes that they
have been uniformly created in the target, with an isotropic angular distribu-
tion, the law describing their angular emission is necessarily of a cosine type.
In our simulation model, Auger electron creation has also been assumed
to be isotropic; however, it is no longer uniform in z. As can be seen Fig. 22,
Auger electron angular distribution obtained by simulation follows a cosine
law very well for all the values of E,, at least for normal primary incidence.
One should note that the experimental results obtained by Matsudaira and
Onchi (1978) in fact show some deviation from the simple cosine law.
Bui Minh Duc (1981a) considered, from a theoretical point of view, a
method for generalizing the form of the angular distribution by including
several physical parameters such as the MFPs of the Auger and primary
electrons, the incidence angle, and the mean parameter of the lattice in order
WJ 9. con 9.
(a) (b)
FIG.22. Dependence of Auger electron emission on the emergence angle.
to interpret the results of Matsudaira and Onchi. Although his generalized

expression still contained some simplifying assumptions, he was able to
reproduce the shape of the experimental distributions obtained for nickel,
either pure or with a sulfur overlayer, by simply adjusting one parameter
called the lateral stopping power of the layer.
4. Extensions to Complex Materials

The direct-simulation method was presented in Section IV,B,2 and
applied to the Auger emission of aluminum in Section IV,B,3. The present
stage of our knowledge of individual cross sections is such that this method
can be applied only in exceptional instances. A second drawback is that the
direct-simulation technique involves large computational times.
Thus, to go beyond the description of a pure target, one has to look for a
simplified simulation method in which the most basic aspects of Auger
emission can nevertheless be incorporated. In this spirit, one has to keep in
mind several conclusions resulting from the above discussion and from the
theoretical analysis which preceded it.
(1) Auger emission is located near the surface of the solid, owing to the
reduced escape depth of the Auger electrons, whereas their source function
can extend rather deeply into the solid. This indicates that some simplifi-
cations which could have been useful for describing the penetration of the
primary beam into the target, for instance, are now inadequate.
(2) The Auger source function is itself composed of various contri-
butions. The primary electrons penetrating the target or turned back towards
the free surface play an important role. However, the backscattered electrons,
which include the primary particles which have lost part of their initial energy
and the high-energy secondary electrons created during the cascade process,
give rise to comparable contributions.
(3) The elastic and inelastic collisions correspond to rather different
angular deflection laws. The various types of inelastic collisions often have
themselves well-separated energy-loss spectra. Although a sharp line can
yield important information about the validity of a model, effects associated
with a broad spectrum are much more difficult to observe.
(4) Many important Auger lines can be excited at moderate primary
energies (in the keV domain) where the usual stopping-power formulas may
become partly inadequate.
These remarks seem to indicate that a multiple-scattering approach is not
useful in the study of Auger emission. One basic question would be how to
describe correctly the cascade process within this method. However, even for
the penetration aspect of the primary beam (which models the source function
of the Auger electrons), a pure continuous slowing-down approach would
certainly be too crude. It would lead to an oversimplified statistical de-
scription. Due to various competing scattering processes, some portions of
an electron path can in fact be quasi-elastic in character, whereas other
portions can be mainly governed by inelastic events. A simulation technique
based upon the “use of some condensed histories’ of an electron (in order to
save computational times) may lack some important aspects (for instance, a
description of the elastic peak).
In this connection, inelastic processes having comparatively small
probabilities of occurence can give rise to important individual energy losses.
These losses can be quite different from a mean energy-loss value. This is
precisely the case for the ionizing collisions by which Auger emission is
initiated. Moreover, the validity of a multiple-scattering theory is certainly
questionable near the surface, that is, in the region probed by Auger analysis.
The basic idea is, consequently, to reduce the step length of the si
portions to an extent comparable with the global MFP so that individual
collisions could in fact be simulated. Elastic and inelastic effects have to be
sharply distinguished. The ionizing collisions themselves have to be treated
separately. Without returning to the pure direct-simulation concept, a
suitable compromise has to be found.
A very satisfying answer, in view of the large variety of problems it was
able to solve, has been given by Shimizu and co-workers (Ichimura et ~ l . ,
1981; Shimizu and Ichimura, 1981). It seems important to retrace, at least
briefly, the main arguments of these authors.
A clear separation is made between elastic collisions and inelastic effects.
A good solution is to break the elastic effects by using a partial wave analysis
of the scattering cross section, starting from a suitable potential. These
authors adopt the Thomas-Dirac-Fermi formulation for the potential for
which Bonham and Strand (1963) have given general analytical expressions.
The partial wave-expansion method, which has also been used by the present
authors (see, for instance, Ganachaud, 1977) was reviewed in Section IV,
where the choice of the potential was discussed.
The inelastic part has to account for the slowing-down effects and for the
central problem in the present context, that is, the creation of Auger electrons.
In general, we remark that all the events in a simulation method which are
associated with a well-defined energy (the creation of an Auger electron in an
ionization process, for instance) or which keep unchanged the energy of a
colliding particle (elastic collisions) have to be described with much care. On
the contrary, collisions giving rise to broad energy-transfer spectra con-

tribute mainly to the “uniform background.” For these events, less precise
descriptions can be sufficient in practice.
It has been recognized by Ritchie et al. (1969) that the result obtained by
summing over the different inelastic contributions to the stopping power
(which can be theoretically evaluated in simple cases) is close to that given by
the general Bethe law. Moreover, the Gryzinski formulation seems to have a
very large domain of validity.
Starting from these conclusions, and following the prescriptions of
Krefting and Reimer (1973), one can define a stopping power, due to the
valence electrons, by the difference
(dElds)va, = (dE/dS)Bethe- (dE/ds)core ( 146)
where the first term of the right-hand side (RHS) is given by Eq. (139), and
the second term is evaluated by Gryzinski’s theory. The contribution
(dE/ds),,,,includes all the inelastic collisions, either individual or collective,
with the delocalized electrons.
The above treatment can be improved, particularly for low energies,
(where the Bethe law becomes questionable when E 5 J ) . According to
Shimizu and Ichimura (1981)a better choice is to adopt the following scheme:
(dElds) = (dE1dsXore + (dE/dsXa, (147)

with
(dEldsk, = (dE/dS)Bethe - (dE/ds)core for (dEldShethe > (dE/ds)core (148)
(dElds) = (dE/ds)core otherwise (147’)
This seems to be a more general approach than that suggested by Shimizu
and Everhart (1978),who proposed an extension of the Gryzinski formalism
to valence electrons.
The choice of s can be accomplished by finding a suitable MFP 1, and by
using the usual random-number-generation procedure described by Eq. (142).
Following Shimizu and Ichimura, the MFP 4to be used is given by
AT1 = 2,’ + 1-’
core (149)
where ,Ie1 denotes the elastic MFP and Lcor, represents the inelastic effects of
all the deep levels j that can be excited by the simulated electron.
Thus, in practice, the separation of elastic and inelastic collisions is
first done according to the standard procedure of Eq. (143):
(1) An elastic collision is described by the corresponding differential
cross section cre,(E,0) which fixes the angular deviation.
(2) An ionizing collision (of which j is first determined) gives rise to an

energy loss AE. Gryzinski’s formula can be used under the differential form
x {g(l - 2) + ;ln[2.7 + rTy’2]} (150)
where n j is the electron population of the subshell considered and E j its

binding energy.
The corresponding angular deflection 0 can itself be approximated by
the simple formula
AE = E sin2 0 (151)
The cascade Auger effect following this ionization can be described by the
standard procedure of Section VI,B,%.
In fact, the MFP A, used in Eq. (149) does not include the valence-
electron contribution to the stopping power. This can be recovered by
incorporating an additional energy-loss term
for the estimated path s.

(A%,, =
l: (dE/dsX,,ds (152)
An important point to note is the great flexibility of this method, which

can be extended to compound materials and alloys by using generalized
relations (Philibert and Texier, 1968; Matsukawa et al., 1973) such as
where p is the density of the compound material where the constituent

labeled k is present with a mass concentration c k and
where NAv is the Avogadro number and njk is the population of the j t h
subshell of the kth constituent.
Calculations made by Ichimura et d.(1980) for a pure aluminum target
can be considered as standard reference calculations. They did this by three
methods : (1) using a direct-simulation technique; (2) describing the con-
tribution of the conduction electrons by the relation proposed by Shimizu and
Everhart (1978); and (3) using the method analyzed in this section. No
marked differences could be found among the results. Method (3) has
subsequently been applied to a large variety of solids (pure materials,
alloys, compound targets) to give, among other results, an evaluation of the
backscattering factor in view of an Auger quantitative analysis. These results
have been extensively presented by Shimizu and Ichimura (1981) in their
review article.
VII. AUGERQUANTITATIVE
ANALYSIS
A. Zntroduction
Auger quantitative analysis aims to determine the various concentrations
in the elements existing at the “surface”of a sample. It is understood here that
“surface” can not and must not be considered in its proper geometrical
meaning; rather, it means the few outermost atom layers under the electron
beam.
There is still a long way to go before being able to relate the detailed
theoretical descriptions presented in Sections I l l and IV with the experimen-
tal observations made on complex targets; even the possibility of realizing
this project completely is open to question.
In order to attain this goal it is nevertheless necessary to follow various
simultaneous complementary approaches. Thus, from a theoretical point of
view, one must be able to give not only a complete quantitative description of
the simplest targets, i.e., pure samples, but also to develop realistic models of
less simple targets such as binary alloys and compounds. In fact, the de-
scription of complex targets seems possible only phenomenologically as
initiated in Section VI.
The same progression must be followed in the experimental studies.
Indeed, it seems fundamental that for the pure targets playing the role of
standards a complete quantitative agreement could be achieved between
theoretical and experimental results. Quite evidently, this agreement must
first be achieved among the results from various experimenters. This im-
plies that a rigorous and indisputable procedure must be defined not only
to control the quality of the conception, fabrication, and adjustment of the
whole apparatus of analysis, but to ensure correct usage of this apparatus
as well.
In this regard, the round robin led by the ASTM Commission 42 (Powell
et al., 1981) revealed the ticklish problem of controlling the validity of
measurements in AES.
The targets used in the round robin were very simple (pure copper and
gold), and a general experimental procedure was defined. However, the

values of the ratios between the intensities of the high- and low-energy
lines for the same element differed by an important factor (120 for gold)
among the extreme measured values. In fact, these intensities were really
measured by the peak-to-peak heights H in the derivative mode. The dis-
agreements noted can particularly arise from some error in manipulation.
However, they point the direction to follow in order to achieve good quanti-
tative agreement even in the case of simple targets and, aJortiori, for complex
samples of a technological nature. The definition and general acceptance of
standard procedures for the adjustment of analysis devices (such as that
proposed by Duraud and Le Gressus, 1981) is a sine qua non condition of
progress. Studies of binary and ternary alloys and compounds are essential
to knowledge of the spurious effects which occur in the preparation of the
sample and in Auger analysis itself.
A serious disadvantage of AES results from the use of an ultrahigh
vacuum (UHV) chamber.
Apart from questions of experimental uncertainties or manipulation
errors, the problem of the interest of the AES is even posed for technological
samples such as those encountered in the study of catalysis, aqueous corro-
sion, and lubrication. Indeed, when a technological sample is introduced in
the UHV system, some disturbances come from the vacuum itself and from
the decontamination treatments carried out before the analysis. Another
means of studying technological problems is to attempt to reproduce in the
UHV system some of the most basic physical processes. This leads to pro-
posing an idealized situation instead of the reality and poses the problem of
making a correct link between them.
The above questions are beyond the scope of the present review article.
There are additional problems in making an AES experiment quantitative.
They can arise, for instance, from the calibration aspects, from the prepara-
tion in situ (i.e., in the UHV system),or from the electron-beam effects during
the analysis. Some of these problems are reviewed in the following sections.
B. Operating Modes
Assuming that all the conditions required to achieve a good analysis of
the target are realized from the point of view of the sample preparation and
the general setup of the apparatus, various operating modes can be used.
The choice depends on the desired information. For a given direction of
incidence of the primary electron beam and for a given acceptance angle of the
collector, the secondary-electron distribution is n(E,, E,, Zp), This is a
function of the secondary-electron energy E,, the energy Ep,and the intensity
I, of the primary beam.
The collector always selects the electrons from between two energy
limits: E,, and E,, . For a high-pass filter such as a grid-retarding analyzer,
E,, is in principle infinite, but in fact E,, = E,. Thus, all the electrons with
an energy higher than the retarding potential E,, = E, are caught by the
collector. Therefore the current detected by the analyzer is
Ep,1,) = [: n(& E p ,Ip)dEs (155)
It is possible to modulate any of the three variables E,, E,, I, and to

detect the modulation induced on N . The brightness (or primary-electron-
beam intensity) modulation technique has been proposed and developed by
Le Gressus et al. (1975).
In the normal AES operating mode, Ep and I , are kept constant, and a
small alternating component e, sin ot is superposed on the retarding
voltage E, . The secondary current can thus be approximated by
N(E, + e, sin at)
n ( E $ )dE$ - dE, 2 ! - n(Es)e,sin ot
By using a lock-in amplifier, it is possible, for instance, to select the o

component of the signal [third term on the RHS of Eq. (1 %)I.
Its amplitude
is n(E,)e, and it gives the secondary-electron energy distribution. It is phase
shifted by .n with respect to the grid potential modulation. In practice, it is
generally desirable to use the second derivative of the current N , i.e.,
dn(E,)/dE,. This is achieved by detecting the second harmonic of the
modulation frequency [fourth term in Eq. (156)l. This operating mode was
developed by Harris (1968), and this began the rapid progress in the
field of AES.
When E, and I , are kept constant by superposing a small sinusoidal
voltage e, sin wt onto the accelerating voltage Ep,one obtains as a first-order
expansion:
N(E,, Ep + e, sin cot, I , )
(157)
The signal corresponding to the second term of the RHS can be detected by
the lock-in amplifier on the collector.
In the integral of the RHS member of Eq. (157) [later termed 6 N ( E , ) ] ,

the lower limit E, must only be lower than Ep.Two particularly interesting
cases are E, = 0 and E, = Ep - 6Ep; 6Epis a positive quantity which has to
be made as small as possible, but it must exceed epin order to keep Eq. (157)
valid. The first case, E, = 0, corresponds to Auger electron appearance
potential spectroscopy (AEAPS) (Gerlach, 1971), and the second case
corresponds to disappearance potential spectroscopy (DAPS) (Kirschner
and Staib, 1973, 1975).
In AEAPS, the detected signal is the sum of the two terms which have
opposite variations when Ep increases:
(1) N ( E , = 0) is the variation of the total secondary-electron current.
Except at high primary-electron energies, 6 N increases with Ep.
(2) n ( E p ,E p ,Ip) represents the elastically reflected beam. It decreases
smoothly with increasing Ep except when its derivative dn/dE, presents a
singularity. Such a singularity can result from the opening of an ionization
channel in the ionic cores or from coherent diffraction effects.
In DAPS, 6 N is nearly zero, and only the variations of the elastically
reflected beam n are detected. In the case of a polycrystalline target, the
diffraction effects are negligible. The rapid variation of n can only come from
the opening of new inelastic channels when the primary-electron energy
becomes equal to a core-ionization threshold.
In DAPS, the decrease of n is not compensated for by an opposite
variation of the secondary-electron current, hence the name disappearance
spectroscopy given to this operating mode. In AEAPS, the decrease of n is
overcompensated for by the appearance of newly emitted low-energy
secondary electrons. These secondary electrons are created via the cascade
process by the Auger electrons which are generated by reorganization of the
excited atoms.
The measurement of the variations of the total X-ray emission of a target
as a function of Ep has led to a similar operating mode called soft X-ray
appearance potential spectroscopy (SXAPS) (Park et al., 1970). In the case
of single-crystal targets, diffraction effects can completely obscure some of
the appearance potential features (Gerlach, 1971; Den Boer et al., 1978). To
separate the diffraction effects from the appearance potential features, Den
Boer et al. proposed a double-modulation (in E, and E p )technique. In such a
case the RHS members of Eqs. (156) and (157) have to be added.
The results obtained by AEAPS and DAPS are not analyzed in detail in
this review, which is devoted more to the quantitative aspects of Auger
analysis. However, one must cite the important pioneering work done by
Park et al. (1970) who showed that it was possible to measure accurately the
binding energies of the core electrons and the work functions of metals by
AEAPS and DAPS.
The accuracy obtained in AEAPS ( 20.2 eV for the binding energies,

f0.04 eV for the work functions) is quite excellent. Therefore for this type
of experiment this technique can compete with much more expensive ones
such as synchrotron radiation photoemission.
Park‘s group has also shown (Cohen et al., 1978; Park, 1979) that the
target current presents an extended fine structure over a few hundreds of
electron volts above the ionization threshold of the inner shells. This ex-
tended appearance potential fine structure (EAPFS) is quite similar to the
extended X-ray absorption fine structure (EXAFS) above the absorption
edge of the X rays in solids.
The EAPFS results from interference effects between the outgoing
spherical wave associated with the Auger electron leaving the atom source
and the waves coherently scattered by the neighboring atoms. For a given
outgoing wave, the phase shifts of the reflected waves depend on the distances
between the central atom and the atoms playing the role of scattering centers.
The results of both EAPFS and EXAFS therefore depend on the inter-
atomic distances and give quantitative information on these distances via a
Fourier transform of the spectra. However, whereas EXAFS characterizes
the bulk of the sample, EAPFS is especially sensitive to the surface (Elam
et al., 1979). Thus, in spite of the complications resulting from the presence
of a background in the spectra, it has been possible to obtain information on
the surface interatomic distances in the cases of clean and gas-covered
metallic surfaces (Elam et al., 1978, 1979;Fukuda et al., 1978; Den Boer et al.,
1980a,b; Laramore et al., 1980). Park has devoted some review papers (1979,
1980a,b) to the core-level spectroscopies by AEAPS and to the studies of
surface spacings. Readers interested in these topics will find much of value
in these papers. Finally, the surface sensitivity of the appearance and
disappearance spectroscopies are now briefly considered.
Nishimori et al. (1980) reported that DAPS is more surface sensitive than
classical AES. This would seem to be intuitively foreseeable by taking into
account the physics of these operational modes. However, to the authors’
knowledge there are no theoretical determinations of the thickness of the
information zone in DAPS, but some information might be obtained with
the help of a simulation model. In this way, the results presented in Section
VI,B concerning the depth at which the elastically reflected primary electrons
penetrate aluminum give an indirect measurement of the DAPS surface
sensitivity.
Diferential or Direct-Energy Spectra in Quantitative AES ?

It was mentioned earlier that the rapid growth in the importance of AES
was started by Harris’ idea (1968) of using a derivative of the secondary-
electron energy spectra. Thus, it was possible to suppress the background and
then to amplify the Auger lines. A quantitative technique was built on these
differentiated spectra. It consists in using the peak-to-peak height H of an
Auger line in the differentiated mode to measure the intensity of the same line
in the nondifferentiated mode.
The differential mode has so far been the most frequently used technique
of quantification, but recently there has been a strong tendency to return to
the use of the energy (or direct) spectra, especially with the development of
automatic data processing. Two major physical arguments favor the use of
the direct-energy spectra. First, the intensity I of an Auger peak is directly
proportional to its area in the energy spectra, whereas the peak-to-peak
height in the derivative spectra only indirectly measures this intensity. Con-
sequently, it is possible to proceed to absolute measurements from the energy
spectra. Thus, comparisons between theoretical and experimental values of
Auger line intensities can be attempted.
Second, the accuracies resulting from the influence of chemical effects on
Auger line shapes are less important to the energy spectra than to their deriv-
atives. For instance, the peak-to-peak height H in the derivative mode can be
strongly disturbed by a change in the shape of the Auger line even if the area
under this line does not change.
For an Auger line having a Gaussian line shape, the proportionality of H
to the intensity I of the line, as measured on the energy spectrum, has been
shown by Hall et al. (1977), who obtained
H = 16 In 2(2.ne)-'l2l/o2
where e is the base of the natural logarithm and w is the half-height width of
the Auger line as measured on the energy spectrum.
In the derivative mode, the Auger line presents a maximum and a mini-
mum which, in the case of a Gaussian line shape, are separated by the distance
It thus appears from Eqs. (158) and (159) that if the Auger line changes in
width but retains its Gaussian shape, the intensity of the line is better
described by H d 2 than by H only.
In fact, the relation of proportionality between I and H d 2 is not limited to
the case of Gaussian line shapes, but as noticed by Hall and Morabito (1979),
it can be extended to other line shapes. This extension is possible because the
origin of the proportionality relation between I and H d 2 is in the abscissa
and ordinate scale factors and not in the Gaussian line shape itself.
In practice, the energy spectra are recorded either directly, generally by
an electron-counting technique, or indirectly. In the latter case the recorded
signal is obtained by a derivation-integration procedure (the dynamical
background subtraction). In this procedure the secondary-electron current is

differentiated as many times as necessary for a complete elimination of the
background. These derivations are then followed by successive integrations
until the energy spectrum is obtained.
If the derivative mode gives a better contrast of the signal, the use of the
direct-energy spectrum gives, at least for low beam currents, a better signal-
to-noise ratio (Seah, 1979a,b).But the use of the direct-energy spectra in view
of quantitative AES raises the essential question of separating the selected
Auger line from the background. The major difficulty arises, however, from
the creation by the energy-loss spectrum of the Auger electrons of a long tail
on the low-energy side of the Auger line. For a quantitative Auger analysis, it
seems desirable if not vital to separate the Auger line from its energy-loss
spectrum. The question is all the more important because computer facilities
will undoubtedly lead to a rapid development of direct-spectrum analyses.
Some progress has been made in this field, as we shall see in what follows.
It was shown in the section devoted to the simulation of Auger emission
that only a very small number of the Auger electrons created in the solid can
reach the free surface with their characteristic energy. The others either
penetrate more deeply into the sample and are lost or undergo inelastic
collisions before reaching the free surface of the target. Taking into account
the total Auger electron creation (obtained by summing both the primary-
and the backscattered-electron contributions), the energy distribution NA(E3)
of the Auger electrons created in the solid when they reach the free surface is
obtained by integrating n$3”(3) (see Section V1,A) over x,, y,, and $2,. In
writing the energy distribution, the superscript prime and x has been dropped
for simplification in what follows.
In principle, NA(E,) is extended over a very large energy range. Indeed,
Auger electrons created in the solid can escape into the vacuum with any
energy value lower than their creation energy. Over the major part of the
energy range, N A ( E ,) is superimposed on the secondary-electron background
B from which it cannot be rigorously separated (see Fig. 23). The secondary-
electron background would be the electron background if the channel of the
Auger emission which is considered here is not open. In fact, far under the
Auger peak NA&) is considered so small that the secondary-electron
background and the total background can be confused.
In Section VI,A it was supposed that true Auger electrons, i.e., those
emitted with a characteristic energy, could be unambiguously separated from
other electrons. In practice, no exact separation is possible and approximate
procedures have been proposed. They consist of extracting from N ( E , ) =
B + N A ( E 3 )the unknown true Auger electron distribution which is denoted
here by A,. The problem is complicated further because the energy distri-
bution N ( E , ) is itself partially unknown. Indeed, the experimental energy
distributions are measured not at the output of the sample but at the output
of the spectrometer.
In a general formulation, the distribution of the electrons at the output
of the spectrometer are given by (see Section VI,A)
nd4) =
s 43)h(4,3)43)
where h(4,3) is the total transmission function of the spectrometer, taking

f 160)
into account the capture of the electrons by the collector as well as their
transmission by the analyzer itself. In the simplest case it is assumed that
h(4, 3) depends only on the energy variables E3 and E4. The integration over
x,, y,, and R3is therefore direct and
WE,) =
s d E 3 W 4 , E3 ) N ( E 3 )
The experimental energy distribution of the electrons at the output of the

(161)
spectromer can be written as

NS(E4) = BS(E4) + A,@,) ( 162)
where B,(E,) is the experimental background and A,(E,) is the experimental
Auger line with its energy-loss tail on the low-energy side. To separate the two
contributions B, and A, on the RHS of Eq. (162), Staib and Kirschner (1974)
suggested that the experimental background could be described by a fitted
spline polynomial. The interpolation is performed by fitting the polynomial
to both low- and high-energy wings of N,(E,) without accounting for the
energy-loss contributions on the low-energy side. In the dynamical back-
ground subtraction, the background disappears provided that sufficient de-
rivatives are taken. However, the rapid variations of the Auger signal are not
suppressed by derivation, and they can be retrieved by successive integra-
tions (Grant et al. 1974).
Background 0
Energy (arbitrary units)
FIG.23. Background and Auger electron peak.

Ramaker et al. (1979a,b) have criticized both these procedures: the

dynamic background subtraction because it is equivalent to subtracting a
background fitted only to the high-energy wing of the energy distribution,
and the Staib and Kirchner procedure because it does not account for the
energy losses. They proposed the following approximative scheme. The
experimental Auger line at the output of the spectrometer is connected to the
unknown true Auger line A,(E) at the output of the sample by the relation
As(E4) = 1 dE3H(E4, E 3 ) A T ( E 3 ) (163)
In practice, the function H ( E , , E 3 ) is very complex because it has to

take into account the following:
(1) the broadening of the Auger line in the sample;
(2) the capture by the collector of the electrons emitted by the target;
(3) the transmission of the analyzer;
(4) the modulation broadening (in the derivative mode);
(5) the efficiency of the detector.
The sample position effects on the energy shifts and on the peak-intensity
changes have been considered by Sickafus and Holloway (1975). The other
factors have been analyzed by Powell (1978) and by Bui Minh Duc (1981a).
Ramaker et al. assumed that H(E4,E 3 )could be obtained by convolution of
the target and spectrometer transmission functions. The target transmission
function HT transforms the true Auger line AT at the output of the sample
into N A ( E ; )according to the relation
NA(r3) =
s dE3H'I'(E!3 9 E3
The spectrometer transmission function H , can be experimentally

( 164)
studied. It transforms N A ( E ; ) into the experimental Auger line A,(E4)

according to
ME41 = f dE3HdE4, E W A ( E 3 ) (165)
Therefore, from Eqs. (163)-( 165), we obtain
For HT, Ramaker et d.proposed the following approximation:

in which each Auger electron source in the solid is assimilated to a primary-

electron source. In this expression Hi and He are the inelastic (energy-loss
spectrum) and the elastic (elastic peak) parts of the experimental energy
spectrum as measured with a primary beam of energy E A , x and of intensity
equal to unity. The factor (1 - E), where t is a numerical constant, takes
account of the differences between the geometries of the Auger electron beam
and the primary beam.
To extract the true Auger electron peak from N,(E,) as given by Eq.
(162), Ramaker et al. have proposed the following iterative procedure:
(1) A solution A,(E) is guessed.
(2) The lower energy side of the experimental Auger peak at the output
of the spectromer can be identified with the experimental energy-loss
spectrum of the true Auger electrons. Therefore in this energy range,
&(E - EL) N (1 - C)Hi(E, EA,X)C ( 168)
where C is obtained by dividing the area of the true Auger peak by the area of
the experimental elastic peak.
-
( 3 ) The background on the lower energy side of the experimental Auger
peak can be estimated by subtracting A,(E EL)from the experimental
N,(E,) according to the relation
B(E - EL) = N,(E - EL) - A,(E - EL) (169)
-
(4) A functional form including adjustable parameters is fitted to N,(E)
above the Auger peak and to B(E EL) below the Auger transition. The
functional form chosen by Ramaker et al. is
B(E) = AE(E - Eo)-l(E + q5)-m
+ B ln[(E, - E)/E,] [(E, - E)/Eb]- n +C 170)
where A, B, C, Eo, 4, Eb are adjustable parameters.
(5) The fitted background is then subtracted from N,(E), giving
A d a = N,(E) - B(E) 171)
(6) A new estimation ofA,(E) is obtained from A,@) by a deconvolution
procedure, i-e., by solving the integral equation (163).
The entire process is then repeated until the solution for A,(E) converges to
an unchanging value.
Sickafus and Kukla (1979) have used the functional form (Sickafus,
1977a,b)
B(E) = BE-" (172)
to describe the background. They used nearly the same analysis as Ramaker
et al.; however, they determined the adjustable parameters B and m by
254 M. CAILLER, J. P . GANACHAUD, AND D. ROPTIN
fitting Eq. (1 71) only to the higher energy wing of the Auger transition.
Sickafus (1980) has simplified Eq. (163) by using the approximate relation
JE*,x
E+€
A,(E) = AT(E) - BE-" AT(E')dE' (173)
where the constant t represents the minimum energy that can be lost by an
electron in an inelastic collision. In metals t = 0.
In studying the Auger line shapes, a solution to Eq. (163) is preferable, but
according to Sickafus, Eq. (173) is acceptable in the case of Auger-yield
measurements. In this latter case the algorithm used to evaluate the Auger
yield is much simpler and faster than that for the deconvolution in Eq. (163).
Moreover, the yield measurements made in this manner show greater
modulation stability than do peak-to-peak measurements in the derivative
mode (Sickafus and Winder, 1981).
By eliminating the long, lower energy tail, the deconvolution procedure
leads to a narrow true Auger line. In principle, the width of the experimental
Auger line would be equal to the natural linewidth of the transition. In view
of the difficulties inherent in the deconvolution procedure, Seah (1979a) has
proposed an alternative scheme to estimate the actual intensity of an Auger
peak without the losses. To obtain this intensity, the true Auger line is
presumed to have a symmetrical shape which is obtained by mirroring the
higher energy part of the experimental line. Such a scheme seems quite
valuable for a single narrow Auger peak, but its use for the multiple lines
such as those obtained in the study of Li,SO, and Li,PO, by Ramaker
et al. (1979a) is inadequate.
Theoretically, the broadening of the natural Auger line during the
transport of the Auger electrons to the free surface of the sample could be
simulated on a computer. The comparison between the so-obtained the-
oretical results and the experimental line shapes would be very interesting;
however making the simulation implies considerable effort to take into
account the natural width of the Auger line and all the individual mechanisms
of inelastic and elastic collisions.
Until there is general agreement concerning the operating conditions
and methods of calculation, Seah (1979a) proposes that the differential
spectrum be used for a quantitative AES. However, the energy-loss disym-
metry on the direct Auger spectra leads to a similar disymmetry on the
derivative spectrum. Giving a symmetrical shape to the direct spectrum by
suppressing the energy losses is equivalent to symmetrizing the negative peak
in the derivative spectrum. Therefore, Seah proposes to quantify the Auger
line by measuring the peak-to-background height of the negative peak in the
derivative mode. To conclude, the question of the use of the direct or
derivative spectra is still not settled.
C. QuantiJication of Auger Analysis

A quantification in absolute values leading, at least for simple targets, to
an agreement between theory and experiment has yet to be accomplished.
The way lies through a theoretical and numerical evaluation of the Auger
yield, from first principles; for instance, by using a simulation model on a
computer. Absolute measurements of the intensity of the Auger lines are
also necessqry.
To the knowledge of the authors, the most remarkable experimental
approach in that sense is that of Staib and Kirschner (1974).They evaluated
the absolute values of the Auger currents for different simple targets from the
measurement of the area under the Auger peak after subtraction of the
background. For this they used a conversion factor obtained by comparing
the area under the energy distribution curve of the electrons emitted by the
target with more than 1 keV and the corresponding measured current.
The number of the Auger electrons, at the output of the spectrometer, is
given by a relation similar to the simplified formula [Eq. (130)l after correc-
tion for the backscattering factor and for the transmission factor H S ( E A , J
of the spectrometer
NA'x = ?A,C,xNP [TP/cOs OPIT , x d ,C (EP ( E A ,x ) N A
x (471)-1 IQccos0 3 d f i 3 H S ( E A , x ) [ 1 + rP(EA,x, (174)
where the dependence on the primary variables of T and r has just been
recalled by the subscript P. Here H s is presumed to depend only on E A , x .
Therefore it is only a multiplying factor. This simplification is valuable only
if the number of the Auger electrons is studied. It does not work for studies
of line shapes.
From Eq. (174), Staib and Kirschner could go back to the numerical
atomic density N A . They worked at normal incidence (cos Op = 1) and set
yA,c,x T and T' equal to unity. The collection and transmission factor of the
spectrometer C(474-l jQ,cos e3 d f i L , H , ( ~ , , , ) ]was evaluated experimentally
by measuring in the same experiment the primary, target, and collected
currents. The cross section ohc was evaluated with Drawin's formula (1961),
the backscattering factor from the results of Gallon (1972) and Meyer and
Vrakking (1972), and the escape depth from the compilation given by Tracy
(1972).
The values N A deduced from the measurements by using Eq. (1 74) agree
with the values usually obtained for Ag, Cu, and C within better than 50%.
The disagreements mainly arise from uncertainties in the evaluation of the
various parameters involved in Eq. (174). They also result from uncertainties
concerning the validity of Eq. (174) itself, and as pointed out by Ramaker
et al. (1979a,b), from an incorrect subtraction of the background and of

the loss peaks.
In spite of the difficultiesencountered, this procedure is strongly requested
by theoreticians, although it is not absolutely necessary for the development
of a quantitative Auger analysis. A more important number of measurements
of this type are necessary in order to set up realistic models. The most
widespread method, in the absence of absolute measurements, is always the
calibration of the apparatus.
Once the problems of adjustment of the apparatus and preparation of the
sample are completely solved, one can distinguish, according to the case,
external calibration methods from internal.
In external calibration methods, the Auger line intensities of either pure
elements or alloys are measured under experimental conditions (primary
current density, accelerating voltage, etc.) as close as possible to those used
during the analysis. The concentrations of the analyzed target in its various
elements are then evaluated by making a comparison with the intensities of
the lines of the reference targets. If the external calibration is performed by
using a series of alloys of various but known compositions, there is generally a
systematic variation of the line intensities with the concentrations. Thanks to
this correspondance, the measurement of the line intensities allows in return
the evaluation of the composition of the unknown target, provided that the
preparation conditions of the sample and the adjustment of the apparatus are
the same in both cases. Because it is somewhat tedious, this method is not
often used.
Generally, one prefers the method of the relative elemental sensitivity
factors evaluated for pure targets (Palmberg, 1973; Davis et al., 1976).
This method is presented in some detail later. For a long time it was criticized
because it neglected matrix effects. These criticisms led Bouwman et al.
(1973a,b; 1976a,b)to propose the internal calibration method. This method is
well suited to the study of a physical or a chemical problem for a given sample.
For instance, it was extensively used to study the elemental surface
enrichment produced by sputtering or by segregation.
The principle of the internal calibration lines in obtaining a virgin surface
(for instance, by fracture of a prenotched test sample) and in measuring just
after the fracture (i.e., before the physical phenomenon one has to study takes
place) the Auger line intensities of the various components.
In general, one uses for a control several preparation modes of the virgin
surface (fracture in DUCUO, scribing, for instance). If the Auger line intensities
do not depend on the preparation mode, one can assume that they charac-
terize well the bulk composition. When this composition can be determined
by other experimental methods [by an electron-probe microanalysis (Ho
et al., 1976,1979)or by absorption spectroscopy (Goto et al., 1978), or by any
other method], it is possible to calibrate the line intensities. It seems that this
method is quite quantitative for compounds and for binary alloys. In this
case, one obtains a linear relation between the surface concentrations
measured by AES and the bulk concentrations. This was pointed out by
Bouwman et al. (1973a) for the three compounds Pt,Sn, PtSn, and PtSn, ;by
Van Santen et al. (1975) for Cu,Au and Au,Cu alloys; by Mathieu and
Landolt (1975) for Ag-Pd and Ni-Pd alloys; by Bouwman et al. (1976b)for
Ag-Au alloys; and by Betz (1980)and Betz et al. (1980a,b) for a large number
of solid solutions.
Similar observations have been reported by Mathieu and Landolt (1979)
for the ratios of the Auger line intensities of Cr and Fe and of Mo and Fe in
Fe-Cr-Mo alloys. Some different observations have been noted for
nonmixable systems such as Ag-Cu and Ag-Au-Cu. If one assumes that the
surface concentrations remain proportional to the line intensities, by
measuring the variations of these intensities one can express the modifications
of the elemental concentrations at the surface which result from the physical
process being studied (Betz, 1980; Betz et al., 1980a,b).
Another method is connected with the internal calibration method of
Bouwman, but it only allows estimation of the variations of the surface
concentrations. It consists in measuring the high- and low-energy peaks for a
given element. The low-energy peak represents well the surface composition,
whereas the high-energy peak gives information about a much thicker layer
and is thus much more connected with the bulk composition (Van Oostrom,
1979; Watanabe et al., 1976, 1977). Whatever the physical process being
brought into play, the low-energy peak increases more than the high-energy
peak (relative variations), and one might conclude that there is a surface
enrichment in this element.
The results obtained by AES could be made more complete with a probe
much more surface sensible than the low-energy Auger peak itself, such as
SIMS (secondary-ion mass spectrometry) or ISS (ion-scattering spec-
trometry). These techniques are respectively based on the study of the mass
distribution of the sputtered particles created by the primary-ion bombard-
ment and on the measurement of the energy of the primary ions elastically
scattered by the solid surface (Yabumoto et al., 1979).
The basic advantage of the internal calibration method compared to that
of external calibration is that matrix effects are automatically taken into
account, at least when the concentration variations remain located close to
the surface. This advantage is largely suppressed by the improvements
brought by Hall and Morabito (1979) to the relative elemental sensibility
factor method to account for the matrix effects in dilute alloys. In this case,
the compensation is no longer automatic, but it can be calculated. If one
assumes that this corrected elemental sensibility factor is exact, no difference
can be found, in principle, between the results obtained by this way and those
obtained by the internal calibration method. Holloway (1977) gives an
example of such an equivalence.
The general formalism for this calibration technique can logically be
applied to the intensities of the Auger currents obtained by integrating the
n(E) spectrum; however, it has been most frequently used together with
derivative spectra. Under these conditions its use can be justified as long as
the peak-to-peak height of the derivative spectrum actually represents a
measure of the Auger line intensity. In the present context, it is not necessary
to show how this intensity is measured. One should admit that the experi-
mental method is adequate. For instance, if the measurement is performed on
the n(E) spectrum, it is assumed that the subtraction of the continuous
background and the loss tail has been exactly achieved. The relative elemental
sensitivity factor of element A compared to a reference element R evaluated
from the Auger lines (x for the element A and r for R) of the target T is
with
CA(T) = NA/NT, CB(T) = NB/NT, .- - 9 CR(T) = NR/NT (176)
where N , is the average number of atoms per unit volume of the analyzed
target and NA,NB,. .., NRare the numerical atomic densities of the elements
A , B,..., R.
In the case of a multicomponent target, the sum of the concentrations
CDCDmust be equal to unity. From the definition relation, one notes that
Pre,(R,r ; A, x ; T) = P,;;(A, x ; R, r ; T) (177)
P,,,(A, x ; B, Y ;T) = P,,,(A, x ; R, r ; T)P,,,(R,r ; B, Y ;T) (178)
In other words, any element can be chosen as a reference.
The elemental concentrations can be expressed in terms of the intensities
N and the relative elemental sensitivity factors according to the relation
1 X ; D, Z ; T)]
CA(T) = NA9”(T)/ [ND,z(T)Pr,,(A,
DET
( 179)
where the sum in the denominator contains all the elements D constituting
the target T, and z is a label identifying the Auger line used to characterize
the D element.
The simplest approximation for the measurement of the relative elemental
sensitivity factors consists in supposing that they are independent of the
composition of the analyzed target. Thus, they can be evaluated from the
Auger line intensities of the pure elements measured under the same experi-
mental conditions as those for the unknown sample. This factor is therefore
approximated by
Pr,,(A, x ; R, r ) = NA~"(A)/NR*'(R) (180)
where the labels A or R indicate that the measurements have been performed
on pure A or R targets. The major disadvantage of this definition is that
matrix effects connected with the composition of the target are neglected.
These effects manifest themselves at least in the backscattering factor and in
the escape depth of the Auger electrons.
For the case of dilute alloys, Hall and Morabito (1979)have proposed eval-
uating the corrective factor, which allows the calculation of PreI(A,x ; R, r ; T)
whenever Prel(A,x ;R, r) is known, according to the relation
Pre,(A, x ;R, r ; T) = Pre,(A,x ;R, r)F(A, x ;R, T) (181)
For this purpose they made use of Eq. (174), which expresses the number of
Auger electrons in the A,x line, at the output of the spectrometer. This
number can be written
In a given target, the backscattering factor r depends not only on the

composition of the target but also on the parameters E p , Rp, Ea,,, and
R,. kA,.+ stands for
and, in practice, depends on the Auger line considered. It depends also on the
parameters of the primary beam, the roughness of the target surface, the
ionization cross section of subshell C, and the characteristics of the
spectrometer.
If one assumes that all the parameters determining the value of k are kept
fixed (except the choice of a particular A,x Auger line), kA,xdepends only on
the selected line, as indicated by the A and x subscripts. In the case of a pure
target A, the number of Auger electrons in the (A,x) line is
where LA,,(A) is the inelastic MFP in the elemental target A of the Auger
electrons of the Auger line A,x which have an energy E A , x , and 1 + rA,JA)
is the backscattering factor for the Auger line A,x in the same target A.
For an arbitrary target T, the number of Auger electrons in the Auger
line B,y is
NB'y(T) = ~ B , Y ~ B ( ~ ) ~+BrB,y(T)]NP
,~(~)[~ (185)
If this target is a dilute alloy of B into A, one has

NB9y(T) NB’y(A) = kB,yNB(T)AB,y(A)[l+ ‘B,Y(~)]~P ( 186)
In Eq. (186), NB(T)is the number of B atoms per unit volume of the target
and N(T) is the average number of atoms, per unit volume, independent of
their nature. The concentration in element B within the explored zone is
cB = NB (T)/N(T) (187)
From Eqs. (179, (1 SO), (1 8 l), (1 84), and (1 86), one can write
It must be recalled here that the backscattering factors depend on the

acceleration energy of the primary electrons.
Hall and Morabito have tabulated, the values of the corrective factor F
for 55 different elements and for two different choices of MFP values (Penn,
1976; Seah and Dench, 1979). For the atomic densities, they used the values
tabulated by Kittel (1953) and for the backscattering factors, the values
deduced from the backscattering coefficients q given by Reuter (1 972)
from the relations
‘B,y = 12.8[1 - 0*9(EB,C/Ek’)]Y (189)
q = -0.0254 + 0.0162 - 0.000186Z2 + 8.3 x 10-’Z3 (190)
where E B , c is the energy of the ionization threshold for the subshell C of the
atom B, and 2 is the mean atomic number of the sample (for an A based
dilute alloy 2 = ZA).
A comparative study of the results obtained from the internal (Bouwman
et al.) and external (Hall and Morabito) calibration techniques would be
particularly interesting. This question is particularly important if, as sug-
gested by Hall and Morabito, their relative elemental sensitivity factors
depend only slightly (or even not at all) on the composition of the binary
alloy. An example of such a comparison has been given by Holloway (1977)
in the case of Cr-Au alloys for which the concentration in Cu varied from 0 to
20%. From results communicated by Hall and Morabito, before they were
published, they could correct the elemental concentrations measured by
AES on a scribed surface. After correction ( 1 5% in the concentration domain
studied), Holloway obtained good agreement between the surface and the
bulk concentrations.
A method of quantification based on a layer model of the volume analyzed
has been developed by Pons et al. (1977a,b,c)and has been applied to Fe-Cr
ferritic steels to the study of passivation films on Fe-Cr and Fe-Cr-Mo steels
(Da Cunha Belo et ul., 1977)and also to Ti and its alloys (Bouquet et al. 1977).
By reference to Eq. (117), the number of Auger electrons created by the
primary beam in the ith atom layer is
N f J = YA,~NA(~)
[Np(i)/(cos 6 > i ] 4 , J E P ) (117')
where Np(i) is the number of primary electrons crossing the ith atom layer.
As can be seen from Eq. (117'), Nfsx depends on the number of atoms A
per unit volume in the ith layer, on Np(i),and on the average value of the
cosine of the propagation angle 6 of the primary electrons.
Neglecting the dependence on i in this last factor, one can write
N f , x = UA,~NA(~)NP(~) (191)
Among all these Auger electrons, only a fraction will reach the surface and
be collected in the spectrometer. Their number will be
Nf$ = aL,,NA(i)Np(i)kfPX ( 192)
In fact, because of the contribution of the backscattered electrons, Eq.
(192) has to be replaced by
+ rA,x(i)]kf'X
N;;; = GY~,,NA(i)NP(i)[l (192')
The product N,(i)[l + TA,&)] can be approximated by an exponential
N exp(id/u) term, where N is the number of efficient electrons (primary or
backscattered) in the first atom layer, d is the thickness of an atom layer, and
a is a length to be evaluated, which depends on the Auger line considered.
For a CMA, can be approximated by a second exponential term:
k f S x
so that N f $ can be expressed by the relation

N;,:' = UL,,NA(i)ki,, (194)
where the factor kA,x (to be raised to the ith power) is the product of two
exponential terms :
and k2 = exp
The total contribution of all the layers is

For example, for a pure sample, NA(l’)is independent of i and equals

NA. Thus
N;.” = UA,~N(A)NA(~
- ICA,~)-’ (1 96’)
where N(A) is the number of efficient primary electrons in the first atom layer.
It must not be confused with NA,which is the numerical density of the pure
element A. The ratio of the Auger intensities due to the pure elements
A and B is
In the absence of experimental observations, the ratio N(A)/N(B) is taken

equal to the ratio (1 + rAJ/(l + T ~ , and~ ) is evaluated from Eqs. (189)
and (190).
By using Eq. (197), Le Hericy and Langeron (1981) have determined the
ratio crb/a;, from measurements on TiO, Ti,03, and TiOz samples. For
that they evaluated kA,xand k,,y from Eq. (195) and the Seah and Dench’s
universal curves for the MFP. They obtained ab/a& N 1,2. They have also
discussed the cases of the heterogeneous samples and reported that in these
cases only approximations can be obtained.
Quantitative Analysis from the Auger Line Shapes

Some papers have lately been devoted to the development of a quantita-
tive analysis method using the Auger line shapes. This new field of study is
very promising. Turner et al. (1980) performed a quantitative analysis of the
surface composition of sulfur-bearing anion mixtures. The anions have a
characteristic LVV line shape which depends only slightly on the nature
of the cation.
To study a binary mixture for instance, the AES spectra of the individual
components have to be made comparable. For that, the AES spectrum of one
component is normalized to that of the other with respect to the XPS area of
the S2p peak and to the beam currents. Then, the component concentrations
are determined by adjusting a linear combination of the component spectra
to the mixture spectrum.
Gaarenstroom (1 98 1) made a very interesting application of the principal-
component analysis method to the study of the Auger line shapes either in
the case of a component mixture or for a depth profiling. The technique is
especially devoted to the analysis of the continuous evolution of the line
shape from one mixture to another (or from one depth to another) when the
Auger lines of the components are in the same energy range. The principles
of the method are the following:
(1) The line shapes of the n different mixtures are digitized for p energy
channels giving a data matrix D with n rows and p columns.
(2) The matrix W which is the product of D by its transpose (W = DD‘)
is of order n and rank m where m is the number of components in the mixtures
(m < n and m < p).
(3) The number rn of components is equal to the number of nonzero
eigenvalues of the matrix W. This result can be proved mathematically.
Owing to the uncertainties inherent in the experimental data, the numerical
results have to be interpreted with some care. For instance, a statistical test
must be used to decide when a given eigenvalue is “physically” different
from zero.
(4) Moreover, it is possible to obtain quantitative information about the
mixture compositions if the Auger spectra of the pure components are known
and included in the data matrix D. In other words, these pure component
spectra are digitized and constitute two rows of the data matrix D. In this
case, and for binary mixtures A-B, the n eigenvectors of the matrix W have
two principal components x and y. Especially normalized eigenvectors
associated with the pure elements A and B have as components (xA, yA),
(xB,ye). They determine in the plane (x, y ) two vectors OA and OB which
represent the pure components. So they determine a system of oblique axes on
which each mixture eigenvector OM can be projected. The lengths of these
projections determine the concentrations CA and CB in this particular
mixture (see Fig. 24).
Although, to the authors’ knowledge no mathematical evidence was given to
this result, the method seems to work well. However, this quantitative
analysis rests on the assumption that the Auger lines are constitutive. In
other words, it is assumed that the Auger line of the mixture can be deter-
mined as a linear combination of the pure component Auger lines. Such an
assumption was also the basis of the analysis by Turner et al. (1980), but
presumably still deserves further confirmation for other systems.
FIG.24. Determination of concentrations: Gaarenstroom’smethod.

D. Sample Preparation
A first method consists in elaborating the sample in a chamber by
evaporation-deposition on a substrate. Such a procedure was used by Goto
et al. (1978) to elaborate a series of Cu-Ni alloys by coevaporation. In the
case of an evaporation onto a liquid nitrogen-cooled substrate, the surface
concentrations given by AES are linearly related with the bulk concentrations
measured by atomic absorption spectroscopy. In the case of a coevaporation
at room temperature, there is an enrichment of the surface in copper, typical
of a thermal segregation. This is only one example of the difficulties
encountered in the preparation of samples.
When the evaporation speed is well controlled and is slow, it is possible to
observe in the time variance of the Auger line intensity of the deposit some
discontinuities typical of a layer-by-layer growth. This was observed by
Guglielmacci and Gillet (1980a) in the case of Ag deposits on a (11l)Pd
surface for a deposition speed of 1 A min-l and a substrate temperature of
20°C. This represents a means of calibrating the thickness of a film, par-
ticularly for submonolayer deposits. More commonly, especially for
technological-like samples, the preparation of the target consists in suppres-
sing the contamination of the surface induced by the gases or the vapors
of the atmosphere or by the pretreatments suffered by the sample.
Bouwman et al. (1978) quote four decontamination procedures:
(1) chemical etching in which a chemical reaction is induced at the
surface of the target as, for example, in the case of the removal of carbon by
exposing the surface to oxygen at 500°C;
(2) plasma etching involving a chemical reaction by intervention of a
reactive plasma so that the etched material is changed into a volatile
compound.
(3) thermal desorption by heating or flashing the sample in uacuo; and
(4) ion etching, which is the procedure most currently used and consists
in sputtering the target with ions, most frequently low-energy inert-gas ions.
Ideally, any method could allow elimination of the contaminating species
without modifying the masked surface which represents the central object
of study.
One often wants to make a surface initially submerged in the bulk of the
sample free. The reason could be that the surface presents some interesting
characteristics, for instance, grain boundaries in the case of the grain-
boundary embrittlement of steels or superalloys. For that, a fracture is
performed in situ in the UHV chamber (McMahon et al., 1977; Briant and
Banerji, 1978, 1979a,b,c). The reason could also be that once the surface is
free of contamination, it can be taken as a standard reference for studying the
physicochemical surface mechanisms (catalysis, gas adsorption, oxidation,
gaseous corrosion, epitaxial growth, surface segregation, etc.). For this, one
usually proceeds to a fracture in uucuo or to removing some matter by

either scribing or lapping.
Finally, one also very often performs a profile analysis in depth. This
technique has as its essential aim the study of the variation of the elemental
concentrations in the bulk of the target. Some parts of the material are
gradually removed so that surfaces which were initially merged at more and
more important depths in the bulk can be analyzed by this method.
In the case considered here, the material to be eliminated is nearly
always sputtered by bombarding the target by ions more energetic than for a
simple decontamination. The analysis is made by AES. Another mode of
etching by a high-power laser technique has been considered by Papagno
et al. (1 980). All these methods for preparing the sample raise some questions
with respect to the quantification of the results. Some of these problems are
now reviewed in the following sections.
1. Preparation by Fracturing in Vacuo

As mentioned earlier, this technique of preparation can be used in two
very different cases. In the first case, the fracture aims to prepare a virgin
reference surface. To achieve this, the rupture must occur either by cleavage,
for brittle materials, or by transgranular fracture, for ductile samples (Van
Oostrom, 1976, Rehn and Wiedersich, 1980).
The assumption that the cleavage surface or the transgranular fracture
domains are representative of the bulk solid is probably satisfactory, except
if voids or particles of a second phase are present. Even in this case, however,
partial analyses leading to the determination of the concentrations of some
elements are still possible (Mulford et al., 1980). Nevertheless, charac-
terization of the fracture surfaces by other techniques must accompany AES.
In the second case, the object of the studies is the embrittlement of
materials by grain-boundary segregation. Therefore, the characterization of
the embrittling solutes which segregate to the grain boundaries requires
that the fracture be intergranular. There are few studies to determine the
dependence of the results of Auger analysis on the characteristics of the frac-
ture surface. To our knowledge, the only studies on this topic were carried
out by Rowe et al. (1978) and by Mulford et al. (1980). By using a high-
resolution scanning Auger microscope they proceeded to a grain-by-grain
analysis of the fracture surface. They made the following observations :
(1) The diffusion along the surface of contaminants from the outside
edges of the fracture surface does not appear to be a problem in the alloys
studied (stainless steels, Fe-Si alloys, low-alloy steels).
(2) At a grain boundary, the impurities are more or less randomly
separated by fracture between the two sides of the fracture surface. On an
average, the quantities of the impurities on each side are equal (50-50%),
but they can vary from one grain to another between the two limits 40 and
60%. Thus, no side is systematically favored.
(3) On a side of the fracture surface, the impurity concentrations vary
smoothly from place to place on an individual facet.
Thus it is shown from observations (2) and (3) that the fracture surface
follows the central region of the boundary. However, Mulford et al. (1980)
caution against generalizing this result to systems other than those they
studied.
(4) The heights of the Auger peaks of the segregated elements vary from
grain to grain because of the variation of the incidence angle; however, this
angular effect can be minimized by means of a normalization technique. For
instance, in the case of steel this technique consists in dividing the height of
the Auger peak of the segregated element by the height of the iron Auger
peak. Under these conditions, the angular effect is no longer dominant, and
the results so obtained give a correct measure of the impurity concentrations
on each grain. It is then possible to study the variations of the elemental
concentrations at the grain boundaries, for instance, during a heat treatment
of the sample.
2. Preparation by Scribing
This mode of preparing a virgin surface was used by Betz (1980)and Betz
et al. (1980a,b) to compete with the method of fracturing in uacuo in order to
characterize the elemental concentrations at the surfaces of binary and
ternary alloys. It is assumed that the results of the Auger analysis are char-
acteristic of the bulk composition of the target if they are the same in both
modes of preparation of the virgin surface. The identification of the surface
concentrations as measured by AES with the bulk concentrations as
measured by an electron microprobe, for instance, requires good calibration
of the Auger technique (Holloway, 1977).
The reference state having been defined in this way, it is then possible to
study the subsequent variations of the concentrations at the surface by
measuring the variations of the intensities of the Auger peaks. Various tools
have been used to scribe or scrape the sample, such as a diamond or a
stainless-steel point or a tungsten carbide blade on a linear rotary-motion
feedthrough (Sliisser and Winograd, 1979) or a tungsten brush (Yu and
Reuter, 1981a,b).
3. Preparation by Lapping
This mode of preparing the sample is intended for depth profiling where
it can compete with sputtering. This method was known for a long time but
was only recently used in relation with AES (Tarng and Fisher, 1978). It
consists in removing part of the sample in order to cause either a spherical
surface or an oblique flat surface to appear.
Lapping has been particularly used to prepare an overlayered sample in
view of the analysis of the interface between the substrate and the deposit.
Then, in the case of spherical lapping, the diameter of the ball used to lap the
sample must be sufficiently large so that the interface area which is available
for the analysis is much larger than the electron beam.
Then, depth profiling is realized by laterally moving the electron beam on
the sample with the help of an electrostatic deflection. However, if the area
which has to be analyzed is so large that the electrostatic deflection will lead
to a wrong use of the analyzer, it will then be necessary to mechanically
translate the sample in front of the electron beam.
In practice, spherical lapping has been realized with the help of a steel
ball freely rotating in contact with the sample while it is fed with diamond
paste (Thompson et al., 1979; Walls et al., 1979; Lea and Seah, 1981). In
some cases (brittle or soft materials), however, the method may not work. In
any case, lapping, as described above, must be followed by cleaning the
sample before introduction into the UHV vessel and by decontamination
treatment after introduction into it. Removal of the contaminants is accom-
plished by a low-energy ion sputtering.
The advantage of the method is that it is possible to obtain very smooth
spherical surfaces (roughness of 100-500 A by using 0.1-pm diamond paste).
Therefore, the lapping-depth resolution Azp, which is limited by the rough-
ness of the lapped surface, can be compared with the sputtering-depth
resolution. Such a comparison was made by Lea and Seah (1981) in the
case of an overlayered sample. Five situations corresponding to different
roughness conditions for the free surface and the interface were distinguished.
From their study, it appears that for deposit thicknesses of the order of
100 nm, sputtering will lead to a better depth resolution than lapping;
however, for thicknesses higher than 1 pm, the interface broadening obtained
by very smooth lapping is smaller than that obtained by ion sputtering.
It should be noted that the theoretical comparison by Lea and Seah of
the lapping and ion-sputtering depth resolutions was made by using the
results of the sequential-layer sputtering (or SLS) model of Benninghoven
(1971) and Hofmann (1976). In this SLS model the contribution A+ to the
depth resolution, inherent in the mechanism of sputtering itself, is considered
to be proportional to the square root of the thickness of the removed layer.
Seah et al. (1981) now doubt the validity of this square-root dependence.
Taking into account the site dependence of the sputtering probability of an
atom, they modified the SLS model. They showed (Section VII,E) that as soon
as the thickness t of the sputtered layer exceeds about ten atom layers, the
depth resolution Azs becomes independent of t and can be neglected, com-

pared with the other interface broadening contributions (atomic mixing,
diffusion effects, etc.). The theoretical comparison of the lapping and
ion-sputtering depth resolutions must therefore be reformulated.
4. Preparation by Sputtering
Sputter etching a target by ion bombardment is the most widespread
method for achieving depth-profile measurements by AES. Most frequently
one uses noble gas ions, especially argon; however, other types of ions have
been used, for instance, 0; and N i , to obtain a better sputtering yield or to
avoid the chemical reduction of the target or to obtain a better resolution in
depth as well. The cleared crater is always of large dimension compared with
the diameter of the Auger probe.
In view of its quantitative elemental analysis, etching a surface by ion
bombardment raises some problems caused by the following:
(1) There are uncertainties about the depth at which the analyzed
surface was located before sputter etching. These uncertainties come partly
from the roughness of the stripped surface, either its initial roughness (before
sputtering) or that induced by the ion bombardment. In the latter case, it may
be due to irregularities in the ionic current density, to inhomogeneities or
anisotropies in the target, or simply to the random character inherent in the
sputtering effect. We may consider that these uncertainties are purely
geometrical in character. They exist prior to any Auger analysis, and their
degree of intervention is partly determined by the transverse dimensions of
the Auger microprobe. Smith (1976) has studied the sputtering effects in
aluminum oxides and has established contour maps of the film thickness by
ellipsometry. The sputtering provoked by the ions is not uniform and there
are more or less hollow domains. The electron beam, which should be
incident on an oblique microsurface, could be deviated to strike the surface
in a hollow part.
(2) There are some uncertainties connected with the finite value of the
zone thickness explored by the probe, that is, by Auger electrons. This
thickness runs from about one to a few Auger electron MFPs. It increases
with increasing electron energies, except at low values (5100 eV). Strictly
speaking, these uncertainties depend on the surface roughness but are
generally considered to be independent of the sputtering result.
(3) There are surface-composition modifications from
(a) preferential sputtering of one or several elements, leading to a
surface depletion in these high-sputtering-yield elements;
(b) implantation of primary ions and introduction of impurities
from the surface;
(c) surface diffusion of some elements;

(d) ion-induced atomic mixing, particularly near the interfaces ;
(e) creation of point defects and the resulting enhancement of the
diffusion ;
(f) chemical reactions induced by the ion beam.
These compositional changes could, in turn, produce some uncertainties
about the value of the initial depth of an interface when used as a means to
observe this interface. In this case, the uncertainties generated in this way
will combine with the uncertainties associated with the roughness, the probe,
and the initial lack of parallelism between the interface and the free surface of
the sample. The consequence of all these effects will be that the variation of
the signal used to characterize the target will not present the expected steplike
variation at the crossing of the interface (supposed to be ideal), but it will
exhibit a continuous variation.
A measure of the interface broadening (Fig. 25) is then given by the
difference Az between the two “depth” values corresponding to signal
amplitudes of 84.15 and 15.85% of the maximum amplitude of the analyzed
signal. The depth values are just mean values, estimated from sputtering
times. It seems that no overall theoretical study of the total interface
broadening has been made to date.
Any of the contributions to the broadening can be related to a separate
term Azi so that
AZ = (1A z ? ) ~ ”
i
if one assumes that they represent independent contributions. This is not

necessarily true, and Eq. (198) gives an upper limit for the total interface
broadening.
The interface broadening AzR due to the initial roughness of the target in
the case of a homogeneous and isotropic overlayer having a thickness zo
along the vaporization and condensation direction has been the object of
’oo%3L
, 84.15%
1565%
0%
FIG.25. Interface broadening.

studies by Seah and Lea, 1981. If evaporation is done normally to the

mean surface of the substrate and if the ionic bombardment is done under an
incidence angle B with respect to this mean surface, the interface broadening
is given by
where z is the sputtered thickness and M , is the standard deviation of the

angular distribution supposed to be Gaussian
B = (f- 1)tanB when fl > m 0 (200)
with
f 3: HSub/(49.5Ali2) (201)
In Eq. (201), A is the atomic weight of the target and Hsubis its molar
sublimation enthalpy (in kilojoules per mole).
In this case the interface broadening would be proportional to the
thickness of the material to be sputtered and the weaker the angle p, the
weaker the broadening. It would then be desirable to operate at normal
incidence to reduce AzR ; however, under these conditions Eq. (199) is no
longer valid and has to be replaced by
AzR = 1.66~;ji (204
We note that according to Eq. (198), the total interface broadening Az
differs from AzR due to the other causes of broadening. Minimizing AzR does
not necessarily minimize Az. For instance, Seah and Lea have considered
what would happen if the interface broadening due to atomic mixing would
compete with that due to initial roughness. Assuming that the former effect
does not depend on fl, they deduced that for M, > lo, it is better to operate
with a normal incidence, whereas for a, < 0.35", an oblique incidence could
lead to a better depth resolution. One may note from Eqs. (199) and (202)
that AzR is determined by the distribution of the microfacet orientations,
which are characterized by a, and not by the height of the rugosities. This
conclusion agrees well with the experimental results of Mathieu et al. (1976).
For elements such as Au, Ag, Pd, and Cu, f is less than unity. With this
condition, application of Eq. (199) raises some problems because its RHS
member is then negative. One would be led to admit that for fl > M , for these
elements, the ion bombardment tends to cause the initial roughness to
disappear. At least it is clear that for Ni-Cr alternate sandwich structures, the
interface broadening is much stronger for an initially rough sample than for a
smooth one (Hofmann et al., 1977).
Roughness can be induced by sputtering. Thus, for example, it has been
observed that a ridge and valley structure (with heights of about 50 A )
appeared by sputtering on amorphous SiO, targets (Cook et al., 1980). For a

crystalline material, roughness can be induced by preferential sputtering
along some special crystallographic directions or by channeling effects.
These effects can be particularly severe for polycrystalline materials.
Roughness can also be induced by inhomogeneities, especially by foreign
materials, at low sputtering rates. According to Hofmann (1977),the interface
broadening due to these defects or to the different orientations in the crystal
lattice would be proportional to the sputtered thickness. According to
Bindell et al. (1976) and Laty and Degreve (1979),very often it is the surface
roughness which represents, in practice, the limitation for the resolution
in depth.
The interface broadening AzA due to Auger analysis is of the order of the
Auger electron inelastic MFP (or a few times this MFP). Thus it is of the
order of a few angstroms, and the weaker the energy of the Auger electrons
used for the analysis, the smaller the broadening. It is usually admitted that
AzA is independent of the sputtered thickness. The influence of surface
roughness upon this interface broadening is not well known, however.
In the literature one usually uses the term “preferential sputtering” in the
case of multicomponent targets in which at least one element is removed
more quickly than the others. Nevertheless, in the case of pure targets one
can also consider that preferential sputtering occurs for some atoms but not
for others, even though they are similar in nature. This greater or lesser
aptitude of the atoms for being ejected is related to position in the surface
layer. The fewer the atoms in a given layer, due to sputtering, the lower the
coordination numbers of the remaining atoms. Consequently, the instan-
taneous binding energy of these atoms decreases as bonds are broken, and
their sputtering is made easier than that of atoms in an intact layer. This
effect has been introduced by Seah et al. (1981) in the model for sequential-
layer sputtering of Benninghoven (1971) and Hofmann (1976).
This model is based upon simple statistical arguments in which the solid
has a layer structure, and sputtering takes place only in that part of the atom
layer exposed to the ion beam, with a rate i. In the basic model, the rate is
constant, but Seah et al. assume that this rate is a linear function of the
fractional coverage 8, between the two limits 6 = 0, where only one atom is
still present, and 0 = 1, where the layer is unaltered. The ratio k of the rates
i(0 = 0) and i(O = 1) is given by the ratio of the coordination numbers of the
atom in these two extreme situations.
In theory, this ratio varies from 1 to co with respect to the orientation of
the atom layers, but in practice it could not exceed a value of 3. For k > 2.5,
the resolution in depth Azs due to the sequential sputtering mechanism
becomes nearly constant beyond 10 evaporated layers. On the other hand,
as soon as k exceeds the value 1.5, which seems to be the usual case [for
instance, k = 3 for the ( 1 11) plane of an fcc system], the influence of the
sequential sputtering quickly decreases so that the resolution in depth is
governed by other causes.
By computer simulation Harrison et al. (1978) have determined the
differences between the sputtering yields of the (loo), (1 lo), and (1 1 I ) faces of
copper for 600-eV argon ions, the sputtering yield being defined as a mean
value in the impact zone of the ion beam of the number of atoms ejected per
incident ion.
Their statistical technique, compared to the statistical model of Benning-
hoven, plays the same role in the description of the intrinsic sputtering
mechanism as that played by a computer simulation of Auger emission
compared to the analytical resolution of the Boltzmann transport equation.
The enormous advantage of computer simulation is that it makes it possible
to take into account more realistic physical assumptions than those obtained
by analytical description. The values obtained are, respectively, 3.93, 3.54,
and 6.48 for the (loo), (1lo), and (1 11) faces. The tendencies one notes from
these results agree well with experimental observations.
Following Hofmann (1977), the interface broadening Az, due to
inhomogeneities in the intensity distribution of the primary ions would be
proportional to the thickness of the sputtered material. On the other hand,
the interface broadening Az, of the knock-on and ion-induced atomic-
mixing effects would be independent of the sputtered thickness and would be
functions of the energy and the mass ratio between the target atoms and the
primary ions (Hofmann, 1977).It would also depend on the angle of incidence
of the ions. Liau et al. (1978, 1979) have established that the thickness of the
mixed layer was approximately proportional to the energy of the ions for
Ar+ in the energy domain from 10 to 160 keV. By extrapolating at 2 keV, one
obtains a mixed-layer thickness of 30 A which is higher than the escape depth
of the Auger electrons (at least those of the low energy).
Broadening by knock-on effect and by atomic mixing can be reduced by
lowering the energy of the ions. Thus for argon ions of less than 1 keV, there
would be no appreciable broadening (Mathieu et al., 1976),and the important
factors would be the surface roughness, the escape depth of the Auger
electrons, and the fact that the interface is itself ill-defined. The broadening
effect by atomic mixing can also be lowered by using heavy ions or grazing
incidences.
To sum up (Hofmann, 1979), the escape depth of the Auger electrons, the
knock-on effect, the preferential sputtering, and the diffusion are determining
for small values of the etched thickness ( c10 nm). Beyond these values, the
irregularities of the beam, the initial roughness of the surface, the sequential
sputtering, the surface diffusion, and the crystalline character of the sample
are the most important factors.
Two examples are now given to illustrate this analysis :

(1) Hofmann and Zalar (1979) have studied the depth resolution in
Ni-Cr sandwich structures, sputtered by argon ions of 1 keV. They
approximated their experimental results by the relation
Az = 2(az + A’)’’’ (203)
with a = 0.3 nm and 1 = 1 n m The values deduced from Eq. (203)and the
experimental results have good agreement when the etched thickness exceeds
50 nm. As an illustration, the experimental depth resolution given by Eq.
(203)would be 11 nm for an etched thickness of 100 nm. The depth resolution
is slightly smaller if one uses N
: ions, but the difference is not significant
(10 nm for an etched thickness of 100 nm).
(2) Roll and Hammer (1979) have studied the Ni-Mo and Co-Mo
sandwich structures and approximated their results for the depth resolution
by the relation
AZ = C I Z ’ ’ ~ + /I (204)
For z = 100 nm, the depth resolution is of the order of 7-12 nm, depending
on the couples of materials and according as crossing the interface leads from
material A to material B or the reverse.
E. Efects of Sputtering on the Surface Composition

of Multicomponent Materials
Multiphase systems should be studied apart from alloys or homogeneous
compounds, which will be considered here. Numerous studies have been
devoted to the analysis of homogeneous binary and ternary targets. The
results obtained show that, generally, surface composition after sputtering
differs from that obtained immediately after fracturing or scribing. Ionic
bombardment in fact creates an altered layer at the surface which spreads
over several atom layers (Ho, 1978) and has a composition which is different
from the mean composition of the bulk. The thickness of this Iayer increases
first with the dose of ions (this is the transient state). It tends to a constant
value when the sputtering times increase.
In a rather general way the values proposed for the altered layer (for the
steady state), at ambient temperature and for argon ions with an energy of
about 1 keV, range from a few angstroms to 30 A (Ho et al., 1976, 1977;
Watanabe et al., 1977; Betz, 1980), which is of the order of the penetration
depth of the ions (Mathieu and Landolt, 1979;Kim et al., 1974);however, the
thickness of the altered layer is much smaller than the projected range of the
ions (Kim et al., 1974). In the case of Al-Cu alloys sputtered by 1-keV Xe
274 M. CAILLER, J. P. GANACHAUD, A N D D. ROPTIN
ions, however, Chu et al. (1976) have reported a value of 300 A for the
thickness of the altered layer.
In any case, the altered-layer thickness values cannot be neglected by
comparison with the MFP and the escape depth of the Auger electrons.
Therefore, the measured Auger signals are highly dependent on the altered
layer. The effect is normally more important for low-energy Auger electrons
than for high-energy Auger electrons. Thus one has a simple means of
collecting information on the altered-layer composition.
Several elementary mechanisms contribute to the variation of the altered-
layer composition:
preferential sputtering
enhanced diffusion
thermal segregation
atomic mixing
precipitation
No global theoretical description has apparently been given for the altered
layer.
The simplest model takes into account the preferential sputtering of some
elements compared to others. When steady-state conditions have been
reached for sputtering, the target is sputtered in a stoichiometric way with
respect to its bulk composition
(1) if S(A) and S(B) are the sputtering yields of elements A and B which
constitute the alloy or the homogeneous compound;
(2) if &,(EA, x) and I,,(&, y ) are the intensities of the Auger lines x and y
of elements A and B before sputtering and Is(EA,x) and Zs(EB,y ) the Auger
intensities of the same lines in the sputtering steady state, one can write
(Shimizu and Saeki, 1977)
By measuring the ratios of the intensities of the high- and low-energy peaks of
elements A and B, one can approximate the ratio S(A)/S(B) so long as the
intensities of the high-energy Auger peaks can be considered, in an
approximate way, representative of the bulk composition of the sample.
In the case of solid solutions or compounds, the steady state is quickly
reached, as soon as the thickness of the eliminated layer becomes comparable
with the penetration depth of the primary ions.
For multiphase systems, preferential sputtering can also take place if the
phases have different sputtering rates. The steady state will be reached only
after the elimination of a layer having a thickness representative of a few
grain diameters (Henrich and Fan, 1974). In this case, the formation of cones
and the accumulation of species difficult to sputter leads to very complicated
sputtering behavior.
Two parameters determine preferential sputtering. The first is related to
the way in which kinetic energy is transferred from the energetic ion to the
atoms. For most compounds, the energy of the ion is transmitted principally
to light atoms which would be preferentially sputtered if this parameter were
preponderant. This kinematic model leads to surface enrichment of the heavy
material (Haff and Switkowski, 1976; Haff, 1977; Kelly, 1978). The second
parameter is the binding energy of the atoms, and its principal effect is to
reduce their recoil energy. For the extreme case where only the binding-
energy effect plays a role in preferential sputtering, the result is enrichment in
the material having the highest binding energy.
The role played by these two parameters appears in the theory of Sigmund
(1968, 1969a,b) who expressed the sputtering yield of pure amorphous or
polycrystalline substances in the form
where f ( M , , M i ) is a function of the atomic weights, M , for the target and

Mi for the incident ions; Eiis the energy of the primary ions; U , is the
binding energy of the target atoms, a measure of which is given by the
enthalpy of sublimation. The product S V, is called the recoil energy density.
If the function f varies slowly with the atomic weight of the target, as
soon as the atomic number of this target exceeds 30, the sputtering yield of a
sample will be lower the stronger its binding energy is. If, on the other hand,
two elements have the same binding energy, the heavier one will have the
higher sputtering yield, the function f increasing with M , .
An empirical relation for the sputtering yield has been proposed by
Okajima (1980):
S = K’(M,/E,)k (207)
in which E, is the cohesion energy, M , is the atomic weight of the target, and k
and K’ are constants. According to this relation, the sputtering yield depends
only on the ratio M , / E , . The value obtained for k is the same for 10-keV
argon ions and for 45-keV krypton ions ( k N 413). For 45-keV krypton
ions, K‘ N 40 when E , / M , is expressed in J kg-l.
The yields for the sputtering of pure elements by argon ions of 500-1000
eV have also been analyzed by Seah (1981), who showed that there was good
overall agreement among the experimental results once they were corrected
for the effects of contamination by oxidation and the theoretical results of
Sigmund. Thus the well-known quasi-parabolic variation of the binding
energy in each of the three transition metal series, related to the filling of the
276 M . CAILLER, J. P . GANACHAUD, AND D . ROPTIN
electronic levels, leads to a systematic but inverse variation for the sputtering
yields.
An examination of Fig. 3 of Seah’s paper, which presents the variations of
the ratio of the measured and predicted yields for 500-eV argon ions as a
function of the enthalpy of reaction with oxygen per gram atom of oxygen,
seems to indicate that there is not just a single relation but several between
these two quantities, a systematic variation appearing, for instance, for each
transition metal series. Taking into account this multiplicity allows a slight
amelioration of the correlation established by Seah between the corrected
experimental yield and the predicted yield.
It is not obvious that the sputtering yields of an element in an alloy or in
a binary or ternary compound are the same as in the pure material. One
might even expect that this will not be the case. It seems, however, that
according to the studies by Betz (1980) for alloys and compounds having
recoil energy densities SU,, nearly equal, the surface is enriched in the element
having the smallest sputtering yields in its pure state. When the components
have quite different recoil energy densities, one can see from Eq. (206) that
the effect of the difference of weights can play an important role in the
evaluation of the sputtering yields, and that enrichment in the heavier
component can take place.
Table IV shows the results obtained for preferential sputtering surface
enrichment. Here, the elements have been classified for both entries along
increasing enthalpies of sublimation. A cross (for an alloy) or a circle (for a
compound) has been placed in the column of an element for which surface
enrichment occurs. Table IV shows that, except in relatively exceptional cases,
there is enrichment by the component which has the strongest enthalpy of
sublimation, and the weight factor plays a secondary role.
The exceptions in Table IV are related to compounds involving silicon,
which is a rather light element presenting covalent bonds in the pure state
and for which the amorphous state is easily reached. This makes uncertain
the validity of the use of a crystalline heat of sublimation to characterize the
binding energy at the surface after sputtering (Sigmund, 1969a). However, it
seems that the effect which is concerned here is principally a difference of
weight effect (Ho et al., 1979).
It it possible to express the sputtering yields of multicomponent targets in
terms of the sputtering yields of pure elements? According to Betz, the
sputtering yield of solid solutions seems to increase linearly with the concen-
tration from the value for the low-yield component to that of the high-yield
component. This seems to indicate that in such a case, each element keeps the
sputtering yield it had in the pure state. For nonmixable systems and for
the major part of the concentration domain, however, the sputtering yield of
the target would first stay near that of the element having the lowest yield and
would thereafter increase rapidly to the yield value of the other element.
Except in the domain of strong concentrations in the high-yield element, one
can assume that the surface bombarded by the ions is covered with atoms of
the component having a low sputtering speed. The observation shows that
after sputtering has ceased, the surface gets covered through a bulk diffusion
process by an overlayer of the high-yield component.
In the case of Ag-Pd alloys, enrichment in Pd would occur by formation
of a segregated alloy layer, rich in Pd (Slusser and Winograd, 1979). The
thickness of this layer would decrease with the bulk concentration in Pd.
Rivaud et a!. (1981) have observed the precipitation of a phase rich in In at
the vicinity of the surface of oversaturated Cu-In alloys.
Preferential sputtering is not the only contribution to the formation of
the altered layer. Liau et al. (1979) have imagined a model in which the
altered layer had a given thickness t, for a given target and a given ion beam
(t, N 700 8, for the Pt-Si system and for 250-keV X: ions). Due to atomic
mixing, the element concentrations are at every moment uniform in the
TABLE IV
PREFERENTIAL
SPUTTERING
SURFACE
ENRICHMENT'
As Pb In Ga Ag Sn Al Be Cu Au Pd Cr Fe Ni Si U Pt Mo Nb
As X
Pb X
In X
Ga
Ag x x x
Sn 0
A1 . o x X
Be
cu x x X X
All x x
Pd X
Cr X X
Fe X
Ni X
Si X 0 0
U X
Pt
Mo
Nb
x , alloy; 0 ,compound
altered layer. Owing to preferential sputtering, however, continuous

depletion in the high-sputtering-yield element takes place (Si in the quoted
example). With the aid of this model and by continously recording, it is
possible to relate the measured profile to the real one.
Ho (1978)has proposed a model of the altered layer based on preferential
sputtering in the very first atom layer at the surface (emission layer) and on
diffusion in the altered layer. This diffusion becomes strongly enhanced by
virtue of the creation of a large number of point defects (vacancies) in the
vicinity of the surface. The model is based on two mass balance equations.
One is related to the outermost surface and takes into account the diffusion
and preferential sputtering fluxes. The second is related to the rest of the
altered layer and takes into account only the diffusion flux.
This model leads to the introduction of an effective thickness parameter:
6 = Du-’ (208)
where D is the diffusivity in cm2 sec-l and u is the sputtering speed. It also
introduces a dependence in z for the composition of the altered layer.
All along the sputtering transient state, 6 varies. At equilibrium, which is
reached after sputtering material about 56 thick, the altered layer extends
over an area about 46 thick. So, for high-diffusivity alloys, the altered layer
can extend up to large depths owing to the enhanced diffusion. This would
explain the high value of the altered-layer thickness in Al-Cu alloys found
by Chu et al. (1976).
Conversely, it is possible, with the aid of Eq. (208) to estimate the diffu-
sivity during ionic bombardment. The values obtained for D in the case of
Cu-Ni systems are much higher than those obtained without sputtering.
Thus, sputter damage seems to be a very efficient process for reinforcing
diffusivity in the altered layer (Watanabe et al., 1977; Ho, 1978).
In Ho’s model the composition of the altered layer varies monotonically
from the surface composition. In the steady state, for a binary alloy, the
concentration of atoms A at a depth z is given by the relation:
CA(z) = CA(0) + CA(O)(1 - C A ) ( b A - dg)[1- ~XP(-Z)] (209)
where
Z = zv/D (210)
and
is the concentration of atoms A in the emission layer; CAis the concentration

of atoms A before the sputtering of the sample, supposed to be homogeneous;
cAand oB are given by the relations

OA = SA/Nal(A), gB = SB/Nat(B) (212)
where N,, represents the numerical atomic density for the pure element.
Equation (209) shows that in the general case CA(Z) exhibits an
exponential dependence starting from the surface z = 0; however, this
dependence disappears as one would expect for a pure target (CA= 1) or
when the reduced sputtering yields 0, and 0, are equal. One can also see from
Eq. (209) that CA(2) > CA(0) when CTA > 0,.
Chou and Shafer (1980) have extended Ho’s model to the case when
primary-ion implantation in the target, initially homogeneous, can no
longer be neglected. They obtained for the concentration at the surface
G(0) = PI/(aWat) (214)

where Na, is the numerical atomic density of the sample before sputtering and
=1 + JPI/(VhT,,) (21 5 )
where J is the primary-ion flow and PI is the fraction of these ions which
impregnates the target. In the absence of ion implantation (PI = 0), CI(0)
becomes null, o( becomes equal to 1, and Eq. (213) reverts to (211).
Hofmann and Zalar (1979) have observed that in the Ni-Cr system
sputtered by l-keV N: ions, the relative amount of implanted ions was
inversely proportional to the sputtering speed. The intervention of the
enhanced diffusion raises the problem of the role played by temperature in
the variation of the composition of the altered layer. Sputtering yields of
pure metals are nearly independent of target temperature up to the melting
points. In the case of bombardment with noble-gas ions (Kaminsky, 1965)
and for coevaporated Cu-Ni systems, there exist no significant differences in
the sputtering mechanism at liquid nitrogen and room temperature (Goto
et al., 1978).
This is not the case for higher temperatures. Various studies by AES and
SIMS have been devoted to the influence of temperature on the sputtering
of Cu-Ni systems (Nakayama et al., 1972; H. Shimizu et al., 1975; Rehn and
Wiedersich, 1980; Yabumoto et al., 1979; Shikita and Shimizu, 1980; Okutani
et al., 1980). The results obtained converge and provide evidence, as early as
200°C, of a surface (or Gibbsian) segregation in addition to preferential sput-
tering, enhanced diffusion, and ion implantation. In the absence of sputtering,
this surface segregation leads to enrichment of the surface in Cu. The effectsof
surface segregation and of preferential sputtering are antagonistic. Therefore,
280 M. CAILLER, J . P. GANACHAUD, AKD D. ROPTIN
overall, the altered layer is enriched in Ni owing to preferential sputtering, but

enrichment of the outermost surface is less important than in the subsurface
region owing to surface segregation. In other words, the concentration in Ni
does not vary in a monotonic way in the vicinity of the surface but reaches a
maximum in the subsurface region. There would even be enrichment in Cu in
the outermost atom layer, indicating that preferential sputtering is weak
for this element.
For very high energy argon ions, of the order of 1 MeV, the profile in
depth of the Cu-Ni system exhibits depletion in Cu (or enrichment in Ni) at
the surface, then enrichment in Cu in the subsurface region (Rehn et af.,
1981),interpreted in terms of a radiation-induced segregation (RIS). The RIS
produces surface enrichment of the undersize solute element (Ni) and
depletion of the oversize solute elements (Cu), the migration of the former
taking place via interstitial fluxes or vacancies. The RIS departs from
enhanced diffusion in that it no longer exists in the absence of radiation.
Kelly (1979) has reviewed thermal effects in sputtering. Two types of
effects are possible: (1) prompt thermal sputtering, where vaporization of the
target results from a local increase of the temperature, and (2) slow thermal
sputtering, which implies no temperature increase. For elemental targets and
metal alloys, prompt thermal sputtering cannot occur under the usual
conditions of temperature, except, perhaps, in some limited cases (Na,
Ga-As, Al-Mg). On the other hand, important stoichiometric changes could
be observed due to prompt thermal sputtering in some oxides (oxygen loss)
for most of the halides and for some organic molecules.
Slow thermal sputtering can take place whether the target (oxides or
halides) is bombarded by ions, photons, or electrons. In this case, the
temperature of the target is well defined. In any case, loss of halogen or
oxygen occurs by electron sputtering (or beam-induced dissociation).
Halogen or oxygen ions are neutralized by electronic interactions (formation
of a relaxed hole, Auger decay, direct ionization), and the neutralized atom is
then either vaporized or ejected (electron sputtering) so that the surface
becomes enriched with the metal. Vaporization of the metal takes place if it
is sufficiently volatile. This happens for compounds containing Cs, Rb, K,
Na, and Cd when they are bombarded at temperatures greater than 100°C.
Bombardment b y Ions Other Than Inert Zons: Chemical Reactions

The secondary-ion yield of metallic targets can be enhanced by several
orders of magnitude with respect to that of inert-gas ions when the ions used
for sputtering lead to ionic bonds with the atoms of the target. The same
occurs when the target is sputtered by noble-gas ions in an oxygen atmo-
sphere (Benninghoven, 1975). This enhancement effect is related to the

formation of oxide.
Yu and Reuter (1981a,b) have studied the emission of singly ionized
positive metal ions from binary alloys A-B under the action of an 0;ion
bombardment or an argon-ion beam, but within an oxygen atmosphere. In
both cases, when A has a stronger oxide bond than B, the presence of B in the
alloy reduces the emission of the A + ions, whereas the presence of A reinforces
the emission of the Bf ions.
This “reactive preferential sputtering,” to be distinguished from the
“nonreactive preferential sputtering” with inert-gas ions (Yu and Reuter,
1981c) leads, in the case of reactive binary alloys, to enrichment of the surface
with the element having the weakest metal-oxygen binding energy. The
Cu-Ni and Ag-Pd alloys, for which the oxidation is weak, do not present
marked differences with respect to nonreactive preferential sputtering
by argon.
It appears that in reactive preferential sputtering, the surface composition
reajusts itself to minimize the energy required for preferential sputtering of
the oxygen, whereas in a simple oxidation there is surface enrichment with
the metal which makes the strongest bond with the oxygen.
Bouwman et al. (1978)used hydrogen particles at 800 eV to clean copper
and steel samples. In the case of copper, the major part of the contaminating
species (C, 0, S) was eliminated within a 6-A depth after 30 min of bombard-
ment. In the case of steel, bombardment by Ar’ eliminates 0 and s, but is
inefficient for C. Bombardment by H eliminates S and part of C, but 0 is
unaffected, which indicates that H attacks only the outermost layer. This
surface cleaning could be a chemical process comparable with plasma
etching, eliminating the contamination species without any appreciable
destruction of the sample under the contaminating layer.
Taylor et a/.(1978) have bombarded Si, SiO, and Si02 targets by 500-eV
N: ions. Before they reach the surface of the sample, these N i ions are
neutralized by charge exchange and dissociated so that bombardment is in
fact achieved by N atoms having virtually half the kinetic energy of the N :
ions. Besides elimination of the contaminating species C and 0 at the surface,
bombardment leads to reactive ion implantation which affects the chemical
nature of the surface by silicide formation. In the case of silicon, the thickness
of the altered layer is of the order of 20 A.
Among the chemical reactions generated by ionic bombardment, the
most frequently quoted is certainly oxide reduction (Kim et al., l974,1976a,b;
Thomas, 1976; Buczek and Sastri 1980). According to Kim et al. (1974), the
fact that an oxide is reduced or not can be linked to its free energy of forma-
tion. The oxides which are reduced are those which have a low free energy
of formation.
The nature of the ions used for sputtering also has some importance. Thus
the reduction of PbO (Kim et al., 1976b) is much stronger with inert-gas
light ions than with heavy ions. So, Kr+ and Xe+ do not reduce the oxide.
Another way to avoid reduction, in some cases, is to use 0; ions.
F . Electron-Beam-Induced Effects in AES

Some disturbances can be induced by an electron beam at the surface of
the sample. Therefore, the results obtained by AES for these damaged targets
have to be interpreted with circumspection. This topic has been reviewed in
papers by Fontaine et al. (1979), Lang (1979), and Fontaine and Le Gressus
(1981). Readers interested in these topics will find much of value in these
papers.
In some cases, the effects induced by an electron beam result only from
the interaction of the sample and the electron beam. However, quite fre-
quently a third system (residual gases in the UHV chamber) can intervene.
The importance of the disturbances depends on the characteristics of each
interesting system. The sample is characterized by its physicochemical
parameters such as: (1) chemical nature for a pure compound, (2) impurity
distribution, (3) thermal conductivity, and (4) electrical conductivity. The
role played by residual gases depends on their composition and on the
partial pressures of their constituents. Finally, the current density, the
acceleration voltage, and the dose used for Auger analysis are the important
parameters for the electron beam.
The main disturbances which can be observed are the following: (1)
physical modifications in the sample (temperature increase and electrical
charging at the impact point of the electron beam), and (2) chemical changes.
However, this classification is rather arbitrary because the physical
modifications induce chemical changes and vice versa.
1. Temperature Increase
According to Roll and Losch (1980)and Fontaine and Le Gressus (1981),
when the acceleration voltage is lower than 10 keV and the diameter of the
electron beam wider than 1 pm, the temperature increase AT of the sample
can be evaluated by the surface-energy dissipation model of Vine and
Einstein (1964). Therefore,
AT = aw/Kd (216)
where W is the electron-beam power, d is the half-height width (fora Gaussian
profile of the electron density), and K is the linear thermal conductivity
coefficient of the sample material. For bulk aluminum, Eq. (216) becomes
AT N 103Wfd (217)
and W is expressed in watts and d in micrometers.
For identical values of the ratio W/d,the temperature increase AT varies
as the inverse power of K . Thus for Si, SiO,, and KCI, AT would be 1.5,
17.5, and 2.3 times higher than AT in aluminum. For a bulk metallic target AT
is small, but for the case of metal films deposited on a glass substrate, the
temperature increase can become important. For instance, for W = 30 mW,
d = 80 p m , AT is of the order of 240°C (Roll and Losch, 1980).
2. Charge Efects
The occurence of electric charge effects under the primary-electron-beam
spot has long been recognized in studies of secondary-electron emission
from insulating targets. For an electrically isolated target, the net variation
of the electric charge per second is
AQ = -eN,(l - 0) (218)
where Npis the primary-electron-beam flow, e is the electron charge, B is the
total secondary yield of the target, and 0 varies as a function of the accel-
eration voltage V,. Usually, for insulators, Q is higher than unity in some
energy range (V,, , Vp2)for the acceleration voltage. In this range, a positive
charge appears on the target, whereas outside this range, 0 is lower than
unity and the target gains a negative charge.
In the case of a conductor, the potential of the sample is kept constant by
connection to an electric source which supplies the charge -AQ, thus
insuring the neutrality of the target.
In studies of secondary-ion emission of insulators, various methods have
been used to reduce charge effects. The first method consists in working with
conditions such that 0 = 1. This is the method most frequently used in AES.
However, the samples are never perfect insulators so that part of the surface-
charge increase AQ is eliminated by electrical conduction if the insulator is
connected to an electric source. For dc, the compensation does not have
enough time to take place so pulse techniques have been used. These tech-
niques have the same result as reducing the number of primary electrons that
knock the target every second.
Another method consists in neutralizing the charge increase by either
depositing positive charges (by ion bombardment) when 0 c 1, or depositing
low-energy electrons when c > 1. In the usual case of a positively charged
surface, one observes a general shift of the Auger spectrum towards low
energies. When the acceleration voltage of the primary electrons is suffi-

ciently important, CJ can become lower than unity and the sample becomes
negatively charged.
The formation of metallic islands (alkaline or alkaline-earth islands on
MgO or Pd on mica), at low secondary yields, on an insulating sample can
partially prevent the secondary emission of the sample by an antagonistic
potential effect. The secondary-electron yield is then lower than unity. In
that case, the Auger peaks are shifted towards high energies.
3. Physicochemical Effects
The main physicochemical modifications induced in the sample by ion
bombardment are (1) desorption of the adsorbed species, (2) dissociation of
the adsorbed molecules and of the sample, and (3) stimulated adsorption. The
electron-beam stimulated desorption (ESD) is a mechanism frequently
encountered in surface studies. It provokes the elimination of species initially
adsorbed at the surface of the sample. When combined with ion mass
analysis, it represents a method for studying chemisorption. Joshi and Davis
(1977) proposed this method to obtain images by scanning-electron
stimulated desorption.
Most frequently, in quantitative AES, the production of the adsorbed
layer is achieved in a more or less uncontrolled way. It occurs as a contami-
nation of the sample to be analyzed in given UHV conditions. Its desorption
does not represent by itself a central aspect of the study but just a stage of the
analysis, which can lead to some artifacts.
Ion-beam stimulated desorption has been interpreted with the MGR
model proposed by Menzel and Gomer (1964) and by Redhead (1964);
however, some aspects in this model have been questioned (Feibelman and
Knotek, 1978; Antoniewicz, 1980). The MGR model works in two successive
steps. In the first step, it is assumed that an electron of the incident beam
collides with a valence electron of the adsorbed species. A bond in this
species is suddenly broken due to the jump of the valence electron from a
bonding to an antibonding level, and thus submitted to the action of a
repulsive potential, the species will desorb.
Gersten and Tzoar (1977) have analyzed the maximum kinetic energy of
the dissociated species and have compared it with the substrate-species
binding energy when these species are either in a molecular state (undis-
sociated)or in an atomic state (dissociated).They have shown that desorption
most frequently takes place in the undissociated state. Two exceptions have
been noted : hydrogen on tungsten, which is dissociated during the desorp-
tion, and oxygen on molybdenum, which can be either dissociated or not.
Feibelman and Knotek (1978)pointed out that the MGR model could not
explain the desorption of 0' ions in the case of transition-metal oxides in

their maximum valence states: TiO,, W 0 3 , V,O, (ionic surfaces). As a
matter of fact, before electron bombardment, oxygen is quite presumably
in a negatively charged state and would have to suffer double ionization to
become positively charged. This is not very consistent with the energy of the
primary beam when it is very low (0-100 eV). Moreover, the stimulated
desorption threshold energy of oxygen is not determined by the atomic levels
of this element but by those of the metal. To account for these two aspects,
these authors built a new model involving an interatomic Auger transition.
In this model, the initial ionization takes place in the core of the metal ion.
The reorganization leads to the creation of two holes in the valence band;
more precisely in oxygen quasi-atomic states. The concerned oxygen
atom is thus forced to desorb because it is surrounded by a repulsive
Madelung potential.
Woodruff et al. (1980) have studied the desorption of F f , Cl', and Of
from W( 100)and found that for the first two ions, part of their results were in
agreement with the model of Feibelman and Knotek, but not all.
Antoniewicz (1980) has considered what occurs following the first step
of ESD in order to answer the question of whether neutral or negative ions
are emitted. He noted that the lifetimes of excited atomic states and of
electron relaxation process are much shorter than the ejection time of the
species after it has been ionized. This poses the problem of the existence of an
antibonding state having a lifetime sufficiently long to cause the ion to gain
enough kinetic energy to desorb. For oxides this is plausible; however it
would not be the case for metals. In any case, the Madelung potential
scheme can hardly account for the desorption of neutral and negative ions.
Antoniewicz has therefore proposed the following explanation for the latter
case. Due to ionization, the atom adsorbed on the substrate becomes reduced
in size so that the new equilibrium position is closer to the substrate after
ionization than before. It starts moving to reach this new equilibrium
position, gaining kinetic energy in this way. During this displacement it
becomes neutralized by electron tunneling from the substrate or by Auger
neutralization. Recovering its initial size, it rebounds on the substrate, flies
in the opposite direction with the kinetic energy it has acquired, and is ejected
into the vacuum. The emission of negative ions implies two tunneling
processes.
Feibelman (1981) has shown that the reneutralization rate of a doubly
ionized surface species was 10-100 times weaker than that of a singly ionized
state. Thus reneutralization is much less likely in Auger-induced desorption
than in MGR desorption.
The kinetics of ESD is described by
N ( t ) = N ( 0 )exp( - N,,ot) (219)
286 M . CAILLER, J . P . GANACHAUD, A N D D . ROPTIN
where N ( t ) is the coverage (in atoms cm-') at time t, N ( 0 ) is the initial

coverage, Np is the flow of incident electrons, and (r is the desorption cross
section. The stimulated desorption cross sections range from lop2, to
cm2 (Fontaine and Le Gressus, 1981) and depend essentially on the
adsorbate-surface binding energy. For the species usually studied in AES,
(r N 10-22-10-23 cm2. This allows the use of current densities of about
10-'-10-2 A ern-,, with no appreciable desorption for the normal duration

of an Auger analysis (1000 sec). To avoid the desorption of weakly bound
species, one has to use current densities much lower than the above values.
4. Electron-Impact-StimulatedAdsorption
The existing studies concern a limited number of systems. One of the most
frequently encountered problems is that of the adsorption enhancement
of CO and 0, on semiconductors.
For silicon (Coad et ai., 1970; Joyce and Neave, 1973; Joyce, 1973;
Kirby and Dieball, 1974; Kirby and Lichtman, 1974) generally, the presence
of carbon is reinforced all over the surface of the sample, whereas for oxygen
this effect occurs only in the impact zone of the electron beam. The mech-
anism invoked to explain these results is based upon a three-step model: (1)
CO and 0, adsorption; (2) dissociation; and (3) surface diffusion of carbon,
electron-induced desorption, and slow diffusion of oxygen into the bulk. The
surface concentration in oxygen still corresponds to one monolayer, but the
Auger signal of oxygen keeps increasing owing to the diffusion of oxygen in Si.
The amplitude of the oxygen signal increases with the primary energy of the
electrons and the pressure of the gases, but decreases with increasing
temperatures of the target.
Fontaine et ai.(1979)have expressed some doubt about the validity of this
model. In fact, they have remarked that no carbon enrichment occured in
silicon whatever the residual pressure was (lo-' or to lo-'' Torr) when the
sample itself contained no carbon. On the contrary, if the target contains
some carbon, this carbon will appear by diffusion under the action of the
beam. It will diffuse towards the surface, up to the boundaries of the
bombarded zone, where it remains frozen.
Under lo-' Torr, the oxygen concentration increases very rapidly at the
surface of the sample in the impact area of the electron beam, whereas at
2 x lo-'' Torr, it remains undetected except when the target has been
previously ion etched in an oxygen atmosphere. At Torr, the oxygen
results from residual gases, whereas at 2 x lo-'' Torr, it results from the
diffusion from the bulk of the sample. Unlike silicon, the oxidation of a
certain number of other semiconductors, reinforced by the electron beam,
leads to the formation of oxides in the impact area. This is the case for
germanium (Margoninski et al., 1975), for AsGa (Ranke and Jacobi, 1975),
and for InP (Olivier et al., 1980). For AsGa and InP, a volatile oxide is formed
(As,O, and phosphorous oxide). Here carbon monoxide plays no role, and
the oxidation is interpreted as resulting from adsorption of molecular
oxygen followed by dissociation at the impact of the beam and by a reac-
tion between the substrate and the atomic oxygen. Enhancement due to
the excitation of oxygen in the gaseous phase plays a negligible role (Ranke,
1978). Oxidation has been observed in InP, even for oxygen partial pressure
as low as 5 x lo-'' Torr, the enhanced adsorption remaining localized in
the irradiated area.
For oxidation of Si, one can remark that the silicon oxides become
dissociated by electron bombardment, leaving an excess of elemental Si on the
surface (Delord et al., 1980). Nickel has also been the object of several studies.
Verhoeven and Los (1976) observed that in an oxygen atmosphere (10-
Torr), an ordered arrangement of chemisorbed oxygen can be formed, up to
0.5 monolayer, followed by a nucleation of NiO, giving a passive film of
about 2 monolayers. Oxidation speed could be increased by a factor of 5 by
the action of the electron beam.
Frederick and Hruska (1977) note that in a C O or CO, atmosphere, the
height of the carbon peak varies with time according to the law
H = A[1 - exp( - t/~)] H, + (220)
where the value of z is in both cases about 30 or 40 min and H , is the value of
H in the absence of the electron impact ( H , N 4'5). The scheme considered
here is one previously encountered : adsorption, CO and CO, dissociation,
and oxygen-stimulated desorption.
Tompkins (1977) noted that H,O interacted with the surface of nickel in
the presence of the electron beam to form, in the bombarded area, a stable
oxide film several tens of angstroms thick; H,O could be physisorbed in
small quantities, but for sufficiently long times, to allow interaction with the
incident electrons.
Lichtensteiger et al. (1980) have observed that a clean, strongly non-
reactive surface of CdS in an H,O atmosphere (even at lo-" Torr), submitted
to electron impact, could adsorb oxygen. This stimulated adsorption could
come from local activation of the surface rather than from dissociation of
previously adsorbed species. In the first step, oxygen becomes bound to the
sulfur atoms. In the second step, oxidation of CdS takes place with increasing
quantities of oxygen atoms bound to the Cd atoms.
The electron beam can also stimulate the adsorption of C, even for
pressures lower than 10- l o Torr of CO, CO,, and CH, (Joyner and Rickman,
1977). At the end of the experiment, the deposit is graphitic, and the saturation
coverages are independent of the carbon source and close to the density of the
(0001) plane of graphite (3.8 x atoms ern-,). The scheme proposed is

the same as for Ni, plus the last phase that adds graphite.
5. Beam-Induced Dissociation
This dissociation mechanism plays an important role as a step for the
desorption and adsorption of species present on a substrate, these processes
being stimulated by the electron beam. The dissociation of these species can
induce an artifactual contamination, principally in carbon, due to the pres-
ence of CO (Hooker and Grant, 1976) and of hydrocarbons in the residual
atmosphere of the chamber.
The higher the resonance energy per n electron, the more delocalized the
71 electrons, and the less the species will be damaged (J. T. Hall et al., 1977).
Moreover, the dissociation cross section varies with respect to the primary
energy E , of the electrons according to the relation (J. T. Hall et al.)
d E P )21 (WEp)WpP) (221)
where this dependence comes mainly from the factor l/Ep. This shows that
the dissociation cross section is essentially inversely proportional to the
primary energy.
The target itself can also become dissociated. Most studies have been
devoted to SiO,. For such a target, the problem of charging is also set.
Carriere and Lang (1977) suggested that secondary electrons were respon-
sible for the creation of surface charges and for bond breaking, producing
elemental silicon and oxygen.
Ichimura and Shimizu (1979) studied the topography modifications and
dissociation provoked in thin SiOz films by a scanning Auger electron
microscope. They showed that the damages were strongly connected with
the amount of energy dissipated in the sample. Bermudez and Ritz (1979)
indicated that small Si clusters of a few atoms appeared in the irradiated area.
In multiplexing, the simultaneous bombardment of a given zone of the
sample by ions and electrons provokes enhancement of the sputtering. In the
case of SiO, ,the electron bombardment alone reduces the concentration in
oxygen, whereas the simultaneous impact of electrons and ions leads to the
sputtering of the Si atoms by the ions and to the elimination of the oxygen
atoms at the surface by the electrons. Consequently, sputtering speed is
enhanced (Ahn et al., 1975).
6 . Migration and Difusion

An electron beam can also provoke migration of some atoms. This is the
case for sodium and potassium, for instance, during analysis of glasses
(Pantano et al., 1975, 1976; Dawson et al., 1978; Malm et al., 1978). This
migration can be substantially slowed by decreasing the temperature.

Diffusion induced by increasing the temperature in the case of targets
deposited on glass substrates have been observed for multilayer thin films
of Cu-Ni (Roll et al., 1979).
VIII. CONCLUSION
In this review article we have described some of the major questions

related to the development of quantitative Auger electron spectroscopy. We
outlined all the steps going from theoretical first principles to a real sample
analysis. As already mentioned, the path is far from complete, but with the
help of very diverse approaches such as the simulation method, phenom-
enological description, calibration and sensitivity factors, matrix effects, and
ion- and electron-beam effects, progress has been made. All these approaches
are essential for improving the quantification of this method of analysis;
however, there is now a need for more rigorous standard procedures, and
this will no doubt result in further progress.
Particular attention was paid here to line-shape analysis. The authors’
interest in this topic was initiated by important work done by Gaarenstroom
(1981)which could lead to the development of a new quantitative procedure.
As mentioned in Section I some topics have been entirely omitted, such
as evaluation of Auger line intensities or the angular aspects of Auger
emission. It was merely a matter of choice and of time.
ACKNOWLEDGMENTS
The authors are indebted to the following copyright owners for permission to reproduce
tables and formulas: McGraw-Hill Book Co., New York (formulas from Slater, 1960), John
Wiley and Sons, Inc., New York (formulas from Kittel, 1967), Cambridge University Press,
London (table from Condon and Shortley, 1970).
REFERENCES
Abraham-Ibrahim, S., Caroli, B., Caroli, C., and Roulet, B. (1978). Phys. Rev. B 18,6702.
Abraham-Ibrahim, S., Caroli, B., Caroli, C., and Roulet, B. (1979). J. Phys. 40,861.
Adler, S. L. (1963). Phys. Rev. 130, 1654.
Adler, S. L. (1966). Plzys. Rev. B 141, 814.
Ahn, J., Perlberg, C. R., Wilcox, D. L., Coburn, J. W., and Winters, H. F. (1975). J . Appl.
Phys. 46,458 1.
Aksela, H., Aksela, S., Vayrynen, J., and Thomas, T. D. (1980). Phys. Rev. A 22, I 1 16.
Antoniewicz, P. R. (1980). Phys. Reo. B 21,381 1.

Arthurs, A. M., and Moiseiwitch, B. L. (1958). Proc. R . SOC.London Ser. A 247,550.
Asaad, W. N. (1 963a). Nucl. Phys. 44,399.
A d , W . N. (1963b). Nwl. Phys. 44,415.
A d , W . N. (1965a). Nucl. Phys. 66,494.
Asaad, W. N. (1965b). Nucl. Phys. 63,337.
Asaad, W. N., and Burhop, E. H. S . (1958). Proc. Phys. Soc. London 71, 369.
Ashley, C. A., and Doniach, S . (1975). Phys. Reo. B 11, 1279.
Ashley, J. C., and Ritchie, R. H. (1974). Phys. Status Solidi B 62,253.
Ashley, J. C., T u g , C. J., and Ritchie, R. H. (1979). Surf: Sci.81,409.
Bambynek, W., Crasemann, B., Fink, R. W., Freund, H.U.,Mark, H., Swift, C. D., Price,
R. E., and Rao, D. V. (1972). Rev. Mod. Phys. 44,716.
Battye, F. L., Jenkin, J. G., Liesegang, J., and Leckey, R. C. 0.(1974). Phys. Rev. 9,2887.
Battye, F. L., Leckey, R.C. G., Liesegang, J., and Jenkin, J. G. (1976a). Thin Solid Films 36,231.
Battye, F. L., Liesegang, J., M e y , R.C. G., and Jenkin, J. G. (1976b). Phys. Rev. B 13,2646.
Baudoing, R. (1971). Solid State Commun. 9, 1231.
Bauer, E. (1970). J. Vac. Sci. Technol. 7 , 3 .
Bauer, E. (1972). 14th Winter School, A.V.C.P. Verbier.
Beni, G., Lee, P. A., and Platzman, P. M. (1976). Phys. Rev. B 13, 5170.
Benninghoven, A. (1971). Z . Phys. 230,403.
Benninghoven, A. (1975). Surf. Sci. 53, 596.
Bermudez, V. M., and Ritz, V. H. (1979). Surf. Sci. 82, L601.
Bernett, M. K., Murday, J. S.,and Turner, N. A. (1977). J. Electron Spectrosc. Relat. Phenom.
1%375.
Bethe, H. A. (1930). Ann. Phys. 5,325.
Betz, G. (1980). Surf. Sci. 92,283.
Betz, G . , Marton, J., and Braun, P. (1980a). Nucl. Instrum. Methods 168, 541.
Betz, G., Marton, J., and Braun, P. (1980b). Nucl. Instrum. Methods 170, 347.
Bindell, J. R., Colby, J. W., Wonsidler, D. R., Poate, J. M., Conley, D. K., and Tisone, T. C.
(1976). Thin Solid Films 37,441.
Bindi, R., Lanteri, H., and Rostaing, P. (1980). J. Phys. D. 13,267.
Bishop, H. E. (1967). Br. J. Appf. Phys. 18,703.
Bishop, H. E. (1976). NBS Spec. Publ. (U.S.) 460,5.
Bonham, R.A., and Strand, T.G. (1963). J . Chem. Phys. 29,2200.
Bouquet, S., Pons, F., Le Hericy, J., and Langeron, J. P. (1977). J . Less-Common Met. 56,243.
Bouwman, R., Toneman, L. H., and Holscher, A. A. (1973a). Ned. Tijhcher. Vacuumtech. 11,
37.
Bouwman, R., Toneman, L. H., and Holscher, A. A. (1973b). Vacuum 23, 163.
Bouwman, R., Toneman, L. H.,and Holscher, A. A. (1973~).Surf. Sci.35,s.
Bouwman, R., Toneman, L. H., and Holscher, A. A. (1976a). Surf. Sci. 57,441.
Bouwman, R., Toneman, L. H., and Holscher, A. A. (1976b). SurJ Sci. 59,72.
Bouwman, R.,Van Mechelen, J. B., and Holscher, A. A. (1978). J . Vac. Sci. Technol. 15,92.
Briant, C. L., and Banerji, S.K. (1978). Int. Merals Reu. 4, 164.
Briant, C. L., and Banerji, S . K. (1979a). Merall. Trans. A 10,123.
Briant, C. L., and Banerji, S . K. (1979b). Metall. Trans. A 10, 1151.
Briant, C. L., and Banerji, S . K.(1979~).Metall. Trans. A 1, 1729.
Buczek, D., and Sastri, S.(1980). J. Vnc. Sci. Technol. 17, 201.
Bui Minh Duc (1981a). These d’Etat, Universitt Claude Bernard Lyon, Lyon, France.
Bui Minh Duc (1981b). “Journbes d’Etudes sur la Spectrombtrie Auger.” Les Arcs, France.
Bui Minh Duc (1981~).Vide, Couches Minces 36,151.
Burhop, E. H. S. (1940). Proc. Cambridge Philos. Soc. 36,43.

Burhop, E.H. S. (1952). “The Auger Effect and Other Radiationless Transitions.” Cambridge
University Press, London.
Cailler, M., Ganachaud, J. P., and Bourdin, J. P. (1981). Thin Solid Films 75, 181.
Callaway, J. (1959). Phys. Rev. 116, 1368.
Carriere, B., and Lang, B. (1977). Surj Sci. 64, 209.
Chang, C. G. (1971). Surf. Sci. 25, 1971.
Chang, C. G. (1974). In “Characterization of Solid Surfaces” (P. F. Kane and G. B. Larrabee,
eds.), p. 509. Plenum, New York.
Chattarji, D. (1976). “The Theory of Auger Transitions.” Academic Press, New York.
Chelikowski, J. R., and Schliiter, M. (1977). Phys. Rev. B 15,4020.
Chen, M. H., Crasemann, B., and Mark, H. (1980a). Phys. Rev. A 21,436.
Chen, M. H., Crasemann, B., and Mark, H. (1980b). Phys. Rev. A 21,442.
Chen, M. H., Crasemann, B., and Mark, H. (1980~).Phys. Rev. A 21,449.
Chou, N. J., and Shafer, M. W. (1980). Surf. Sci.92, 601.
Chu, W. K., Howard, J. K., and Laver, R. F. (1976). J. Appl. Phys. 47,4500.
Chung, M. S., and Everhart, T. E.(1977). Phys. Reo. B 15,4699.
Cini, M. (1976). Solid State Commun. 20,605.
Cini, M. (1977). Solid State Commun. 24, 681.
Chi, M. (1978a). Phys. Rev. B 17,2486.
Cini, M. (1978b). Phys. Rev. B 17, 2788.
Citrin, P. M., Rowe, J. E., and Christman, S. B. (1976). Phys. Rev. B 14,2642.
Coad, J. P., Bishop, H. E., and Riviere, J. C. (1970). Surf:Sci. 21,253.
Cohen, P. J., Einstein, T. L., Elam, W. T., Fukuda, Y., and Park, R. L. (1978). Appi. Surf.Sci.
1,538.
Collins, R. A., and Gould, R. D. (1975). Thin Solid Films 30, L1.
Condon, E. U., and Shortley, G. H. (1970). “The Theory of Atomic Spectra.” Cambridge
University Press, London.
Cook, Jr., C. F., Helms, C. R., and Fox, D. C. (1980). J . Vuc. Sci. Technol. 17,44.
Crotty, J. M., and Larkins, F. P. (1976). J. Phys. B 9,881.
Crotty, J. M., and Larkins, F. P. (1977). Phys. Lett. 60A, 193.
Da Cunha Belo, M., Rondot, B., Pons, F., Le Hericy, J., and Langeron, J. P. (1977). J. Electro-
chem. SOC.124, 1317.
Davis, L. E., MacDonald, N. C., Palmberg, P. W., Riachi, G. E., Weber, R. E. (1976). “Hand-
book of Auger Electron Spectroscopy” (2nd ed.). Physical Electronics Ind., Inc., Eden
Prairie, Minnesota.
Dawson, P. T., Heavens, 0. S., and Pollard, A. M. (1978). J. Phys. C 11,2183.
Defrance, J. E. (1976). Rev. Quest. Sci. 147,339.
Delord, J. F., Schrott, A. G., and Fain, Jr., J. C. (1980). J. Vac. Sci. Technol. 17, 517.
Den Boer, M. L., Cohen, P. I., and Park, R. L. (1978). J . Vuc. Sci. Techno[. 15,502.
Den Boer, M. L., Einstein, T. L., Elam, W. T., Park, R. L., Roelofs, L. D., and Laramore, G. E.
(1980a). Phys. Rev. Lett. 44,496.
Den Boer, M. L., Einstein, T. L., Elam, W. T., Park, R. L., Roelofs, L. D., and Laramore, G. E.
(1980b). J. Vac. Sci. Technol. 17,59.
Drawin, H. W. (1961). Z. Phys. 164,513.
Drawin, H. W. (1963). Z. Phys. 172,429.
Dufour, G., Mariot, J. M., Nilson-Jatko, P. E., and Karnatak, R. C. (1976). Physica Scripta
13,370.
Duke, C. B., and Tucker, Jr., C. W. (1969). Surf: Sci. 15, 231.
Duraud, J. P., and Le Gressus, C. (1981). Vide, Couches Minces 36, 1.
Elam, W. T., Cohen, P. l., Einstein, T. L., Fukuda, Y . , and Park, R. L. (1978). J . Vac. Sci.
Technol. 15, 655.
Elam, W. T., Cohen, P. I . , Roelofs, L., and Park, R. L. (1979). Appl. Surf: Sci. 2, 636.
Emerson, L. C., Birkhoff, R. D., Andersen, V. E., and Ritchie, R. H. (1973). Phys. Rev. B 7,
1798.
Engelsberg, S. (1961). Ph.w. Rev. 123, 1130.
Evans, S., Pritchard, R. G., and Thomas, J. M. (1977). J . Pliys. C 10, 2483.
Falk, D. S. (1960). Phys. Rev. 118, 105.
Feibelman, P. J. (1973). Surf. Sci. 36, 558.
Feibelman, P. J. (1974). Surf. Sci.46, 558.
Feibelman, P. J. (1981). Surf: Sci. 102, L51.
Feibelman, P. J., and McGuire, E. J. (1977). Phys. Rev. B 17, 690.
Feibelman, P. J., and Knotek, M. L. (1978). Phys. Rev. B 18,6531.
Feibelman, P. J., McGuire, E. J., and Pandey, K. C. (1977). P h y . ~Rev.
. B 15, 2202.
Fisher, B., Pollak, R. A., Distefano, T. H., and Grobman, W. D. (1977). Phys. Rev.B 15,3193.
Fitting, H. J., Glaefeke, H., and Wild, W. (1978). Surf. Sci.75, 267.
Fontaine, J. M., and Le Gressus, C. (1981). Vide, Couches Minces 36, 11.
Fontaine, J. M., Duraud, J. P., and Le Gressus, C. (1979). Sur$ Interface Anal. I, 196.
Frederick, P. J., and Hruska, S. J. (1977). Surf. Sci. 62,707.
Friedel, J. (1952). Phil. Mag. 43,2392.
Fry, J. L. (1969). Phys. Rev. 179, 892.
Fukuda, Y., Elam, W. T., and Park, R. L. (1978). Appl. Surf. Sci. 1, 278.
Gaarenstroom, S. W. (1981). Appl. Surf. Sci. 7, 7.
Gallon, T. E. (1969). Surf. Sci. 17,486.
Gallon, T. E. (1972). J. Phys. D 5, 822.
Ganachaud, J. P. (1977). These d'Etat, Universite Nantes, Nantes, France.
Ganachaud, J. P., and Cailler, M. (1979). Surf. Sci. 83, 498.
Gerlach, R. L. (1971). Surf. Sci.28, 648.
Gerlach, R. L., Houston, J. E., and Park, R. L. (1970). Appl. Phys. Lett. 16, 179.
Gersten. J. I. (1970). Phys. Rev. B 2, 3457.
Gersten, J. I., and Tzoar, N. (1973). Phys. Rev. B 8, 5671.
Gersten, J. I., and Tzoar, N. (1977). Phys. Rev. B 16,945.
Goto, K., Koshikawa, T., Ishikawa, K., and Shimizu, R. (1978). J. Vuc. Sci. Technol. 15, 1695.
Grant, J. T., Haas, T. W., and Houston, J. E. (1974). J . Appl. Phys. Suppl. 2 2, 81 1.
Grant, J. T., and Hooker, M. P. (1976). J. Electron Spectros. Relat. Phenom. 9, 93.
Gryzinski, M. (1965a). Phys. Rev. 138, 305.
Gryzinski, M. (1965b). Phys. Rev. 138,322.
Gryzinski, M. (1965~).Phys. Rev. 138,336.
Guglielmacci, J. M., and Gillet, M .(1980a). Surf. Sci. 94,407.
Guglielmacci, J. M., and Gillet, M. (1980b). Surf. Sci.94, 424.
Haff, P. K. (1977). Appl. Phys. Lett. 31, 259.
Haff, P. K., and Switkowski, Z. E. (1976). Appl. Phys. Lett. 29,549.
Hall, P. M., and Morabito, J. M. (1979). Surf: Sci.83, 391.
Hall, P. M., Morabito, J. M., and Conley, D. K. (1977). Surf: Sci.62, 1.
Hall, J. T., Hansma, P. K., and Parikh, M. (1977). Surf: Scz. 65,552.
Harris, L. A. (1968). J. Appl. Phys. 39, 1419.
Harrison, Jr., D. E., Kelly, P. W., Garrison, B. J., and Winograd, N. (1978). Surf. Sci. 76,311.
Hattori, T. (1977). Thin Solid Films 46,47.
Hedin, L., and Lundquist, S. (1969). Solid State Phys. 23, 1.
Henrich, V. E., and Fan, J. C. C. (1974). Surf: Sci. 42, 139.
Hink, W., and Ziegler, A. (1969). 2. Phys. 226,22.

Ho, P. S. (1978). Surj. Sci.72, 253.
Ho. P. S., Lewis, J. E., Wildman, H. S., and Howard, J. K. (1976). Surf: Sci. 57, 393.
Ho, P. S., Lewis, J. E., and Howard, J. K. (1977). J . Vuc. Sci. Teclmol. 14, 322.
Ho, P. S., Lewis, J. E., and Chu, W. K. (1979). Surf: Sci. 85, 19.
Hofmann, S. (1976). Appl. Phys. 9, 59.
Hofmann, S. (1977). Appl. Phys. 13,205.
Hofmann, S . (1979). Vide, Couches Minces March, 259.
Hofmann, S., Erlewein, J., and Zalar, A. (1977). Thin Solid Films 43, 275.
Hofmann, S., and Zalar, A. (1979). Thin Solid Films 60,201.
Holloway, P. H. (1977). Surj: Sci. 66,479.
Holloway, P. H. (1978). In “Scanning Electron Microscopy” Vol. 1, p. 241. SEM, OHare,
Illinois.
Holloway, P. H. (1980). Adv. Electron. Electron Phys. 54,241.
Holloway, P. H., and Holloway, D. M.(1977). Surj: Sci. 66,635.
Hoogewijs, R., Fiermans, L., and Vennik, J. (1977). Surf. Sci. 69, 273.
Hooker, M. P., and Grant, J. T. (1976). Surf: Sci. 55, 741.
Hubbard, J. (1955). Pruc. Phys. Soc. London A68,976.
Ichimura, S., and Shimizu, R. (1979). J . Appl. Phys. 50,6020.
Ichimura, S . , Aratama, M., and Shimizu, R. (1980). J . Appl. Phys. 51,2853.
Ichimura, S . , Shimizu, R., and Ikuta, T. (1981). In “Scanning Electron Microscopy.” SEM Inc.,
OHare, Illinois.
Ing, B. S., and Pendry, J. B. (1975). J . Phys. C 8, 1087.
Inokuti, M. (1971). Rev. Mod. Phys. 43, 297.
Inokuti, M., Itikawa, Y., and Turner, J. E. (1978). Reu. Mod. Phys. 50,23.
Jablonski, A. (1978). Surf. Sci.74,621.
Jackson, D. C., Gallon, T. E., and Chambers, A. (1973). Surf: Sci.36, 381.
Jardin, C. (1981). These d’Etat, Universite Claude Bernard, Lyon, Lyon, France.
Jennison, D. R. (1978a). Phys. Reu. Leu. 40, 807.
Jennison, D. R. (1978b). Phys. Reu. 18,6865.
Jennison, D. R. (1978~).Phys. Reo. 18,6996.
Jennison, D. R., Madden, H. H., and Zehner, D. M. (1980). Phys. Rev. B 21,430.
Joshi, A., and Davis, L. E. (1977). J. Vuc. Sci. Technol. 14, 1310.
Joyce, B. A. (1973). Surf: Sci. 35, I .
Joyce, B. A., and Neave, J. H. (1973). Surj. Sci.34, 401.
Joyner, R. W., and Rickman, J. (1977). Surf. Set. 67,351.
Kaminsky, M. (1965). “Atomic and Ionic Impact Phenoma on Metal Surfaces,” p. 163. Springer-
Verlag, New York.
Kanter, H. (1970). Phys. Rev. B 1, 522.
Kelly, R. (1978). Nucl. Instr. Meth. 149,553.
Kelly, R. (1979). Surf: Sci. 90,280.
Kim, K. S., Baittinger, W. E., Amy, J. W., and Winograd, N. (1974). J. Electron. Spectrosc.
Relat. Phenom. 5, 351.
Kim, K. S., Gaarestroom, S. W., and Winograd, N. (1976a). Phys. Rev. B 14,2281.
Kim, K. S., Baittinger, W. E., and Winograd, N. (1976b). Surf: Sci. 55,285.
Kirby, R. E., and Dieball, J. W. (1974). Surf: Sci.41, 467.
Kirby, R. E., and Lichtman, D. (1974). Surf. Sci. 41,447.
Kirschner, J., and Staib, P. (1973). Ph-vs. Lett. A 42, 335.
Kirschner, J., and Staib, P. (1974). Appl. Phys. 3,421.
Kirschner, J., and Staib, P. (1975). Appl. Phys. 6, 99.
Kittel, C. (1953). “Introduction to Solid State Physics” (2nd Ed.), Wiley, New York.
Kittel, C. (1967). “Quantum Theory of Solids” (4th Ed.). Wiley, New York.
Kleinman, L. (1971). Phys. Rev. B 3, 2982.
Kolbenstvedt, H. (1967). J. Appl. Phys. 38,4785.
Koval, I. P., Krynko, Yu. N., Melnik, P. V., Nakhodkin, N. G., and Shaldervan, A. I. (1978).
Surf.Sci. 11,40.
Kowalczyk, S. P., Pollak, R. A., McFeely, F. R., Ley, L., and Shirley, D. A. (1973a). Phys.
Rev. B 8,2387.
Kowalczyk, S. P., Pollak, R. A., McFeely, R. R., Ley, L., and Shirley, D. A. (1973b). Phys.
Rev. B 8, 3583.
Kowalczyk, S. P., Pollak, R. A., McFeely, F. R., Ley, L., and Shirley, D. A. (1974). Phys.
Rev. B 9,38 1.
Krefting, E. F., and Reimer, L. (1973). In “Quantitative Analysis with Electron Microprobe
and Secondary Ion Mass Spectrometry” (Preuss, ed.), p. 114. Zentral Bibliotekder K. F. A.,
Julich.
Lang, B. (1979). Vide, Couches Minces March, 61.
Laramore, G. E. (1980a). Phys. Rev. Lett. 44,496.
Laramore, G . E. (1980b). J. Vac. Sci. Technol. 17,59.
Laramore, G . E.,Einstein, T. L., Roelofs, L.D., and Park, R. L. (1980). Phys. Rev. B21,2108.
Larkins, F. P. (1971a). J. Phys. B 4 , 1.
Larkins, F. P. (1971b). J . Phys. B4, 14.
Larkins, F. P. (1973a). J. Phys. B6, 1556.
Larkins, F. P. (1973b). J. Phgs. 86,2450.
Larkins, F. P. (1974). J. Phys. B I, 37.
Larkins, F. P. (1975). I n “Atomic Inner Shell Processes: Ionization and Transition Probabili-
ties” (B. Crasemann, ed.), Vol. I, p. 377. Academic Press, New York.
Larkin, F. P. (1976a). J . Phys. B9, 37.
Larkins, F. P. (1976b). J . Phys. B 9,47.
Larkins’ F. P. (1977a). J . Phys. B 10,2453.
Larkins, F. P. (1977b). J . Phys. B 10,2461.
Larkins, F. P. (1977~).J. Electron. Specirosc. Relat. Phenom. 12, 11 1.
Larkins, F. P. (1978a). Chem. Phys. Lett. 55,335.
Larkins, F. P. (1978b). J. Phys. C 11, 1965.
Larkins, F. P., and Lubenfeld, A. (1977). Phys. Lett. A 60, 193.
Lasser, R., and Fuggle, J. C. (1980). Phys. Rev. 22,2637.
Laty, P., and Degreve, F.(1979). Vide, Couches Minces, March,
Lea, C., and Seah, M. P. (1981). Thin Solid Films 75, 67.
Lee, P. A., and Beni, G. (1977). Phys. Rev. B 15,2862.
Le Gressus, C. (1978). Vide 191,58.
Le Gressus, C., Massignon, D., and Sopizet, R . (1975). C . R . Hehd. Seances Acad. Sci. 280,439.
Le Hericy, J., and Langeron, J. P. (1981). Vide, Couches Minces 36, 37.
Ley, L., Kowalczyk, S. P., McFeely, F R.,Pollak, R. A., and Shirley, D. A. (1973). Phys.
Rev. B 8, 2387.
Lezuo, K. L. (1972). Z . Physik 254, 189.
Liau, 2.L., Mayer, J. W., Brown, W. L., and Poate, J. M. (1978). J. Appl. Phys. 49, 5295.
Liau, 2.L., Tsaur, B. Y., and Mayer, J. W . (1979). J . Yac. Sci. Technol. 16, 121.
Lichtensteiger, M., Webb, C., and Lagowski, J. (1980). Surf. Sci. 97, L375.
Lindhard, J. (1954). K. Dan. Vidensk. Selsk. Mat. Fys. Medd. 28, 3.
Losch, W. (1979). J. Vuc. Sci. Technol. 16, 865.
Lotz, W. (1967). Z. Phys. 206,205.
Lotz, W. (1968). 2.Phys. 216,241.

Lotz, W. (1969). Z . Phys. 220,466.
Lotz, W. (1970). Z . Phys. 232, 101.
Lundqvist, B. I. (1967). Phys. Kond. Materie 6, 193, 206.
Lundqvist, B. I. (1968). Phys. Kond. Materie 7, 117.
Lundqvist, B. I. (1969a). Phys. Status Solidi 32,273.
Lundqvist, B. I. (1969b). Phys. Kond. Materie 9, 236, 48.
Madden, H. H., Zehner, D. M., and Noonan, J. R. (1978). Phys. Rev. B 17,3074.
Malm, D. L., Vasile, M. J., Padden, F. J., Dove, D. B., and Pantano, Jr., C. G . (1978). J . Vac.
Sci. Technol. 15,35.
Margoninski, Y., Segal, D., and Kirby, R. E. (1975). Surf. Sci. 53,488.
Mariot, J. M. (1978). These d’Etat, Universitt P. et M. Curie, Paris, France.
Mariot, J. M., and Dufour, G. (1977a). Chem. Phys. Lett. 50, 218.
Mariot, J. M., and Dufour, G. (1977b). J. Phys. C 10, L213.
Martin, P. C., and Schwinger, J. (1959). Phys. Rev. 115, 1342.
Mathieu, M. J., and Landolt, D. (1975). Surf. Sci. 53,228.
Mathieu, M. J., and Landolt, D. (1979). Appl. Sure Sci. 3, 348.
Mathieu, M. J., McGlure, D. E., and Landolt, D. (1976). Thin Solid Films 38,281.
Matsudaira, T., and Onchi, M. (1978). Surf. Sci. 72, 53.
Matsukawa, T., Murata, K., and Shimizu, R. (1973). Phys. Status Solidi 55,371.
McGuire, E. J. (1970). Phys. Rev. A 2, 273.
McGuire, E. J. (1971a). Phys. Rev. A 3, 587.
McGuire, E. J. (1971b). Phys. Rev. A 3, 1802.
McGuire, E.J. (1977). Phys. Rev. A 16, 2365.
McGuire, G. E. (1978). Surf. Sci.76, 130.
McMahon, Jr., C. J., Briant, C. L., and Banerji, S. K. (1977). Fracture 1, 363.
Melles, J. J., Davis, L. E., and Levenson (1974). Phys. Rev. A 9,4618.
Menzel, D., and Comer, R. (1964). J. Chem. Phys. 41, 3311.
Mermin, N. D. (1970). Phys. Rev. B 1,2362.
Meyer, F., and Vrakking, J. I. (1972). Surj. Sci. 33, 27.
Minnhagen, P. (1974). J. Phys. C 7,3013.
Minnhagen, P. (1975). J . Phys. CS, 1535.
Mulford, R. A,, Briant, C. L., and Rowe, R. G. (1980). In “Scanning Electron Microscopy,”
Vol. 1, p. 487. SEM Inc., O’Hare, Illinois.
Miiller, K. (1975). In “Springer Tracts in Modem Physics,” Vol. 77, p. 97. Springer-Verlag,
Berlin.
Nagel, D. R., and Witten, Jr., T. A. (1975). Phys. Rev. B 11, 1623.
Nakayama, K., Ono, M., and Shimizu, H. (1972). J . Vac. Sci. Technol. 9, 749.
Nishimori, K., Tokutaka, H., and Takashima, K. (1980). Surf. Sci. 100,665.
Norman, D., and Woodruff, D. P. (1977). Solid State Commun. 22, 711.
Norman, D., and Woodruff, D. P. (1978). Surf. Sci. 75, 179.
Nozieres, P., and De Dominicis, C. T. (1969). Phys. Rev. 178, 1097.
Okajima, Y. (1980). J . Appl. Phys. 51, 715.
Okutani, T., Shikata, M., and Shimizu, R. (1980). Surj: Sci.99, L410.
Olivier, J., Faulconnier, P., and Poirier, R. (1980). J. Appl. Phys. 51,4990.
Oxley, D. P., and Thurstans, R. E. (1975a). Thin Solid Films 26, 157.
Oxley, D. P.,and Thurstans, R. E. (1975b). Thin Solid Films 30, L5.
Palmberg, P. W. (1973). Anal. Chem. 45, A549.
Palmberg, P. W., and Rhodin, T. N. (1968). J. Appl. Phy.~.39, 2425.

Pantano, Jr., C. G., Dove, D. B., and Onoda, Jr., G. Y. (1975). J. Non-Cryst. Solids 19,41.
Pantano, Jr., C. G., Dove, D. B., and Onoda, Jr., G. Y. (1976). J . Vuc. Sci. Technol. 13,44.
Papagno, L., Scarmozzino, R., and Simoni, F. (1980). Thin Solid Films 67, 157.
Park, R. L. (1979). Surf: Sci. 86, 504.
Park, R. L. (1980a). Appl. Sci. 4, 250.
Park, R. L. (1980b). In “Methods of Experimental Surface Physics” (R. L. Park, ed.). Academic
Press, New York.
Park, R. L., Houston, J. E., and Schreiner, D. G. (1970). Rev. Sci. Instrum. 41, 1810.
Payling, R., and Ramsey, J . A. (1977). In “Proceedings of the Seventh International Vacuum
Congress and Third International Conference on Solid Surfaces,” p. 2451.
Peisner, J., Aszodi, G., Nemeth-Sallay, M., and Forgacs, G. (1976). Thin Solid Films 36,251.
Pendry, J. B. (1974). “Low Energy Electron Diffraction,” p. 73. Academic Press, London.
Penn, D. R. (1975). Phys. Rev. B 13, 5248.
Penn, D. R. (1976a). J. Vuc. Sci. Technol. 13, 221.
Penn, D. R. (1976b). J . Electron. Spectrosc. Relat. Phenom. 9, 29.
Pessa, V. M., and Newell, W. R. (1971). Phys. Sci. 3, 165.
Philibert, J., and Texier, R. (1968). In “Quantitative Electron Probe Microanalysis” (K. F. J.
Heinrich, ed.), p. 13. NBS Washington Special Publication 298, Washington, D.C.
Pierce, D. T., and Siegmann, H. C. (1974). Phys. Rev. B9,4935.
Pons, F., Le Hericy, J., and Langeron, J . P. (1977a). Surf. Sci.69, 547.
Pons, F., Le Hericy, J., and Langeron, J. P. (1977b). Surf: Sci. 69, 565.
Pons, F., Le Hericy, J., and Langeron, J. P. (1977~).J. Microsc. Spectrosc. Electron. 2,49.
Powell, C. J. (1974). Surj: Sci. 44, 29.
Powell, C. J. (1976). Rev. Mod. Phys. 48, 33.
Powell, C. J. (1978). In “Quantitative Surface Analysis of Materials” (N. S. McIntyre, ed.),
p. 5. ASTM, Philadelphia.
Powell, C. J., Stein, R. J., Needham, Jr., P. B., and Driscoll, T. J. (1977). Phys. Rev. B 16,1370.
Powell, C. J., Erickson, N. E., and Madey, T. E. (1981). In press.
Price, R. E., and Rao, P. V. (1972). Rev. Mod. Phys. 44, 716.
Quinn, J. J., and Ferrel, R. A. (1958). Phys. Rev. 112,812.
Quinn, J. J. (1962). Phys. Rev. 126, 1453.
Ramaker, D. E. (1980). Phys. Rev. B 21,4608.
Ramaker, D. E., and Murdey, J . J. (1979). J . Vuc. Sci. Technol. 16, 510.
Ramaker, D. E., Murdey, J. J., andTurner, N . H. (1979a). J. Electron Spectrosc. Relat. Phenom.
17, 45.
Ramaker, D. E., Murdey, J. J., Turner, N. H., Moore, G., Legally, M. G., and Houston, J.
(1979b). Phys. Rev. B 19,5375.
Ranke, W. (1978). J. Phys. D 11, L87.
Ranke, W., and Jacobi, K. (1975). Surf. Sci. 47, 525.
Rao, P. V., and Crasemann, B. (1965). Phys. Rev. A 139, 1926.
Rasolt, M., and Davis, H. L. (1980). Phys. Rev. B 21, 1445.
Redhead, P. A. (1964). Can. J . Phys. 42,886.
Rehn, L. E., and Wiedersich, H. (1980). Thin Solid Films 73, 139.
Rehn, L. E., Wagner, M., and Wiedersich, H. (1981). Scr. Metall. 15, 683.
Reimer, L. (1968). Optik 27, 86.
Reuter, W. (1972). In “Proceedings of the Sixth International Conference on X-Ray Optics
and Microanalysis” (G. Shinoda, K. Kobra, and T. Ichinokawa, eds.), p. 121. University
of Tokyo Press, Tokyo.
Ritchie, R. H., and Howie, A. (1977). Phil. Mug. 36,463.
Ritchie, R. H., Garber, E. W., Nakai, M. Y.. and Birkhoff, R. D. (1969). Adv. Radiut. Biol. 3, 1.
Ritchie, R. H., Tung, C. J., Anderson, V. E., and Ashley, J. C. (1975). Radiat. Res. 64, 181.
Rivdud, L., Ward, I. D., El Toukhy, A. W., and Greene, J. E. (1981). Sur$ Sci. 102,610.
Roll, K., and Hammer, C. (1979). Thin Solid Films 57,209.
Roll, K., and Losch, W. (1980). In “Proceedings of the Eighth International Congress on
Vacuums” (F. Ahels and M. Crosit, eds.), Vol. 1, p. 251.
Roll, K., Losch, W., and Achette, C. (1979). J. Appl. Phys. 50,4422.
Rosler, M . , and Brduer, W. (1981a). Phys. Status Solidi b101, 161.
Rosler, M.,and Brauer, W. (1981b). Phys. Status Solidi b104, 575.
Rowe, R. G., Briant, C. L., and Bacon, F. (1978). General Electric Report No. 78.
Rudge, M. R. H., and Schwartz, S. B. (1966). Proc. Phys. Soc. London 88,563.
Sawatzky, G. A. (1977). Phys. Rev. Lett. 39, 504.
Sawdtzky, G. A,, and Lenselink, A. (1980). Phys. Rev. B 21, 1790.
Schou, J. (1980). Phys. Rev. B 2 2 , 2141.
Seah, M. P. (1972). Surf Sci. 32, 703.
Seah, M. P. (1973). Surf.Sci. 40, 595.
Seah, M. P. (1979a). Surf; Interface Anal. 1, 86.
Seah, M. P. (1979b). Surf. Interface Anal. 1,91.
Seah, M. P. (1981). Thin Solid Films 81,279.
Seah, M. P., and Dench, W. A. (1979). S w f . Interface Anal. 1,86.
Seah, M. P., and Lea, C. (1981). Thin Solid Films 81, 257.
Seah, M. P., Sanz, J. M., and Hofmann, S. (1981). Thin Solid Films 81,239.
Sevier, K. D. (1972). “Low Energy Electron Spectrometry.” Wiley, New York.
Shelton, J. C. (1974). J. Electron Spectrosc. Relat. Phenom. 3,417.
Shikata, M., and Shimizu, R. (1980). Sure Sci. 97,L363.
Shimizu, H., Ono, M., and Nakayama, K. (1975). J. Appl. Phys. 46,460.
Shimizu, R., and Everhart, T. E. (1978). Appl. Phys. Lett. 33,784.
Shimizu, R.,and Ichimura, S. (I98 I). In “Quantitative Analysis by Auger Electron Spectros-
copy.” Toyota Foundation Research Report No. I , 006.
Shimizu, R., and Saeki, N. (1977). Surf. Sci.62,752.
Shimizu, R., Ikuta, T., and Murata, K. (1972). J . Appl. Phys. 43, 4233.
Shimizu, R., Kataoka, Y., Matsukawa, T., Ikuta, T., Murata, K., and Hashimoto, H. (1975).
J . Phys. D 8, 820.
Shirley, D. A. (1973). Phys. Rev. A 7 , 1520.
Sickafus, E. N. (1977a). Phys. Rev. B 16, 1436.
Sickafus, E. N . (1977b). Phys. Rev. B 16, 1448.
Sickafus, E. N. (1980). Surf. Sci. 100, 529.
Sickafus, E. N., and Holloway, D. M. (1975). Surf: Sci. 57, 131.
Sickafus, E. N., and Kukla, C. (1979). Phys. Rev. B 19, 4056.
Sickafus, E. N., and Winder, D. R. (1981). J . Vtrc. Sci.Techno/. 18, 598.
Sigmund, P. (1968). Cun. J . Phys. 46, 731.
Sigmund, P. (1969a). Phys. Rev. 184, 383.
Sigmund, P. (1969b). Phys. Rev. 187,768.
Singwi, K. S., Tosi, M. P., Land, R. H., and Sjolander, A. (1968). Phys. Rev. 176, 589.
Singwi, K. S., Sjolander, A., Tosi, M. P., and Land, R. H. (1970). Phys. Rev. B 1, 1044.
Slater, J. C. (1960). In “Quantum Theory of Atomic Structure” (L. I. Schiff, ed.), p. 244.
McGraw-Hill, New York.
Sliisser, G. J., and Winograd, N. (1979). Su<$ Sci. 84, 21 1.
Smith, T. (1976). Surf. Sci.55,601.
Smkka, L. (1970). Czech. J . Phys. B20,291.
Staib, P., and Kirschner, J. (1974). Appl. Phys. 3,421.

Stein, R. J. (1976). Surf. Sci. 60,436.
Steiner, P., Reiter, F. J., Hochst, H., and Hiifner, S. (1978). Phys. Status Solidi 90,45.
Stern, R. M., and Sinharoy, S. (1972). Surf: Sci.33, 131.
Szajman, S., Liesegang, J., Leckey, R. C. G., and Jenkin, J. G . (1978). Phys. Rev. B 18,4010.
Tagle, J. A., Martinez Saez, V., Rojo, J. M., and Salmeron, M. (1978). Surf.Sci. 79, 77.
Tarng, M. L., and Fisher, D. G. (1978). J. Vuc. Sci. Technol. 15,50.
Taylor, J. A,, Lancaster, G. M., Ignatiev, A., and Rabalais, J. W. (1978). J . Chem. Phys. 68,
1776.
Thomas, S. (1976). Surf: Sci.55, 754.
Thompson, V., Hinterman, H. E., and Chollet, L. (1979). Surf. Technol. 8,421.
Tompkins, H. G. (1977). Surf: Sci. 62,293.
Tosatti, E., and Parravicini, G. P. (1971). J . Phys. Chem. Solids 32, 623.
Tracy, J. C. (1972). Lecture Notes. Nato, Summer School Gent.
Treglia, G., Ducastelle, F., and Spanjaard, D. (1980a). Phys. Rev. B 21, 3729.
Treglia, G., Ducastelle, F., and Spanjaard, D. (1980b). J . Physique 41,281.
Treglia, G., Desjonqueres, M. C., Ducastelle, F.,and Spanjaard, D. (1981). J . Phys. C 14,4347.
Tung, C. J., and Ritchie, R. H. (1977). Phys. Rev. B 16, 4302.
Tung, C. J., Asley, J. C., Birkhoff, R. D., Ritchie, R. H., Emerson, L. C., and Anderson, V. E.
(1977). Phys. Rev. B 16, 3049.
Tung, C. J., Ashley, J. C., and Ritchie, R. H. (1979). Surf: Sci. 81,427.
Turner, N. H., Murday, J. S., and Ramaker, D. E. (1980). Anal. Chem. 52,’84.
Van Oostrom, A. (1976). J. Vac. Sci. Technol. 13,224.
Van Oostrom, A. (1979). Surf: Sci. 89,615.
Van Santen, R. A,, Tongman, L. H., and Bouwman, R. (1975). Surf.Sci. 47,64.
Vashista, P., and Singwi, K. S . (1972). Phys. Rev. B 6, 875.
Vayrynen, J., Aksela, S., Kellokumpu, M., and Aksela, U. (1980). Phys. Rev. A 22, 1610.
Verhoeven, J., and Los, J. (1976). Surf: Sci. 58, 566.
Vine, J., and Einstein, P. A. (1964). Proc. IEEE 111,921.
Von Barth, U., and Grossmann, G. (1979). Solid Commun. 32,645.
Walls, J. M., Hall, D. D., and Sykes, D. E. (1979). Surf: Interface Anal. 1, 204.
Watanabe, K., Hashiba, M., and Yamashina, T. (1976). Surf. Sci.61,483.
Watanabe, K., Hashiba, M., and Yamashina, T. (1977). Surf: Sci. 69,721.
Wiser, N. (1963). Phys. Rev. 129, 62.
Woodruff, D. P., Traum, M. M., Farrell, H. H., Smith, N.V., Johnson, P. D., King, D. A.,
Benbow, R. L., and Hurych, 2. (1980). Phys. Rev. B 21,5642.
Worthington, C. R., and Tomlin, S. G. (1956). Proc. Phys. Soc. London A69,401.
Yabumoto, M . , Kakibayashi, H., Mohri, M.,Watanabe, K., and Yamashina, T. (1979). Thin
Solid Films 63, 263.
Yin, L. I., Tsang, T., Adler, I., and Yellin, I. (1972). J. Appl. Phys. 43, 3464.
Yin, L. I., Adler, J., Tsang, T., Chen, M. H., Ringers, D. A,, and Crasemann, B. (1974). Phys.
Rev. A 9, 1070.
Yip, K. L., and Fowler, W. B. (1974). Phys. Rev. B 10, 1400.
Yu, M. L., and Reuter, W. (1981a). J. Appl. Phys. 52, 1478.
Yu, M. L., and Reuter, W.(1981b). J. Appl. Phys. 52, 1489.
Yu, M. L., and Reuter, W. (1981~).Appl. Phys. Lett. 38, 525.
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS. VOL. 61
The Wigner Distribution Matrix for the Electric Field in a

Stochastic Dielectric with Computer Simulation
D. S. BUGNOLO*
Department of’ Electrical and Computer Engineering
Florida Institute of’ Technoloyy
Melbourne. Florida
H. BREMMER
31 Bosuillaan, Flargebouw Houdringe
Bilthoven, The Netherlands
I. Introduction ........................................................... 300

11. The Differential Equation for the Electric Field Correlations ... . . . . . . . . . . . 303
A. Reductionof(EijCjk) ...............................
B. Evaluation of the Functional Derivative in Eq. (9). ......
C. The Ensemble Average of eijCjk,a Most Relevant Parameter.. . . . . . . . . . . . . . 310
D. The Equation for the Ensemble Average of the Electric Field Correlations . . , 312
111. Derivation of the Equations for the Wigner Distribution Functions . . . . . . . . . . . . 313
A. Introduction of the Wigner Functions and Derivation of Their Equations. ... 313
B. Preparation for Further Analysis.
IV. Related Equations for the Wigner Distribution Function ..................... 325
V. Asymptotic Equations for the Wigner Distribution Function . . . . . . . . . . . . . . . . . . 330

A. The Forward-Scattering Approximation ................................ 330
VI. Equations for Some Special Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
A. The Case of a Plane Wave Incident on an Isotropic Stochastic
Dielectric Half Space.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
B. Monochromatic Waves in an Isotropic Time-Invariant Stochastic Dielectric. . 342
C. Monochromatic Waves in an Anisotropic Stochastic Dielectric ............ 345
VII. A Brief Review of Other Theoretical Methods . . . . . . . .
VIII. The Coherent Wigner Function ....................
A First Approximation for the Coherent Wigner Function.. . . . . . . . . . . . . . . . . . . 348
IX. Computer Simulation of the Stochastic Transport Equation
for the Wigner Function in a Time-Invariant Stochastic Dielectric. . . . . . . . . . . . . . 354
A. An Integral Equation for a Very Narrow Beam
B. An Algorithm for the Computer Simulation of
Equation for the Case of a Very Narrow Beam
C. Variables Used in the FORTRAN Programs . ............ 359
D. Experimental Program Number One . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
* Present address: Department of Electrical and Computer Engineering, University of

Alabama in Huntsville, Huntsville, Alabama 35899.
299
Copyright 0 1983 by Academic Precs, Inc.
All rights of reproduction in any form reserved.
ISBN 0-12-014661-4
300 D. S. BUGNOLO AND H. BREMMER
E. Experimental Program Number Two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365

F. Conclusions and Suggestions for Future Work ........................... 378
X . Conclusions.. .......................................................... 382
Appendix 1. A Listing of Experimental Program Number Two for the Case
of an Exponential Space Correlation Function .................. 383
Appendix 2. A Sample of a Computer Simulation .......................... 386
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
I. INTRODUCTION
In this work we shall make extensive use of the Wigner distribution

function. This function was first defined by E. Wigner in 1932 in connection
with a problem in quantum mechanical systems. Using present-day notation,
the one-particle Wigner function is defined by Reickl(l980) as
In effect, the Wigner function is the Fourier transform of a special kind of

quantum mechanical density. Applications of the Wigner function to quan-
tum theory have been discussed by many authors, including Mori et al.
(1962). Because our applications are limited to stochastic but classically
posed propagation problems, we do not attempt to present a complete
summary of the applications to quantum theory.
We begin by noting that the Wigner function of quantum theory is
defined as the Fourier transform of the expectation of a density matrix. Our
definition is somewhat more general. Let the Wigner distribution function
of the time function h(t)be defined by
We stress that this is a “distribution” since the function is not yet ensemble
averaged. It is also evident that the ensemble average of Eq. (2) defines at
most a nonstationary spectral density function; i.e., a spectral density func-
tion that may be a function of the time of observation. It is also evident that
for the general nonstationary case given by
(W,
0))= - dt‘ exp(iwt’) (h* (t + g) (t - I;))h (3)
a time-averaged determination of the time correlation function of h(t) may

not be exchanged or substituted for the ensemble average. This point may be
of importance when theoretical results are to be compared to experimental
results.
STOCHASTIC DIELECTRIC WITH COMPUTER SIMULATION 30 1
Wigner functions as defined by Eq. (3) have proved useful in propagation

problems and have thus occurred in Tatarskii’s (1961, 1971) theories, al-
though the reference to Wigner was not mentioned. We further observe that
the Wigner function has been introduced with great success by Bastianns
(1978, 1979a,b) for the case of optical signals and optical imaging con-
siderations (Bartelt et al., 1980).
Proceeding in a straightforward manner by way of the Wigner function,
Bremmer (1979) has used these assumptions to obtain many of his results.
However, the analysis was concerned with the scalar Helmholtz equation.
Nevertheless, Bremmer (1979) was able to obtain a stochastic transport
equation for wave propagation in a stochastic medium in a rigorous manner.
In effect, this has served to give a more solid foundation to an earlier and
more heuristic analysis by Bugnolo (1960a,b, 1972a).
In this article we extend the method first proposed by Bremmer (1979) to
the complete vectorial Maxwell’s equations. This is achieved by first deriving
the wave equation in the stochastic medium in terms of a tensor of the second
rank, the elements of which consist of correlation functions associated with
the three components of the electric field. At a later stage it will be possible
with the aid of a simple Fourier transform to pass from the matrix of these
correlation functions to that of the Wigner distribution functions. The latter
are of special physical interest because they are connected with local (running,
i.e., nonstationary) energy distributions over the various wave-number vec-
tors and over different frequencies. We also obtain a set of stochastic trans-
port equations for the Wigner function in the tensor formulation.
We have made every effort to apply our most general result to a number
of special cases. The most general of these special cases is that of forward
scattering by a space-time variable stochastic dielectric. Our Eqs. (119) and
(120) are transport-like equations for the real and imaginary parts of the
ensemble-averaged Wigner tensor. Since the Wigner tensor is Hermitian, it
follows that these equations contain nine independent equations from which
it is theoretically possible to obtain a complete discription for the Wigner
tensor.
In view of the complexity of our general results for forward scattering,
we have also obtained some approximate transport equations for the case of
a plane wave incident on an isotropic stochastic dielectric, and for a mono-
chromatic plane wave incident on a time-invariant stochastic dielectric. We
have compared this last result to some other results previously reported in
the literature by Howe (1973).
Since the stochastic transport theory is just one of many methods used
to address this class of propagation problems, we have also included a brief
review of the parabolic equation method, comparing the physical assump-
tions of this method to our own. We conclude that the parabolic equation
method and our method do not differ a great deal in their physical assump-
tions, although differences do exist of perhaps fundamental importance.
Any application of stochastic transport theory must eventually address
the problems associated with the physics of the fluctuations of the medium.
When these fluctuations are due to turbulence, it is then necessary for com-
pletness to address the problem of the space-time correlation or spectra of
the additives that determine the statistics of the dielectric. In the lower
atmosphere, passive additives such as temperature and humidity are of
importance. In a weakly ionized gas or in a plasma, reacting additives such
as the electron and ion densities are of importance. In view of the present
fundamental limitations of the theory of turbulence and the theory of addi-
tives, caution is required, particularly if the dissipation regions are of im-
portance to the problem at hand. This is particularly true at optical fre-
quencies where the field is affected by the dissipation region under certain
conditions, since the inner scale size ld is large compared with the wavelength
of the radiation. Some problems may also exist at millimeter wavelengths.
In conclusion, if the theory of propagation in stochastic media is to be placed
in a solid theoretical framework, it is essential that the theory of turbulence
be closely examined, particularly in wave-number regions affected by the
dissipation of reacting and nonreacting additives.
We have also made a considerable effort to present the results of many
computer simulations using a VAX-11 system and have developed an
algorithm for the special case of a narrow beam of monochromatic radiation
normally incident on a time-invariant stochastic dielectric. The computer
simulation is a somewhat idealized example for propagation in the wave-
length range of 1 cm to 1 mm in an atmosphere characterized by very weak
to strong turbulence. We have presented our results for the ensemble average
of the principal diagonal component of the Wigner tensor, both as a function
of range to about 20 km and as a function of wavelength for various fixed
ranges. Since our algorithm is limited to a single component of the Wigner
tensor, the range of propagation is limited. These limitations are also dis-
cussed in detail.
We have also included a number of our computer programs, written in
the FORTRAN language, for the special case of our narrow-beam algorithm.
Our programs have been written to call integration subroutines, using the
cautious adaptive Romberg extrapolation, by way of IMSL (1979). Double-
precision computations have been used throughout the programs, thus
ensuring good numerical accuracy. Our programs may be used to simulate
the effects of any valid, self-consistent, spacewise spectrum for the dielectric
fluctuations.
Of the various wavelengths of interest for propagation in the earth’s
atmosphere, we should like to suggest that the millimeter-to-optical regions
are of the greatest interest. The recent activities of other authors, particularly
STOCHASTIC DIELECTRIC WITH COMPUTER SIMULATION 303
Bastiaans (1979b), have suggested a connection between the Wigner scalar

function and the ray concept of geometrical optics for the case of an in-
homogeneous but nonstochastic medium It should indeed be possible to
find a simular connection for the case of a stochastic dielectric described by
way of our general results as given by Eqs. (119) and (120); however, it is
self-evident that the connection is not simple for the case of a complete tensor
Wigner function. Indeed, the Wigner tensor may at most have a constant
value along a geometrical optical light ray, in a nonstochastic medium.
In our conclusions we have made a number of suggestionsfor future work,
such as the problem of very long range propagation at both millimeter and
optical frequencies. In this case we believe that a single component of the
Wigner tensor is insufficient and a more complete discription is required.
The complete tensor description is also required for the case of strong
turbulence with strong fluctuations in the dielectric and for the case of a
stochastic anisotropic dielectric. Some examples are atmospheric turbulence,
ionospheric turbulence, or wake turbulence. It is evident that a great deal of
work remains to be done in this relatively new field of stochastic transport
equations for the ensemble-averaged Wigner tensor in a stochastic space-
time-variable dielectric. Some thought should also be given to the per-
formance of well-controlled laboratory experiments, since the statistics of
the atmosphere are usually imperfectly known in the large, and may indeed
be quite variable from point to point.
11. THEDIPFERENTTAL
EQUATION FOR THE ELECTRIC
FIELDCORRELATIONS
In order to ensure that the results of our theory may be applied to one of
the most general discriptions of the media, we shall characterize the sto-
chastic dielectric by the nondispersive tensor to(l + i). The fluctuations of
the dielectric from its constant background value of co may be a function of
both position r and time t ; however, we shall assume that (Z) = 0. Although
this may at first appear restrictive, media such as partially or totally ionized
plasmas may still be included by a simple extension in the anisotropic case
and by including any average background in co .
In such a general medium, Maxwell's equations are
a
V xE + PO- aH
at
= 0, V x H- €0 -
at
((1 + B)E) =J (4)
with the following wave equation for the electric field intensity:
We now define a space-time “correlation function” for the electric field by

Cik(T,t ; I’, t ‘ )= ET(r + r’/2, t + t’/2)Ek(r- d/2, t - t’/2) (5b)
where the asterisk denotes the complex conjugate and i, k = 1,2,3. By using
a systematic derivation which parallels that previously introduced for the
Helmholtz equation by Bremmer (1979), we can show after considerable
algebra that the “correlation function” c i k satisfies the two independent wave
equations, each defined by either the upper or lower notation throughout
Eq. (6):
Given a deterministic, tensor-like dielectric, Eq. (6) could in principle be

solved for a likewise deterministic “correlation function.” However, given a
stochastic dielectric tensor, we must ensemble average Eq. (6) over all
possible realizations of the tensor E. Defining the correlation function of the
electric field by such an average, we may write
(Cik(r, t ; r’, t ’ ) ) = (ET(r + r’/2, t + t’/2)Ek(r- r‘/2, t - t’/2)) (7)
We then obtain the wave equation for the correlation function (Cik) in a
stochastic tensor dielectric
Ensemble averages such as (&ijCik)present us with some difficulty. This, or

a similar difficulty, seems to exist in all formulations of wave propagation in
stochastic media. In the following we shall introduce a number of simple

assumptions which may be used to relate terms such as (cijCjk)to the various
elements of (C i k ) .
A. Reduction OJ’(&ijCjk)
1. First Physical Assumption: A Gaussian Distribution for E

In order to reduce the cross correlation (cijCjk),we shall limit our
discussion to media fluctuations that may be characterized by a joint Gaus-
sian stochastic process. By this we mean that the elements of Eik(r,t ) observed
at a point in space and time and the product of their values at different points
in space and time may both be characterized by a joint Gaussian probability
density function. Although this may appear restrictive, we note that the
Gaussian process is characterized by the condition of maximum entropy for
a fixed standard deviation and as such is found to occur repeatedly in nature.
This is true in ionized as well as nonionized media and has been observed
experimentally.
With regard to plasmas, we note that whereas the electron density may
be Gaussian in its fluctuations under the turbulent stress of the gas, the tensor
B will only be Gaussian when the frequency of the probing electromagnetic
wave is well under the critical frequency of the plasma. The Gaussian re-
striction permits us to apply the Novikov (1965) relation previously used
by Bremmer (1979) in connection with the Helmoltz equation. We may
write
(E(r, t)F{E(r,t ) } )=
s dr’ dt’(E(r, t)E(r’,t’))
where the last term on the right-hand side (RHS) is a functional derivative of
F with respect to changes in the stochastic function E. Functional derivatives
are to the calculus of variations what partial derivatives are to the ordinary
calculus (see Gelford and Formin, 1963). Equation (9) also constitutes an
extension of the Novikov relation as previously used by Bremmer (1979)
without the time dependence. The inclusion of both t and t‘ will later permit
us to apply this result to a much larger class of problems.
2. Second Physical Assumption

In order to calculate 6Ej(r, t)/6Eik(r’, t’), we must begin with Eq. ( 5 ) for
the jth component of the electric field:
Introducing variations in both E j and E and subtracting the resulting equation

from Eq. (10a) yields
(8" - $); a (V.6E)
6 E j - axj
We shall proceed by neglecting second-order products in the 6 terms, a

reasonable assumption at all frequencies. However, if we are to proceed in
a simple manner, we must also neglect the first term on the RHS, E~~ 6Er.
We shall address the assumption in a moment. With this final assumption,
Eq. (lob) reduces to
This form of the variational equation may be addressed directly.

By neglecting terms of the type
1 a2
-- ejr 6Er
c2 a t 2
we have, in effect, neglected the effect of the stochastic medium on the field
E itself, as is apparent from Eq. (11). This amounts to the use of a first Born
approximation for the evaluation of the functional derivative. Such a pro-
cedure is similar to that used by Bugnolo (1960b) in evaluating one of the
integrals of a proposed stochastic transport equation for the spectral density
of the electromagnetic field. The first Born approximation for the effects
produced by a local perturbation is useful when the volume involved is
small compared with a mean free path (MFP)for scattering by the stochastic
dielectric. Over such a bound volume the probability of more than a single
scattering is very small. Using models for the spacewise spectrum of the
dielectric fluctuations originally introduced by Norton (1960), we obtain the
following expression for the MFP d,:
A result for the MFP of this form was first obtained by Bugnolo (1960a) in
connection with the solution of the approximate stochastic transport equa-
tion for the spectral density of the field which was proposed at that time. It
should also be noted that the MFP d, should be on the order of ( k & , ( ~ ~ ) ) - ' ,
independent of the model used to characterize the stochastic dielectric. This
latter point was first obtained by Bremmer (1964), who also later noted the
same in Bremmer (1974).
Returning to Eq. (12), we note that the parameter p = 1/3 for the Kolmo-
goroff spectrum and that such a spectral region occurs frequently in both
ionized and nonionized media when the fluctuations are the result of a fully
developed turbulence of the gas (see Bugnolo, 1972b).
If we limit the range of the spacewise variation of the 6 E terms to a
distance on the order of the mean scale size lo, and if the wavelength A0 is
such that d, >> l o , then it follows that we may use Eq. (11) provided that we
make a second physical assumption:
Although this places a physical lower bound on the wavelength, or an upper

bound on the frequency, we note that our overall theoretical model may be
on weak ground when Eq. (13) is not satisfied. This is not due to the ap-
proximations made in obtaining Eq. (ll), but to other reasons related to
the models used to characterize the spacewise spectrum of the dielectric
fluctuations.
In cases where the dielectric fluctuations result from turbulence in the
medium, the higher frequencies will interact with scale sizes beyond those
controlled by the viscous subrange of the turbulent velocity spectrum. We
must at these higher frequencies deal with the inertial diffusion region or
perhaps even with the dissipation region. A typical example of an isotropic
Kolmogoroff spectrum is illustrated in Fig. 1 for the very simple case of an
Kolmogoroff spectrum in the

inertial region: ~ - 1 1 / 6
--
Y
p.
Dissipation
region
FIG. I . A typical example of an isotropic Kolmogoroff spacewise spectrumfor the dielectric

fluctuations: (a) I,, mean scale size of the dielectric fluctuations; I,, characteristic scale size of
the dissipation region; (b) K = k - k’.
inertial and dissipation region. Using the root-mean-square (RMS) scatter-

ing angle previously defined by Bugnolo (1960a),
ORMS E [ ( E ~ ) R / ~ , ] ~ / ~
where R is the range or distance traveled by the wave in the stochastic
dielectric. It follows that the wave number K will lie within the inertial
region of the spacewise spectrum of the dielectric fluctuation if and only if
the wavelength of the electromagnetic (EM) wave A, is such that
It follows that Eq. (13) will be most relevent when the range R is less than
(W3.
In any case, a complete solution of the stochastic propagation problem
will require a detailed knowledge of the dielectric fluctuation spectrum.
Although Bugnolo (197213)has already addressed the problem of turbulence,
it is our opinion that more work is required in the theory of turbulence if
progress in the corresponding propagation problem is to be forthcoming.
For this and other reasons, such as the answer to the pointed question,
What is actually really known on a sound statistical basis about a complete
propagation path?, we shall restrict ourselves for the time being to wave-
lengths which satisfy Eq. (13). Some numerical examples are in order.
For a typical propagation path in the earth's troposphere, we might use
the values 1, E 10 m and (c2) z consequently A, >> lo-' m, a con-
dition that is most certainly satisfied at frequencies below the far infrared.
As our second example we shall use the turbulent interstellar plasma
which is of interest in the study of pulsar pulse shapes (Bugnolo, 1978). Here
we take I , = 3 x 10" m and d , = 1014 m at 300 MHz. Similar considera-
tions indicate that we may use our theory in the frequency range below about
10 GHz.
Prior to concluding this section we should like to note that whereas the
region bounded by the condition dB >> 1, is well posed and easily calculated
when I,, may be determined experimentally, a large body of literature exists
for wave propagation at optical frequency that clearly violates this simple
condition unless 1, is assumed to be very small.
B. Evaluation of the Functional Derivative in Eq. ( 9 )

According to our second physical assumption, SE should propagate as
if the E disturbance were absent almost everywhere; therefore we may proceed
with the solution of Eq. (11). We shall solve this equation for the vector SE
with the aid of operational calculus. We introduce the four-dimensional

“image functions”
ej(P, 4) = PlP2P34
s dr dt exp[ -(P * + qt)] GEj(r,t,
f
(16)
aj(P, 4) = PlP2P34 dr dl exp[ -(p’r f 4t)]Xj(r,
where p is a vector with components (pl,p 2 , p3). The vector equation corre-
sponding to Eq. (11) is now “imaged” into
[ p 2 - (q2/c2)]e - p(p-e) = q2a, xj E 1
c - ~ ~E~,E,
r
(17)
The solution for the three components ej of e may be found with the aid of
a determinant to yield
In view of the “operational images”
and the transformation rule for the convolution products, we obtain the
following “operational original” of Eq. (18):
This involves the explicit expression

1 J {a2xj(rff,t”)
GEj(r, t ) = - - dr” dt”
4n atif2
6(t - t” - Ir - r”l/c)
-
c2 F
m a2
Ir - r”l
After applying partial integrations, we obtain
6(t - tf’ - Ir r”I/c)

- c2 1Xr(r”, t”) ~
-
r Ir - rnI
Substituting the definition of the components of x , we find

a2
6Er,(r“, t“)Es(r”,t“)
ax: ax:,
~
4.n
- -
1
c2 r
SEjr(r”, t”)Er(rrf,
t”) -
a2
atn2 I ~ (-t
Remembering that the disturbances BE,, are confined to an infinitesimal

t” - Ir - r”l/c)
(r - r n )
(22)
domain around r” = r’, t“ = t’, we obtain the functional derivatives
The dependence of the functional derivative on the second-order partial

derivative may be explained physically by noting that the dielectric constant
normally enters this analysis by way of a second time derivative in the wave
equation [see Eq. (5)].
C . The Ensemble Average of &;jCjk,a Most Relevant Parameter

The third-order correlation function involving the’dielectric fluctuations
and the electric field intensity is required if we are to reduce Eq. (8). A
similar problem must be addressed in the parabolic equation formulation
of the mutual coherence function (for example, see Ishimaru, 1978, Vol. 2,
pp. 412-414).
It is best to begin by writing the ensemble average under consideration in
its complete form:
(Eij(r + r‘/2, t + t’/2)Cjk(r,t ; r’, t ‘ ) )
This may be transformed by subtracting r‘/2 from r and t‘/2 from t to yield
(Eij(r,t)Er(r, t)Ek(f - r‘, t - t’)) (24)
We now proceed with the aid of the Novikov relation [Eq. (9)] and the
property that the functional derivative of a product can be determined in
the same way as the derivative of an ordinary product:
(Eij(T, t)Ej*(r,t)Ek(r - r’, t - t’))
Next, applying Eq. (23) and its complex conjugate and using the definition
(Eik(f, t)&ik(r’,t‘)) = hik(r - r’, t - t’)
we obtain
s
(cij(r, t)Ej*(r,t)Ek(r- r’, t - t’))
1
=- dr” dt“hij(r - r”, t - t“) (E,(r - r‘, t - t‘)ET(r“,t“))
4n
d{ a2 h(t - t” - [r - r”[/c)
+
x(+---) axi! c2 a t r r 2 Ir - r“I
(EF(r, t)Ej(r”,t”))
b(t - t’ - t” - Ir - r’ - r”I/c)
ax; ax; c2 atfr2 Ir - r’ - rrrl
By a shift of the arguments, while making use of the definition for the Ciis
[Eq. (5b)], we further obtain
=
4n
j drf’dt”h,, (r + r‘ - r”, t + -t’2 - t”
r‘ rft t - t‘ t” r‘ t‘
+ ?; -r + - + r’f, -t + + t”
x {(Cjk(i - +T, 2 2
-
The other quantity needed for the evaluation of Eq. (8)’

(&kj(r- r’/2, t - t’/2)Cij(r,t ; r’, t’))
is obtained from the preceding one by, in succession, (1) interchanging i and
k, (2) replacing r’ and t’ by -r’ and - t’ and (3)taking the complex conjugate.
We have yet to apply the identity
Cg(r, t ; r’, t’) = cki(r, t ; -r’, -t’) (28)
After some algebra, the resulting equation reads

r'
=-
1 dr" dt"hkj(r -
r'
-
t'
r", t - - - t"
47t 2
-r + -r'+ - r" -t + - +t' - - ;t"r + - - rr'" , t + - - t " t'

2 4 2'2 4 2 2 2
r r' + -
r" -t - -
t' t"
2'2 4 2
+ -; -r + r'2 + r", -t + -t'2 + t"
~
D. The Equation for the Ensemble Average of

the Electric Field Correlations
If we substitute Eq. (29) in Eq. (8) and replace r" and t" by new variables
of integration defined by s = r f r'/2 - r'' and z = t t'/2 - t", we obtain
after some algebra
where CS(;,~) marks the Kronecker 6 function 6!, henceforth represented by

6(a,b) should this representation become simpler. In the present case, a = i
or k and fl = j .
This is our equation for the ensemble average of the electric field corre-
lation (C i k )under the condition of the two physical assumptions.
FOR THE WIGNER

OF THE EQUATIONS
111. DERIVATION
DISTRI~UTION
FUNCTIONS
A. Zntroduction of the Wigner Functions and Derivation

of Their Equations
The matrix for the Wigner function is related to the matrix for the electric
field correlations by
Wik(r, t ; k, W ) - dr‘ dt‘ exp[i(k r‘ - - ot‘)]Cik(r,t ; r’, t’) (31)
The inverse of the Fourier transform reads

n
Cik(r, t ; r’, t’) =

J dk’ do‘ exp[ - i(k’* r’ - dt’)]Wik(r,t ; k , o r ) (32)
In terms of the Wigner function, Eq. (30) may be written
x 1 dk d o exp[ - i(k * r‘ - at’)]( Wik(rrt ; k, 0))

x dk do exp[ - i(k. r’ - cot’)]( Wjk(r, t ; k, o))

ij
- (? 2
4ac2 at - at
+ 4y ds drh,(s, z)
k
x {J dk do exp{ -i[k-(r‘ T S) - d t ’ T .)I>
+
s ((I;; t’ z
dk do exp[Ti(k.s - oz)] W j j r 2 - - -, t k - - -; k, co
b(~
f t’ - (ST r‘l/c)
2 2
(33)
Evaluating the effect of the operators a/&’ and d/dt’, we obtain
x 1 dk do exp[ - i(k * r’ - at’)]( Wik(r,t ; k, co))
x J m( do exp[ - i(k.r’ - cot’)]( Wjk(r, t ; k, 4)

ij
1
In order to recognize in all terms a Fourier transform with respect to the

variables r’ and t’, we yet introduce a corresponding Fourier integral for the
operant at the end of the second term on the RHS, as well as for the last term
on the RHS. Thus the new RHS reads
ds dz h,(s, z)
k s -
dk do exp[ - ik (r‘ f s) + iw(t’ T T)]
~ I+)
x ( ij( 4
Wj, r - -, t - -; k, w))
2
z
(- a2asj 2 -)d2
as,
k
6’
- c2 az2
6(-
IS[
+-6 4 1n 5 c 2 F sdsdzh..(s,z)
s dkdoexp(Tik.sf i o z )
X { dk’ dw‘ exp( - ik‘ r’ a + iw‘t’)

s dh d p exp(ik‘ - - iw’p)
x [dk’ dw‘ exp( - ik’ r’ + iw’t’)

s dh d p exp(ik’. h - iw’p)
We next replace the operator a/at k 2a/at’ by a/at & 2io‘ and omit the
operator sdk do exp(-k.r’ + iwt’) in order to get rid of the variables r’
and t’; however, we first have to interchange k and k’, as well as o and ID’,
316 D. S. BUGNOLO AND H.BREMMER
in the second term on the RHS and then replace k by k‘ and o by o’in the
last term. This leads to
{(i T 2iky - $(: 2 i o y ] ( Wik(r,t ; k, w ) )
- (& T 2ikJ (& T 2ikj) ( Ydr,

ij t ; k, a))
k
- 1
- 471c2
ds dthy(s, T ) exp(fik s T iwz) -
(it-
+ 2iw >’
S
1
+-
dh d p exp(ik-h -
We must further apply the substitutions s = - 2r’ and z = - 2t’ in the first
term on the RHS; h = + s & 2r’ and p = + T f 2t‘ in the second term on
the RHS; and d(2r) = $6(r). We then obtain, after some reordering,
d2 a2 6(t’ + Ir’l/c)
+ r ’ ~ + t’;
(Wjkcr
ij
ky
(d.:a.;
k
- 7 p) lril
1
+-64n5c2 7s s
dk’ do‘ dr exp[fi(o’ - o)z]
s dr’
exp( f 2ik r’) ds hij(s,z) exp[ k(k - k’) s]

S k S dt‘
( Wjj(r+ r’, t + t’; k’, 0‘))
exp[i(k.h - op)] ($(r T 5, t T !)$i(r $

- ?,
h t & i)) (37)
B. Preparation for Further Analysis

According to the Wiener-Khintchine theory it is customary to pass, in
the treatment of stochastic media, from the correlation function for the
permittivity to a power function, connected to the former by a Fourier
transform. In our vectorial treatment we have to introduce a new matrix
F, the elements of which are given by
‘S
Pik(k, 0) - da dz exp[i(k a
16z4
+ wz)]hik(a,z) (38)
We may consider Pi, to be a component of the spatial frequency spectrum of

the dielectric fluctuations. The inverse relation reads
h,(a, z) =
s dk’do‘exp[ - i(k‘ a - + 0‘7)]Pik(k’,
The space-time correlation function h, is a real even function in a and z;
0’) (39)
namely, h,(a, z) = & ( - a , -7). As a consequence, Pik is also real and

even, i.e.,
Pik(k, a)= Pik( -k, - 0) (40)
An inspection of Eq. (37) indicates that the correlation matrix and inte-
grals of the form of Eq. (38) occur in both the first and second terms on the
RHS. If the h, terms are replaced by integrals of the form of Eq. (39), this in
turn will lead to integrals of the form
Integrals of this type occur frequently in stochastic propagation theory.

They may be reduced in the following novel manner. Let us introduce a set
of coordinates in the r‘ space such that the polar axis is taken in the direction
318 D . S. BUGNOLO AND H. BREMMER
dr‘exp(Tiar‘)[exp(TilVlr‘) - exp(+i(~lr‘)]
Hence
dr‘ exp[ f i ( a + IVJ)r’]

-:j dr‘ exp[ f i(a - (Vl)r’]
In view of the condition Im a 5 0, both integrals converge at r’-+ 00,

yielding
P(a; V) = &(2.ni/lV1){T[i/(a+ Iv~)] & [i/(a - Iv~)]} (44)
or P(a; V) = 4z/((Vl2 - a’), Im a 5 0. On the other hand, if a is real we
must apply the formula
J: dr’ exp( T ipr’) = n6(p) f i/b (45)
where the 6 function gives use an additional contribution. Recalling Eq. (43),
we may write for the &function contribution to P(a; V),
If we use the identity
we obtain the contribution in question:

-
+4n2i sgn a6(1VI2- a’)
Consequently, for a real we find,
Since we must in our analysis deal with the real quantities k, w, etc., it
would appear that the more complex result of Eq. (49) must be used rather
than the more simple result given by Eq. (44).We shall prove that
In fact,
4a
lim
&-.+I) IV(’ - (a T i&)2
= lim 4a
((V(’ - a’ + 2 )T 2 i a ~
E’+O (lV12 - az + &2)2 + 4a2E2
Taking t = IVl’ - a’, we must evaluate

T = lirn [ ~ / ( (+t E’)’ +~ C X ~ E ~ ) ]
e++O
the limit of which may be obtained with the aid of the following reduction,
assuming provisionally that a > 0:
& 1
(t + &’)’ + 4a2E2 2aE - i(t + &’) + 2aE -k i(t + E 2 )
+m:J
W
do exp[(2as
do exp[i(t
+ it + is’)o]
+ &’)a - 2acloll
I
(53)
Hence, taking E = 0, we have,

71
.
=‘
,f
a do exp(ito) = -
201 d(t) (54)
4a -w
Finally, substituting for t, we have derived from Eqs. (50)-(52), for the case
of a positive a,
Since the second term in Eq. (55) is odd in a,a factor sgn a may be added
to it such that the result may be valid for all real a.This concludes our proof
that Eq. (50) is the equivalent of Eq. (49). We shall use as the abbreviation
for the positive value of E + +0, simply +O. We thus have proved the
following:
Before proceeding further, we note that the following requires the intro-
duction of the Fourier integrals for the Wigner functions, and these are
defined as follows with the aid of a transform F:
Wik(r,t ; k, co) =
s dh dp Fik(h,p ; k, co) exp[i(h * r + pt)]
All of the above considerations are used to further the reduction of the
(57)
RHS of Eq. (37). The first term on the RHS may be termed the “radiation”
contribution since the argument of the W, terms is independent of k , and
consequently is not associated with the mutual interference between different
k waves. This type of term is indicative of a general emanation or radiation
that results in an attenuation of the wave. It may also be attributed to some
absorption effect.
On the other hand, the integrations over k’, w’ in the second term on the
RHS of Eq. (37) reveal an “interference” between the waves of differing wave
numbers. Finally, the last term on the RHS of Eq. (37) may be associated
with a source effect.
C. Reduction of the Radiation Contribution R,

As explained above, the first term on the RHS of Eq. (37) shall be termed
the “radiation” contribution. Returning to Eq. (37), we proceed by sub-
stituting Eq. (39) for the correlation function hij and the Fourier integral
Eq. (57)] for the Wigner function. This leads to the following expression
for R,;
x exp[ f 2i(k * r‘ - wt’)] -

(it -
+ 2iw y d l dp Fjk(I.,p ; k, w )
ij
x exp[iI.*(r
k
or
R, = -
X
exp[i(i * r + pt)]
s
mC. do”Pij(k,of’) dr’ dt’
exp[i(l T 2k - 2k”)’r’ + i ( p & 2w

k s
+ 2w’‘)t’I
X (59)
The last fourfold integral, a, say, over r‘, t’ may be reduced by partial
integrations over the coordinates x’, y’, z’,and t‘. Since each of these integra-
tions extends from - co to + 00, the integrals tending to zero at these
boundaries, we may write the following after applying partial integrations :
-(5 T 2k - 2 k ) i - ( 5 f 2k - 2k”)j + a ( p
k C’
20 + 2~”)’
x exp[i(l f 2k - 2k”)- r‘ + i ( p k 20 + 2w”)t’l (60)

Next, we integrate over t‘ to obtain
= { exp[i(I. T 2k - 2k”)*r’- i ( p f 2 0
Ir‘l
+ 20”)lr’l/c] (61)
To this result we next apply Eq. (56). Hence,

~~
[-(I. T 2k - 2k”)i, (I. T 2k - 2k”)j + d c ,

* j)
(pk 2 0 + 2w”)2
0’
c
@=4n‘
p + 20’’ - iO)2
((I. T 2k - 2k”)(’ - ~ - ~ &( 2w
Substituting this result into Eq. (59) for R , , we find
-(A T 2k - 2 k ) i k ’ ( 1T 2k - 2k”)j + S(i,j)

(p f 20 + 2o”)2
c2
2k - 2k12 - ~ - ’ ( pf 20 + 20” - i0)’
X
11 f
x Fjk(k,
ij
p ; k , o)exp[i(l*r + pt)]
This may be put into a more convienient form by again using operato
to remove the term in brackets from the integrations on h’ and p. Since
then follows that the remaining integration over these variables equals tl
Wigner function, we may write, taking k“ =. k ,
x W$r, t; k , a>
V
This concludes our present reduction of the radiation term.
D. Reduction of the InterJerence Contribution R,

The second term on the RHS of Eq. (37) was identified as the interferen
term. It shall be evident later that this term, in the limit of single scatterir
yields a first Born approximation for the Wigner function. Again, using Ec
(38),(39),and (57), we obtain
R, = -
1
647csc2
x 1
fdk’do’ f dz exp[ fi(o’- o ) z ]
d k do” exp[ - i(k” s - + o”z)]Pij(k”,w”)

f dr’ exp[ f2ik.r’l
Ids
x exp[ k i(k - k ‘ ) s]
s dt‘ exp(T 2iot‘) -
(it-
+ 2io >’
x [dh dp Fjj(I,p ; k’, w ’ )
The t’ integration yields 2nS[o” T (w‘ - w)], enabling the integration.

0’’
This yields
R2 =
1
7J dk’ do‘ J dr‘ exp(k2ik-r’)
J ds
x jdk”P$k; +(a’- w)] exp[i(-k” f k T k)-s]

s dt‘
x exp( f 2iwt’) -
(:t-
+ 2io>’J d I dp Fjj (I, p; k’, w‘)
By following a procedure similar to that of the previous section, we may

eliminate the integrations over r’, t’, s, and k” to yield
1
R2 = 2 1
I
dk‘ d o ’ P i j [f ( k - k’); f(of- o)]
k
x
(it-
- + 2io>’j
d I d p Fjj(h,
p ; k’, 0’)
f 2k)k + (Sf/c2)(p f
x exp[i(h-r + pt)] -(I f(h2k)i*(I
f 2k)’ - ( p T 2 0 i0)’/c2
-
20)’
(67)
Again, I may be replaced by the operator -a/& and p by - ia/at, so that

in view of Eqs. (57) and (40),we obtain the desired result:
“J j
dk’
J dw’Pij[(k - k’); (o- o’)]Wjj(r, t ; k‘, m’)
k
We note that the “interference” term samples the spectrum of the dielectric
fluctuations at the wave number (k - k ) and the frequency (w - 0’).
E. Final Operational Equation f o r the Wigner Function

This may be obtained by substituting Eqs. (64) and (68) into Eq. (37).
After a minor transformation of variables by way of k” = k - k‘ and
w” = w - w‘ in R , , we obtain the desired result:
{:( f 2ik)i - $ (i+ 2iw)2}( Wik(rrt ; k, a))

- (&
k
T 2ik;) 7(& f 2ikj) ( W,k(r,
ij t ; k, w ) )
- - A(w T 4 2)i 7j dk‘ dw’Pij(k - k’, w - a’)

C2 2 at k
x {a;(k‘,
CO’; a/&, a/&)( Wjk(r,t ; k, w ) +
k ij
x ( k , CO; a/&, slat) ( W j j ( r ,t ; k’, a))} + S: (69)

where
is a useful operator, and where the source term is given by
x exp[i(k.h - up)] (E;(r f

h
2, t f ;)(r
h
f 2,t k ;)) (71)
In Eq. (69), (i, k ) = 1, 2, 3. As a consequence we have a total of nine

equations for the ensemble average of the Wigner function. These individual
equations are elements of the matrix equation for ( W ) .
Equation (69)may be put into a more compact form as follows: Let
Vc = (a/&) T 2ik, V; = (a/&) rfl 2iw (72)
and let the matrix A’ be defined by
Finally, let the elements of the R H S of Eq. (69) be used to define the matrix
R i . It then follows that Eq. (69) with the upper sign may be represented by
{(Vk+)’ - C-’(V:)’}( Wik) - C (A’)ij( Wjk) = Ri (74)
j
whereas that with the lower sign may be represented by

{(VL;)’ - C-’(V;)’}( Wik) - C ( x - ) k j ( Wij) = RLk (75)
j
From the properties of the V and A functions defined above, it is self-

evident that the two equations are not independent, one being the complex
conjugate transpose of the other, since the Wigner function is itself Hermitian.
Iv. RELATED
EQUATIONS
FOR THE WIGNER DISTRIBUTION
FUNCTION
A. An Integral Equation of the Second Kind for the Wigner Function

Equation (69)is a complete operational equation for the ensemble average
of the Wigner function. In its present form it is striking that even the terms
remaining in the case of a homogeneous medium, i.e., for P = 0, are rather
complicated. One possible way around this difficulty is the following.
We begin by writing Eq. (74) in matrix form:
[((VZ)2 - c-2(V,+)2}1- A + ] ( W ) = R + (76)
The matrix solution of Eq. (76) reads
( W )= [{(VZ)2 - c-2(V:)2}1 - ,+I-%+ (77)
In the above, I is the unit matrix. In order to further reduce Eq. (77), we
must use the general formula which may easily be proved:
Applying Eq. (78) to Eq. (77), we obtain
Once again using the definition for V: and A+, it follows that the ele-
ments of the ( W )terms are
Ri
C2
[ ( a p t ) + 2ioI2
2iki 1 - 2ikj Ri
(”-)+ )
-
(L
~
axi
(w)ik =
2 i a
(a:j 2 (80)
- 2ik) - ;;Z(’ 2io)
This, in turn, yields the following integrodifferential equation for the Wigner
function:
{(z dr
- 2ik)i- ?(%
i a + 2iw)i} (Wik)
- 2iki) 5: (& - Zik,)

+ 2ioI2 (2
C2
=Ri - R$ (81)
[(a/&) axi
Although this equation is somewhat simpler in form than Eq. (69), it is
still quite complicated in a homogeneous medium. This difficulty may be
removed as follows. Let the RHS of Eq. (81) be denoted by the function G
such that
{(: - 2ik)i - f (E + 2io)i} (Wik(r,t; k, =

0)) G(r, t ; k, w ) (82)
We can transform this to a more tractable form by defining a function 4

such that
-
( Wik(r, t; k, w ) ) = exp[2i(k r - wt)]$J(r,t ; k, o) (83)
Substituting Eq. (83) in Eq. (82) we find that
[(a/&) - 2ikI2(Wik) = exp[2i(k*r - at)] (84)
A similar relation may be obtained for the operator (a/& + 2io),
[(apt) + 2io]’(Wik) = exp[2i(k.r - wt)]d24/dt2 (85)
We thus arrive at the following inhomogeneous four-dimensional wave
equation for 4:
(..- f $)4(r, -
t; k, a)= exp[ - 2i(k r - wt)]G(r, t ; k, w ) (86)
In the absence of an inhomogeneous scattering medium, G = 0 and

Eq. (86) becomes homogeneous with the solution for the homogeneous
background 4pr.By treating the RHS as a source function, we may arrive
at a formal ‘‘solution’’:
M,t; k, o)= &(r, ‘s

t ; k, w ) - -
471
dr’ dt’
d(t - t’
Ir
- Ir - r’l/c)
- r‘(
x exp[ -2i(k*r‘ - ot‘)]G(r’, t‘; k, a) (87)
This is a formal integral equation of the second kind for 4.
It is of some importance to develop such an integral equation for the
ensemble average of the Wigner function itself. There are many reasons for
such an approach. The first is that a first Born approximation to the integral
equation is very easily obtained. Another is that the first-order smoothing
approximation by way of the diagram method may be used to obtain a
solution valid in the multiple-scattering region. For examples of the smooth-
ing approximation, see Frisch (1968) and Bugnolo (1972a).
s
Integrating Eq. (86) over t’ we obtain
4@,t ; k, 4 = 4pr-
exp(2iot)
dr‘
expi-2iCk.r‘ +
(o/c)lr - r’l])
4n Ir - r‘l
Now according to Eq. (81), the G function can be written as
G(r, t ; k, o)= 1
j
- zitti)(& - 2ikj)]Rf (89)
In order to remove the integration over r’, we introduce the Fourier integral
Rj+k(r,t ; k, w ) =
s ds d t b j k ( s , t;k, 0)exp[i(r *s + tt)] (90)
1
Substituting and applying the operators yields
4b.7 t ; k, 4 = 4pr-
exp(2iot)
dr’
exp{ -2i[k-r’ +
(o/c)(r - r’l]}
471 Ir - r‘l
x
j s ds dz bj&, 7; k, 0)
c2
(isi - 2iki)(isj- 2ikj)
+
(it 2i0)~
In order to apply Eq. (56),we must change the order of integration as follows:
c2
x
{
exp(i7t) 8; -
(t + 242 - 2 k J ~ j- 2kj)
( ~ i
+
exp(i[(s - 2k)ar’ - (t 2co)lr - r‘l/c]>
(92)
)r - r’)
328 D. S. BUGNOM AND H. BREMMER
From this point we follow the usual procedure. Using Eq. (56), we then
translate the { } outside of the integral over S and z. We then use the defining
equation for Q [Eq. (70)]. This yields
M . 3 t ; k, 0) = 4pr- 7
-
c2 exp[ - 2i(k r - wt)]
[w - +i(d/dt)I2
x 1Q$(k, co; d/dr, d/dt)R$(r, t ; k, w )

j
(93)
Returning to the Wigner function by way of Eq. (83),

C2
( W i k ( r , t ; k, @I)
= wikpr(r, t ; k, 0) -
410 - 3i(d/dt)12
x C O; a/ar, d/dt)Rj+k(r,t ; k, 0) (94)
.i
This may be put into final form by using the RHS of Eq. (69) for R i k with i
replaced by j :
( W i k ( r , t ; k, O)>
= Wik,,(C, t ; k, 0 ) -k 1a&&,CD; a/&, d/dt)
i
x [F / dk’ do’Pi,(k - k’; w - w’){Q;(k’, d/dr,

0’; d/dt)
It is self-evident that we may obtain a first Born approximation for

the ensemble-averaged Wigner function by replacing (Wlk) and (W,,) on
the RHS by Wpr. This result may prove to be of use in certain special
circumstances.
B. The Transport Equations for the Wigner Function

In our preceding analysis we considered two simultaneous equations
for the ensemble average of the Wigner function which, however, are not
independent. This dependence is connected by the requirement that ( W i k )
be Hermitian. That this is the case is self-evident from the definitions given
by Eqs. (31) and (7).
We may construct a stochastic transport equation for the ensemble

average of the Wigner function by taking the difference of the equations with
the upper and lower signs. The same result may be obtained by using the
imaginary part of the other equation. For example, from Eq. (81) we obtain
This result is complicated by the fact that the RHS depends on the other
components of the Wigner function. We shall return to this later.
A number of comments are in order. Stochastic transport equations of
the type described above constitute a generalization of the transport concept
to the case of stochastic media in the large. Whereas their most simplified
forms appear to be similar in a superficial manner to the classical transport
equations of astrophysics (see, for example, Chandrasekhar, 1950), they are
quite dissimilar in detail as a result of the fact that we are dealing here with
stochastic media in the large. In fact, any similarity may be superficial and
some caution is advised.
On the RHS of Eq. (96), the operator L where
indicates that the rate of change of ( Wii) in the fi direction is to be measured

while moving in this direction with the velocity
Vobs = (C2/4k (98)
In fact, the rate of change per unit time is given by
D/Dt = (c2/w2)L (99)
On the other hand, the phase velocity of the special plane waves associated
with ( Wik),i.e., exp(k r - wt), must satisfy the relationship
V,.k = w or k.Vph= w/k (100)
It is evident that since lVp,,l may exceed c, we must satisfy the relativistic
constraint
We shall return to the stochastic transport equation later.

v. ASYMPTOTIC EQUATIONS FOR THE WIGNER DISTRIBUTION

FUNCTION
A. The Forward-Scattering Approximation

Forward scattering of electromagnetic waves by stochastic dielectrics is
a special case of interest with many applications. By this we mean such
applications where the forward-scattered component of the Wigner tensor
is of the utmost importance. Before proceeding, it is best to review some
fundamentals.
Given an isolated volume of space characterized by the stochastic dielec-
tric E, and given that the volume is of linear extent L such that L is small
compared to the MFP for scattering by the stochastic dielectric [our dB of
Eq. (12)] and yet large compared to the mean scale size of the dielectric
fluctuations, we may proceed as follows.
Let the dielectric fluctuations be anisotropic so that the mean scale size
is a tensor related to the correlation function hij by
I-
loij E J hij(s)ds
By invoking conventional approximations, we may define a scattering cross
section per unit volume, per unit solid angle by
a(K) = 2nk: sin’ xP(K) (103)
In this result a(K) is obtained by way of a first Born approximation for the
scattered field, and P(K) is defined by Eq. (38), z = 0. In effect, in this particu-
lar formulation the fluctuations are characterized by a time-invariant space
correlation function. The geometry for this approximation is illustrated in
Fig. 2. The details of the above may be found in Ishimaru (1978, Vol. 2,
Direction of
scattering
FIG.2. Geometry for scattering for the case of a plane wave incident on a stochastic dielectric
of Volume V : K r 2k, sin(O/2).
Section 16-2). In order to proceed we require a model for the dielectric

fluctuation spectrum P(K) defined by Eq. (38).
Before continuing it may be judicious to indicate the differences between
our formulation and that commonly used by authors in the U.S.S.R. Some
time ago they noted that a stochastic dielectric such as the earth’s tropo-
sphere might be characterized in terms of a “structure function’’ defined by
Df(4 = ( I f ( r 1 + r) - f ( r , ) I 2 ) ( 104)
(See Ishimaru, 1978, Vol. 2, Appendix B or Tatarskii, 1961.) As an example,
consider the case of an isotropic turbulence producing, in turn, an isotropic
P(K) and Df(r). Existing theories indicate that the spectrum for such a case
will be Kolmogoroff in form. This yields for the structure function the form
where 1, is the mean or outer scale size and Id is the dissipation scale size.
For this case, Tatarskii (1961, 1971) obtains a spectrum of the form
P ( K ) = 0.033C,2K-”’3 exp( - K2/KH) (106)
for K > 271/1, where K , = 5.91/ld. We note that although this form may
only be applicable for K > 27&, it does not depend on the outer scale size
1, explicitly. This is usually justified by noting that the outer scale size I,,
is difficult to measure and is dependent on the geometry, whereas 1, is some-
what independent of the geometry.
Instead of the above we have used a formulation originating with Norton
(1960). In our notation, Norton defined a space correlation function for the
dielectric fluctuations of the form
where r ( p ) is the gamma function and K , the modified Hankel function. If

this formulation is used, then
We note that this formulation depends on the outer scale size 1, explicitly.
It does, however, have the disadvantage of being restricted to wave numbers
less than 27&. This latter problem has been addressed by Bugnolo (1972b)
for the case of a weakly turbulent gas or plasma.
Now the upper bound on the frequency which resulted from the second
physical assumption of Section II,A,2, i.e., Eq. (13), is based on the North
model for the case when the parameter p = 113, the Kolmogoroff case [Eq.
(108)], rather than Eq. (105). We suggest that since the outer scale size 1,
can indeed be measured in the troposphere and in the ionosphere, this seems
to be a better approach.
1. Complete Equations for the Wigner Function

Returning to Eq. (103) and using Eq. (107), we note that the scattering
will be well directed in the forward direction, i.e., small 8, and in fact will be
concentrated within a cone of half-angle
ezq24, (109)
It follows that the scattering will be primarily in the forward direction when
the wave number of the electromagnetic field is such that
(kl0)-' << 1 (110)
We are now in a position to proceed with a forward-scattering approximation
for Eq. (69).
Our formulation for the operational equation for the Wigner function
[Eq. (69)] should prove to be most useful because of the occurrence of the
operators d/dr and a/&, which lead to a set of reasonably well-defined con-
ditions for conventional approximations. We only need observe that notice-
able changes in ( W ) are to be expected over distances of the order of the
outer scale size 1,. This may be used as a further justification for explicitly
using this scale size in the propagation theory. As a consequence, the order
of magnitude of the effect of the operator a/ar may be given by
a/ar z 1;' (111)
Similarly, the time dependence of the fluctuations of the medium will be
associated with a characteristic time scale, to be denoted by z. In practice,
this is due to the motion of the irregularities with the mean wind and to their
internal decay across the wave-number spectrum. The order of magnitude
of this effect is connected with the operator a/&. Hence it follows that
a/at z 7-1 (112)
If we now heuristically invoke a disturbance velocity 'Vdist defined by
I/dist
then it may be shown that the operator R of Eq. (69) reduces to

provided that the following inequalities are satisfied:
Since the phase velocity of the waves vph is on the order of c, and since Vdis,
as defined by Eq. (113) is very small compared to the velocity of light, it
follows that Eq. (114) will be the forward-scattering approximation for the
operator R. This follows from Eq. (110).
Theoretically, from inspection of Eq. (108) we not only deduce the con-
dition of Eq. (110), but also note that the eddies sampled by the wave as it
proceeds through the stochastic medium will depend on the scattering angle
8, the size of the sampled eddy decreasing as 8 increases.
If the above is now applied to the RHS of Eq. (81), we obtain
Returning to Eq. (114) and comparing the denominator with that occurring
in Eq. (56) enables us to write
Qi’,(k, U ;a/&, a/&) -+ (- kikk + ($/c2)U2)
Substituting this result into the expression for R$ on the RHS of Eq. (69)
yields our result:
{(L - 2ikT - -$(: + 2 i w Y ] ( Wik(r,t ; k, u))

dk‘ do’Pj,(k - k , w - w’)
x ( WIk(r,t ; k, w ) )
x
+
1
]
x ( Wll(r, t ; k’, w ’ ) ) - __ S +
Jk]]
As before, the independent second equation may be obtained from the com-
plex conjugate of the above, taking into account the Hermitian property of
the Wigner function.
2. Transport Equations for the Wigner Function

Since according to Eq. (96), the operator
k . -a + o_2 _
a
dr c2 at
is characteristic of a “stochastic transport equation,” we may obtain such
an equation for the Wigner function by writing in succession the real and
imaginary part of Eq. (118), taking into account the fact that P and W,,
are
real. We obtain
Re( wik(r, t ; k, 0 ) )
+4
( ir +
k.- -
it)
Im( Wik(r,t ; k, w ) )
dk‘ ddPj,(k -k , co - 0‘)
x {(-k>ki+2cor2 ” )[Re( Wlk(r,t ; k, 0)) a 6(k‘ - ] o ’ ( / c )

k‘2 - 012/c2 2 +- w’/c
x Im( Wlk(r,t ; k, w ) )
X - -Re Si
+ --
(
- 4 k.-
r: it)
Re( Wik(r,t ; k, 0))
= - 4 1
j (YZ26.i
- - kikj
)[[F dk’ dw’Pjl(k - k’, o - 0’)
Im( Wlk(r,t ; k, 0))_ -.n 6(k‘ - (o’(/c)

k’2 - u 1 2 / c 2 2 w’/c
x ( WJr, t ; k', 0'))

I1
Equation (120) is a stochastic transport equation for the real part of the
Wigner function, whereas Eq. (119) is such an equation for the imaginary
part of (Wik).At present we are of the opinion that Im( Wik) is of less im-
portance then Re( Wik).However, this observation may require further
study.
Equations ( 1 19) and (120) are some of the most important results of this
work. We note that our stochastic stochastic transport equations are valid
in the absence of stochastic fluctuations in the medium as well as in the most
general case, which includes dielectric fluctuations moving with the mean
wind and subject to internal decay as predicated by the theory of turbulence.
3. Integral Equation of the Second Kind for the Wigner Function

Finally, we write the forward-scattering approximation for the integral
equation of the second kind for the Wigner function [Eq. (95)]. By following
a procedure similar to that above, we obtain
(W i k ( r , t ; k, o))
VI. EQUATIONS
FOR SOME SPECIAL CASES
In order to reduce our general results [Eqs. (1 19) and (120)] in the forward-
scattering region to some special cases of interest, we must first address the
stochastic media itself and extend our characterization of Section V,A. We
be& by notiqg that our general result €or forward scattering contains a
matrix expression for the space-time spectrum of the dielectric fluctuations.
In view of this it may be applied to the most general cases of an anisotropic

stochastic medium such as those found in many places as a result of natural
phenomena. The effects of propagation through such a medium may be
illustrated in a physically quantitative manner by the following description.
Let us begin with a plane-wave source of monochromatic spectrum at
some distance from a semi-infinite, time-variable stochastic medium. As this
wave begins to penetrate the medium it will first suffer from a slight loss of
phase coherence, an effect that may easily be described by a first Born ap-
proximation or by the WKBJ approximation of our Eq. (121). Such phase
perturbations of course require that the original spectrum of the wave, say
6(0 - w,), be spread somewhat. As our wave moves further into the medium,
the phase fluctuations of the wave will increase and its amplitude will begin
to fluctuate. In addition to this, we will begin to find an appreciable amount
of energy in wave numbers concentrated in a cone around the original
direction of propagation k, . The angular width of the initial cone will be on
the order of 342.111,. Following our wave further, we soon encounter the
distance marker d,, the MFP for stochastic scattering as given by Eq. (12).
At this point the phase spread will be exactly equal to 2.11. This point marks
the beginning of strong multiple-scattering effects. In fact, we may actually
estimate the number of such events (see, for example, Bugnolo, 1960~).As
a result of this multiple scattering, our wave will also be spread in wave-
number space itself, a fact that will become evident later in this work. Be-
fore proceeding, let us briefly illustrate the event probability for multiple
scattering.
Let us assume an isotropic process for the dielectric fluctuations because
this does not affect the fundamental characteristic of the discussion. Let us
further assume that all frequency shifts due to eddy Dopplers are negligible.
Under these conditions, it follows that the spectral function P, defined by
Eq. (38), may be written as
P,(k - k , w - 0’)
= 6(0 - d)P(lk - k’l) 6: (122)
For this case we may easily define a total cross section for scattering by the
stochastic dielectric (Bugnolo, 1960a,b, 1972a) such that
Q, E
s dkP(lk1) = dB1 (123)
Substituting for P(k) from Eq. (108) and proceeding in a manner similar to
that used in reducing Eqs. (41), we obtain in the high-frequency limit defined
by 2.11lo/,l > 10:
t
1.01
*
-0 1 2 3 4 5
( x ) Range in mean-free paths (dB) ( x = Range/d8)
FIG.3. Probability of multiple scattering by an isotropic scattering dielectric. Curves: A,
at least once; B, at least twice; C, at least three times.
In the forward scattering of interest we may also proceed here, as in

Bugnolo (1960c), to show that the probability of any ray of the incident
plane wave being scattered at least N times in the distance R is given by
P,(r I R) = [Q:/r(n)] r"-l exp(-Q,r) dr (125)

jOR
Integrating, we obtain the following:

PI = 1 - exp(-Q,R), P , = 1 - (1 + Q,R)exp(-Q,R),
(126)
f
', = f'z - t(Q,R)' ~ X P-(Q,R)
Equations (126) have been plotted in Fig. 3. We note that P , crosses the
equiprobability line (1/2) at a distance of Q,R = 1.7, or at 1.7 MFPs. At
greater distances it is only reasonable to expect extensive multiple-scattering
effects. We shall next address this interesting special case in greater detail
by way of Eqs. (119) and (120).
A. The Case of a Plane Wave Incident on an Isotropic

Stochastic Dielectric Half Space
We shall consider the case of an isotropic stochastic half space free of
sources such that the effects of the source are contained in the boundary
conditions fork, positive (i.e., into the half space). The geometry is illustrated
by Fig. 4. In accordance with the discussions of Section V, we begin with
Eq. (120) for the ensemble average of the Wigner function. In view of the
Hermitian property of the Wigner function, it follows that Im( Wii) = 0 and
Eq. (120) thus reduces to the following for the diagonal terms of the Wigner
matrix, i.e., for i = k ,
71 6(k’ - IO’I/C)
-- Re( Wji(r,t ; k, 0))
2 O‘/C
The source term has been retained for possible later reference. We have also
used the rather obvious condition for the isotropic dielectric:
= 6:P(lk - kl, o - w ‘ )
P,(k - k‘; o - 0’) (128)
It is clear that the left-hand side (LHS) of Eq. (127) follows directly
without approximation from Eq. (120) under the condition that i = k.
However, we are still faced with components of the imaginary part of the
Wigner tensor on the RHS of Eq. (127). We must still consider the effects
of the Im( Wji)’swithin the integral and compare this contribution to that
of the Re( Wji)’s.In order to do so, it is convenient to introduce the concept
of “blurred” and “sharp” resonances.
Y
i
4
Stochastic dielectric
* X
FIG.4. Geometry for a plane wave incident on a stochastic half space.

Blurred resonances are here defined as those obtained from the integral
dk’ dw’
P(lk
k’2
- k’l, cu - 0’)
- 0’2/cz Im( (5
Wji(r, t ; k, 0)) - k;’) (129)
whereas sharp resonances are here defined as those obtained from the integral
s dr’ dw‘P((k - k‘l, o - a’)
X
6(k’ - Iw’I/c)
Re( WJr, t; k‘, w‘))
W’/C
Since 6(k’ - w’/c) = c6(o‘ - k‘c), the latter yields after integration over w’,
c Re( Wji(r, t; k, w))

s dk‘
P((k - k l , o - 0’)
whereas the blurred resonances depend on the integral,

k
(k’ - k:’) (131)
Integrals such as these must always be considered as “principal values” if

they are not well defined otherwise (Whittaker and Watson, 1952). This
follows from the method that was initially used to construct the transport
equations, namely, Eq. (50). In view of this, Eq. (133) wil! 5e real if P is real,
which is always the case in this work. In addition, it will vanish as a “principal
value.”
Having so disposed of the contributions of the blurred resonances we
may rewrite Eq. (127) as follows:
+ I1
c2
x (Wjj(r,t ; k’,0’)) - Q Irn sji (134)
Equation (134) is our final, complete stochastic transport equation for the
case of a plane wave incident on an isotropic stochastic dielectric half space,
when observed within the latter.
In order to compare some of the results of our theory to those obtained
by other theoretical methods it is convenient, when possible, to make another
assumption. Let us specialize our result to the case of a normally incident,
monochromatic plane wave of the form
E = TE0 exp[i(k,x - coot)], Ex = E, = 0 (135)
propagating in the 2 direction. It is evident that the only component of the
Wigner tensor at the boundary that does not vanish is given by
W,,(O, t; k, 4 = 6(w - o 0 ) W o - k , ) W q 6 ( k z ) ( G ) (136)
If the stochastic dielectric is sufficiently weak, then the Wigner tensor within
the stochastic dielectric will also approximately be given by the single element
W,, . This suggests that a self-consistent approximation could well be ob-
tained by neglecting all terms in Eq. (134) for whichj # i. It follows that the
Wigner tensor for x > 0 will also be approximated by
(k - 2 +
ar c2 at
4) ( Wii(r, t ; k, w ) )
-(-
= - 71 o2-
2 c2
k i ) ( Wii(r,t ; k, 0))
6(k‘ - Io‘I/c)
x SdL’ do’($ - ki”)P(lk - k‘l, o - 0‘)
o’lc
o2
+ -271 6(k - Iwl/c) (7
@/C
- k f ) ’ S dk’ dw’P(1k - k’l, o - o‘)
x ( Wii(r, t ; k’, w ’ ) ) - (c2/4w2) Im S z (137)

If we finally take ki = 0 and k! = 0, which implies that the wave-number
vectors in the direction of the primary field y have a negligible effect on
-
(Wii), we are left with the following equations when we take, possibly,
o’ o = ck, if o occurs in a special factor without other terms:
71k4
= - -(Wii(r, t ; k, o>>
2
x s dk’ dW’P(1k - k’l, w - 0’)

6(k’ - Io’l/c)
WlfC
+ $zk36(k - ~o(/c)
1
x dk’dw‘P(1k - k’l, o - w’)( Wii(r,t ; k’, w’)) - ?Im St (138)
4k
Our final result for the approximate stochastic transport equation for the
ensemble average of the Wigner function is similar in form to the stochastic
transport equation previously proposed by Bugnolo (1960a,b). This previous
result was defined for the related function, the power spectral density S i i ,
which in our present notation may be written as
Sii(r, t ; k7 W)
1
dr‘ dt’ exp[i(k.r’ - wt’)](ET(r,t)Ei(r + r’, t + t’)) (139)
In order to illustrate this similarity we must first define a new function F

such that F and W are related by
Wii(r,t ; k, w) = 6(k - lwl/c)Fii(r,t ; k, w ) ( 140)
Substituting Eq. (140) into Eq. (138), we obtain the following stochastic
transport equation for the related function F :
nk4_
- -_
X
(Fii(r, t ; k, 0))
2
6(k’ - IW’ ~ / C )
+ z k2 3
~
s1 dk’dw’P(lK(,o - w’)
dk‘ d~’P(lK1,w - w‘)S

of/c
x (Fii(r, t ; k‘, 0’)) - (1/4k2)Im SiT (141)

where
SiT = 6(k - ~w~/c)S;
This approximate form for the stochastic transport equation may be com-
pared to that of Bugnolo [(1960a,b), Eq. (36)]. Let us first consider the
differences. Our present result includes a time dependence on the LHS and
a source term on the RHS. These results are new. The two integral terms may
be reduced to those proposed previously by noting that the presence of the
6 function within the integrals reduces the integration over k’ to an integra-
tion over dn’k. Although a similarity in mathematical form does indeed
exist, we propose that our present result is an extension of the previous one
in that the integrations over k' are now well defined and indicate that the
wave number k of the incident wave will be affected in both direction and
magnitude. This will be particularly the case when we must deal with strong
fluctuations of the dielectric.
B. Monochromatic Waves in an Isotropic Time-Invariant

Stochastic Dielectric
This is another special case with a possible hint of an number of interest-
ing applications. We may consider the stochastic dielectric to be time in-
variant if the time constant associated with its change z is very long compared
to the time constants associated with the wave itself. For this special case we
may substitute in all of our equations
Ei(r, t) = exp(-jq,t)Ei(r) (142)
Since the Wigner function is defined by
(Wik(r, t ; k, 0))
dr' dt' exp[i(k * r' - cot')]
x (ET (r + f.t + i)E, (r - f.t ;)- (143)
1
it follows that we may reduce this by way of the following steps:
(Wik) = dr' dt' exp[i(k- r' - at')] exp[io,(t + t'/2)1

exp[ - io,(t - t'/2)]E,
1
s
( Wik)= 16.n4 dt' exp[i(o, - o)t'] dr'
s
x exp(ik * r') (ET (r + g)E, (r - g)
(Wik) = 6(o - ad( Wik(r, k))
We also require a reduced form for the P function of Eq. (38), since for this
case &(a, T) + hik(a).A procedure similar to the above yields
Pik(k, O) = b(w)Pik(lk1)6F (145)
and likewise for the source term we have

PO
S; = i - o o 6 ( o - w 0 )
2A3 s dlexp(t-h)
where we have also taken the time dependence of the source current dis-
tribution to be
Ji(r, t) = e x d - icoot)I;(r) (147)
The equivalent of Eq. (127) for this case may also be obtained from Eq.
(120). Taking i = k, we have
a
k * - ( Wii(r, k))
dr
d5 exp(ik * 5)(E J 7 )
I1
Using our usual arguments to minimize the effects of the blurred reso-
nances we may reduce Eq. (148) to a more useful result by removing the
term in Im( W), hence,
n
x sdk’P(1k - k’I)S(k’- k,)(kt - ki2) + -S(k - k,)
2kO
x
j
(6jk; - kikj)’
i dk’P(1k - k‘l)( Wjj(r, k’))
_ _ _ _1
8n3~,oO s
(bjkg - kikj)Re dh exp(ik * 5)( E J f )
Again we argue that a self-consistant equation may be obtained for the case
(149)
when the stochastic medium is sufficiently weak by neglecting all terms in

Eq. (149) for whichj # i. This yields

a
k * - ( Wii(r, k))
ar
J
71
- - _ dk‘P(1k - k’1)6(ko - k’)(ki - k12)
( k i - k:)( Wii(r, k))
2kO
+ ~
2kO
71
6(k - k,)(ki -
s(
kZ)2 dk‘P(1k - k’l)( Wii(r,k’))
dh exp(ik.1) Ei (r - ’;>IT (r + i)) (150)
If we once again take ki = 0 and kf = 0, we are left with Eq. (151):
-
- -~
nki
2 ( Wii(r, k)) s dk’P(1k - k‘1)6(ko - k‘)
rcki
+ 26(k - k,) s dk’P(1k - k’l)( Wii(r, k’))
With the exception of the source term, Eq. (151) has previously been
obtained by Howe (1973) by a completely different method, starting with a
very general Lagrangian density. In mmparing his result with ours, we
must take
6(k2 - k;) = 6(k - k,)/2ko (152)
On the other hand, a derivation based on the wave equation has previously
been presented by Barabanekov et al. (1971) for a transport equation similar
to Eq. (151).
We again note that our result as given by Eq. (151) may be transformed
by defining a function F such that
Wii(r, k) = 6(k - ko)Fii(r,k) (153)
Substituting this into Eq. (151), we obtain the result, omitting the source term:
a
k . - (Fii(r, k))
ar
= - 9
71k3
(Fii(r, k))
s dk’P(IK1)6(ko- k‘)
+2
2 71k3 s dk‘P(IKI)G(K - ko)(Fii(r, k’)) (154)
STOCHASTIC DIELECTRlC WITH COMPUTER SIMULATION 345
We may compare Eq. (154) to Howe’s (1973) result particularly his Eq.
(5.25). Our result is similar to his if we set k = k , in his Eq. (5.25) and use the
definition for the 6(k2 - k;). Setting k = k , is, in our opinion, proper in the
case of weak dielectric fluctuations. However, this is not the case for strong
fluctuations of the dielectric in that the wave number of the incident wave
will be affected in both magnitude and direction by the fluctuations. Again,
we also note that setting k = ko in Eq. (154) results in a stochastic transport
equation which is similar to that first proposed previously by Bugnolo
(1960a,b).
By this we seek to indicate that Eq. (154) in its present form does include
the effects of the medium on both the magnitude and direction of the vector
k, whereas the previous result obtained some time ago was limited to the
effects of the medium on the direction of the vector k alone and thus con-
sequently limited to the case of weak fluctuations in the medium.
C. Monochromatic Waves in an Anisotropic Stochastic Dielectric

By following a procedure similar to that of Section VI,B above, we may
obtain a self-consistent transport equation for the ensemble average of the
Wigner function. We find that this may be written as
-
-
xki
--
2 <Wii(r,k)) J dk’dw’P(k - k’, o - co’)6(k0 - k‘)
+ nk3
2
6(k - k,)
f dk’ do’P(k - k‘, o - o’)(Wii(r, k’, co’))
This concludes our discussion of special cases. We shall return to these later
(155)
in this work.
VII. A BRIEFREVIEW METHODS

OF OTHERTHEORETICAL
A number of other theoretical methods have been proposed for dealing

with the problem of strong fluctuations of the field in the multiple-scattering
region. In brief these are the diagram method, the integral equation method,
and the parabolic equation method. Of these three, the parabolic equation
method has shown the most progress and popularity to date. The interested
reader is referred to Ishimaru (1978) for an introduction to the latter method
as well as to the extensive list of references. It is not our intention here to
duplicate this effort.
346 D. S . BUGNOLO AND H. BREMMER
More recently, the method of Feynman’s path integrals has been applied
to the solution of the parabolic equation for scattering by a stochastic
dielectric (Dashen, 1979).This is probably one of the more advanced works
to be published as of this date.
The parabolic method contains four physical assumptions (Ishimaru,
1978). The first of these, which we shall label assumption 1, is contained in
the approximation used to formulate the parabolic equation. For a wave
propagating in the x direction, we must take
Assumption I
Ik(dv/&c)l >> ld2u/8x21 as long as 1, >> il (156)
(see Ishimaru, 1978, Eq. 20-6b, p. 408). In effect this is a high-frequency con-
dition similar to condition (i) of Dashen (1979).
The second physical assumption is similar to our own, namely, that the
stochastic dielectric must be characterized by a Gaussian random process
at any space point. We shall label this assumption 2, which is given in
Ishimaru [1978, Eq. (20.8)] and is used later in a conditional manner by
Dashen (1979, Section 2):
Assumption 2
{p} is Gaussian or kL(p’)“’ << 1 (157)
The third physical assumption of the parabolic method (assumption 3) is
that the stochastic dielectric is 6 correlated in the direction of propagation
of the wave, which in the case of our example is the x direction:
Assumption 3
<EI(X? P)&’, P’)) = &- X ’ M P - P’) (158)
[See Eq. (20.9) of Ishimaru (1978) and Eq. (1.2) of Dashen (1979).]
Dashen’s condition (ii) is really a fourth assumption and is similar to
our condition on the velocity of propagation of a disturbance within the
stochastic dielectric.
Assumption 4
His kL << T is similar to our Vdist= l,/t << c (159)
It is not our purpose here to criticize the parabolic method. Each method
has its advantages and disadvantages, and the reader is certainly encouraged
to read the literature, which for the case of the parabolic method is extensive.
In the Soviet literature, this is often referred to as the Markov approximation.
Our primary purpose here is to indicate that the parabolic method and our
method for the Wigner transport equations do not differ a great deal in
their physical assumptions.
Dashen’s criterion (iii), that the RMS multiple-scattering angle, i.e.,
KP2)(WW’* << 1 ( 160)
is small, differs from any of our physical assumptions; however, the method
used to solve the resulting transport equations of our method is certainly
affected by this condition. Yet the stochastic transport equations can be
used when the RMS scattering angle is not small, as is the case in ordinary
transport theory.
Of greater interest is Dashen’s alternative for the Gaussian assumption
for I. This alternative is that kL( p 2 ) 1/2 is small. This is the equivalent of our
condition on the stochastic MFP dB as given by our Eq. (13). In view of this,
we may write this condition as
dB >> 1, (161)
This concludes our discussion of the physical assumptions of the parabolic
method. We now return to the principal purpose of this work.
VIII. THECOHERENT FUNCTION

WIGNER
Of all the field quantities that may be calculated by any of the existing
theories, the coherent part of the field is the most straightforward. Physically,
we here define the coherent part of the Wigner function as that part of the
original coherent incident wave remaining after interaction with the sto-
chastic dielectric. In this section we shall address this from the point of view
of our stochastic transport equation for the ensemble-averaged Wigner
function. We shall also compare our results to the results of some of the
other methods.
In keeping with our interest in the problem of multiple scattering, we
limit our discussions to path lengths greater than one mean-free part in the
medium; i.e., greater then our dB, as defined by Eqs. (13), (123), and (124).
For a discussion of the coherent field at short distances, the reader is referred
to the early work of Wheelon (1959).
Our most general result for the coherent part of the Wigner function
may be obtained from Eq. (120)by neglecting that part of the result obtained
by way of the term R,. Following this procedure, we obtain
o2
= C ( F Sj - kikj dk’ do’Pjl(k- k’, o - 0‘)
J
This result contains the effect of the blurred resonances and all of the cross-
polarization terms. Equation (162) must be used where the dielectric is
anisotropic or where the fluctuations of the dielectric are strong. We now
proceed to a simple case.
Let us assume that the dielectric fluctuations are sufficiently weak such
that we may use our self-consistent approximation and that the effects of
the blurred resonances are removed by our usual arguments. For this case,
Eq. (162) reduces to the following:
a nk3
k . - ( Wii(r,k ) ) = - 2( Wii(r,k))
ar 2 s dk’P(IK1)G(ko- k’)
with IKI = Ik’ - kl. The above could also have been obtained by way of
(163)
Eq. (151). This follows from a consideration of the details of the second term
on the RHS of Eq. (151), which is the term obtained by way of R,.
A First Approximation for the Coherent Wigner Function

We may proceed with our solution for the coherent part of the Wigner
function by obtaining a solution of Eq. (163). The integral on the RHS may
be written as follows:
The spacewise spectrum of the dielectric fluctuations for this case may be
written as
4n dR exp(iK.R)(c(r)E(r + R)) (165)

Historically, our function P ( K ) was first defined as S(K), where these are
related by way of
P(IK1) = (8n3)-’S(K) (166)
This should be noted in comparing our results to those presented previously
by Wheelon (1959) for the case of a single scattered short path, or by Bugnolo
(1960a,b) for the case of multiple scattering on a long path.
The vector length IKI in the above is related to the vectors k and k by
way of IKI = Ik’ - kl. We shall approximate the length of this vector by
K E 2k sin 38 (167)
It is evident that by so relating K to k, we have neglected magnitude changes
in the wave number k resulting from the effects of the stochastic dielectric.
Since we are dealing here with our Eq. (151), which is valid in the time-
invariant case, we may neglect all doppler effects as well as all shifts in k due
to the internal decay of the turbulence; however, by neglecting the effects
of the dielectric on the magnitude, we have indeed made an approximation.
It is self-evident that we have included the effects of the stochastic dielectric
on the direction of k.
The differential scattering cross section per unit volume per unit solid
angle may be obtained by way of a first Born approximation as discussed
previously. This is here related to our P ( K ) by
o ( K ) = *(nki)P(K) (168)
We put Eq. (164) into a more familiar form by noting that
dR’ = sin 8’ do‘ dcp‘ (169)
Integrating Eq. (164) over k‘ and using Eqs. (168) and (169), we obtain
k .-dra (Wii(r,k)) = - ko( Wii(r,k))

f
dQ’a(K)
We next define a first approximation for the total stochastic cross section
for scattering by the stochastic dielectric in a manner similar to that first
introduced by Bugnolo (1960a,b), namely,
Q,(ko) =
f
with K = 2k0 sin @‘, so as to obtain the result
dR’o(K)
FIG. 5. Geometry for a plane wave normally incident on the stochastic dielectric half
space.
Before proceding we note that we have here three equations for the three
diagonal components of the coherent Wigner function. To this approxi-
mation, the equations are independent of each other. In view of this we may
search for a solution for any given component without any loss of generality.
Let use consider the case of a monochromatic plane wave normally
incident on a stochastic half space as illustrated by Fig. 5. It is evident that
the only component of the Wigner function that does not vanish at the
boundary is given by Eq. (136), which we repeat here for convenience:
W,,(O>t ; k, 0)= 6(0 - o,)W, - k x ) 6 ( k , ) W , ) ( G ) (136)
We propose that the solution of Eq. (172) for this geometry and boundary
condition is
(wY,(x,ko)) = wyy(O, t ; k, o)exp[:-Q,(ko)x] (173)
We suggest that this is the correct self-consistent result for the coherent
component of the Wigner function to a first approximation in that we have
neglected the effects of the dielectric fluctuations on the magnitude of k
(hence the use of k , in our result). [A detailed derivation of Eq. (173) may be
found in Bugnolo (1983).]
We may specialize our result so as to make numerical evaluation possible
by beginning with Eq. (108) for P ( K ) and by using the definition for Q,(k,)
as given by Eq. (171). Hence we obtain
This may be integrated directly without difficulty:
2n
k 0 -- - (175)
10
We note that within the constraints of our first approximation and our
model for P ( K ) , Eq. (175) is exact and may be used to determine Q,(ko) for
any wavelength.
There are two models which we believe to be of particular interest. These
are the exponential model for the spacewise spectrum and the Kolmogoroff
model. The exponential model may be obtained from the above by taking
the parameter p equal to 1/2, whereas we obtain the result for the
Kolmogoroff model by taking p equal to 1/3 in Eq. (175). For the exponential
case we may write
whereas for the Kolmogoroff case, we may write

Q,(ko) = A ( 2 ~ ~ ~ l o / i l ~1) - +
( ~[I~ ) { 4(2do/&)2]-5’6} (177)
with A = 0.7468342005. In obtaining this value for A we have used the
numerical data for the r function given in the Handbook of Mathematical
Functions (Abramowitz and Stegun, 1964). These tables are given to 11
significant figures and should be adequate for our present task.
Our numerical examples of Eqs. (176) and (177) have been plotted as
Figs. 6 and 7. At very high frequencies we may neglect the factor within the
brackets. At low frequencies this factor, which is labeled 2 in Figs. 6b and
7b, may be used to define a correction factor which is presented as a function
of the ratio lo/l0. From these results we may conclude that the high-fre-
quency approximation may be used with little error when the ratio Zo/lo 2 10.
Later in this work we present the results of computer simulations of the
stochastic transport equation in wavelength ranges of 1 cm to 1 mm, with
lo values of 10 to 1 m and ( E ’ ) from 1 x lo-’’ to 1 x lo-”. In Figs. 6a
and 7a we have plotted the function exp[ - Q,(ko)R] for the exponential and
Kolmogoroff space-correlation functions. We have done this for four differ-
ent intensities of the turbulent fluctuations, ranging from weak to very
strong. We note from these results that the coherent component will be
larger in an atmosphere discribable by way of a Kolmogoroff spectrum for
the dielectric fluctuations. Although this is evident from the equations, Figs.
6 and 7 provide us with examples of these differences at a wavelength of 3 mm.
t
Range (km)
(a)
FIG.6 . (a) Plot of exp( - Q,R)for exponential case: A,, = 3 mm, .I = 10 m. (b) Low-frequency correction factor,
exponential correlation: Q, = [ ( 2 d o ) / ~ ~ ] ( & (m-').
*)z
Curve Strength (EZ) MFP (km)

A Weak 1 x 10-12 45.59
B Mildly strong 2 x lo-'* 22.195
C Stronger 5 x 10-12 9.118
D Very strong 1 x lo-" 4.559
1 .I
O.!
0.:
0.
t
0.0
0.0
0.c *
0.02 0.1 1.o 10
Ickfko
b)
FIG.7. (a) Plot of exp(-Q,R) for Kolmogoroff case: 1, = 3 mm; 1, = 10 m. (b) Low-frequency correction
factor, Kolmogoroff correlation: Q, = 2.344[(2n1,,)/A~](e2)z(m-').
Curve Strength (4 MFP (km)
A Weak 1 x 10-'2 61.6

B Mildly strong 2 x lo-'' 30.55
C Stronger 5 x 12.22
D Very strong 1 x lo-" 6.11
354 D . S. BUGNOLO AND H . BREMMER
This concludes our discussion of the first approximation for the total
stochastic cross section for scattering by an isotropic dielectric contained in
the half space defined by x > 0. Any further work on this subject must
address the problems inherent in the solution of the more complete Eq.
(162). Suggestionsfor further work in this field are contained in Section IX,F.
Ix. COMPUTER SIMULATION OF THE STOCHASTIC TRANSPORT

EQUATION
FOR THE WIGNER FUNCTION IN A TIME-INVARIANT
STOCHASTIC DIELECTRIC
In order to better understand the results of Section VI, we propose to

simulate the stochastic transport equation for the Wigner function for a
monochromatic wave in a time-invariant stochastic dielectric. Such a simu-
lation is made possible by the large-scale mainframe computers of today
and by the availability of fast integration subroutines. We hope that this will
prove to be a much better method of solution than that proposed some years
ago by Bugnolo (1960a,b).
We begin by recalling Eqs. (153) and (154) for the special case under
consideration:
a
k . - (Fii(r,k))
dr
= -
nk3
(Fii(r,k))
s dk’P((KI)G(k,- k’)
dk‘P(IK1)G(k’- ko)(Fii(r,k ) ) (154)
with
(Wii(r9k)) = d(k - b)(Fii(r, k)) (153)
We propose to simulate Eq. (154) under the following conditions:
(1) The source is monochromatic and consists of an antenna array such
that the initial wave is launched within a narrow beam of width BETA,
normally incident on the stochastic half space.
(2) The time-invariant stochastic dielectric is everywhere homogeneous
and isotropic within the half space and characterized by way of the spacewise
spectrum P(IK1) and weak fluctuations.
(3) The solution of Eq. (154)is observed by an ideal device that is
designed to collect all components of (Fii) or ( Wii) that are contained
within the beam of angle BETA. All other components are rejected.
Of these three conditions, the third is probably the most restrictive. However,
conversion of our results to a more practical receiver is not a simple matter,
because this would depend on the details of the receiving array.
A. An Integral Equation for a Very Narrow Beam

We next address the problem of finding an integral equation equivalent
to Eq. (154)for the geometry of Fig. 8. We begin by noting that the field
launched by the source is initially contained within the cone of angle BETA
and is characterized by an electric field vector E , in the R direction, propa-
gating with the wave number k , in the 9 direction such that k , = q,/c. We
observe that the first term on the RHS of Eq. (154) is just that considered
previously in connection with our discussion of the coherent wave. As we
again address the case of weak fluctuations of the dielectric, we expect that
our integral equation will contain terms of the type given previously by
Eq. (173), as applied to the geometry of Fig. 5.
The launched wave will again have the approximate form
F X X ( 0 , k,) = - ~ o ) W x ) W , ) ~-( k,)E%
~, ( 178)
We note that this wave is deterministic and is a reasonable approximation
for the case of a very narrow beam. A wide beam would require all three
initial tensor components.
With these conditions we write Eq. (154)as follows:
We next address the problem of the integration of Eq. (179) for the
geometry of Fig. 8. We again invoke our very narrow beam condition so as
to permit the approximation (R- r ( r R - r, thus placing all distance
measurements on the z axis. We must also consider the boundary condition at
z = 0 in the (x, y ) plane. Here our source is an “antenna” of effective aperture
size A, radiating a uniform beam of width BETA into the stochastic medium
X
A
FIG.8. Narrow-beam geometry: [ R - r’l z R - r‘.

z > 0. In order to simplify our notation, we have placed the extensive

mathematical details into our model for the function (F,,(O, k , ) ) such that
the integral over the first term in our solution is well behaved as R + 0. For
the details in a somewhat similar case, see Bugnolo (1960b). We therefore
obtain as our approximate result for a very narrow beam
+ $ JoR ~XPC
dr'r'2 - Qs(ko)(R - 1'11
16z2(R - r')'
x J dk'P(IKl)6(k' - ko)<Fxx(r',k')) (180)
We now proceed with the integration over k'. Let this integral be denoted
by CD where
s
CD = *(zkg) dk'P(IKl)G(k' - ko)(Fxx(r',k'))
dk' = 'k d k sin 8' do' d4'
(181)
Integrating over k' by way of the 6 function yields
0 = $(zk:)
s dQLfYIK/)(f'xx(r',ko)) (182)
where d o ; = sin 8' do' d4'. Using Eq. (180) and our definition for the differ-
ential cross section per unit volume per unit solid angle o, as well as Eq.
(182), it follows that the integral equation may be written as
(183)
For the stochastic differential cross section o ( K ) ,we shall use the models
Once again two models are of interest. The exponential space correlation for
the stochastic dielectric, which may be obtained from the above by taking
the parameter p = 1/2, and the Kolmogoroff space correlation function,

which may be obtained from the above by taking p = 113. We shall obtain
numerical results for these two cases for an effective aperture A,, = 1 m.
B. An Algorithm for the Computer Simulation of the Stochastic

Integral Equation for the Case of a Very Narrow Beam
Before entering into a discussion of our computer algorithm or method
for programming the computer, it is necessary to call the reader’s attention
to a number of interesting and important properties of our integral equation
for the Wigner function. First, we note that the integrals on the RHS contain
the solution itself in the form of (F,,(r’, ko)). Because of this, the integral
equation cannot be integrated directly because (Fx,.(r’, k,)) is unknown for
all r > 0. At the boundary r = 0 or z = 0, F,, must take its proper value at
the boundary, that form given by Eq. (178). We must therefore begin our
simulation at the boundary z = 0 where F,, is known and proceed into the
stochastic half space step by step, evaluating the RHS of Eq. (183) at each step.
Any proposed method of numerical integration requires that the volume
V illustrated by Fig. 10 be subdivided into a number of sections. This requires
that we subdivide the cone of angle BETA, into, say, M sections, as illustrated
by Fig. 9. However, we note that Eq. (183) is no ordinary integral equation,
but a stochastic integral equation in the large. By this we mean to emphasize
the conditions used to obtain a model for the differential cross section per
unit volume per unit solid angle o ( K ) .The extent of the volume covered by
the integrals used to define this function must be of extent L such that
L >> lo, the mean scale size of the dielectric fluctuations. We note in passing
that this problem does not exist in the problem of multiple scattering of a
wave by very small particles.
From the previous discussion, it follows that we may write
1, << L = RIM (185)
On the other hand, the extent of each section must be bounded in the large
such that the wave within any given section does not undergo multiple
scattering with any degree of probability. We therefore propose to bound
FIG.9. The cone of volume V partitioned into M sections of linear extent RIM.
the linear extent of each section, denoted by the length L, by the condition
1, << L << Qs-'(ko) = dB1 (186)
where, once again, dB is the stochastic MFP in the right half space z 2 0.
We next address the details of our simulation. Let us consider the ith
section. The function (Fxx(r',k,)) within the integral on the RHS of Eq.
(183) must be estimated by way of a numerical simulation. We propose to
do so by summing the contributions from all previous sections, i.e., from
m = 0 to i - 1. In summing these incoherent contributions, we scatter by
way of the empty space characterized by the dielectric constant z0. These
contributions add directly because they are statistically independent by
virtue of their separation in space. (We recall that the extent of each section
must be large compared to lo, the mean scale size.) To the sum of the in-
coherent components we propose to add the coherent component, calculated
by way of Eq. (173).
The first integration on the RHS of Eq. (183) is over the variable dQ;.
Let us consider this integration by way of the coordinate point at the center
of the ith section. We recall that dQL = sin 8' dd' d4'. It is self-evident that the
angle 4' varies from 0 to 2n in this geometry. On the other hand, the range
for the angle 8' is not as obvious. A number of possibilities suggested by the
integral are to bound 8' by the angle to the receiving device or by the initial
angle of the beam BETA/2 or by, say, the 10- or 20-dB points in the forward-
scattering angle of o(K).We propose to use both the 20-dB point and the
angle of the beam BETA/2 as our choice for what will be denoted by the angle
B in our programs. The details of the conditions for the angles in the case of
the 10- or 20-dB-point bounds will depend on the spectrum used for the
dielectric fluctuations. For the case of the exponential spectrum we obtain
8' = 3i1,/4nlO 10-dB point (187)
8' = 3L0/2n10 20-dB point (188)
whereas for the case of a Kolmogoroff space correlation function, we obtain
8' = 3.17L0/4w10 10-dB point (189)
8' = 3.366~,/2d0 20-dB point (190)
A study of the integral equation (184) will indicate to the reader that the
effects of our choice for the angle B will be frequency dependent. At low
frequencies, where the angle of scattering may be large, the 20-dB point
could easily exceed the angle BETA/2. On the other hand, at very high
frequencies, BETA/2 will exceed the 20-dB angle for any antenna beam
widths of practical interest. In view of this, we expect that our results will be
of greater practical significance. However, we note that any reduction of
our numerical results in practice must consider the effects of the receiving
antenna.
The last integration in the RHS of Eq. (184) is over r'. This has the effect
of taking the results of the integration over dQ; section by section and
transferring the results of each section to the end of the path. We propose
to approximate this part of the integration by multiplying the results of the
integration over dokby the volume of the ith section V(i) and the function
~XPC-QAR- Y - W2)I
16nZ(R- y - H/2)2
This result must be summed over all sections by way of integrations (DO
loops in FORTRAN).
There remains the problem of the first section. We begin by noting that
the length of each section L is small compared with the stochastic MFP d,.
We propose to estimate the magnitude of the incoherent component in
section zero (see Fig. 11) by way of a first Born approximation. The magnitude
of the coherent field is to be estimated in the usual manner. In all coherent
field estimations, we use the center of each section as our point of calculation.
It is our expressed hope that the above-detailed discussions of our method
of approximation for all components of the integrals will enable our readers
to follow the computer programs. Before proceeding to the first program,
we list all computer variables in alphabetical order.
C . Variables Used in the FORTRAN Programs

B The angle for integration over 8' in the subroutine INT(IS, B)
BETA The angle of the narrow beam
CH The coherent field at the end of the path
DELTA (4
FREE The free-space attenuation from the source to the end of the
path: 1/16n2R2
F(N) A subscripted variable used to evaluate the double integral
WN) The coherent field in the Nth section
WN) The incoherent field in the Nth section
LAMBDA The wavelength in meters
LO The mean scale size in meters
MFP The stochastic mean free path in meters: MFP = dB
RANGE The range in meters from the source to the path end
QS The total stochastic cross section in meters-'
The programs contain some other notation as well, but that notation should
be obvious after a study of Section IX,B and the programs.
D. Experimental Program Number One

Our first program was written in an effort to understand how the number
of sections M , used to evaluate the double integrals, affects the magnitude
of the Wigner function at the end of the path. In view of the condition given
by Eq. (186), we expect that our results for the Wigner function will depend
on the number of sections. For small M , we may violate the condition
RIM << d,, whereas for large M , we may violate the condition RIM >> I , .
Intermediate between these two bounds, we expect to obtain a result that
approximates the correct solution for each simulation.
Although we have performed numerous computer simulations in order
to gain an understanding of these effects, the results are always similar. We
next present a listing of the FORTRAN program for Experiment Number
One. Note that the program contains a number of WRITE commands that
have been c'd out. These may be used to trace the program should the user
wish to do so. The program was written using double-precision Real *8 in
an effort to minimize roundoff error in summing the incoherent components.
The integration subroutine was written using available subroutines on the
VAX-11 obtained from the IMSL (1979). Subroutine DCARDE: numerical
integration of a function using cautious adaptive Romberg extrapolation
was used. We note that although our program was written using the FUNC
above, any other model may be used for P ( K ) , including future models
based on more complete and rigorous theories of turbulence.
I N P L I C I T KEALtH(A-2)
C EXPERIUENTAL PROGRAU NUUBER 1
C YRITTEN BY DR. D.S. BUGNOLO
INTEGER NrPrLpUrS
DIUENSION 1 1 ~ 0 ~ 1 0 0 0 ~ r I Z ~ O ~ 1 0 0 0 ~ r I % S ~ O ~ 1 0 0 0 ~ r V ~ O ~ 1 0 0 0
COUUON /FUNARG/ PIrPHvOELiA-LAUBDA
TYPE Sr'EXPERIUENTAL PROGRAM NUNI3f.R 1:'
TYPE tr' NAWHUW B E A M ALGORITHM'
TYPE tr' KOLUOGOROFF CORRELAIION'
TYPE LOr'ENTER RANGE'
ACCEPl t r R
TYPE I O r ' E N T E R BETA'
ACCEY i * I BETA
TYPE I O r ' E N T E R UAX U '
ACCEPT t r M A X
TYPE 1 0 r ' E N T E R M I N N'
ACCEPT S r M I N
TYPF I O r ' E N T E R STEP SlZE I N M '
ACCEPT * * S T E P
TYPE 10s'CNTER ( E P S l L U N t t 2 ) '
ACCEPi **DELTA
PI~3*1415926535897932
TYPE 1 0 r ' E N T E R UAUELCNCilH I N METERS'
ACCEPT trLANBDA
TYPE 1 0 r ' E N T E R UEAN SCALE S I Z E '
ACCEPT t r L O
PH=(2SPISLO/LAMBUA)
U R I T E ( 5 r S ) 'EXPERINENIAL PROGRAM NUMBER ONE'
URITE(5rt) ' NAKROU BEAM ALGOHITHH'

URITC(5rt) ' KOLHUGORUFF CORNFLAllON'
URI'TE(Sv I)'RANGE.: ' t R
URIlE(5rX) 'BETA= ' , B E T A
URITE(5rt) 'EPSILON**S= ',DELTA
URITE(Sr8) 'LAHBDAr 'rLAHBDA *
URITE(Srt) 'LO= 'rLO

BEPI
C4LL INTl(1SrB)
0S=IS
URITE(5rt) 'OS= ' 9 0 6
URITE(5rt) 'HFPs 'rl/RS
C
C ANGLE FOR INCOHERENT FOWUAKU SCATTER
B=BETA/2
URITE(5tt) 'BETA OF ci= 'PB
CALL INTl(KSt8)
G=IS
YRITF.C5rIl 'G= '90
DO 500 S=HINrHAXrSTEY
URITE(5t40) S
H=R/S
I=O
u=s-1
DO 400 N l O r U
C URITE (5rt) 'N= ' t N
Y=N*H
C
C CALCULATE THE COHERFNT COHPClNhNT
C
C
C
C
C
IF
100
C
150
C
200
250
300
C
C
C
C
C
400 CONTINUE
C
C PRINT THE RESULTS
C
CH=(EXP(-UStR))/((4tPI*~)$*~)
WRITE ( S p Y I ) 'COHERENT COMPUNENT='rCH
U R l T E ( 5 v X ) 'INCOHERENI CUMf'UNENT~'v1
SUM-ItCH
U R I T E ( 5 9 8 ) 'SUM='vSUH
FREE=1/(14tPI*R)**2)
*
WR1TE ( S 9 ) 'FREESPACE=: ' r FREE
ATTEN=lOXLOGlO (SUM/FREE)
U R I T E ( S v X ) 'ATTENUATION I N OB RELATXVE TO
1 FREESPACE UIGNER I N T E N S I T Y = ' r A T T E N
500 CONTINUE
STOP
10 FORMAT ( ' S ' ? A ? ' ' )
40 FORHAT(/,' n= ' 9 13 )
50 FORHAT(' ' v A 2 r 1 3 v A Z r l P E 2 4 . 1 6 )
END
SUXKOUTINE I N T l ( 1 S v L )
I M P L I C I T REhLtB(A-Z)
E X T E RNAL FUNC
INTEGER IERR
COMMON /FUNARG/ PIvPHvUELTAvLAMHUA
ACCl=lE-S
ACCZ=O
LOW=O
HIGH-8
IS=llCADRE( FUNCILOW r H I G H ?A C C l PACCZIEKKOR~ IERK)
I F ( I E R H .NE. 0 ) THEN
U R I l E ( 5 v X ) 'ERROR NUMBER ' v I E R R
WRITE(J,YI) 'IS= ' r ~ ~
U R I T E ( 5 v t ) 'ERROR BOUND I S ',ERROR
END I F
RETURN
END
FUNCTION FUNC(THETA)
I M P L I C I T REALXB(A-2)
COMHON /FUNARG/ P I ~ P H ~ D E L T A I L A M B I M
A=.6217870772
FUNC=(AtZ. $PI/LAMBIIA) t ( P H S X 3 )*IIEL'I A * ( DSIN( 1HETA) )/
1 ((1~+4.S(PH1$2)$(~SIN(THETA/2.))1$2)X1(11~/6~))
RETURN
END
1. Results of a Computer Simulation for the Case of a KolmogorofS

Space Correlation Function
This program has been run at many different wavelengths, but we have
elected to present some typical results for 3 and 1 mm. These are within the
range of wavelengths used later in this work, namely, 1 cm to 1 mm. We have
used two values for the mean scale size lo, i.e., 10 and 1 m, whereas ( E ~ )
ranged from to This range is typical of weak to very strong
fluctuations of the dielectric in the earth's troposphere. Very strong fluctu-
ations may occur as a result of some natural or unnatural phenomena. For
our propagation range in these examples, we have used the interesting case
of 10 km.
Our numerical results for these eight sample computer runs have been
plotted as Fig. 10a and b. It is essential that these results be interpreted with
an understanding of the approximations. We begin by noting that the con-
dition as given by Eq. (186) must be satisfied. However, these bounds are
“weak,” in the sense that << or >> are “weak.” We therefore expect that our
results for the attenuation of the Wigner function relative to its free-space
value (computer by way of 1/167c2R2)will be affected by the number of
sections M or by RIM, the actual physical length of each section. A study of
our examples indicates that this is indeed the case here.
Consider for example our Fig. 10a for the case where the wavelength is
3 mm, the mean scale size I , is 10 m,and ( E ’ ) ranges from lo-’’ to lo-”.
Our “weak” bounds for these cases become
10 m << 10,00O/M << MFP (192)
where the MFPs are given for each run in the Fig. 10a. In order to understand
our proposed procedure, some reference must be made to the character of
Eqs. (183) and (184).The final integration is over the variable r’. It is over this
integration that the parameter M is used to subdivide the integration over
the volume of the cone of angle BETA. As is the case for ordinary numerical
integration, we expect the numerical result to approach its correct value as
the number of subdivisions increases. However, since this is a stochastic
problem we are also faced with the weak upper bound of Eq. (191). We have
used the double-integration subroutines in the IMSL (1979) to study the
effects of a finite volume of integration on our estimates for the differential
scattering cross section o(K)and have found that this effect is weak for the
ranges in question. By this we mean that an integration volume of ten mean
scale sizes in extent will not greatly affect the numerical value.
In view of the weak bounds on the parameter M and the fact that all
subsequent runs of our programs will take I, = 10 and 1 m, we have elected
to vary the parameter M in all runs such that the LENGTH variable of
Experimental Program Two is always equal to 100 m. This will ensure that
the length of any section will always be very small compared to the ap-
plicable MFP, whereas the lower weak bound is still satisfied, even for the
cases when I , = 10 m.
It should be noted that this choice for the variable LENGTH is not to
be taken lightly, since the final values obtained for the attenuation will be
weakly dependent on this choice for propagation ranges of 20 km or less.
This very weak dependence is apparent from the study of our two samples,
Fig. 10a and b. This conclusion is based on the results of over 100 com-
puter runs; however, we should like to note at this time that our choice of
-2O [ -
I
A
L t
0 10 20 30 40 50 60 70 80 90 100
Number of sections, M
(a)
-m -2
O r
-
- A
B
C
-
-0
5-4-
+-
3
D
E -6 -
2
I I
0 10 20 30 40 50 60 70 80 90 100
Number of sections, M
(b)
FIG.10. Variation in attenuation of the Wigner function relative to the free-space Wigner
function as a function of the number of sections M (Kolmogoroff case):
MFP (km)
A 10-12 61.1 67.896

B 2 x lo-'' 30.553 33.948
C 5 x 10-12 12.221 13.576
D 1 x lo-" 6.11 6.789
I, = 10 m ; I,, = 3 m m ;range, 10 km.

I, = 1 m ; I , = 1 m ;range, 10 km.
LENGTH equal to 100 m does lead to difficulties at very long ranges of, say,
100 km for reasons that are related to other assumptions inherent in our
narrow-beam algorithm. These effects at very long ranges are discussed in
Section IX,F of this work since that section is devoted to problems worthy
of future study.
In conclusion, our extensive numerical study of the validity of our
narrow-beam algorithm by way of over 100 computer runs indicates that
our choice for the new variable LENGTH equal to 100 m is good for propa-
gation ranges of 20 km or less in the earth‘s atmosphere.
E . Experimental Program Number Two

This is the principal program for our computer simulations of Eq. (183).
In this program we seek to develop an understanding of how the Wigner
function varies with the range of the propagation path for the case of a very
narrow beam. We have used the same algorithm as that used for Program
Number One, modifying the first program so as to input our new variable
LENGTH. This variable is used within the program to fix the parameter
M , given any particular range, by way of the Eq. (193):
M = S = JIDNNT(R/LENGTH) (193)
This takes the integer value of the division of the two double-precision
numbers.
The variable LENGTH will be taken as 100 m in all subsequent runs,
with the exception of some further work reported in Section IX,F. This is
justified once again by the fact that the mean scale size 1, will always be
taken equal to or less than 10m. It follows that the MFP, d g , must be at
least 1000 m if this choice for LENGTH is to be justified by way of Eq. (186).
This is the case for all of our sample runs in this section.
We again present a listing for the case of a Kolmogoroff spacewise
spectrum for the dielectric fluctuations. This is given by the variable FUNC
in the subroutine. However, we note that our program has been written so
as to call an integration subroutine of great “power” and that it may be used
for any other spectrum in the future. Suggestions for future work are con-
tained in Section IX,F of this work.
We again choose the angle B equal to half the antenna beamwidth. This
will ensure that our result includes all scattered components that remain
within the main beam of half-angle BETA/%.
Returning to the program itself, we should like to call the reader’s
attention to the PRINT THE RESULTS section, beginning at the bottom of
p. 367. It should be noted that this section of the program contains both
TYPE and WRITE commands. When running this program, the TYPE
commands are printed to the screen, while all WRITE commands are printed
to a file on the disk. In this way, we may observe the results of a program
run on the screen and delay any decision to print a hard copy on the printer
until the run is complete. The program also contains a number of WRITE
commands that have been c'd out. These may be used to TRACE the pro-
gram. A listing of Experimental Program Number Two is now presented
(a listing of this program for the case of an exponential space correlation
function may be found in Appendix 1).
c ALC1,F F O R I N C O H E R E N T FORNARD S C A T T E R
C
C CNEh
C
c
t
r
100
r
150
200
250
300
C
c
c
C
400
c
c
c
1. Results of the Computer-Range Simulations for the Case of

an Exponential Space Correlation Function
We have performed a number of simulations using the exponential space
correlation function for the stochastic dielectric because this correlation
function has received a great deal of attention in the literature. Our Experi-
mental Program Number Two is based on our narrow-beam algorithm. In
1 *
0 10 20 30 40
Range (km)
(a)
a
-20 -
1
MFP (km)
Fig. Curve A Curve B 1. 10

part ( E ~ )= 1 x 1 0 - l ~ (2)= 1 x lo-'' (m) (mm)
(a) 506.563 50.662 10 1 (cm)

(b) 45.594 4.559 10 3
(c) 455.934 45.593 1 3
(4 5.066 2.533" 10 1
(e) 50.661 10.132* 1 1
a (EZ) = 2 x 10-12.
(EZ) = 5 x 10-12.
m
x
.U.
._
6 -1
4-
W
C
3
c
a
-2 -
t
0 10 20 30 40
Range (km)
(C)
-25 I
0 2 4 6 8 10 12 14 16 18 20
*
Range (km)
(d)
I I I *
0 2 4 6 8 10 12 14 16 18 20
Range (km)
(e)
FIG.11. (continued)
view of this we have taken the beamwidth BETA equal to 0.2 rad for all
simulations. This corresponds to a beamwidth of 11.4591559", a value easily
obtained by antennas in this frequency range, which ranged from 1 cm to
1 mm. For these simulations we let the mean scale size I , range from 10 to
1 m, whereas ( E ' ) ranged from 1 x to 1 x 10-l'. The above condi-
tions are typical of those to be found in the earths troposphere. A ( E ~ )of
1 x lo-'' corresponds to the case of a weak turbulence, whereas a ( E ~ )of
1 x lo-'' may be used as an example of strong turbulence. We may expect
to find strong turbulence in regions where the normal atmosphere is disturbed
by some natural or unnatural phenomena.
It is of importance to note that our simulations do not include the effects
of molecular absorption. Since this does occur in the earth's atmosphere in
the frequency range used for our simulations, this effect must be included in
any practical application. This was not included at this time because we are
primarily interested in studying the effects of the stochastic dielectric on the
propagation path. We note that the effects of molecular absorption may
easily be added to our program should the need to do so arise. The absorption
will affect both the coherent and incoherent components of the Wigner
function.
Our numerical results for the exponential correlation function have been
plotted as Fig. 1la-e. We have plotted the attenuation of the Wigner function
relative to the free-space Wigner intensity in decibels. The latter is calculated
by way of 1/16x2R2.From a study of the results, it is self-evident that the
attenuation increases with frequency and the mean scale size 1,. We have also
included our numerical result for the MFP dB. We shall not dwell on this
case further because our principal results are the computer simulations for
the Kolmogoroff space correlation function.
2. Results of the Computer-Range Simulations for the Case of

a KolmogoroffSpace Correlation Function
We have performed a number of simulations using the Kolmogorof
space correlation function for the stochastic dielectric since this function i:
a useful model for cases where the turbulence causing the dielectric fluctu
ations is fully developed. This model for o ( K ) may be obtained from ou
Eq. (184) by taking the parameter p equal to 1/3. We note again that ou
model does not include the effects of dissipation because the wave-numbe
dependence in this region is still open to question.
Since our Experimental Program Number Two is based on our narrow
beam algorithm, we have taken the beamwidth equal to 0.2 rad for all ou
simulations. This corresponds to a beamwidth of 11.4591559",a value easil
obtained by antennas in this frequency range. Once again the frequenc
372 D . S. BUGNOLO AND H.BREMMER
-1 -
-
m
E 0 4 8 12 16 20 24 28 32 36 40
.-cE (a)
a
a2
c
3 0
-5 -
-10 -
I
0 2 4
I
6 8
a
10 12
I
14 16 18 20
*
Range (km)
(b)
FIG. 12. Attenuation of the Wigner function relative to the free-space Wigner intensity
(Kolmogoroff case) for BETA = 0.2 rad; p = 1/3; B = BETA/2; length 100 m:
Fig. Curve A Curve B 10 20

part (E2) = ( 2 ) = 1 x lo-" (m) (mm)
~~
(4 678.973 67.899 10 1 (a)

(4 61.107 6.110 10 3
(c) 61 1 61.1 1 3
(4 6.789 3.395" 10 I
(e) 67.899 13.579b 1 1
~ ~~
(&2) = 2 x 1o-I2.
* (&*) =5 x 10-12.
0 10 20 30 40
Range (km)
(C)
I w
0 2 4 6 8 10 12 14 16 18 20
Range (km)
(d)
Range (km)
(el
374 D. S. BUGNOLO AND H . BREMMER
range of the simulations will be bounded by 1 cm and 1 mm. We have taken

the mean scale size 1, = 10 and 1 m, whereas ( E ' ) ranged from 1 x to
1 x lo-". The above conditions are typical of those found in the earth's
troposphere. A ( e 2 ) of 1 x corresponds to the case ofweak turbulence,
whereas a (c2) of 1 x lo-' may be used as an example of strong turbulence.
We may expect to find strong turbulence in regions where the normal
atmosphere is disturbed by some natural or unnatural phenomena.
It is important to note that our simulations do not include the effects of
molecular absorption. Since this does occur in the earth's atmosphere in the
frequency range used for our simulations, these effects must be included in
any practical application. This was not included at this time because we are
primarily interested in studying the effects of the stochastic dielectric on the
propagation path. We note that the effects of molecular absorption may
easily be added to our simulation program should the need to do so arise.
The absorption will affect both the coherent and incoherent components of
the Wigner function.
We have performed many simulations in the frequency range of 1 cm to
1 mm. From these we have selected ten range runs which have been plotted.
These may be used to study the effects of propagation range on the attenu-
ation of the Wigner function (see Fig. 12a-e). In Fig. 12a-e we have included
our result for the MFP d, in kilometers. From a study of these results and
those of other computer simulations, we propose to conclude the following
with regard to the attenuation of the Wigner function relative to its free-
space value of 1/16z2R2for propagation paths of 20 km or less.
(1) Weak turbulence: (c2) = 1 x 10-l2. For 1, = 1 m (Fig. 13a), the
attenuation will be less than 1 dB in the wavelength range of 1 mm or greater.
For 1, = 10 m (Fig. 13b), the attenuation can be appreciable at wavelengths
of about 2 mm or less.
(2) Mildly strong turbulence: (c2) = 2 x For 1, = 1 m (Fig.
13c), the attenuation will be less than about 2 dB in the frequency range of
interest; wavelengths of greater than 1 mm. For 1, = 10 m (Fig. 13d), the
attenuation will be appreciable and exceed 3 dB at wavelengths of less than
3 mm.
(3) Stronger turbulence: (c2) = 5 x For 1, = 1 m (Fig. 13e),
the attenuation will be less than 3 dB for wavelengths greater than about
1 mm. For 1, = 10 m (Fig. 13f),the attenuation will be appreciable and exceed
3 dB at wavelengths of less than about 5 mm at a range of only 10 km.
(4) Verystrong turbulence: (c2) = 1 x lo-". For 1, = 1 m(Fig. 13a),
the attenuation will be appreciable and exceed 3 dB at wavelengths of less
than 4 mm. For 1, = 10 m (Fig. 13h), this example is the most severe of those
(a1
-1.2 I
-
m
9
1.o 2 3 5
Wavetength (km)
10 1a 2 3 5
Wavelength (mm)
10
C
.-c
a
p
(dl
1 .o 2 3 5 10 *
Wavelength (mml Wavelength (mml
FIG.13. Attenuation as a function of wavelength for frequency ranges of 10 and 20 km and

5 and 10 km [see pp. 376 and 377 for parts (e)-(h)]:
v
1 .o 10
Wavelength (km)
(el
Wavelength (mm)
(f)
FIG.13. (continued)
1 .o 2 3 5 10
Wavelength (mm)
(9)
0-
-5-
-10 -
-m
-
XI
C
._
g -15-
3
C
c
2
FIG.13. (continued)
included in this work. The attenuation will exceed 3 dB at wavelengths of

about 5 mm or less. The attenuation at 2 mm will exceed 10 dB at 20 km.
From the results of our simulation it is possible to conclude that attenu-
ation due to scattering by a time-invariant stochastic dielectric is of some
consequence in the wavelength range of 1 cm to 1 mm. At some wavelengths
this can easily exceed the attenuation due to molecular absorption, which
has not been included in this work. For example, at a wavelength of 1 mm,
the molecular absorption will result in an attenuation of about 1.0 dB/km, or
10 dB in a path of 10 km. This is exceeded by the attenuation due to scattering
when the turbulence is only mildly strong (Fig. 13d) or stronger (Fig. 13f-h).
This concludes our discussion of the results of the simulations by way of our
experimental programs.
F. Conclusions and Suggestions for Future Work

In this Section IX we have presented our results for the computer simu-
lation of our Eq. (184) in an effort to prove that computer simulation may
be used to provide us with a solution of the stochastic transport equation.
In order to proceed we have written an algorithm for the special case of a
very narrow beam launched into a time-invariant isotropic stochastic di-
electric free of other sources. The results of numerous simulations, some of
which were presented in this work, indicate that computer simulation is
indeed feasible on a VAX-11 computer system, given the availability of the
system CPU time required for the runs. This amounts to some 2 min per
range run. Our results indicate that the attenuation due to scattering is of
some consequence in the wavelength range of 1 cm to 1 mm.
Our simulations have been limited to those considered by way of our
narrow-beam algorithm and any improvement would require a much larger
amount of computational time. It is evident that there is ample opportunity
for further work in this area. We should like to suggest the following as
worthy of further study by way of the large digital computer.
In this work, we have limited our simulations to an approximate form of
the stochastic transport equation, namely, that given by our Eqs. (1 53) and
(154), for the case of a time-invariant isotropic stochastic dielectric. At very
long ranges, or in regions of very strong scattering, we expect that it may be
necessary to consider Eq. (149). This equation includes the effects of the
interactions between the various components of the Wigner tensor. We note
that although our simulations are valid at ranges of about 20 km or less, the
assumption that the LENGTH variable may always be taken as equal to
100 m can lead to difficulties at very large ranges. In Figs. 14a-f we have
presented the results of simulations to very large ranges at a wavelength of
3 mm, a mean scale size of 10 m, a beamwidth of 0.2 rad, and a Kolmogoroff
spectrum with mean squared fluctuation levels of 1 x 10-l2, 5 x 10-l2,
and 1 x lo-". In order to demonstrate the effects of changes in the length
parameter on attenuation, we have performed all simulations for a LENGTH
of 100, 150, and 200 m. From an inspection of Fig. 14a,c, and e, it is evident
that changes in the LENGTH parameter do not have a large effect on the
attenuation for ranges of 20 km or less; however, the effects can be ap-
preciable at large ranges. In these figures, we have also included the coherent
component so that the reader may study the differences between the magni-
tude of the coherent and incoherent components of the normalized Wigner
function.
We attribute the apparent failure of our narrow-beam algorithm at very
large ranges to our neglect of two effects. As the range increases, the width
of the beam becomes appreciable and requires partitioning. At large ranges
the coupling between the various components of the Wigner tensor may not
be negligible. This is worthy of further study; however, the amount of com-
putational time will increase significantly.
Another interesting case is that given by our Eq. (155) for the case of a
time-invariant but anisotropic dielectric. Since anisotropic dielectrics are
to be found in nature, the solution of this case could be of some practical
interest. However, we have reason to believe that an equation of the form of
Eq. (149) must be used, since the anisotropic dielectric will scatter energy
into other Wigner components.
We again note that our computer programs have been written so as to
call integration subroutines for the integrals involving o(K),i.e., P ( K ) , the
spacewise spectrum of the dielectric fluctuations. Another problem of interest
is that of the effects of the dissipation region of the dielectric fluctuations on
the results of our simulations. These effects will most likely be of significance
at optical frequenciesand, perhaps, to a lesser degree in the millimeter region.
This, of course, will depend on the wave number of the eddies characterizing
the dissipation region (see Fig. 1). In the millimeter range, we expect to find
an effect when the dielectric fluctuations are strong, or perhaps when the
turbulence is not fully developed, but strong. Such fluctuations may be ex-
pected to occur when the environment is stressed in times of conflict or by
some natural phenomenon. This concludes our discussion of the possibilities
for future work by way of the large digital computer. Perhaps the reader can
easily formulate plans for some additional work of practical as well as
theoretical significance.
1 .o
LENGTH
0.5 -
03
0.05 -
0.03-
0 20 40 60 80 100 120 140

la)
c
a
-2.5
O
h
-
-5- LENGTH
-7.5 -
-10 -
-12.5 -
-15 - 10 20 30 40 50 60
(C)
FIG.14. (a, c, e) Effects of variation in length parameter on the attenuation; (b) normalized
coherent and incoherent components; (d) effects of variation in length parameter on the mag-
nitude of the incoherent component; ( f ) effects of variation in length parameter on the magni-
tude of the coherent and incoherent components: p = 1/3; BETA = 0.2 rad; 1, = 3 mm; I, =
10 m.
1 x lo-'* 61,106 5 x 10-12 12,221

(b)" 1x 61,107 (km) 1 x 6,110
(c) 5 x 10-12 12,221 (f) 1 x lo-" 6,110
a fIiol6= 1.607 x rad; Kolmogoroff case.

0 10 20 40 50
\I
L ~ G T H
100 rn
-Y LENGTH
Range (km)
(f)
382 D . S. BUGNOLO AND H . BREMMER
X. CONCLUSIONS
In this work we have attempted to present the derivations of our principal

results [Eqs. (119) and (120)] for the stochastic transport equations for the
Wigner function, in a manner that may be followed by the reader. These
equations are obtained for the most general case of a time-variable aniso-
tropic stochastic dielectric which may contain sources. Equation (119) is a
transport equation for the real part of the Wigner function, whereas Eq. (120)
is our result for the imaginary part. At present, we are of the opinion that the
Im( Wik)part is of less importance than the Re( Wik) part. We have also
obtained an integral equation of the second kind for the Wigner function
in a general stochastic dielectric [Eq. (121)].
In addition, we have tried to obtain equations for the Wigner tensor
function in a number of interesting special cases, namely, that of a plane
wave incident on an isotropic dielectric [Eq. (141)] and that of a mono-
chromatic plane wave in an isotropic time-invariant dielectric [Eq. (151)].
We have compared our fundamental physical assumptions with those of the
parabolic equation method and found that there are many simularities, as
well as some differences.
In order to obtain some estimate of the magnitude of the effects of scat-
tering as predicted by our transport equations, we have developed an algo-
rithm and written a number of programs in the FORTRAN language. These
have been written for the special case of a very narrow beam normally
incident on a half space containing an isotropic time-invariant stochastic
dielectric. In our effrot to model the stochastic dielectric we have used a
function based on the exponential model and one based on the Kolmogoroff
model for the spacewise correlation function of the stochastic dielectric. We
have presented the results of many computer runs for the, hopefully, interest-
ing wavelength range of 1 cm to 1 mm, taking (&’) ranging from the case of
weak turbulence, i.e., (2)= 1 x lo-”, to the case of very strong turbulence,
i.e., ( 6 ’ ) = 1 x 10-l1.The latter could occur in an atmosphere stressed by
some natural or unnatural phenomenon. In our simulations, we have taken
the mean scale size of the dielectric fluctuations 1, (sometimes referred to as
the outer scale size in the literature) as either 10 or 1 m This range is expected
to cover any typical value found in the earth’s troposphere. We have per-
formed the majority of our simulations for propagation ranges equal to or
less than 20 km.
As a result of these simulations for the cases under study, we have con-
cluded that multiple scattering is significant in the wavelength range of 1 cm
to 1 mm, and that this may exceed the attenuation due to molecular absorp-
tion at the shorter wavelengths. However, we also note that our algorithm,
based as it is on a single component of the Wigner tensor, is probably not
very useful at very large ranges such as, say, 100 km. This is due to our
neglect of the other components of the Wigner tensor and to the width of the
narrow beam itself at very large ranges.
We have also made a number of suggestions for future work which we
hope to pursue given the support that such extensive computer simulations
would require. We note that the CPU time requirements for the simulation
of stochastic transport equations can be very large, depending on the ap-
proximations used in writing the algorithm and the number of Wigner com-
ponents required for the simulation. It is our hope that this work will be of
some value to the scientific and engineering communities.
APPENDIX 1 . A LISTINGOF EXPERIMENTAL

PROGRAM
NUMBERTwo
FOR THE CASE SPACECORRELATION
OF AN EXPONENTIAL FUNCTION
TMPLTCTT k E A L * 8 t A - Z )
C F I P F R l M F N f A L PROGRAM NUMBFR T h o
C M R T T T E N a y DR. D.s.. BUGNOLC
TNTFGVR N,P,L,H,RH,RN
~IMFNSTI)~ 11(o:1000),12r0:1000)rr2S~o:lOoo~,v~o:looO~,
I Frn:1nOO)
C m w m / F U N A R C / PI,PH,DELTA,LAMBDA
T Y P F *,*LXPEI?I*EWTAL PRnGRAM NUMRFR THO'
TYPF *,' NARRr)W REAN A I . G O R I T H M '
TYPF *,' EXPOWENTIAL CO RRE LA TI ON '
T Y P F 10,'LNTCR YAXIMUM RANGE'
n r r F P T *,HHAX
TYPF IO.'ENTER M INIM UM R A N G E '
ACCFFT * r R Y I N
T V P F 10,'ENTER NUMBER OF STEPS I N RANGE'
PCPFPT *,HN
I)Fl,74RtfRMAX-RPIIN)/RN
TYPF in,'CNTER BFTA'
ACPFPT *,BETA
TYPF I0,'ENTFR (EPSILON**?)'
bCCCPT *,DELTA
P1=~.141502h5)5X979J2
T V P C i n . * m ~ wU~AVELENCTHN I CIETFRS'
ACrFPT *,LAMRnA
TVPV 10,'F.NTFD MEAN SCALE SIZE'
DCPFPT * , L O
'rvpF i n , * E N T F R SECTION LENGTH'
ACPFFT * , L E W G T t {
p H = f 2* P T *LO/LAMBDA )
W P T T F(5,* ) ' E X P CHIM E Y TA L PROGRAM NUMRER T N O'
HRlTFr5.*) ' NARROW PPAM ALGORITHM'
WRTTFlS,*) *EXPOVENTIAL CO RRE LAT ION '
WQTTF(5,*) ' R MAX='+RMAX
WRTTFf5,*) ' R YIN=',PMIN
WQlTF(S,*) 'NO OF S TE P S IL RANGE=',RN
*
W ~ T T F ( 5 , 1 ' RFTA Z' ,@ETA
WWTTFf5,*) 'FPSILON**Zr *.GELTA
IVFTTF(5, * 1 'l,Af+RnA=' ,LPHBOA
WRTTF(S,*) 'LO= '.LO
! d P T TF(S .* ) 'LENCTH=',LENGTH
r
c
C
C
C
C
c
c
TF
100
c
150
C
200
250
300
C
C
C
C
C
STOCHASTIC DIELECTRIC WITH COMPUTER SlMULATION 385
400
C
C
C
son
1v
40
50
C
C
2. A SAMPLE
APPENDIX OF A COMPUTER
SIMULATION
I1G Nk H I NI EN SI 7 Y =
REFERENCES
Abramowitz, M., and Stegun, 1. A., eds. (1964). “Handbook of Mathematical Functions,”
Appl. Math. Ser. 55, June. Natl. Bur. Std., Washington, D.C.
Barabanenkov, Yu. N., Vigorradov, A. G., Kravtsov, Yu. A,, and Tatarskii, V. I. (1971). Sou.
Phys.- Usp. (Engl. Transl.) 13,551.
Bartelt, H. O., Brenner, K. H., and Lohmann, A. W. (1980). The Wigner distribution and its
optical production. Opt. Commm. 32,32-38.
Bastianns, M. J. (1978). Wigner distribution function applied to optical signals and systems.
Opt. Commun. 25,26-30.
Bastianns, M. J. (1979a). Transport equations for the Wigner distribution function. Opt. Acta
26,1265-1272.
Bastianns, M. J. (1979b). Transport equations for the Wigner distribution function in an in-
homogeneous and dispersive medium. Opt. Acta 26, 1333- 1344.
Bremmer, H. (1964). Semi-geometrical-optical approaches to scattering phenomena. M . R. L.
Symp. Proc. 14,415-437.
Bremmer, H. (1974). Propagation through inhomogeneous and stochastic media. AGARD
Conf. Proc. 144 (March).
Bremmer, H. (1979). The Wigner distribution function and transport equations in radiation
problems. J. Appl. Sci. Eng. A 3,251-160.
Bugnolo, D. S. (1960a). Transport equation for the spectral density of a multiple scattered
electromagnetic field. J. Appl. Phys. 31, 1176-1 182.
Bugnolo, D. S . (1960b). Dissertation, Columbia Univ., School of Engineering and Applied
Science, New York.
Bugnolo, D. S. (1960~).On the question of multiple scattering in the troposphere. J . Geophys.
Res. 65,879-884.
Bugnolo, D. S. (1972a). Phys. Rev. A 6,477-484.
Bugnolo, D. S. (1972b). Turbulence in weakly ionoized plasma. J . Plasma Phys. 8, 143-158.
Bugnolo, D. S. (1978). Multiple scattering of pulsar-signals by turbulent interstellar plasma.
Sou. Astron. (Engl. Transl.) 22(1), 38-42. [Astron. Zh. 5569-75, Jan.-Feb. (1978).]
Bugnolo, D. S. (1983). The coherent electromagnetic field in a stochastic dielectric. Trans.
IEEE Southeastcon, Orlando 1983. (In press.)
Chandrasekhar, S. (1 950). “Radiative Transfer.” Oxford Univ. Press, London and New York.
Dashen, R. (1979). Path integrals for waves in random media. J. Math. Phys. (NY)20,894-920.
Frisch, A. (1968). In “Probabilistic Methods in Applied Mathematics” (A. T. Bharucha-Reid,
ed.), Vol. 1. Academic Press, New York.
Gelford, I. M., and Formin, S. V. (1963). “Calculus of Variations.” Prentice-Hall, New York.
Howe, M. S. (1973). O n the kinetic theory of wave propagation in random media. Philos.
Trans. R . SOC.London 274% 523-549.
IMSL (1979). “FORTRAN Subroutines in the Areas of Mathematics and Statistics.” Inter-
national Mathematical & Statistical Libraries, Inc., Houston, Texas.
Ishimaru, A. (1978). “Wave Propagation and Scattering in Random Media,” Vols. 1 and 2.
Academic Press, New York.
Mori, H., Oppenheim, J., and Ross, J. (1962). Some topics in quantum statistics: The Wigner
function and transport theory. In “Studies in Statistical Mechanics” (J. de Boer and G. E.
Uhlenbeck, eds.), Vol. 1, pp. 262-298. North-Holland Publ., Amsterdam.
Norton, K. (1960). “Carrier Frequency Dependence of the Basic Transmission Loss in Tropo-
spheric Forward-Scatter Propagation.” Mem. Rept. PM-83-21, Natl. Bur. Stand., Wash-
ington, D.C.
Novikov, E. A. (1965). Functionals and random-force methods in turbulence theory. Sou.

Phys.-JETP (Engl. Transl.) 20, 1290-1294.
Reickl, L. E. (1960). “A Modern Course in Statistical Physics.” Univ. of Texas Press, Austin,
Texas.
Tatarskii, V. J. (1961). “Wave Propagation in Turbulent Media.”
Tatarskii, V. J. (1971). “The Effects of the Turbulent Atmosphere on Wave Propagation.”
Natl. Tech. Inf. Serv., Springfield, Virginia.
Whelon, A. D. (1959). Radio-wave scattering by tropospheric irregularities. J . Res. Nutl. Bur.
Stand. Sect. D-Radio Propagation 63D.
Whittaker, E. T., and Watson, G. N . (1952). “Modern Analysis.” Cambridge Univ. Press,
London and New York.
Wigner, E. (1932). On the quantum corrections for thermodynamic equalibrium. Phys. Reu.
40, 749.
Author Index
Numbers in parentheses are reference numbers and indicate that an author’s work is
referred to although his name is not mentioned in the text. Numbers in italics indicate the
pages on which the complete references are given.
A Ashley, K. L.,134, 135,143

Aszodi, G., 171, 296
Abdalla, M. I., 125, 142
Abraham-lbrahim, S., 188,289
Abrahams, M. S., 127, 142
B
Abramowitz, M., 14,59, 351,388
Achete, C., 289, 297 Bachelet, G. B., 143
Acket, G . A,. 131,142 Bacon, F., 265,297
Adler, I., 195, 298 Badawi, M. H., 116, 143
Adler, J., 195, 298 Bahraman, A,, I 1 3, 143
Adler, S. L., 174, 180, 289 Baittinger, W. E., 195, 200, 273, 281, 282,293
Adler, R., 5,61 Baldereschi, A., 144
Agaev, A. M., 106, 107, 147 Baliga, B. J., 123, 132, 143
Ahn, J., 288,289 Bambynek, W., 193, 195,290
Akita, K., 125, 142 Baneji, S. K., 264,290,295
Akkerman, Z. L., 126, 142 Barabanenkov, Yu. N., 344,388
Aksela, H., 195,289 Bartelt, H. O., 301, 388
Aksela, S., 195,289, 298 Bastianns, M. J., 301, 303,388
Aksela, U., 195, 298 Bastow, B. D., 149
Aleksandrova, G. A., 126, 130,142 Battye, F. L., 171, 290
Alferov, Zh, I., 131, 142 Baudoing, R., 186,290
Allen, G. A,, 101, 142 Bauer, E., 176, 184, 186, 290
Allen, J. W., 142 Bazhenov, V. K., 102, 107,143
Allred, W., 109, 121, 142, 157 Beal, S. W., 143
Allred, W. P., 113, 119, 159 Bebb, H. B., 126,159
Amato, M. A., 94,143 Bekmuratov, M. F., 103,143
Amy, J. W., 273, 281, 293 Benbow, R. L., 297
Andersen, V. E., 181, 185,292,296,298 Beni, G., 172, 18 I , 290,294
Anderson, C. L., 116,143 Benninghoven, A., 267,271,28 I , 290
Andrews, A. M., 115,143 Berg, N. J., 82, 90, 143, 152
Andrianov, D. G., 106, 107,143 Berger, R. A., 83, 149
Antell, G. R., 121, 143 Bergmann, Ya. V., 132,143
Antoniewicz, P. R., 290 Bernett, M. K., 196, 212, 290
Aoki, K., 116, 143 Berz, F., 133, 143
Aratama, M., 243,293 Bethe, H. A., 176, 179,290
Arthur, J. R., 128, 145 Betz, G., 257, 266, 273, 276, 290
Arthurs, A. M., 179,290 Bhattacharya, P. K., 75,80,136,137,143,152
Asaad, W. N., 193,290 Biermann, R., 126, 155
Ashen, D. J., 116, 117, 143 Bimberg, D., 121, 144
Ashirov, T. K., 112, 143 Bindell, J. R.,271, 290
Ashley, C. A,, 172, 290 Bindi, R.,223,290
Ashley, J. C., 176, 184, 290,296, 298 Birkhoff, R. D., 181, 185,242,292,296, 298
391
392 AUTHOR INDEX
Bishop, H. E., 225, 237, 238,286, 290, 291 Chakravarti, A. N., 132, I45
Bishop, S. G., 92, 116,144 Chambers, A., 171,293
Biswas, S. N., 132, 144 Chan, P. C. H., 132, 145
Blackmore, G. W., 127, 144 Chandrasekhar, S., 329, 388
Blatte, M., 113, 114, 144 Chang, C. C., 130,145
Blakemore, J. S.,77,91, 106, 144 Chang, C. G., 163,291
Blashku, A. I., 128, 144 Chang, L. L., 69, 145
Bludau, W., 132, 144 Chantre, A,, 81, 113, 145
Blurn, J. M., 126,154 Chapman, R. A,, 104, I45
Boborykina, E. N., 112, 144 Chattarji, D., 162, 193,291
Bois, D., 81, 127, 144, 145, 158 Chelikowski, J. R., 196, 197, 291
Boltaks, B. I., 104, 105, 110, 113, 144 Chen, M. H., 195,291,298
Bonfig, K. W., 2,59 Cheng, L. J., 72, 145
Bonham, R. A., 241,290 Chi, J.-Y., 133, 145
Borisova, L. A., 127, 144 Chiang, S. Y., 68, 145
Bouquet, S.,261, 290 Chiao, S.-H. S., 11 1, 145
Bourdin, J. P., 181, 185, 291 Cho, A. Y., 100, 128,145,154
Bourgoin, J., 75, 83, 85, 88,155 Chollet, L., 267, 298
Bouwman, R., 163,256,257,264,281,290,298 Chou, N. J., 279,291
Brauer, W., 223, 297 Christman, S. B., 196, 198, 291
Braun, P., 257, 266,290 Chu, W. K., 256, 274,276,278,291,293
Brebbia, C., 18,59 Chung, M.S., 223,291
Brehm, G. E., 84,144 Chwang, R., 15, 36,59
Bremmer, H., 301, 304, 305, 307, 388 Cini, M., 207, 291
Brenner, K. H., 301,388 Citrin, P. M., 196, 198,291
Briant, C. L., 264,265,266,290,295,297 Clark, D. J., 91,159
Brillson, L. J., 75, 145 Clerjaud, B., 92, 93, 145
Brown, I. C., 18, 59 Coad, J. P., 286, 291
Brown, M., 114,155 Coates, R., 86, 88, 145
Brown, W. J., 106, 145 Coburn, J. W., 288,289
Brown, W.L., 272,294 Cohen, P. I., 247,248,291,292
Brozel, M. R., 120,145 Colby, J. W., 271,290
Bube, R. H., 94, 126, 158 Collins, R. A,, 170,291
Bublik, V. T., 67, 145 Comas, J., 128, 145
Buczek, D., 281,290 Condon, E. U., 192,205,291
Bugnolo, D. S., 301, 306, 307, 308, 327, 331, Conley, D. K., 248, 271, 288, 290,292
336,337,341,345,349,350,354,356,388 Conway, E. J., 86, 158
Bui Minh DUC,237,239,252,290 Cook, Jr., C. F., 271,291
Bulman, W. E., 2,59 Cooke, C. J., 114,145
Burhop, E. H. S., 162, 179, 193,290,291 Corelli, J. C., 84, 154
Bykovskii, V. A., 105,145 Courant, R., 52,60
Crasemann, B., 193, 195, 291,296, 298
Crawford, J. H., Jr., 113, 114, 146
C
Cronin, G. R., 106, 120, 148
Cailler, M., 181, 185,291,292 Crotty, J. M., 193,291
Callaway, J., 181, 291 Crowell. C., 15, 36,59
Caroli, B.,188,289
Caroli, C., 188, 289
D
Carriere, B., 213, 288, 291
Casey, H. C., Jr., 127, 131, 134, 145 Da Cunha Belo, M., 261,291
AUTHOR INDEX 393
Damestani, A., 107, 146 Elliott, C. R., 91, 146

Das, B. K., 149 Elliott, K. R., 86, 146
Dashen, R., 346,388 El Toukhy, A. W., 277,297
Davidson, R. S., 3, 60 Emerson, L. C., 181, 185, 292,298
Davies, D. E., 91, 146 Emori, H., 127, 146
Davis. H. L., 182, 296 Engelsberg, S.,292
Davis, L. E., 192, 256, 291,293,295 Ennen, H., 106, 107, 146, 130
Dawson, P. T., 288,291 Epifanov, M. S., 133, 147
De Dominicis, C. T., 207, 295 Erickson, N. E., 244,296
Defrance, J. E., 163,291 Erlewein, J., 270, 293
Degen, P. L., 91, 128, 146 Ettenberg, M., 135, 136, 147
Degreve, F., 271,294 Euthymiou, P. C., 82, 132, 150,155
Delord, J. F., 287,291 Evans, S., 171, 292
De Mey, G., 5, 11, 12, 13, 18, 21, 23, 39, 49, Everhart, T. E., 223, 242,243, 291,297
57, 59,60,61 Ewan, J., 90, I47
Demidov, E. S.,105, 146
Den Boer, M. L., 247,248,291
Dench, W. A., 170, 183,260,297
Desjonqueres, M. C., 209,298 F
Deveaud, B., 93, 125, 146 Fain, Jr., J. C., 287, 291
de Visschere, P., 18, 60 Falk, D. S., 175, 292
Dieball, J. W., 286, 293 Fan, J. C. C., 275,292
DiLorenzo, J. V., 127, I46 Farmer, J. W., 82, 86, 88, 147
Distefano, T. H., 196, 292 Farrell, H. H., 297
Dobson, P. A., 72, 128,146 Faulconnier, P., 287,295
Dobson, P. S., 113, 127,149 Favennec, P. N., 91, 125,147
Doniach, S., 172,290 Feibelman, P. J., 172, 182, 207, 212, 285, 292
Donnelly, J. P., 91, 146 Ferrel, R. A,, 296
Dove, D. B., 288,295,296 Fiermans, L., 200,293
Drawin, H. W., 180, 183,255,291 Fink, R. W., 193, 195,290
Driscoll, C. M. H., 75, 146 Finstad, T. G . , 123, 147
Driscoll, T. J., 172,296 Fisher, B., 196,292
Drowley, C., 133, I48 Fisher, D. G., 267,298
Ducastelle, F., 209, 298 Fistul’, V. I., 105, 106, 107, 147
Dufour, G., 192, 195,291, 295 Fitting, H. J., 173, 292
Duke, C. B., 172,291 Flat, A., 133, 147
Dunlap, H. L., 116, 143 Fontaine, J. M., 282, 286, 292
Duraud, J. P., 245,282, 286, 292 Forbes, L., 107, 146
Dzhafarov, T. D., 112,146 Forgacs, G., 171,296
Formin, S. V., 305,388
E Fowler, W. B., 197, 298
Fox, D. C., 271,291
Eaves, L., 92,93, I46 Fransen, F., 18,60
Eberhardt, J. E., 131, 156 Frederick, P. J., 287, 292
Edmond, J. T., 120, 146 Freund, H. U., 193, 195,290
Edwards, T. W., 18,60 Friedel, J., 200, 292
Einstein, P. A,, 282,298 Frisch, A., 327,388
Einstein, T. L., 248, 291, 292, 294 Fry, J. L., 181, 292
Eisen, F. H., 73, 91, 128, 146 Fuggle, J. C., 21 1,294
Elam, W. T., 248,291,292 Fujimoto, M., 116, I47
394 AUTHOR INDEX
Fukuda, Y., 248,291,292 Haase, S., 155

Fuller, C. S., 108, 147 Haeusler, J., 2, 11, 33, 34, 60, 61
Haff, P. K., 275,292
G Haisty, R. W., 106, 120, I 4 8
Hall, D. D., 267, 298
Gaarenstroom, S. W., 165,195,200,213,262, Hall, J. T., 248, 249, 288,292
282,292,293 Hall, P. M., 166, 248, 249, 257, 259, 292
Gallon, T. E., 171,255,292, 293 Hall, R. N., 132, 148
Gamo, K., 128, 147 Hamilton, B., 141, 148
Ganachaud, J.P., 176, 181, 184,185,186,241, Hammer, C., 297
291,292 Hansma, P. K., 248,292
Ganapol’skii, E. M., 105, 147 Harris, L. A,, 246, 248, 292
Garber, E. W., 242,296 Harrison, Jr., D. E., 272, 292
Garbuzov, D. Z., 131, 147 Hasegawa, F., 75, 110, 148, 156
Garrison, B. J., 272, 292 Hasegawa, H., 106, 148
Gatos, H. C., 133, 145 Hashiba, M., 257, 273, 278, 298
Gelford, I. M., 305, 388 Hashimoto, H., 225,279, 297
Gerlach, R. L., 164,247,292 Hatch, C. B., 126, 148
Gersten, J. I., 182, 292 Hattori, T., 171, 292
Ghandhi, S. K., 123, 143 Heavens, 0.S., 288,291
Ghosh, S., 9,60 Hedin, L., 176,292
Gibbon, C. F., 123, 147 Helms, C. R.,271,291
Gibbons, J. R., 91, I56 Hemment, P. L. F., 91,148
Gillet, M., 219, 264, 292 Hennel, A. M., 92,94,148
Gippius, A. A., 92, 130, 147, 158 Henrich, V. E., 275,292
Glaefeke, H., 173,292 Henry, C. H., 133, 139, I48
Glinchuk, K. D., 92, I47 Hiesinger, P., 115, 148
Goldstein, Y., 4,61 Hilbert, D., 52,60
Gomer, R., 284,295 Hilsum, C., I48
Goodman, A. M., 132,147 Hink, W., 183,293
Goodwin, A. R., 70,147 Hinterman, H. E., 267, 298
Goto, K., 256, 264, 279, 292 Hjalmarson, H. P., 103, 148
Could, R. D., 170, 291 Ho, P. S., 256, 213, 276, 278, 293
Gourlay, R. D.. 3, 60 Hobgood, H. M., 157
Grant, J. T., 213, 251, 293 Hochstadt, H., 52, 61
Gray, R. M., 60 Hochst, H., 192, 298
Greene, J. E., 277, 297 Hofmann, S., 267, 270, 271, 272, 273, 279,
Greene, P. D., 127, 147 2Y3,2Y7
Grimm, A., 108, 147 Holloway, D. M., 219, 220, 252, 293; 297
Grobman, W. D., 196,292 Holloway, P. H., 163,219,258,266,293
Grossmann, G., 207,298 Holonyak, N., Jr., 115, 143
Grover, N. B., 4.61 Holscher, A. A., 163,256,257,264,281,290
Grutzmann, S., 15, 60 Hoogewijs, R., 200,293
Gryzinski, M., 165, 176, 180,292 Hooker, M. P., 213,293
Guglielmacci, J. M., 219, 264, 292 Houston, J., 196,213,252,254,255,256,2%
Guinolds, H. R., 129, 148 Houston, J . E., 164, 247,251, 292, 296
Gutkin, A. A., 103, 148 Howard, J. K., 256,273,274,278,291,293
Howe, M. S., 301,344,345,388
H Howie, A,, 180, 296
Hruska, S. J., 287, 292
Haas, T. W., 251,292 Hu, C., 133, 148
AUTHOR INDEX 395
Huang, C. I., 100, 126, 152 Joyce, B. A,, 123,159,286,293

Hubbard, J., 176, 293 Joyner, R. W., 287,293
Huber, A. M., 122, 125, 148
Hiifner, S., 192,298
K
Humbert, A,, 135, I48
Hume-Rothery, W., 114, I45 Kachurin, G. A,, 123,149
Hunsperger, R. G., 91, I48 Kadhim, M. A. H., 116, I49
Hurle, D. T. J., 66, 67, 69, 70, 72, 122, 123, Kakibayashi, H., 257,279,298
149,152 Kalma, A. H., 83, I49
Hurych, Z., 297 Kamath, G. S.,90, 147
Hutchinson, P. W., 113, 127,149 Kamejima, T., 132, I49
Hutchinson, W. G., 104, I45 Kaminska, M., 125, 149
Huth, F., 118,149 Kaminsky, M., 279,293
Hwang, C. J., 132, 149 Kamm, J. D., 132, I49
Kanter, H., 170, 171,293
I Karamalikis, A., 2, 59
Karelina, T. A,, 128, I50
Ichimura, S., 163,213,241,242,243,244,288, Karnatak, R.C., 192,291
293, 297 Kataoka, Y., 225, 279, 297
Ignatiev, A,, 281, 298 Kaufmann,U.,86,100,101,103,107,150
Ikoma, T., 75, 80, 106, 125, 149, 156, 157 Kawasaki, Y., 116, 154
Ikuta, T., 225, 241, 279, 293, 297 Kellokumpu, M., 195,298
Ilic, G., 91, 149 Kelly, P. W., 272, 292
Il’in, N. P., 67, 103, 149 Kelly, R., 275, 280, 293
Ing, B. S., 181,293 Kendal1,D. L., 113, 114, 118, I50
Inokuti, M., 176, 293 Kennedy, T. A,, 86, 150, 159
Ishikawa, K., 256, 263, 278, 292 Ketchow, D. R., 123, 147
Ishimaru, A., 310, 330, 331, 345, 346,388 Kheifets, V. S.,120, 155
Itikawa, Y., 176, 293 Kim, H. B., 115, 150
Ito, S., 15, 61 Kim, K. S., 195,200,273,281,282,293
Iwasaki, H., 136,149 Kimerling, L. C., 83, 84, 151
King, D. A., 297
J Kirby, R. E., 286, 287,293,295
Kirschner, J., 247, 255, 293, 297
Jablonski, A,, 237,238,293 Kitahara, K., 106, I50
Jackson, D. C., 171,293 Kittel, C., 175, 260, 294
Jacobi, K., 287,296 Kladis, D. L., 82, 150
Jacobs, B., 18,60 Klein, P. B., 92, 104, 150
Jain, G . C., 114, I49 Kleinman, L., 184, 294
Janak, J., 5,61 Klerk, M., I58
Jardin, C., 209,210,293 Knechtli, R. C., 90, 147
Jaros, M., 103,149 Knotek, M. L., 292
Jastrzebski, L., 132, I49 Kobus, A., 2, 3, 61
Jaswon, E., 18,59 Kolbenstvedt, H., 180,294
Jaswon, M. A,, 18, 54,61 Kolesov, B. A., 150
Jenkin, J. G., 171,290,298 Kornilov, B. V., 103, 107, 129, 130, 150
Jennison, D. R., 210, 212, 293 Koshikawa, T., 256, 263, 278, 292
Jesper, T.. 96, 149 Koval, 1. P., 172, 294
Johnson, P. D., 297 Kowalczyk, S. P., 195, 200, 294
Jordan, A. S., 122, 125, 158 Kozeikin, B. V., 122, 150
Joshi, A., 293 Kravchenko, A. F., 132, 150
396 AUTHOR INDEX
Kravtsov, Yu. A., 344, 388 Leheny, R. F., 132, I51

Krebs, J. J., 92, 150 Le Hericy, J., 260, 261, 262,290, 291, 294, 296
Krefting, E. F., 242,294 Lehovec, K., 82, 160
Kressel, H., 121, 123, 127, 134, 150 Lender, R. J., 132, I51
Kroger, F. A., 66, I50 Lenselink, A., 207, 208,296
Krynko, Yu. N., 172,294 Leung, P. C., 109,151
Kudo, H., 73, I50 Levenson, 192,295
Kiinzel, H., 120, 151 Lewis, J. E., 256,273, 276, 293
Kuhrt, F., 11,21,61 Ley, L., 195,200,294
Kuiken, H. K., 133,143, I51 Lezuo, K. L., 176,294
Kukla, C., 253,297 Li, G. P., 94,151
Kumar, V., 107, 113, 115,151 Li, S. S.,82, 89, 100, 126, 140, 152
Kung, J. K., 121,151 Liau, Z. L., 272,277,294
Kushiro, Y., 132, 151 Lichtensteiger, M., 287, 294
Kwun, S. I., 116, I51 Lichtman, D., 286,293
Kyser, D. F., 132, 159 Lidow, A., 128, 152
Lieberman, A. G., 143
L Liesegang, J., 171, 290, 298
Lindhard, J., 176, 177, 184, 294
Lasser, R., 211,294 Lindquist, P. F., 95, 152
Lagowski, J., 86, 94, 141, 151, 287, 294 Lippmann, H., 2, 11, 21, 33, 34,61
Laithwaite, K., 118, 121, I51 Littlejohn, M. A., 91, I52
Lancaster, G. M., 281,298 Logan, R. A., 76,151
Land, R. H., 176,297 Logan, R. M.,66,72,152
Landolt, D., 257,270,272,273,295 Lohmann, A. W., 301,388
Lang, B., 213,288,291,294 Lomako, V. M., 132,152
Lang, D. V., 76,82,83,84,85, 87,88,90, 133, Look, D. C., 82, 86, 88, 91, 94, 95, 125, 147,
148, 1-71 152
Langeron, J. P., 260, 261, 262, 290, 291, 294, Lorimor, 0. G., 109,152
296 Los, J., 287,298
Lanteri, H., 223, 290 Losch, W., 213, 282,283,289,294,297
Laramore, G. E., 248,291,294 Lotz, W., 180, 183, 294, 295
Larkins, F. P., 188, 193, 200, 201, 202, 206, Loudliche, s., 89, 152
291,294 Luberfeld, A., 294
Lastras-Martinez, A., 133, 151 Ludman, J. E., 90, 152
Laty, P., 271, 294 Lundquist, S., 292
Laver, R. F., 274, 278, 291 Lundqvist, B. I., 176, 295
Lea, C., 267,270,294,297
Leach, M. F., 126, 151 M
Lebedev, A. A., 106, I51
Leckey, R. C. G., 171,290,296,298 Ma, Y. R., 94, 151
Ledebo,L.-A., 107, 113, 115,150 MacDonald, N. C., 291
Lee, K. S., 116,151 Madden, H. H., 209,210,293,295
Lee, P. A., 172, 181,290,294 Madelung, O., 5,151
Leedy, H. M., 109, I51 Madey, T. E., 244,296
Legally, M. G., 196, 213, 252, 254, 255, 256, Magee, T. J., 115, 116,152
296 Maier, M., 122, 152
Le Gressus, C., 163, 245, 246, 282, 286, 292, Majerfeld, A., 80, 152
294 Makram-Ebeid, S., 81, 94, 153
AUTHOR INDEX 397
Malm, D. L., 288,295 Mohri, M., 257,279,298

Many, A., 4,61 Moiseiwitch, B. L., 179,290
Margoninski, Y., 287,295 Monemar, B., 126,154
Mariot, J. M., 192, 193, 194, 195, 291, 295 Moore, G., 196, 213, 252, 254, 255, 256, 2%
Mark, H., 193, 195,291 Moore, J. A,, 91, 154
Marsh, 0. J., 91, I48 Morabito, J. M., 166, 248, 249, 257, 259, 288,
Martin, G. M., 77, 78, 93, 94, 95, 96, 97, 98, 292
99, 100, 103, 122, 153 Morgan, D. V., 82, 157
Martin, P. C., 175,295 Morgulis, L. M., 113, 127, 154
Martinelli, R. U., 133, 153 Mori, H., 300, 388
Martinez, G., 92, 93,94, 146, 148, 153 Morkoc, H., 100, 154
Martin62 Saez, V., 213,298 Morrison, S. R., 118, 121, 154
Marton, J., 257, 266,290 Morrizumi, T., 121, 154
Massignon, D., 246, 294 Miiller, K., 163, 295
Masterov, V. F., 67, 103, 149 Mulford, R. A., 265,266, 295
Matart, H. F., 132, 153 Mullin, J. B., 95, 154
Mathieu, M. J., 257, 270, 272, 273, 295 Murata, K., 225, 243, 219, 295, 297
Mathur, V. K., 132, 153 Murday, J. S., 196,212,213,262,290,298
Matino, H., 128, I53 Murdey, J. J., 196,213,252,254,255,256,296
Matsudaira, T., 235, 237,239,295 Murygin, R. I., 107, 154
Matsukawa, T., 225, 243,279,295,297 Murygin, V. I., 103, 143
Matveenko, Yu A., 107,153
Mayadas, A. F., 2,62
N
Mayburg, S., 132, 153
Mayer, J. W., 272, 277,294 Nagel, D. R., 176, 180, 295
McFeely, F. R., 195, 200, 294 Nahory, R. E., 132, 151
McGlure, D. E., 270,272,295 Nakai, K.,106, 154
McGuire, E. J., 180, 193, 207, 209, 212, 292, Nakai, M. Y., 242,296
295 Nakayama, K., 279,295,2Y7
McGuire, G. E., 295 Nakhodkin, N. G., 172,294
McLevige, W. V., 116, 152 Nalecz, W., 3,61
McMahon, Jr., C. J., 264,295 Narayanan, G. H., 108, 154
McNichols, J. L., 90, 152 Neave, J. H., 286,293
Mei, K., 18, 61 Needham, Jr., P. B., 172,296
Melles, J. J., 192, 295 Nelson, R. J., 129, 154
Melnik, P. V., 172,294 Nemeth-Sallay, M., 171,296
Menzel, D., 284,295 Neumann, H., 122, I54
Mermin, N. D., 176, 179,295 Newell, W. R., 180,296
Meyer, F., 255,295 Newsome, J. P., 1 5 6 1
Miki, H., 132,153 Nilson-Jatko, P. E., 192,291
Miller, M. D., 75, 136, 153 Nishimori, K,, 295
Milnes, A. G., 114, 115, 133, 147, 159 Nishizawa, J. I., 71, 128, 154
Mimizuka, T., 15,61 Nojima, S., 116, I54
Minnhagen, P., 295 Noonan, J. R., 209, 210,295
Mircea, A., 77, 81, 127, 132, 153, 156 Norman, D., 171,295
Mitchell, E. W. J., 86, 88, I45 Norris, B., 108,154
Mitonneau, A., 77, 78, 80, 81, 102, 127, 136, Norton, K., 306,331,388
153,154 Novikov, E. A., 305,389
Monch, W., 94,154 Nowak, W. B., 90,152
398 AUTHOR INDEX
Nowicki, R. S., 129, 154 Petit, G. D., 122, 156

Nozaki, T., 122,154 Philibert, J., 243, 296
Nozieres, P., 207, 295 Picoli, G., 93, 155
Pierce, D. T., 172,296
0 Pilkuhn, M. H., 132, 155
Pinard, P., 127, 144
OBrien, J. K., 84, 154 Pitfield, R. A., 18, 61
Okajima, Y . ,295 Platzman, P. M., 181,290
Okurnura, T., 140, 154 Plew, L., 128, 145
Okutani, T., 279,295 Ploog, K., 120, 122, 123, 151, 155
Oldham, W. G., 113,143 Poate, J. M., 271, 272,290,294
Olivier, J., 287, 295 Poirier, R., 287, 295
Olowolafe, J. O., 129, 154 Pollack, M. A., 132, I51
Omel'yanovskii, E. M., 106, 154 Pollak, R. A,, 195, 196,200,292,294
Onchi, M., 235,237, 239,295 Pollard, A. M.,288,291
Ono, M., 295,297 Pons, D., 75, 77, 83, 84, 85,86,88, 155
Onoda, Jr., G. Y . ,288,2Y6 Pons, F., 260, 261,290,291,296
Oppenheirn, J., 300,388 Powell, C. J., 163, 170, 172, 183, 244, 252,
Oxley, D. P., 170, 295 296
Ozeki, M., 75, 119, 136, 154 Price, R. E., 193, 195,290,296
Pritchard, R. G., 171, 292
P Putley, E. H., 59,61
Padden, F. J., 288,295

Palfrey, H. D., 114. 155 Q
Palmberg, P. W., 163, 171, 256, 291, 295,296 Quichaud, G., 2, 3,61
Pandey, K. C., 207,212,292 Quinn, J. J., 175, 184,296
Panish, M. B., 113, 114, 155
Pantano, Jr., C. G., 288, 295,296 R
Papagno, L., 265,296
Papaioannou, G. I., 132,155 Rabalais, J . W., 281, 298
Parikh, M., 248, 292 Rachmann, J., 126, 155
Park, R. L., 164, 247. 248, 291,292, 294,296 Ramaker, D. E., 196,213,252,254,255,256,
Park, Y. S., 116, 156, 160 262,296,298
Parravicini, G. P., 181, 298 Ramsey, J. A., 172,296
Partin, D. L., 102,108,131,132,133,138,139, Ranke, W., 287,296
155 Rao, P. V., 193, 195, 290,296
Parui, D. P., 132, I45 Rao-Sahib, T. S., 133, 155
Pashley, R. D., 128, 155 Rasolt, M., 182, 296
Payling, R., 172,296 Redhead, P. A., 284,296
Pearson, G. L., 68, 84, 131, 144, 145 Rehn, L. E., 265,279,280,296
Peisner, .I.171.
, 296 Reickl, L. E., 300,389
Peka, G. P., 132, 155 Reirner, L., 225, 242,294,296
Pendry, J. B., 181, 186,293, 296 Reiter, F. J . , 192, 298
Peng, J., 115, 152 Reuter, W., 260, 266, 281,296,298
Penn, D. R., 176, 183, 184, 260, 296 Rheinlander, B., 123, 155
Perlberg, C. R., 288,289 Rhodin, T. N., 171,295
Pessa, V. M., 180, 296 Riachi, G. E., 291
Peters, R. C., 158 Richman, D., 131, 155
AUTHOR INDEX 399
Rickman, J., 287, 293 Schlachetzki, A., 106, 156

Ridley, B. K., 126,148, I51 Schliiter, M., 196, 197,291
Ringers, D. A., 195, 298 Schneider, J., 75, 86, 100, 101, 103, 150, 158
Ritchie, R. H., 176, 180, 181, 184, 185, 242, Schou, J., 223, 297
290,292,296,298 Schreiner, D. G., 247, 296
Riuaud, L., 277,297 Schrott, A. G., 287,291
Riviere, J. C., 286, 291 Schwartz, S. B., 180,297
Robinson, G. Y., 129, 148 Schwinger, J., 175, 295
Roll, K., 282, 283, 289, 297 Seah, M. P., 170, 171, 183,260,267,270,271,
Roelofs, L. D., 248,291,292,294 275,294,297
Rogers, S., 153 Sealy, B. J., 91, 127, 156, 157
Rogers, S. C., 132, 159 Segal, D., 287, 295
Rojo, J. M., 213,298 Sekela, A., 132, 139, 140, 156
Romanenko, V. N., 120, 255 Seki, Y., 132, 156
Romankiw, L. T., 2,62 Seltzer, M. S., 104,156
Rondot, B., 261,291 Sevier, K. D., 162, 193,297
Rose-lnnes, A. C., 148 Shafer, M. W., 279,291
Rosler, M., 223, 297 Shaldervan, A. I., 172, 294
Ross, J., 300, 388 Share, S., 107, 156
Rossin, V. V., 133, I56 Shaw, R., 18,61
Rostaing, P., 223, 290 Shelton, J. C., 171, 176, 184, 297
Roulet, B., 188, 289 Shifrin, S. S., 127, 156
Roussel, A,, 81, I56 Shih, K. K., 122, 156
Rowe, J. E., 196, 198,291 Shikata, M., 279,295, 297
Rowe, R . G., 265,266,295,297 Shimizu, H., 279,295,297
Rozgonyi, G. A,, 130,156 Shimizu, R., 163, 213, 225, 241, 242, 243,244,
Rubin, V. S., 107, 154 256,263,274,278,279,288,293,294,295,
Rudge, M. R. H., 180,297 297
Ryan, R. D., 131, 156 Shin, B. K., 121, 156
Shirley, D. A., 195, 199, 200, 295,297
S Shishyanu, F. S., 113, 115, 144,157
Shortley, G. H., 192,205,291
Sadana, D. K., 149 Sickafus, E. N., 220, 252, 253, 254, 297
Saeki, N., 274, 297 Siegmann, H. C., 172,296
Sah, C. T., 132, 145, 156 Sigmund, P., 275,276,297
Saito, T., 75, I56 Simoni, F., 265, 2%
Sakai, K., 80, 156 Singwi, K. S., 176, 298
Salmeron, M., 213,298 Sinha, A. K., 130, 157
Samuelson, L., 137, 156 Sinharoy, S., 172,298
Sankey, 0. F., 103, 125,156 Sjolander, A., 176,297
Sansbury, J. D., 91, 156 Slater, J. C., 192, 297
Sanz, J. M., 267,271,297 Slifkin, L. M., 113, 114, 146
Sastri, S., 281, 290 Sliisser, G. J., 266, 277, 297
Savua, M.A., 128,156 Smith, A., 5,61
Sawatzky, G. A., 207,208,297 Smith, B., 15, 36, 59
Scarmozzino, R., 265,296 Smith, N. V., 297
Schairer, W., 122, 156 Smith, T., 297
Scheer, J. J., 131, 142 Smrcka, L., 186,297
Schemer, M., 141, 156 Sokolov, V. I., 115, 157
400 AUTHOR INDEX
Solov, H., 106, 156 Thomas, S., 281,298

Solov’ev, N. N., 102, 143 Thomas, T. D., 195,289
Sopizet, R., 245,294 Thompson, D. A., 2,62
Sorvina, F. P., 105, 158 Thompson, V., 267,298
Spanjaard, D., 209,298 Thurmond, C. D., 69,158
Spicer, W. E., 73,74,75,157 Thurstans, R. E., 170,295
Spitzer, W. G., 109, 121, 150, 152, 157 Tisone, T. C., 271,290,297
Springthorpe, A. J., 122, 157 Tokutaka, H., 295
Staib, P., 247,255,293,298 Tomlin, S. G., 180,298
Stall, R. A., 122, 157 Tompkins, H. G., 287,298
Stegun, I. A., 14, 351, 59, 388 Toneman, L. H., 163,256,257,290
Stein, H. J., 90, 157 Tongman, L.H., 257,298
Stein, R. J., 170, 171, 172,296,298 Tosatti, E., 181,298
Steiner, P., 192, 298 Tosi, M. P., 176,297
Stern, R. M., 172,298 Tottenham, H., 52,62
Stevens, K., 23,61 Tracy, J. C., 255,298
Stillman, G. E., 70, 119, 123, 157, 159 Traum, M. M., 297
Stoneham, E. B., 122,157 Treglia, G., 209,298
Strack, H., 106, 157 Tsang, T., 195,298
Strand, T. G., 241,290 Tsaur, B. Y.,272,277,294
Strauss, G. H., 86,92, 150, 157 Tuck, B., 116,149
Stringfellow, G. B., 120, 157 Tucker, Jr., C. W.,172,291
Su, J. L., 118, 157 Tung, C. J., 176, 181, 184, 185,290,296,298
Suchkova, N. I., 107,157 Turner, J. E., 176,293
Sugibuchi, K., 136, 149 Turner, N. H., 196, 212, 213, 252, 254, 255,
Surridge, R. K., 91, 122, 157 256,262,290,296, 298
Swift, C. D., 193, 195,290 Tzoar, N., 182,292
Switkowski, Z. E.,275,292
Sykes, D. E., 267,298
U
Symm, G., 18,52,54,61
Szajman, S., 171,298 Ushakov, V. V., 92,130, 147, 158
Uskov, V. A., 105, 158
T
Ta, L. B., 86, 157
v
Tagle, J. A., 213, 298 Vayrynen, J., 195,289,298
Takahashi, K., 121, 127, 154, 157 Vaitkus, J., 94, 158
Takai, M., 91, 157 Vallin, J. T., 92, 158
Takashima, K., 295 Van Blade], J., 18,60,61
Takikawa, M., 75, 149 Van Der Pauw, L. J., 7, 12,36,62
Tandon, J. L., 122, 157 Van Mechelen, J. B., 264, 281, 290
Taniguchi, M., 125, 157 Van Oostrom, A., 163,257,265,298
Tarng, M. L.,267, 298 Van Opdorp, C., 137,158
Tatarskii, V., 301, 331,344,388,389 Van Santen, R. A., 257,298
Taylor, J. A., 281,298 Van Vechten, J. A., 67, 68, 69, 72, 158
Taylor, P. D., 82, 157 Vashista, P., 176, 298
Texier, R., 243,296 Vasile, M. J., 288, 295
Thomas, J. M.,171,292 Vasil’ev, A. V., 103, 158
Thomas, R. N., 91, 157 Vasudev, P. K., 81,94, 126, 158
AUTHOR INDEX 401
Vennik, J., 200, 293 Willmann, F., 113,144,159

Verhoeven, J., 287,298 Willoughby, A. F. W., 75, 114,146, 155
Vigorradov, A. G., 344,388 Wilsey, N. D., 86, I59
Vincent, G., 80, 127, 144, 158 Wilson, R. G., 105, 159
Vine, J., 282, 298 Winder, D. R., 254,297
Vitale, G., 133, 158 Winograd, N., 195, 200, 266, 272, 273, 277,
Von Barth, U., 207,298 281,282,292,293,297
von Roos, O., 133, 158 Winters, H. F.. 288, 289
Vrakking, J. I., 255,295 Wirth, J. L.,82, 159
Wiser, N., 176, 180, 298
Witten, Jr., T. A., 176, 180,295
W
Wittry, D. B., 133,155,159
Wada, O., 76,158 Womer, R., 86, 159
Wagner, E., 132,144 Wolfe, C. M., 70, 119, 120, 121, 122,123,159
Wagner, M., 280,2% Wolff, G. A., 132,153
Walker, G. H., 86, 158 Wolford, D. J., 124, 159
Wallis, R. H.,89, 125, 158 Wolfstirn, K. B., 108, 147
Walls, J. M.,267, 298 Wonsidler, D. R., 271,290
Wang, C. C., 133,153 Wood, C. E. C., 123, 159
Wang, I., 129, 154 Woodcock, J. M., 91,159
Wang, K. L., 94, I51 Woodruff, D. P., 171,295 297
Ward, I. D., 217,297 Worthingon, C. R., 180,298
Warsza, Z. L., 3, 61 Wu, C. J., 133, 159
Watanabe, K., 257,273.278,279,298
Watkins, G. D., 89,158 Y
Watson, G. N., 339,389
Webb, C., 287,294 Yabumoto, M., 251,219,29
Weber, E. R., 75,88, 158 Yamashina, T., 257,273,278, 219,29
Weber, R. E., 291 Yamazaki, H., 123,159
Weiner, M. E., 122, 125, 158 Yan, Z.-x., 114, 115, 159
Weinert, H., 155 Yellin, I., 195, 298
Weiser, K., 92, 150 Yin, L. I., 195, 298
Welch, B. M., 128, 146, 155 Yip, K. L., 197,298
Whelon, A. D., 347, 349,389 Yu, M. L., 266,281, 298
White, A. M., 93, 126,159 Yu, P. W., 91, 116,160
Whittaker, E. T., 339, 389
Wick, R. F., 11, 34,62 Z
Wieder, H. H., 59,62
Wiedersich, H., 265,279, 280,296 Zakharova, G. A., 112,160
Wight, D. R., 133, 159 Zalar, A,, 270,273, 279,293
Wigner, E., 300,389 Zehner, D. M., 209, 210, 293, 295
Wilcox, D. L., 288,289 Zelevinskaya, V. M., 91, 160
Wild, W., 173,292 Zemon, S., 123, 160
Wildman, H. S., 256, 273,293 Zhu, H.-z., 137,160
Willardson, R. K., 113, 119, 159 Ziegler, A., 183, 293
Williams, E. W., 126,159 ZOU,Y-x., 75, 76,160
Williams, P. J., 92, 93, 159 Zucca, R., 95,160
Williamson, K. R., 116,159 Zuleeg, R., 82,160
Subject Index
A semiconductors and insulators, 13 1- 182

theoretical evaluations and experimental
AEAPS, see Auger electron appearance po-
data, comparison of, 182-185
tential spectroscopy
introduction, 162-167
AES, see Auger electron spectroscopy
mean escape depths, 173
Aluminum
mean free path, 168- 170
Auger emission, 227-240
quantitative analysis, 244-289
elastic collisions, 186-187
electron-beam-induced effects, 282-289
inelastic collisions, 177-179, 184
operating modes, 245-254
Aluminum-copper alloys, sputtering, 273,
quantification of Auger analysis, 255-263
278
sample preparation, 263-273
Aluminum oxide, inelastic collisions, 18I
sputtering, effects of, 273-282
Anistropic stochastic medium, propagation
Auger transition, 162
through, 336,345,379,382
in a solid, 187-213
Appearance potential spectroscopy (APS), in
classification, 188-198
Auger electron spectroscopy, 166
intraatomic, 188-196
Arsenide vapor, behavior of GaAs in contact
kinetic energy of Auger electrons, 198-
with, 66-67
206
Attenuation length, measurement of, 170- 172
line-shape analysis, 206-2 13
Auger electron appearance potential spec-
lines involving valance bands, 196-1 98
troscopy (AEAPS), 247-248
Auger electron spectroscopy (AES), 161-298
attenuation length, 170-172
B
Auger emission, quantitative description of,
213-244 Backscattered electron, in Auger electron
theoretical analysis, 222-244 spectroscopy, 164, 221-222, 227, 231 -
Auger transition in a solid, 187-213 232,234-235,238
classification, 188- 198 Backscattering factor, in Auger electron spec-
kinetic energy of Auger electrons, 198- troscopy, 221-222, 234, 237-239, 259-
206 260
line-shape analysis, 206-21 3 BEM, see Boundary-element method
basic experiment, 163-164 Beryllium
calibration techniques, 166, 256-258, 260 Auger line shape analysis, 210
cross sections, 167-168 in gallium arsenide, 116-117, 120
definitions, 167-1 73 Boron, in gallium arsenide, 118
elastic collisions, 185-187 Boundary-element method (BEM), Hall-plate
inelastic collisions, 173- 185 potential calculations, 18-38
d-electron metals, 174, 180-181 cross-shape geometry, 25-27, 30-33,45-46
depth dependence and anisotropy of direct calculation of geometry correction,
mean free path, 182 27-29
general equations, 174- 176 functions, calculation of, 40-43
normal metals, 176-180 improvement, 38-48
402
SUBJECT INDEX 403
integral equations, 18-20, 39-40 Conformal mapping, Hall-plate problems,

numerical solution, 20-21 10-15
numerical calculation of current through a basic ideas, 10-1 1
contact, 24-25 cross-shaped geometry, 1 1 15-
octagonal Hall plate, 34-36, 46-48 Copper

rectangular Hall plate, 21-22, 29-30, 33- Auger quantitative analysis, 255, 257, 260,
35,43-49 270
unsymmetrical cross-shaped form, 37-38 Auger transition line shapes, 193-194, 209
zeroth-order approximation, 22-24 in gallium arsenide, 100, 102-103, 106,
Bridgman-grown gallium arsenide, diffusion 109-113, 120
lengths in, 139-141 inelastic collisions, 184-185
Copper-indium alloys, sputtering, 277
C Copper-nickel alloys
Cadmium, in gallium arsenide, 116-1 18, 139 Auger quantitative analysis, 263-264
Cadmium sulfide, electron-beam-induced ad- diffusion, by increasing temperature, 289
sorption of oxygen in H,O, 287 sputtering, 278-281
Carbon Cross section, in Auger electron spectroscopy,
Auger line shape analysis, 2 13 167- 168
Auger quantitative analysis, 255,286 Cylindrical mirror analyzer (CMA), in Auger
in gallium arsenide, 116-1 17, 120-121 electron spectroscopy, 164
Carbon dioxide, adsorption enhancement on
semiconductors, 287 D
Carbon monoxide, adsorption enhancement
on semiconductors, 286-287 DAPS, see Disappearance potential spec-
Cascade process, in Auger electron spectros- troscopy
copy, 164,241,243 Deep-level optical spectroscopy (DLOS), gal-
Cassette recorder, Hall-effect applications, 2 lium arsenide studies, 81
Chromium Deep-level transient spectroscopy (DLTS),
Auger quantitative analysis, 257, 260-261, gallium arsenide studies, 76-81, 90, 96,
270,213 98, 100
in gallium arsenide, 91-100, 102, 120 Desorption, in Auger quantitative analysis,
CMA, see Cylindrical mirror analyzer 284-286
Cobalt, in gallium arsenide, 100, 102-103, Disappearance potential spectroscopy
106-107,120 (DAPS), in Auger electron spectroscopy,
Coherent Wigner function, 347-354 166,241-248
Computer simulation, stochastic transport DLOS, see Deep-level optical spectroscopy
equation for the Wigner function, 302, DLTS, see Deep-level transient spectroscopy
354-381
algorithm for integral equation for very
E
narrow beam, 357-359, 378, 382
exponential space correlation function, EAPFS, see Extended appearance potential
368-37 1 fine structure
FORTRAN program, 359,382 Elastic collision, in Auger electron spec-
integral equation for very narrow beam, troscopy, 164-165, 185-187, 226-229,
355-357 231,241-242
Kolmogoroff space correlation function, Electric field in a stochastic dielectric, Wigner
362-365,371-378,382 distribution matrix for, see Wigner dis-
listing of experimental program, 383-385 tribution matrix for an electric field in a
sample, 386-387 stochastic dielectric
404 SUBJECT INDEX
Electron-beam-induced effects in Auger elec- irradiation, levels produced by, 8 1-91

tron spectroscopy, 282-289 minority-carrier recombination, genera-
charge effects, 283-284 tion, lifetime and diffusion length, 13l -
dissociation, 288 141
electron-impact-stimulated adsorption, molybdenum, 128- 129
286-288 oxygen, 123-127
migration and diffusion, 288-289 palladium, 129- 130
physicochemical effects, 284-286 platinum, 130-131
temperature increase, 282-283 possible native defects and complexes, 65-
Electron-beam stimulated description (ESD), 16
in surface studies, 284-285 rare-earth studies, 13I
Electron-irradiated gallium arsenide, 82-90 ruthenium, 129
Electron paramagnetic resonance (EPR), semi-insulating GaAs with and without
GaAs studies, 73, 86 chromium, 91-100
ESD, see Electron-beam stimulated descrip- shallow acceptors, 116-1 18
tion transition metals, 100- 108
EXAFS, see Extended X-ray absorption-edge traps and nomenclature, from DLTS
fine structure experiments studies, 76-8 1
Extended appearance potential fine structure tungsten, 130
(EAPFS), in Auger quantitative analysis, Gaussian stochastic process, 305
248 Germanium, in gallium arsenide, 116-1 17,
Extended X-ray absorption-edge fine struc- 119-120, 122
ture (EXAFS) experiments Gold
attenuation lengths, measurement of, 170, Auger quantitative analysis, 270
172 in gallium arsenide, 115- 116, 139
in Auger quantitative analysis, 248
H
F
Hall-effect applications, 2-3
Fast-neutron-irradiated gallium arsenide, 86 photovoltaic cell, 57-58
FEM, see Finite-element method Hall-effect measurements, gallium arsenide
Field-effect transistor (FET),effects of radia- with and without chromium, 94-99
tion on, 82 Hall plates, potential calculations in, 1-62
Finite-element method (FEM),potential cal- boundary-element method, 18-38
culations in Hall plates, 18 cross-shaped geometry, application to,
Forward scattering by stochastic dielectrics, 25-27,30-32,45-46
30 1 functions, calculations of, 40-43
asymptotic equations for the Wigner dis- geometry correction, direct calculation
tribution function, 330-335 Of, 27-29
improvement, 38-48
G integral equation, 18-20, 39-40
integral equations, numerical solutions
Gallium arsenide, impurity and defect levels, of, 20-21
63- 160 numerical calculation of current through
concluding discussion, 141 142
- a contact, 24-25
group I impurities, 108- I 16 octagonal Hall plate, application to, 34-
group IV elements as dopants, 118-123 36,46-48
group VI shallow donors, 127-128 rectangular Hall plate, 21-22,29-30,33-
introduction, 64 35,43-41
SUBJECT INDEX 405
unsymmetrical cross-shaped form, appli- Insulator

cation to, 37-38 charge effects in Auger electron spectros-
zeroth-order approximation, 22-25 copy, 283
conclusions, 48-49 dielectric theory of inelastic collisions, 181 -
conformal mapping techniques, 10-1 5 182
basic ideas, 10-11 Interstellar plasma, 308
cross-shaped geometry, 11- 15 Inverse mean free path (IMFP), in Auger elec-
current’s contribution to magnetic field, tron spectroscopy, 170, 182
58-59 Ion-scattering spectrometry (ISS), 257
finite-element method, 18 Iron
Fredholm integral equations, 52-54 Auger quantitative analysis, 257,260-261
fundamental equations, 3-7 in gallium arsenide, 100-102, 104-106, 120
approximations for thin semiconducting Iron-silicon alloys, Auger quantitative analy-
layer, 4-5 sis, 265
externally applied magnetic field, bound- Irradiation, effect on GaAs and GaAs devices,
ary conditions with, 6-7 81-91
externally applied magnetic field, con- Isotropic stochastic dielectric
stitutive relations, 5-6 half space, plane wave incident on, 337-
general equations for semiconductors, 342, 382
3-4 scattering effects, 336-337
geometry, influence of, 9- 10 time invariant, monochromatic waves in,
Green’s theorem, 54-57 342-345, 378, 382
introduction, 2-3 ISS, see Ion-scattering spectrometry
relaxation methods, 15-17
three-dimensional Hall effect, 49-52
Van Der Pauw method, 7-9
Hall probe, applications of, 2 K
Hall voltage, 2
Kinetic energy, Auger electrons, 198-206
Kolmogoroff space correlation function, com-
I puter simulation for, 362-365, 371-378,
IMFP, see Inverse mean free path 382
Indium phosphide, electron-beam-induced Kolmogoroff spectrum, 307,351
oxidation, 287
Inelastic collision, in Auger electron spectros-
COPY,164-165, 173-185, 226, 228, 231,
L
241 -242
d-electron metals, 174, 180-181 Lead, in gallium arsenide, 1 19-120, 123
depth dependence and anisotropy of the LEED, see Low-energy electron diffraction
mean free path, 182 Liquid-phase epitaxy (LPE) gallium arsenide,
general equations, 174-176 64,76, 132-135, 140
normal metals, 176-180 Lithium
inner shell electrons, collisions on, 178- Auger line shape analysis, 210-21 1
180 in gallium arsenide, 108-109
plasmon mode and individual transitions, Low-energy electron diffraction (LEED),
176-178 attenuation length, measurement of, 170,
semiconductors and insulators, 181- 182 172
theoretical evaluations and experimental LPE gallium arsenide, see Liquid-phase epi-
data, comparisons of, 182-1 85 taxy gallium arsenide
406 SUBJECT INDEX
M Auger quantitative analysis, 285

in gallium arsenide, 123-127
Magnesium. ingalliumarsenide, 116-1 17,120
Magnetic bubble memory, Hall-effect applica-
tions, 2 P
Magnetic field, measurement with Hall-effect
Palladium, in gallium arsenide, 129- 130
components, 2
Parabolic equation method, for propagation
Manganese, in gallium arsenide, 100, 102-
problems, 301, 345-347
104, 120
Photoconductivity, gallium arsenide: chro-
Mean free path (MFP)
mium, 94
in Auger electron spectroscopy, 165, 168-
Plasma, stochastic transport equations for
170,218
Wigner function, 303, 305, 308
elastic collisions, 186- 187
Platinum, in gallium arsenide, 130- 131
inelastic collisions, 182- 185
Potassium, electron-beam-induced migration
stochastic dielectric, 306
of atoms, 288
Mercury, in gallium arsenide, 118
Potential calculations in Hall plates, see Hall
MFP, see Mean free path
plates, potential calculations in
Minority carrier, recombination, generation,
Preferential sputtering, 271-272, 275-278
lifetime, and diffusion length in gallium
reactive and nonreactive, 281
arsenide, 131-141
Proton-irradiated gallium arsenide, 89-90
Molybdenum
Push buttons, Hall-effect applications, 2
Auger quantitative analysis, 257, 261
in gallium arsenide, 128-129
Monte Carlo simulation method, in Auger Q
electron spectroscopy, 223-227
Quantitative analysis, Auger, see Auger elec-
tron spectroscopy, quantitative analysis
N
R
Neodymium, in gallium arsenide, I3 1
Neutron-irradiated gallium arsenide, 86, 89- Randium-jellium model, in Auger electron
90 spectroscopy, 165, 186,224
Nickel Rare-earth elements, in gallium arsenide, 131
Auger emission, 237 Relaxation method, Hall-plate potential cal-
Auger quantitative analysis, 270, 273 culations, 15-17
electron-beam-induced oxidation, 287 Russell-Saunders coupling, 190
in gallium arsenide, 100, 102, 106-108, 120, Ruthenium, in gallium arsenide, 129
137-139
Nickel-chromium system, sputtering, 279
S
Nickel-palladium alloys, Auger quantitative
analysis, 257 Scandium, in gallium arsenide, 102
Secondary electron, in Auger electron spec-
0 troscopy, 164
Secondary-ion mass spectrometry (SIMS), 257
Optical imaging, Wigner functions, 301 Selenium, in gallium arsenide, 119-120, 127
Optical signals, Wigner functions, 301 Semiconductors, see also Gallium arsenide;
Oxidation of semiconductors, 286-287 Hall plates
Oxygen dielectric theory of inelastic collisions, 181
~
adsorption enhancement on semiconduc- I82

tors, 286 electron-beam-induced oxidation, 286-287
Auger line shape analysis, 2 13 n-type, general equation for, 3-4
SUBJECT INDEX 407
size effects, 9 Tin, in gallium arsenide, 69-70, 120, 122-123

thin-film, approximations for, 4-5 Titanium
Sequential-layer sputtering (SLS), 267 Auger quantitative analysis, 261-262
Silicon in gallium arsenide, 102-103
Auger line shape analysis, 21 1-2 13 Transition metals, effects in gallium arsenide,
Auger quantitative analysis, 265, 286 100- 108
electron-beam-induced oxidation, 287 Troposphere. stochastic propagation, 308,382
in gallium arsenide, 116-122, 141 Tungsten, in gallium arsenide, 130
inelastic collisions, 181 Turbulence theory
sputtering, 281 computer-range simulations for Kolmo-
Silicon devices, effects of radiation on, 82 goroff space correlation function, 371,
Silicon dioxide 374, 378, 382
Auger quantitative analysis, 271 Wigner function, 302-303, 307-308
Auger transition, 197-198
electron-beam induced dissociation, 288
U
Silver
Auger quantitative analysis, 255, 263, 270 Ultrahigh vacuum (UHV) system, in Auger
in gallium arsenide, 1 13-1 15 electron spectroscopy, 245,264
Silver-palladium alloys
Auger quantitative analysis, 257
V
sputtering, 277,281
SIMS, see Secondary-ion mass spectrometry Vanadium, in gallium arsenide, 100, 102-103
SLS, see Sequential-layer sputtering Van Der Pauw method, Hall-mobility mea-
Sodium surements, 7-9
Auger line shape analysis, 21 1 Vapor-phase epitaxy (VPE), gallium arsenide,
electron-beam-induced migration of atoms, 64,76, 135 - 139
288
Soft X-ray appearance potential spectroscopy
W
(SXAPS), 247
Solar cell, radiation damage, 82 Wave propagation, Wigner functions, 301 -
Sputtering, in quantitative Auger analysis, 303, 308
268-273 Wiener- Khintchine theory, 317
multicomponent materials, 273-282 Wigner distribution matrix for an electric
Steel, Auger quantitative analysis, 261, 265- field in a stochastic dielectric, 299-389
266 asymptotic equations for the Wigner dis-
Stochastic dielectric, Wigner distribution ma- tribution function, 330-335
trix for electric field, see Wigner distribu- coherent Wigner function, 347-354
tion matrix for an electric field in a computer simulation, 354-38 I
stochastic dielectric algorithm for stochastic integral equation
Sulfur for very narrow beam, 357-359,378,
Auger spectra, 213 382
in gallium arsenide, 120, 127-128 exponential space correlation function,
SXAPS, see Soft X-ray appearance potential 368-371
spectroscopy FORTRAN program, 302, 359, 382
integral equation for very narrow beam,
T 355-357
Kolmogoroff space correlation function,
Tellurium, in gallium arsenide, 69-72, 127- 362-365,371-378,382
128, 141 listing of experimental program, 383-385
Thulium, in gallium arsenide, 13 1 sample, 386-387
408 SUBJECT INDEX
conclusions, 382-383 related equations for the Wigner distribu-

derivation of the equations for the Wigner tion function, 325-329
distribution functions, 313- 325 theoretical methods, review of, 345-347
differential equation for the electric field
correlations, 303-313
equations for special cases, 335-345
Z
forward-scatteringapproximation, 330-
335 Zinc, in gallium arsenide, 1 16- 1 18, 120, 139,
introduction, 300-303 141

Adv Electronics Electron Physics v61 Volume 61 Adv

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Adv Electronics Electron Physics v61 Volume 61 Adv

Uploaded by

Copyright:

Available Formats

ADVANCES IN ELECTRONICS AND

United Kingdom Edition published by

ISBN 0-1 2-014661 -4

PRINTED IN THE UNITED STATES OF AMERICA

Potential Calculations in Hall Plates

Impurity and Defect Levels (Experimental) in Gallium Arsenide

XI . Group VI Shallow Donors: S. Se. Te . . . . . . . . . . . . . i27

Quantitative Auger Electron Spectroscopy

H. BREMMER,31 Bosuillaan, Flatgebou’w Houdringe, Bilthoven, The

*Present address: Department of Electrical and Computer Engineering, University of

The boundary between electron physics and solid state physics is by

Waveguide and Coaxial Probes for Nondestructive

Potential Calculations in Hall Plates

Appendix 4. The Hall-Effect PhotovoltaicCell.. ............................ 57

Hall plates are thin semiconducting layers placed in a magnetic field.

A . General Equations f o r a Semiconductor

(substitute [(n - no)/r,,] for a p-type layer);

B. Approximations f o r a Thin Semiconducting Layer

n = ND and p<< n (5)

A thin semiconducting layer can be seen simply as a sheet with a constant

C. Constitutive Relations with an Externally Applied Magnetic Field

where the index i denotes the ith component in a rectangular coordinate

Equation (14) is derived from the Jones-Zener expansion (12) neglecting

D. Boundary Conditions with an Externally Applied Magnetic Field

FIG. 1. Hall-plate configuration for outlining the stream potential.

I = - JAB' J . u , d l = JAB V$*u,dl = $(B') - $ ( A ' ) = -$o (21)

111. THEVAN DERPAUW METHOD

(3) The layer must be homogeneous

V2 = = pl:J.dr + p p H I I ( Jx B).dr (23)

with B = Bu,(u, directed perpendicular to the Hall plate):

FIG.2. Hall generator of arbitrary shape placed in a magnetic field B.

where d represents the thickness of the layer. Equation (25) enables us to

FIG.3. Hall generator with contacts placed at the ends of strips.

calculations presented in the next sections, it is shown that disturbances due

B. Approximate Analysis of the Cross-Shaped Geometry

h/l = {2[(k/h) - l]}-'

FIG.5. Conformal mapping of a cross-shaped geometry (b) onto a parallelogram (a).

where F is the hypergeometric function. In the case that c sz 0, i.e., the

C . Exact Analysis of the Cross-Shaped Sample

From Eq. (37), the values of h and I can be found:

x (z + U)-+'~/~(Z + C ) - ~ ~ / ~+( Zd " ) ( z + e)-'+'lln(z + g)-'+'2/n (39)

D . Properties oJ the Cross-Shaped Hall Plute

FIG. 8. Boundary conditions for the finitedifference approximation.

FIG.9. Geometry correction as a function of M :p H B= 0.1 ;h/l = 1.

4i- a boundary condition leading to a reduced accuracy and hence never

VII. THEBOUNDARY-ELEMENT METHODFOR POTENTIAL

B. Integral Equation f o r the Potential Distribution in a Hall Plate

FIG. I I . Rectangular Hall plate.

The solution of our problem 4(r) is now written as

V4(r) = (271.)-’ p(r’)[(r - r’)/(lr - r’l’)] dc‘ (49)

C. Numerical Solution of the Integral Equations

Aci, has to be calculated otherwise. Replacing pi by a uniform constant

The integral equations are then replaced by

= 0, riE AA', BB'

D. Application t o a Rectangular Hall Plate

The integral equations (53) and (54) now read

1" p j g1I A c J l n l r -i r ' I d c ' =

FIG. 15. Complex plane used to evaluate some integrals.

If one uses a computer with complex-number facilities, all integrals appearing

F. Numerical Calculation of the Current through a Contact

because the tangential component of E is zero at a metallic contact. From

The current I is then given by

FIG. 16. Use of the angle a to calculate current through AB

approximations will be carried out. It is more convenient to interchange the