Ellipsometry in
Ellipsometry in
Ellipsometry in
REVIEWS Further
Ann. Rev. Mater. Sci. 1981. 11:97-122
Quick links to online content
Copyright © 1981 by Annual Reviews Inc. All rights reserved
ELLIPSOMETRY IN +8659
J. B. Theeten
Laboratoires d'Electronique et de Physique Appliquee, 3 avenue Descartes,
94450 Limeil Brevannes, France
-
by Duke University on 06/13/12. For personal use only.
D. E. Aspnes
Bell Laboratories, Murray Hill, New Jersey 07974
INTRODUCTION
97
0084- 6600/81 /0801-0097$01.00
98 THEETEN & ASPNES
obtain full spectra, samples must usually have reached a steady state.
Consequently, "final products" are examined. In contrast with the kinetic
data obtained in the previous case, which are usually analyzed in a
3-phase (ambient/film/substrate) model, one is interested here in ex
by Duke University on 06/13/12. For personal use only.
BACKGROUND
Basic Equations
Only results are given here; derivations of these results can be found in
(1). Ellipsometry is based on the fact that the reflectance of a polarized
incident beam is different for polarization perpendicular (s; German
senkrecht) or parallel (p) to the plane of incidence. The orientations are
defined in Figure 1. An ellipsometric measurement determines the com
plex reflectance ratio
1.
between the corresponding complex reflectances rs and rp' The angles 1/1
ELLIPSOMETRY 99
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
by Duke University on 06/13/12. For personal use only.
and !J. are used for historical reasons and correspond respectively to the
differences in amplitude and phase of the two reflections. At a sharp
interface between a medium of refractive index no and one of n. (Figure
1), the following relations (from Descartes and Fresnel) hold:
nosinOo =n. sinO. 2.
.
rp =ro•p =(n. cos 00 -nocos 0.)( n. cos 00 +nocos 0.)- 3.
5.
Note that in this "2-phase" model, the two parameters of p are used to
determine uniquely and exactly, without a Kramers-Kronig transform or
additional assumptions, the two parameters of the complex quantity E. in
terms of the ambient no and angle of incidence f)o.
Figure 2 shows a 3-phase system consisting of an ambient medium, a
uniform film, and a substrate. Here, the multiple interference regime that
develops within the film can be described by using an effective reflection
100 THEETEN & ASPNES
AMBIENT
eo
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
FILM
by Duke University on 06/13/12. For personal use only.
SUBSTRATE
coefficient,
a=p or s 7.
Instrumentation
The schematic block diagram of an ellipsometer is given in Figure 3. The
various possible combinations of optical elements were reviewed recently
by Hauge (6). We give only the most frequently encountered configura
tions.
The null ellipsometer uses a polarizer and a compensator or quarter-wave
plate to produce a phase shift just opposite to the one induced upon
reflection from the sample. This phase shift is obtained by varying the
polarizer azimuth with respect to the fast and slow axes of the compensa
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
tor. The light reflected from the sample is then linearly polarized and can
be completely extinguished by properly adjusting the azimuth of the
analyzer. This type of ellipsometer is mechanically simple and can be
operated manually. Consequently, it was the first to be used. But it
by Duke University on 06/13/12. For personal use only.
requires much care and time to adjust to adequate accuracy: of the order
of several minutes to achieve 0.0 I. 0 in lJt and �. By using stepping motors
to rotate the polarizer and the analyzer, or Faraday cells to rotate the
planes of polarization directly, it is possible to obtain a self-compensating,
automatic instrument.
The rotating analyzer ellipsometer (RAE) is a photometric configura
tion that does not rely on a null but rather imposes a time dependence on
a transmitted beam that can later be decoded for the desired information.
This is accomplished in the RAE by a rotating analyzer prism that
generates a sinusoidal output signal from the detector, representing the
"unrolled" polarization ellipse. The polarizer is kept fixed. When (typi
cally) no compensator is used, the system is completely achromatic; the
only wavelength-dependent element is the sample itself. Here, the ellipso
metric angles are given by the relations
tan 1f;=tan p[ ( c+a )( c-a ) - T/2 9.
cosA=b ( c2 _a2 ) -1/2 10.
REFLECTING
SA MPLE
L-----.JH �1
SOURCE J� H'-----J
POLARIZER COMPEN
(optional)
TO ANALYSER DETECTOR
and P and A are the polarizer and analyzer azimuths, respectively. This
type of ellipsometer is ideally suited for a minicomputer since the
sinusoidal detector output can be processed digitally with an analog-to
digital converter and Fourier analysis. Because it operates as a modula
tion spectrometer with high optical efficiency, it can be used with weak
continuum sources and hence is ideally suited for spectroscopic applica
tions. It can also follow reactions as long as they are slower than the
rotation rate of the analyzer, which is usually of the order of 50 turns .per
second.
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
All remarks made concerning the RAE apply equally to the rotating
polarizer ellipsometer (RPE), which differs from the RAE only in that
source and detector are interchanged. Placing the dispersing optics
between sample and detector instead of between source and detector
by Duke University on 06/13/12. For personal use only.
offers advantages when the sample and/or ambient may generate light,
as in a plasma reactor.
If faster rates of measurement are needed, another photometric config
uration, the phase modulation ellipsometer, can be used. It can operate in
the kHz range. The design is similar to that of a null ellipsometer except
that the compensator is replaced by a variable retardation device whose
phase is modulated by electro-optical or piezo-optical techniques, for
instance a piezobirefringent silica plate typically driven at 50 kHz. If
P=90°, A=45°, C (the compensator azimuth axis)=45°, and the phase
modulation of the compensator is of the form Osinwt (instead of 90° as
with the null ellipsometer), and if 8 is adjusted to
i.e. (J= 137.8° 12.
where Jo is the Bessel function of zero order, then the detected signal can
. be analyzed with lock-in amplifiers centered at w and 2w and the
ellipsometric angles obtained from the following relations:
Rw =I"!od = 2(sin2o/ )(sin t:. )J1 ((J), 13.
Traditional Applications
Thin-film technologists have long used null ellipsometers to determine
the thickness of various dielectric films deposited on semiconductors (e.g.
8). Cleaning procedures for Si or III-V compounds (9), qualification tests
of polished surfaces ( lOa, b, 11), and routine analysis of the composition
of oxynitride layers deposited onto Si (12) are typical examples of
traditional applications where an ellipsometer is used as an improved
refIectometer. Here, the analysis is restricted to the 3-phase model where
the film is nonabsorbing and only its average thickness and refractive
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
REAL-TIME ANALYSIS
DELTA
270
180
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
90
by Duke University on 06/13/12. For personal use only.
DELTA
BARE SUBSTRATE
b PSI
Figure 4 Use of the (>fI,�) plane in connection with the 3-phase model. (a) Typical
trajectories corresponding to the deposition of a dielectric (n\ J.14-iO) and a weakly
=
absorbing (n\ =2-;0.5) film onto GaAs (00 =70°, .\=6328 A). The crosses along the
trajectories correspond to loo-A increments of the film thickness. (b) Definition of the
residual used in Eqs. 15 and 16 of the text.
ELLIPSOMETRY 105
15.
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
where CJ; is the minimal squared distance between the experimental point
i and the trajectory calculated for n=ntest' That is,
by Duke University on 06/13/12. For personal use only.
16 .
where [1[;(dJ, Ll(dJJ is the point calculated along the ntest trajectory
corresponding to a thickness d; of the film (Figure 4b). For each i, the
thickness d is varied until the individual CJ; is minimal, i.e. SCJ;/Sd=O
when d= d;. The "best" value of n test is that for which the total residual
F( n test ) is minimal. It is clear that d is only a dummy variable in this
ca lcu la tion. Once the best value has been found for n tes!' the correspond
ing d; are the thicknesses of the film at the various experimental points.
In other words, the assumption of the 3-phase model allows both real
and imaginary parts of the refractive index of the film to be determined
even though the film thickness is not known a priori at any single point.
Furthermore, changes in the refractive index during deposition or other
deviations from the assumptions of the 3-phase model can readily be seen
from changes in the shape of the trajectory. This is particularly useful for
monitoring the compositional change of a film as for example in the case
of III-V ternary alloys.
Cahan (13) proposed using the relative reflectance change SRIR as a
third indepe nde nt ellipsometric parameter for real-time measurements.
Although this parameter cannot be measured to the same accuracy as I[;
and Ll and because it is more sensitive to experimental artifacts such as
increased scattering from surface roughness, it can be used as an addi
tional tool to test the applicability of the 3-phase model in specific
situations. A detailed discussion is given in (13). This is especially useful
for the difficult case of electrochemical reactions, where one does,
however, have the advantage of being able to cycle repeatedly.between
bare and reacted conditions and' thus to determine SRIR by modulation
spectroscopic techniques.
106 THEETEN & ASPNES
Applications
claimed for GaAlAs (14) and 0.4% for Ge-Si alloys (15) from the fact
that the components of the alloy, for instance GaAs and AlAs, have
markedly different dielectric functions. With a spectroscopic capability,
by Duke University on 06/13/12. For personal use only.
190r----
DELTA 9=710
(deg.) A=6328A
180
17
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
160
by Duke University on 06/13/12. For personal use only.
150
6 8 10 12
a PSI (deg.)
4.3 1.2
n k
4.1 to
n
3.9 ·8
3.7 .6
3.5
3.3
... .. a ••••••
k
·4
.2
GaAs (Ga,AIlAs
3.1 0
a 500 1000
b thickness (AJ
Figure 5 GaAs-->GaAlAs transition in a metallorganic CVD system. (a) Experimental
trajectory in the (,p, ll) plane. The calculated trajectory assuming a 3-phase model is given
for comparison (solid line), with crosses corresponding to loo-A increments in thickness.
(b) Refractive index profile calculated from the data.
108 THEETEN & ASPNES
----�
180--
DELTA
(deg.)
170
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
160
by Duke University on 06/13/12. For personal use only.
150
8 10 12 14
a PS I (deg.)
4.3 1.2
n ... n k
...... ...
4.1 1.0
3.9 .8
3.7 .6
• •
k
3.5 ·4
33 .2
Ga,AI)As GaAs
I 0
3.1
0 500 1000
0
b thickness (A)
Figure 6 GaAlAs ..... GaAs transition in the same metallorganic CVD system (continuation
of the experiment shown in Figure 5) (a) Experimental trajectory and comparison with
3-phase model (solid line), with crosses corresponding to loo-A increments in thickness. (b)
Refractive index profile calculated from the data.
ELLIPSOMETRY 109
to the data. This is confirmed in Figure 6b, where the refractive index
profile calculated from the data of Figure 6a is shown as a function of
film thickness. The GaAIAs� GaAs transition occurs in less than 200 A.
Real-time ellipsometry is thus capable of determining the width of the
GaAIAs� GaAs and GaAs� GaAlAs transition regions in CVD deposi
tion. By repeating the experiments of Figures 5 and 6, one can directly
estimate the influence of various growth parameters such as temperature,
As/metallic species ratio, hydrodynamic regimes, etc upon the sharpness
of the transition. This procedure is obviously much faster than for
example a posteriori ion milling and Auger analysis, in addition to being
nondestructive and free from sputterin g artifacts.
KINETIC STUDIES The in situ capability can be very useful for studying
small changes in the kinetics of a reaction and for establishing weak
correlations of a reaction to parameters otherwise undetectable. This is
illustrated in Figure 7, where the rate of oxidation of Si under 02 and
H 2 0- 0 2 atmospheres is analyzed (17). The ellipsometer measures the
thickness of the growing oxide, and the influence of a small amount of
H2 0 in 02 is demonstrated. Oxidation curves corresponding to ambients
of dry 02 and 02 +2000 ppm H2 0 are first obtained. In another
experiment, H2 0 is turned on and off sequentially. It is seen in Figure 7
that when the ambient is changed from dry O2 to 02 + H2 0, the
oxidation behavior proceeds in parallel to the H2 0-02 curve. When the
H2 0 component is absent, the rate reverts to that for dry 02 . Real-time
analysis proves that the presence of H2 0 influences the oxidation and
that this effect is reversible. The effect can be evaluated quantitatively by
changing the amount of H2 0 during the on-off sequence so that several
dilutions can be tested in one run. Because of the accuracy of the
11 0 THEETEN & ASPNES
1 ppm H20 in 02
H20-02off
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
1000
by Duke University on 06/13/12. For personal use only.
500
t (min)
Figure 7 Thermal oxidation of Si under dry O2 and H20-containing O2 atmospheres [after
(17»).
Single crystal
46
<I
48
36.5
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
35.5
40
�
by Duke University on 06/13/12. For personal use only.
o
I ! ! I
100 150 200
Time, sec
---------------,-------�
I I I
I
0.5
Ir=----� I
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
I I
I I
fl./Pa
o 0.2 0.4 0.6 0.8
by Duke University on 06/13/12. For personal use only.
°-
. __
__ �4
__,_-0�6
----�._��
0
-0.04
-0.08
T=97'K
-0.12
Figure 9 Variatioo of the ellipsometric angles of reflected light (A = 6328 A) from a (000 I)
graphite surface exposed to Xe [after (21»).
and SiH4 molecules reach the Si3N4 surface and decompose, leaving Si
atoms at the surface. The (1/;, d) point moves along a trajectory that
cOI:lverges to a location corresponding to amorphous Si. The fact that the
thick deposit is effectively a-Si was checked by in situ RHEED and Auger
measurements. Of interest here is the dashed calculated curve, which
supposes a uniform deposition of a-Si on the substrate. Clearly, there is
gross initial disagreement although the calculated and experimental curves
tend to converge as the thickness increases. This disagreement is due to
the growth mechanism, which proceeds rather through the formation of
small Si nuclei since the sticking probability of Si atoms on Si is much
higher than that of Si atoms on Si3N4• The nuclei grow independently
and finally coalesce. By assuming a distribution of hemispherical
amorphous Si nuclei with an average separation dN, the trajectory can be
recalculated as shown in the dotted-dashed curves of Figure 1 0. Good
agreement is obtained with dN � 100 A. The optical response of nuclea
tion growth is markedly different from that of layer-by-layer growth
because, when the nuclei are small, the growth layer "seen" by the
ellipsometer is an average mixture of voids and absorbing materiaL
Initially only a small fraction of absorbing material is present and the
average behaves as a dielectric instead of the absorbing film that one
would have for homogeneous deposition.
Limitations
Real-time ellipsometry is well adapted to kinetic studies. Its use with thin
film deposition systems can have a direct impact on the control of the
process. This in-situ methodology, however, cannot at present analyze
more complex structures in which not only the film but also the underly
ing substrate is changing with time. This has been pointed out by Muller
114 THEETEN & ASPNES
200
---- ---
180 --
...
--
__ CALC.
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
- ...
--
o --
- ._ 50 A .......
...... ...
. ·
.\
,. '...
.\ EXP.
,. .\, "" ,
<J ,
by Duke University on 06/13/12. For personal use only.
160 �\ "-100 A ,
,
\
\.
� ,
'. �
'.
-,, "
'.... ... ,r
'0' ....... .... .. ... -'
140 200A .. -.... --_ ....
11 13 15 17 19
�\cleg)
Figure 10 Determination of nucleation mechanism and center spacing for a-Si grown by
CVD on Si3N4 [after (26)].
(18) and Cahan (13) for electrochemical systems where surface roughen
ing occurs simultaneously with the formation or dissolution of a passivat
ing film. This can be seen in Figure 8 where a general decrease of 1/1 is
observed as a background to the modulation due to current changes.
Sometimes not only the vapor-film interface but also the film-substrate
interface is modified during film growth. This effect is seen for plasma
oxidation of GaAs (27) where unoxidized arsenic tends to pile up at the
oxide-GaAs interface during the plasma exposure. Spectroscopic el
lipsometry and other complementary techniques are necessary to in
vestigate these more complicated situations.
Methodology
interface, without etching the film. This assumes that the film can be
treated as a succession of layers of variable thicknesses and of composi
tions consisting of mixtures of known materials. The effective medium
approximation (EMA) is used to evaluate the dielectric functions from
by Duke University on 06/13/12. For personal use only.
cally isotropic materials show that the simple expressions can be rela
tively accurate. The physically most appealing simple expression, which is
symmetric in the constituents and already adapted to multicomponent
by Duke University on 06/13/12. For personal use only.
17.
18.
where VA' VB" " are the volume fractions of constituents A, B, .... Equa
tions 1 7 and 18 give the effective medium approximation.
N
F= � I pcaIc(Ei)-Pexp;1 2, 19.
;=1
where E; are the energies at which the Pexp i are measured. In linear
regression analysis F is minimized by standard least-squares techniques
(35), which also give the 90% confidence limits of the best-fit parameters.
These confidence limits have to be small; otherwise the model is irrele
vant or an excessive number of parameters have been used.
Applications
SI/SI0 INTERFACE
2
Taft & Cordes (3 6) recently showed that a 3 -phase
model analysis of ellipsometric data of oxides grown thermally on Si does
not give consistent results. Measurements were made as a function of
oxide thickness, using HF in a step-by-step chemical etching process. The
lack of consistency is shown in Figure II, where the index of refraction
of the film varies as a function of thickness. By adding a 4-8-A interface
ELLIPSOMETRY 117
2
Q 1.48 'x
�
()
oCt �x 820°C
c:::
u.. �x xOXIDE ONLY
W -0--0--0"- 0'---'0- 0 8A INTERFILM
x ----
x
w --1{= -- 0--
�
0 xOXIDE ONLY
Cl _0___'0 --- 0
y----x__
0
2 1.46 04AINTERFILM
by Duke University on 06/13/12. For personal use only.
film of refractive index 2.8 at 5461 A between Si02 and Si, the top film
index was found to be independent of thickness.
A similar conclusion was obtained when spectroscopic ellipsometric
data were analyzed (3 7 ). Figure 12 gives the differences in tan t/; and cos
tl between experiment and various models for a 13 60-A oxide overlayer
on Si (100). The 3 -phase model with its single parameter, the oxide
thickness, deviates strongly from experiment. By allowing some voids to
exist in the Si02 layer in a 2-parameter model, a better fit is obtained.
However, the fit is not satisfactory, especially for cos tl in the 3 -5-eV
range. By incorporating an interface of chemically mixed Sio75(Si02 )o25
with a thickness of 6 . 6 A in a 3 -parameter model, far bette; agreeme� t
results. In the latter case the residual is six times lower than with the
I-parameter model. Similar analyses on (110) and (Il l) Si substrates
show that an interface of width 7 �±: 2 A consisting of atomically mixed Si
and ° of average stoichiometry SiOo.4±o. is always present. The spectro
scopic ellipsometer has a high sensitivity to the interface region because
of the high optical contrast between Si and its mixed oxides.
In this case the interface is extremely sharp and a 4-phase model is
sufficient. However, interface regions can be much wider. In order to
illustrate the capabilities of ellipsometry, we discuss now an extremely
wide interface: an ion-implantation-induced defect profile, where a
Gaussian spatial variation is expected.
118 THEETEN & ASPNES
0
,.
,
,.I I \,
\ . '��.
, ,
�-0.1 \ '
� . I
c
E
I· Si- SiOz
.I (100) INTERFACE
(;() I· 0
��
dox 1360 A
·1
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
-0.2
\. :--- = PARAMETER
·1 MODELS
I· 3-
'I
by Duke University on 06/13/12. For personal use only.
- 0.3
\.
V
E (eV)
Figure 12 Spectroscopic ellipsometric analysis of a 1360·11. Si02 film grown on Si (100)
[after (37)].
ION IMPLANTATION DEFECTS Kim & Park (38) determined the effective
refractive index of ion-implanted GaAs wafers as a function of depth,
using the chemical layer-stripping technique. The imaginary part of the
refractive index is very sensitive to the presence of defects induced by the
ion bombardment with doses as low as 1012 ions cm -2 being detected.
Figure 13 gives the results for lOI4 S and Cr ions cm -2. A comparison
with L SS and damage theory allows one to draw interesting conclusions
about the ion mass and implant dose dependence of the defect creation
mechan ism, and also the annealing mechanism when a similar study is
done after thermal treatment.
Again, comparable results can be obtained using spectroscopic el
lipsometry (39) assuming that the implanted ions locally transform the
crystalline GaAs (c-GaAs) substrate into amorphous GaAs (a-GaAs).
The defect profile can then be described by � series of layers of fixed
thickness, typically 50 A , and variable composition, typically a physical
ELLIPSOMETRY 119
1.4 GaAs:Cr
00 C (10 ions/cm .120KeV)
,4 2
::t.
<l - - - - Damage Theory
1.2 --- LSS Profile
'-"'
I-
Z
!:!:!
(.)
i:1: 1 .0
u.
w
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
0
(.)
z 0.8
0
by Duke University on 06/13/12. For personal use only.
I-
(.)
�
I-
><
W
Z
W
C-' •
z
ct
J:
(.)
01L----�--.-L-----�--��
1000 2000 3000 4000
DEPTH (}�)
Figure 13 Dependence of the imaginary part of the index of refraction on depth for
implanted GaAs wafers using combined ellipsometry/chemical layer stripping [after (38)].
100 -- -
%
a
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
GaAs
100
by Duke University on 06/13/12. For personal use only.
50
Figure 14 Defect profiles for lOO-keV Ar ions implanted into GaAs as deduced from
spectroscopic measurements [after (39)].
Limitations
The above analyses rely on the validity of the EMA and on reference
dielectric functions measured separately on standard samples. Standard
spectra are needed, of course, for any spectroscopic method using either
photons or electrons; owing to the newness of the technique very few
accurate standard spectra are yet available for use in spectroscopic
ellipsometry. Physical mixtures are usually assumed, because "first
principles" calculations required to obtain the dielectric response of
chemical mixtures may be inaccurate and difficult to perform in general.
The analysis of surface roughness is also difficult since dangling bonds,
surface relief, and the inherent anisotropy of the surface region are
boundary conditions not easily described in the above assumptions.
Although some success has been obtained in the case of microscopic
surface roughness on Si layers (3 4), complementary techniques such as
grazing incidence x-ray reflection (40) should be used for a detailed
study.
ELLIPSOMETRY 121
CONCLUSION
Literature Cited
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
29. Bruggeman, D. A. G. 1 935. Ann. Phys. 36. Taft, E., Cordes, L. 1 979. J. Elec
(5 Folge) 24;636 trochem . Soc. 1 26; 1 3 1
30. Granqvist, C. G., Hunderi, O. 1 976. 37. Aspnes, D . E., Theeten, J . B . 1 979.
Phys. Rev. B 1 6 ; 3513 Phys. Rev. Lett. 43 ; 1 046
31. Sheng, P . 1 980. Phys. Rev. Lett. 45:60 38. Kim, Q., Park, Y. S. 1 980. J. Appl.
32. Aspnes, D. E. Theeten, J. B., Hottier, Phys. 5 1 : 2024
F. 1 979. Phys. Rev. B 20:3292 39. Theeten, J. B., Simondet, F., Erman,
33. Bergman, D. J. 1 980. Phys. Rev. Lett. M., Pemas, J. 1 980. Proc. 4th Int. Conf.
44; 1 285 on Solid Surfaces, Cannes. Supp\. to Le
34. Milton, G. W. 1 980. Appl. Phys. Lett. Vide, 1es couches minces, No. 20 1 , p.
37;300 107 1 . Paris; Soc. Fran�se du Vide
35. Keeping, E. S. 1 962. Introduction to 40. Nevot, L., Croce, P. 1 980. Rev. Phys.
Statistical Inference, Ch. 12, Prince Appl. 1 5 ; 76 1
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org
CONTENTS
PREFATORY CHAPTER
A Perspective on Martensitic Nucleation, G. B. Olson and M. Cohen
ExPERIMENTAL AND THEORETICAL METHODS
Higher Order Elastic Constants of Solids, Yosio Hiki 51
Weak-Bearn Electron Microscopy, D. J. H. Cockayne 75
Ellipsometry in Thin Film Analysis, J. B. Theeten and D. E. Aspnes 97
Annu. Rev. Mater. Sci. 1981.11:97-122. Downloaded from www.annualreviews.org