Elastic properties of zinc-blende and wurtzite AlN, GaN, and InN

A. F. Wright
Citation: J. Appl. Phys. 82, 2833 (1997); doi: 10.1063/1.366114
View online: http://dx.doi.org/10.1063/1.366114
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v82/i6
Published by the American Institute of Physics.

Related Articles
Surface elasticity effect on the size-dependent elastic property of nanowires
J. Appl. Phys. 111, 083506 (2012)
Modeling of the microstructure of carbon nanotube with two nonlocal elasticity theories
J. Appl. Phys. 111, 034315 (2012)
X-ray interference effects on the determination of structural data in ultrathin La2/3Sr1/3MnO3 epitaxial thin films
Appl. Phys. Lett. 99, 221901 (2011)
A variable constraint tube model for size effects of polymer nano-structures
Appl. Phys. Lett. 99, 191910 (2011)
Existence of a giant hypersonic elastic mirror in porous silicon superlattices
Appl. Phys. Lett. 99, 171901 (2011)

Additional information on J. Appl. Phys.

Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

Elastic properties of zinc-blende and wurtzite AlN, GaN, and InN

A. F. Wrighta)
Sandia National Laboratories, Albuquerque, New Mexico 87185-1415

~Received 12 May 1997; accepted for publication 5 June 1997!

Elastic constants for zinc-blende and wurtzite AlN, GaN, and InN are obtained from
density-functional-theory calculations utilizing ab initio pseudopotentials and plane-wave
expansions. Detailed comparisons are made with the available measured values and with results
obtained in previous theoretical studies. These comparisons reveal clear discrepancies between the
different sets of elastic constants which are further highlighted by examining derived quantities such
as the perpendicular strain in a lattice-mismatched epitaxial film and the change in the wurtzite c/a
ratio under hydrostatic pressure. Trends among results for the three compounds are also examined
as well as differences between results for the zinc-blende and wurtzite phases. 1997 American
Institute of Physics. @S0021-8979~97!00418-0#

I. INTRODUCTION

II. ZINC BLENDE

Virtually all epitaxial AlN, GaN, and InN films are expected to contain some amount of residual strain due to differences in the film and substrate lattice constants and in
their thermal expansion coefficients. This unintentional
strain can modify the films electronic structure, and accounting for this effect is important when interpreting optical
spectra.1,2 Alternatively, a semiconductor film may be grown
intentionally strained in order to modify the performance of
devices such as laser diodes or light-emitting diodes. Compressive strain, for example, has been found to change the
TE/TM gain ratio in laser diodes grown from other IIIV
compounds. Regardless of whether strain is introduced intentionally or unintentionally, predicting its effect on electronic
properties requires knowledge of the materials elastic properties, and specifically the elastic constants which describe
the response to an applied macroscopic stress.
In this article, we report elastic constants for zinc-blende
and wurtzite AlN, GaN, and InN obtained from densityfunctional-theory calculations utilizing ab initio pseudopotentials and plane-wave expansions. Detailed comparisons
with the available measured values and with results obtained
in previous theoretical studies reveal clear discrepancies between the different sets of elastic constants. These differences are further highlighted by examining derived quantities
such as the perpendicular strain in a lattice-mismatched epitaxial film and the change in the wurtzite c/a ratio under
hydrostatic pressure. We also examine trends among results
for the three compounds and between results for the zincblende and wurtzite phases showing, for example, differences in their behavior under shear strain along the wurtzite
c axis ~the @111# direction in zinc blende!. The remainder of
the article is organized as follows. We describe the zincblende calculations and their results in Sec. II and the wurtzite calculations in Sec. III. In Sec. IV we compare the elastic
properties of the zinc-blende and wurtzite phases and then
summarize our results in Sec. V.
a!

Electronic mail: afwrigh@sandia.gov

J. Appl. Phys. 82 (6), 15 September 1997

Elastic properties of zinc-blende crystals are described in

terms of three elastic constants denoted C 11,c , C 12,c and
C 44,c . 3 ~The subscript c for cubic is used to distinguish
zinc-blende and wurtzite values.! The first two were extracted from calculations for the bulk modulus, B c , and the
shear modulus, C s,c , using the expressions B c 5 (C 11,c
1 2C 12,c )/3 and C s,c 5 (C 11,c 2 C 12,c )/2. The bulk modulus was determined by fitting the total energy versus volume
to the Murnaghan equation.4 The shear modulus was determined by fitting the energy versus volume-conserving tetragonal strain to the expression
DE56V 0 C s,c d 2 ,

~1!

where DE is the energy change ~relative to equilibrium!, V 0

is the unit-cell volume, and the strain parameter is defined as
d [ ( e' 2 e i )/3 where e' is the strain along the @001#
direction and e i is the strain in the ~001! plane. The other
shear modulus, C 44,c , was likewise determined by fitting the
energy versus volume-conserving trigonal strain to the expression
DE56V 0 C 44,c d 2 ,

~2!

where the strain parameter has the same definition d

[ ( e' 2 e i )/3, but e' is the strain along the @111# direction
and e i the strain in the ~111! plane. The calculations for C 44,c
required the energy to be minimized with respect to the separation of the cation and anion sublattices along the @111#
direction. This relaxation is specified by the Kleinman parameter, 0 < z c < 1, where the lower limit corresponds to
a separation consistent with the macroscopic strain and the
upper limit to the separation in an unstrained crystal.5 For the
shear strain calculations, we used values of d between 20.02
and 10.02 except for C 12,c of InN where we increased the
range to 60.03 in order to resolve the shallow energy curve.
Our density-functional-theory6 ~DFT! calculations were
performed within the local-density approximation ~LDA! using the plane-wave pseudopotential method. General features
of this technique as applied to AlN, GaN, and InN are given
elsewhere.7,8 For zinc blende, we used a 480 Ry energy cutoff for the plane-wave expansion in the GaN calculations and
a 200 Ry cutoff in the AlN and InN calculations. Brillouin-

0021-8979/97/82(6)/2833/7/$10.00

1997 American Institute of Physics

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

2833

TABLE I. Calculated zinc-blende elastic constants ~GPa! and Kleinman

parameters.
AlN

GaN

InN

TABLE II. Directly calculated zone-center TO phonon frequencies (cm21)

and estimated values ~in parentheses! using Eq. ~5! and the quantities given
in Table I. Also given are results from direct measurements and experimental estimates using Eq. ~5! along with measured wurtzite frequencies.

Present Other calc. Present Other calc.b Present Other calc.b

a

C 11,c
C 12,c
C 44,c

304
160
193

C 44,c,0
zc

237
0.55

348, 304
168,a 152b
135,a 199b
230b
0.6b

293
159
155

296
154
206

187
125
86

184
116
177

200
0.61

225
0.5

138
0.80

209
0.7

Present measurement
Kim et al.a
direct measurement
Expt. estimate

AlN

GaN

InN

666 ~682!
680 ~523!

550 ~560!
603 ~444!

470 ~467!
540 ~418!

654,d 667,e,f 668g

556,b 552c
556,h 551g,i,j,k

462l

a
a

zone sampling was accomplished using special k points generated from MonkhorstPack9 parameters $6,6,6%. For the
wurtzite calculations ~see below!, we used MonkhorstPack
parameters $5,5,3% and the same energy cutoffs.
Table I shows our zinc-blende elastic constants together
with results from two earlier theoretical studies. Ruiz et al.10
calculated elastic constants for AlN using a method based on
HartreeFock theory. The noticeable differences between
our results can likely be traced to differences between DFT
and HartreeFock theory. The only other complete set of
elastic constants is from the calculations of Kim, Lambrecht
and Segall11,12 who utilized DFT implemented via the fullpotential linear muffin-tin orbitals ~FP-LMTOs! method. The
agreement for AlN is good with the largest discrepancy being 5% in C 12,c . The agreement is also good for C 11,c and
C 12,c of GaN and InN. However, our GaN C 44,c values differ
by about 33% and our InN C 44,c values differ by a factor
of 2.
To resolve these discrepancies, we examined predictions
for the zone-center transverse optical ~TO! phonon frequency, v TO . This quantity can be calculated directly using
the expression
DE5

1
2
m v TO
Dx 2 ,
2

~3!

where Dx is the displacement of the cation and anion sublattices along the @111# direction, DE denotes the change in
energy for this displacement, and m is the reduced mass for
the cationanion system. However, the frequency can also
be estimated using the expression
C 44,c 5C 44,c,02

m
a0
v TOz c
V0
4

lists measured zinc-blende frequencies and the fourth gives

estimates for zinc blende using the expression,

v TO5 @~ v 2Al 12 v 2El ! /3# 2 ,

~5!

where v Al and v El are measured zone-center TO phonon

frequencies for wurtzite. Our values are within 2% of experimental estimates and within 1% of the direct measurements.
Before leaving this section, we note several trends
among the zinc-blende elastic constants that also occur
among the wurtzite results given below. First, the C 11,c and
C 12,c values for AlN and GaN are similar while the corresponding InN values are smaller in magnitude. This is consistent with the general observation that AlN and GaN are
more robust chemically and thermally than InN. Second, the
C 44,c values decrease in magnitude from AlN to GaN to InN
and the Kleinman parameters increase in this series. We also
note that, while no measurements are available for the zincblende elastic constants, Sherwin and Drummond14 have estimated values by inverting Martins transformation15 and
using measured wurtzite elastic constants. Our agreement
with their results is poor, likely due to inaccuracies in the
earlier measured wurtzite values and the nonuniqueness of
the inverse transformation.

~4!

utilizing calculated values for the shear elastic constants with

(C 44,c ) and without (C 44,c,0) internal relaxation, the Kleinman parameter, and the cubic lattice constant a 0 . 8 The first
two rows of Table II illustrate that Eqs. ~3! and ~4! yield
nearly the same frequencies in our study, whereas they gave
different values for all three compounds in the study of Kim
et al.12 This disagreement suggests that there may be some
problem with the C 44,c obtained by Kim et al. This suggestion is further supported by recent FP-LMTO calculations
which yield zinc-blende elastic constants in good agreement
with our values.13 For completeness, the third row in Table II
2834

DFT-LDA, FP-LMTO method, Refs. 11 and 12.

Raman spectra, Ref. 24.
c
Infrared spectra, Ref. 25.
d
Raman spectra, Ref. 17.
e
Raman spectra, Ref. 26.
f
Raman spectra, Ref. 27.
g
Raman spectra, Ref. 28.
h
Infrared spectra, Ref. 29.
i
Infrared spectra, Ref. 30.
j
Raman spectra, Ref. 31.
k
Infrared spectra, Ref. 32.
l
Raman spectra, Ref. 33.
b

HartreeFock, Ref. 10.

b
DFT LDA, FP-LMTO method, Refs. 11 and 12.

III. WURTZITE

There are five elastic constants for the wurtzite structure

denoted C 11 , C 12 , C 13 , C 33 , and C 44 . These quantities
were determined from energy calculations for five different
strain configurations of the form D5 @ e 1 ,e 2 ,e 3 ,e 4 ,e 5 ,e 6 #
where the strain parameters, e, are related to the usual strain
components, e, via e 1 5 e xx , e 2 5 e y y , e 3 5 e zz , e 4
5 e zx 1 e xz , e 5 5 e zy 1 e yz , and e 6 5 e yx 1 e xy with
the z direction taken along the wurtzite c axis.3 The configurations used in our calculations along with the corresponding
strain-energy densities, U, are given below in terms of the
strain variable d.

J. Appl. Phys., Vol. 82, No. 6, 15 September 1997

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

A. F. Wright

TABLE III. Wurtzite AlN elastic constants ~GPa!. Values calculated using Martins transformation are shown
in parentheses.
C 11

C 12

C 13

C 33

C 44

Measurements by Tsubouchi et al. ~Ref. 16!

Measurements by McNeil et al. ~Ref. 17!

345
411610

125
149610

120
9964

395
389610

118
12565

201
210

Calculations by Ruiz et al. ~Ref. 10!

Calculations by Kim et al. ~Refs. 11 and 12!
Present calculations

464
~398!
396

149
~140!
137

116
127
108

409
382
373

128
~96!
116

228
218
207

D 1 5 ~ d , d ,0,0,0,0 ! U 1 5 ~ C 111C 12! d 2 ,

~6!

D 2 5 ~ d , d ,22 d ,0,0,0! U 2
5 ~ C 111C 1224C 1312C 33! d 2 ,

~7!

D 3 5 ~ 0,0,d ,0,0,0! U 3 5 21 C 33d 2 ,

~8!

D 4 5 ~ 0,0,0,0,0,d ! U 4 5 41 ~ C 112C 12! d 2 ,

~9!

D 5 5 ~ 0,0,0,d , d ,0! U 5 5C 44d 2 .

~10!

For small strains, the energy relative to the equilibrium

structure is given by E 5 V 0 U. To determine the elastic
constants, we therefore calculated energies for d between
20.02 and 10.02, including full relaxation of the internal
degrees of freedom, and fit the results to a second-order
polynomial. Our results are given in Tables IIIV together
with the available measured values and results from previous
theoretical studies. In comparing values from two different
sets, we give the root mean square ~rms! of the differences
and in some cases point out which moduli differ the most.
For each set in Tables IIIV, we also give the bulk modulus
calculated via the expression
B5

C 33~ C 111C 12! 22 ~ C 13! 2

.
C 111C 1224C 1312C 33

~11!

Experimental results for AlN were derived from a 1981

surface-acoustic-wave study by Tsubouchi et al.16 and a
more recent Brillouin scattering study by McNeil et al.17 The
rms difference between the two sets is 33 GPa with most of
this occurring in C 11 . The three sets of measured values for
GaN were estimated from temperature-dependent broadening
of powder x-ray diffraction spectra by Sheleg and
Savastenko,18 and Brillouin scattering by Polian et al.19 and
Takagi et al.20 Results from the two Brillouin scattering
measurements differ from each other by 26 GPa rms, and
from the x-ray measurements by 81 and 87 GPa. These
rather large discrepancies with the GaN results of Sheleg and

Savastenko suggest that their values may not be accurate.

Examination of derived quantities ~see below! further supports this suggestion. Only one set of measurements is currently available for InN from the temperature-dependent
x-ray broadening study of Sheleg and Savastenko.18
Our AlN results differ from those of Tsubouchi et al. by
26 GPa rms and from those of McNeil et al. by 13 GPa rms.
Of note is that our C 11 is close to the value found by McNeil
et al. suggesting that their results are more accurate than
those of Tsubouchi et al. For GaN, our results agree well
with the values found by Polian et al. ~13 GPa rms!, somewhat worse with the results of Takagi et al. ~23 GPa rms!,
and poorly with the earlier results of Sheleg and Savastenko
~80 GPa rms!. The InN results of Sheleg and Savastenko are
in somewhat better agreement ~32 GPa rms! than was the
case for GaN, but not as good as was found for the AlN and
GaN results from Brillouin scattering.
In terms of previous theoretical results, our AlN results
differ from those of Ruiz et al.10 by 35 GPa rms with the
largest difference being 68 GPa in C 11 . As noted above, one
should expect some intrinsic differences between the two
sets of results simply due to the different theoretical foundations. More important, however, Ruiz et al. did not relax
internal degrees of freedom in the structures used for calculating C 44 , and only relaxed them for the largest strain values ~d560.02! otherwise. This could lead to noticeable differences and, in particular, one would expect their moduli to
be larger than ours which is what is observed.
The theoretical foundation utilized in the wurtzite calculations of Kim et al.11,12 was the same as used by us ~DFT!,
however they implemented it via the FP-LMTO method.
They performed direct calculations for two of the wurtzite
elastic constants, C 13 and C 33 , estimating the others from
Martins transformation15 using their previously calculated
zinc-blende elastic constants. Martins transformation is
specified by the expressions

TABLE IV. Wurtzite GaN elastic constants ~GPa!. Values calculated using Martins transformation are shown
in parentheses.
C 11
Measurements by Sheleg and Savastenko ~Ref. 18!
Measurements by Polian et al.. ~Ref. 19!
Measurements by Takagi et al. ~Ref. 20!
Calculations by Kim et al. ~Refs. 11 and 12!
Present calculations

J. Appl. Phys., Vol. 82, No. 6, 15 September 1997

C 12

C 13

C 33

C 44

296618 130610 15865 267617 2462 195

390615 145620 106620 398620 105610 210
374
106
70
379
101
180
~396!
367

~144!
135

100
103

392
405

~91!
95

207
202

A. F. Wright

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

2835

TABLE V. Wurtzite InN elastic constants ~GPa!. Values calculated using Martins transformation are shown in
parentheses.

Measurements by Sheleg and Savastenko ~Ref. 18!

Calculations by Kim et al. ~Refs. 11 and 12!
Present calculations

C 11

C 12

C 13

C 33

C 44

19067

10463

12167

18266

1061

139

~271!
223

~124!
115

94
92

200
224

~46!
48

147
141

C 115 61 ~ 3C 11,c 13C 12,c 16C 44,c ! 23D 2 / ~ C 11,c 2C 12,c

1C 44,c ! ,

~12!

~13!

C 135 61 ~ 2C 11,c 14C 12,c 24C 44,c ! ,

~14!

C 335 ~ 2C 11,c 14C 12,c 18C 44,c ! ,

~15!

1
6

C 445 ~ 2C 11,c 22C 12,c 12C 44,c ! 26D 2 / ~ C 11,c 2C 12,c

1
6

14C 44,c ! ,

~16!

e 3 5 e zz [

D[

~ C 11,c 2C 12,c 22C 44,c ! .

c 8 2c
,
c

c 8 5c 122

~17!

~21!

S DG

C 13 a 8 2a
C 33
a

~22!

or, in terms of the strain components,

The second term in the expressions for C 11 , C 12 , and C 44

accounts for differences between trigonal and hexagonal systems. The overall agreement between the two sets of DFT
results is quite good with the AlN values differing by 13 GPa
rms, the GaN values by 15 GPa rms, and the InN values by
a somewhat larger 24 GPa rms. We should note, however,
that the observed differences in C 11 and C 12 for GaN and
InN are mostly due to our different C 44,c values. To illustrate
this, we note that adding Eqs. ~12! and ~13! yields
C 111C 1252B c 1 32 C 44,c

~18!

using the relationship B c 5 (C 11,c 1 2C 12,c )/3. Substituting

our zinc-blende InN B c and C 44,c values yields C 11 1 C 12
5 350 GPa, close to our directly calculated value 337 GPa.
Kim et al. found a similar zinc-blende bulk modulus, B c
5 139 GPa so that their larger value for C 11 1 C 12 ~395
GPa! is mostly due to their different C 44,c value. Similar
discrepancies arise for the other GaN and InN transformed
elastic moduli including transformed values of C 13 and C 33 .
In the remainder of this section, we discuss the strain in
a wurtzite crystal resulting from two common forms of applied stress: ~1! Biaxial stress resulting from epitaxial film
growth on a lattice-mismatched substrate ~c axis perpendicular to the substrate!, and ~2! hydrostatic pressure such as
occurs in a diamond anvil cell. The starting point for this
discussion is the expression for the strain energy per unit
cell,
E ~ e 1 ,e 3 ! 5

1
2

V 0 ~ 2C 11e 21 12C 12e 21 14C 13e 1 e 3 1C 33e 23 ! ,

~19!

where we assume that the hexagonal symmetry is preserved

~off-diagonal elements in the strain matrix are zero!. The
strain parameters are defined by the expressions
2836

~20!

where a and c are the unstrained bulk lattice constants with

a 8 and c 8 the strained values. For a thin epitaxial film grown
on a thick substrate, the in-plane lattice constant is strained
to match the value in the substrate. The perpendicular lattice
constant is then determined by minimizing the energy with
respect to the perpendicular strain which yields

where

A2

a 8 2a
a

and

C 125 61 ~ C 11,c 15C 12,c 22C 44,c ! 13D 2 / ~ C 11,c 2C 12,c

1C 44,c ! ,

e 1 5 e xx 5 e y y [

e zz 522

C 13
e .
C 33 xx

~23!

Since it is the strain components that typically appear in

equations describing the strain-induced splitting near the top
of the valence band, we focus on the second expression. In
Table VI, we show our values for 2C 13 /C 33 along with those
derived using the other sets of elastic constants. The GaN
result of Sheleg and Savastenko clearly stands out.18 It is
more than twice as large as others and predicts that the perpendicular strain will be larger in magnitude than the inplane strain. Considering this discrepancy and those noted
previously among the GaN elastic constants, it seems reasonable to avoid using the measured values of Sheleg and
Savastenko. The GaN result of Takagi et al.20 also differs
significantly from the value of Polian et al.19 due to their
smaller C 13 value.
The strain components under hydrostatic pressure can be
determined by minimizing the energy @Eq. ~19!# with respect
TABLE VI. Predicted values of 2C 13 /C 33 .
AlN
0.607,a 0.509b
0.665,f 0.567,g 0.579h

GaN

InN

1.184,c 0.533,d 0.369e

0.510,g 0.509h

0.665c
0.940,g 0.821h

Measurements by Tsubouchi et al. ~Ref. 16!.

Measurements by McNeil et al. ~Ref. 17!.
c
Measurements by Sheleg and Savastenko ~Ref. 18!.
d
Measurements by Polian et al. ~Ref. 19!.
e
Measurements by Takagi et al. ~Ref. 20!.
f
Calculations by Ruiz et al. ~Ref. 10!.
g
Calculations by Kim et al. ~Refs. 11 and 12!.
h
Present calculations.

J. Appl. Phys., Vol. 82, No. 6, 15 September 1997

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

A. F. Wright

TABLE VII. e xx /(DV/V 0 ), e zz /(DV/V 0 ), and 2 (D h / h 0 )/(DV/V 0 ) under hydrostatic pressure.

AlN

e xx /(DV/V 0 )

GaN

0.353, 0.308

0.295,a 0.384b

10.058,a 20.076b
20.091,f 20.037,g
20.061h

0.253,g 0.316h

0.335,c 0.356,d
0.355e
0.368,g 0.329h

0.394,f 0.358,g 0.374h

2(D h / h 0 )/(DV/V 0 )

0.351c

0.332, 0.322,
0.323e
0.316,g 0.336h

0.303,f 0.321,g
0.313h

e zz /(DV/V 0 )

InN
d

20.003,c 20.034,d
10.032e
20.052,f 10.007h

0.299c
0.494,g 0.368h
10.052c
20.241,g 20.053h

Measurements by Tsubouchi et al. ~Ref. 16!.

Measurements by McNeil et al. ~Ref. 17!.
c
Measurements by Sheleg and Savastenko ~Ref. 18!.
d
Measurements by Polian et al. ~Ref. 19!.
e
Measurements by Takagi et al. ~Ref. 20!.
f
Calculations by Ruiz et al. ~Ref. 10!.
g
Calculations by Kim et al. ~Refs. 11 and 12!.
h
Present calculations.

to one of the strain components for a given fractional change

in volume DV/V 0 . This leads to expressions relating the
strains and the fractional volume change,

S D

~24!

S D

~25!

e xx 5

DV
C 332C 13
C 111C 1224C 1312C 33 V 0

e zz 5

DV
C 111C 1222C 13
,
C 111C 1224C 1312C 33 V 0

and

where the constraint 2 e xx 1 e zz 5 DV/V 0 was used in the

derivation. In Table VII, we show the values for
e xx /(DV/V 0 ) and e zz /(DV/V 0 ) derived from our set and
other sets of elastic constants. In discussing these values, we
note that they should likely be close to 1/3, the value found
in zinc blende. As expected, the predictions are all close to
this value except for the 0.253 and 0.494 found by Kim
et al.11,12 for InN.
Perhaps a better way to interpret the strains under hydrostatic pressure is in terms of the expression given below relating the fractional change in the c/a ratio, h, to the fractional volume change:
Dh

h0

S D

C 111C 122C 132C 33 DV

.
C 111C 1224C 1312C 33 V 0

~26!

In Table VII, we give predictions for 2(D h / h 0 )/

(DV/V 0 ) derived from our set and other sets of elastic constants. We note that a positive ~negative! value indicates that
the c/a ratio initially increases ~decreases! with applied pressure. Also, the predicted values should be close to zero since
this would be the result for zinc blende. The values given in
Table VII are all close to zero except for the 20.241 found
by Kim et al.11,12 for InN. The change in c/a as a function of
pressure was actually measured for all three compounds by
Ueno et al.21 While there is considerable scatter in their data,
D h / h 0 appears to increase slightly with pressure in GaN and
J. Appl. Phys., Vol. 82, No. 6, 15 September 1997

decrease with pressure in AlN and InN, consistent with our

results. In addition, we note that Christensen and Gorczyca
have directly calculated D h / h 0 vs DV/V 0 for AlN using
the FP-LMTO method.22 The initial slope of their curve
was roughly 20.06, in quantitative agreement with our
prediction.
Before comparing our wurtzite and zinc-blende results,
we note trends among the wurtzite values for the three compounds. As noted above for zinc blende, the wurtzite elastic
constants obtained for AlN and GaN are similar in magnitude whereas those found for InN are smaller. There is also a
similarity in the values of C 13 /C 33 for AlN and GaN that
differ noticeably from the value found for InN. However, the
values found for 2(D h / h 0 )/(DV/V 0 ) are similar for AlN
and InN, but different for GaN. This may reflect the less
ionic character of GaN as compared with AlN and InN since
its value for 2(D h / h 0 )/(DV/V 0 ) is closer to zero.
IV. WURTZITEZINC-BLENDE COMPARISONS

We now compare elastic properties of the zinc-blende

and wurtzite phases. To facilitate this, we transform the usual
cubic elastic constants into a consistent ~trigonal! form using
Eqs. ~12!~16! with D set equal to zero ~Table VIII!. The
rms differences between these transformed values and our
wurtzite results are 51 GPa for AlN, 8 GPa for GaN, and 21
GPa for InN. The ordering of these differences suggests that
wurtzite GaN is most similar to zinc-blende GaN, AlN is
TABLE VIII. Cubic elastic constants transformed into the trigonal system
~GPa!.

AlN
GaN
InN

C 11

C 12

C 13

C 33

C 44

425
381
242

120
130
107

79
101
89

465
411
261

112
96
49

A. F. Wright

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

2837

TABLE IX. Comparison of wurtzite and zinc-blende elastic properties

~moduli in GPa!. For clarity, wurtzite quantities are identified with subscript
h and zinc-blende quantities with subscript c.
AlN
u (C 11 1 C 12) 2 (C 13 1 C 33) u h

(2C 13 /C 33) c
(2C 13 /C 33) h

52
0.340
0.579

GaN

InN

22

0.491
0.509

0.682
0.821

@ C 11 1 C 12 2 2C 13(C 13 /C 33) # c
@ C 11 1 C 12 2 2C 13(C 13 /C 33) # h

518
470

461
450

289
262

(C 11 1 C 12 2 4C 13 2 2C 13) c
(C 11 1 C 12 2 4C 13 2 2C 13) h

1158
847

930
900

516
418

zc
zh

0.55
0.76

0.61
0.66

2838

b5b 0 1b 0 e' ~ 12 z h ! ,

0.80
0.88

~27!

~28!

where e' is the strain along the c axis and

z h[

least similar, and InN is in between. This is consistent both

with calculated zinc-blende/wurtzite energy differences23
and with deviations of the wurtzite c/a ratio from the ideal
value 1.633.8
Additional zinc-blende/wurtzite comparisons involving
the two forms of applied stress discussed above are given in
Table IX. We noted earlier that under hydrostatic pressure
the quantities e xx /(DV/V 0 ) and e zz /(DV/V 0 ) equal 1/3 in
zinc blende. A different way of stating this is that the quantity (C 11 1 C 12) 2 (C 13 1 C 33) is equal to zero in zinc
blende. The magnitude of this difference in wurtzite therefore provides another measure of its similarity to zinc
blende. As can be seen in Table VIII, the ordering of these
quantities is consistent with that found above for the elastic
constants themselves, i.e., wurtzite GaN is most similar to
zinc-blende GaN and wurtzite AlN is least similar. In terms
of biaxial stress, we compare the quantity (2C 13 /C 33), the
ratio of the in-plane and perpendicular strain magnitudes,
and @ C 11 1 C 12 2 2C 13(C 13 /C 33) # which yields the strainenergy density when multiplied by the square of the in-plane
strain component. The strain ratios are larger for the wurtzite
structure in all three compounds indicating that the perpendicular strain will be larger in epitaxial wurtzite films than in
zinc-blende films. Consistent with these results the energy
densities are smaller in the wurtzite structure. We also note
that the trend observed previouslythat wurtzite GaN is
most similar to zinc-blende GaNis also seen here both in
the strain ratios and in the energy densities.
Another quantity compared in Table IX is (C 11 1 C 12
2 4C 13 2 2C 13), the elastic modulus for strain configuration D 2 . When multiplied by the square of the in-plane strain
component, this yields the strain-energy density for constantvolume trigonal distortions in zinc blende or the strainenergy density with changes in the c/a ratio of wurtzite at
constant volume. As was the case for biaxial stress, this distortion takes less energy in wurtzite than in zinc blende. Another aspect of these distortions worth examining is the
change in length, with strain, of the anioncation bond along
the @111# direction in zinc blende and along the c axis in
wurtzite. In zinc blende, this change is specified by the
Kleinman parameter5 via the expression
b5b 0 1b 0 e' ~ 12 z c ! ,

where b and b 0 are bond lengths in the strained and unstrained crystals and e' is the strain along the @111# direction. An analogous expression for wurtzite crystals is

3 h0
j
2 u0 0

~29!

is the wurtzite form of the Kleinman parameter with h 0 the

c/a ratio at equilibrium, u 0 the internal parameter at equilibrium, and j 0 the rate of change of u with h at equilibrium.12
The calculated Kleinman parameters are given in Table IX
and we note that the wurtzite values are larger than those for
zinc blende. This indicates that wurtzite is more resistant to
changes in its bond length than zinc blende or, conversely,
that wurtzite is less resistant to changes in its bond angles.
This behavior is consistent with the wurtzite structure being
more ionic than zinc blende. In addition, GaN exhibits the
largest difference, followed by InN and then AlN. Again this
suggests that wurtzite and zinc-blende GaN are the most
similar, while wurtzite and zinc-blende AlN are the least
similar.

V. SUMMARY

We have obtained elastic constants for zinc-blende and

wurtzite AlN, GaN, and InN from DFT calculations utilizing
the plane-wave pseudopotential technique. Our wurtzite values are in fairly good agreement with the available results
from Brillouin scattering measurements and in rather poor
agreement with other measurements. The GaN elastic constants obtained by Sheleg and Savastenko, in particular,
should be avoided since they appear to greatly overestimate
the perpendicular strain in lattice-mismatched epitaxial films.
We note that having an accurate value for this quantity is
important in modeling optical gain in strained nitride films
since it affects the splitting at the top of the valence band. In
addition to the perpendicular strain, we examined several
other derived quantities that reveal how the wurtzite structure behaves under hydrostatic pressure. From this examination, the c/a ratios of AlN and InN are predicted to initially
decrease with pressure while the c/a ratio of GaN is predicted to remain nearly constant. We also compared elastic
constants and various derived quantities for the wurtzite and
zinc-blende phases. Wurtzite GaN is consistently shown to
be the most similar to zinc-blende GaN, wurtzite and zincblende AlN are the least similar, and InN is in between.
Note added in proof: Two recent measurements of the
GaN wurtzite elastic constants have been published utilizing
Brillouin scattering @M. Yamaguchi, T. Yagi, T. Azuhata, T.
Sota, K. Suzuki, S. Chichibu, and S. Nakamura, J. Phys.:
Condens. Matter 9, 241 ~1997!# and resonant ultrasonic spectroscopy @R. B. Schwarz, K. Khachaturyan, and E. R. Weber,
Appl. Phys. Lett. 70, 1122 ~1997!#. The Brillouin scattering
results are in good agreement with previous Brillouin scattering values. The value of C33 from the ultrasonic spectros-

J. Appl. Phys., Vol. 82, No. 6, 15 September 1997

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

A. F. Wright

copy study is small ~209 GPa! compared to previous values

leading to anomolous results for C13/C33 and other derived
quantities.
ACKNOWLEDGMENT

This work was supported by the United States Department of Energy under Contract No. DE-AC04-94AL8500.
B. Gil, O. Briot, and R. L. Aulombard, Phys. Rev. B 52, R17 028 ~1995!.
S. Chichibu, A. Shikanai, T. Azuhata, T. Sota, A. Kuramata, K. Horino,
and S. Nakamura, Appl. Phys. Lett. 68, 3766 ~1996!.
3
C. Kittel, Introduction to Solid State Physics ~Wiley, New York, 1971!.
4
F. D. Murnaghan, Proc. Natl. Acad. Sci. USA 30, 244 ~1944!.
5
L. Kleinman, Phys. Rev. 128, 2614 ~1962!.
6
P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 ~1964!; W. Kohn and
L. J. Sham, Phys. Rev. 140, A1133 ~1965!; D. M. Ceperley and B. J.
Alder, Phys. Rev. Lett. 45, 566 ~1980!; J. P. Perdew and A. Zunger, Phys.
Rev. B 23, 5048 ~1981!.
7
A. F. Wright and J. S. Nelson, Phys. Rev. B 50, 2159 ~1994!.
8
A. F. Wright and J. S. Nelson, Phys. Rev. B 51, 7866 ~1995!.
9
H. J. Monkhorst and J. D. Pack, Phys. Rev. B 13, 5188 ~1976!.
10
E. Ruiz, S. Alvarez, and P. Alemany, Phys. Rev. B 49, 7115 ~1994!.
11
K. Kim, R. L. Lambrecht, and B. Segall, Phys. Rev. B 50, 1502 ~1994!.
12
K. Kim, W. R. L. Lambrecht, and B. Segall, Phys. Rev. B 53, 16 310
~1996!.
13
M. Van Schilfgaarde ~private communication!.
14
M. E. Sherwin and T. J. Drummond, J. Appl. Phys. 69, 8423 ~1991!.
15
R. M. Martin, Phys. Rev. B 6, 4546 ~1972!.
16
K. Tsubouchi, K. Sugai, and N. Mikoshiba, in 1981 Ultrasonics Symposium, edited by B. R. McAvoy ~IEEE, New York, 1981!, Vol. 1, p. 375.
17
L. E. McNeil, M. Grimsditch, and R. H. French, J. Am. Ceram. Soc. 76,
1132 ~1993!.
1
2

J. Appl. Phys., Vol. 82, No. 6, 15 September 1997

18

A. U. Sheleg and V. A. Savastenko, Izv. Akad. Nauk SSSR, Neorg. Mater.

15, 1598 ~1979!.
19
A. Polian, M. Grimsditch, and I. Grzegory, J. Appl. Phys. 79, 3343
~1996!.
20
Y. Takagi, M. Ahart, T. Azuhata, T. Sota, K. Suzuki, and S. Nakamura,
Physica B 219&220, 547 ~1996!.
21
M. Ueno, M. Yoshida, A. Onodera, O. Shimomura, and K. Takemura,
Phys. Rev. B 49, 14 ~1994!.
22
N. E. Christensen and I. Gorczyca, Phys. Rev. B 47, 4307 ~1993!.
23
C.-Y. Yeh, Z. W. Lu, S. Froyen, and A. Zunger, Phys. Rev. B 46, 10 086
~1992!.
24
S. Murugkar, R. Merlin, T. Lei, and T. D. Moustakas, Bull. Am. Phys.
Soc. 37, 556 ~1992!.
25
M. Giehler, M. Ramsteiner, O. Brandt, H. Yang, and K. H. Ploog, Appl.
Phys. Lett. 67, 733 ~1995!.
26
O. Brafman, G. Lengyel, S. S. Mitra, P. J. Gielisse, J. N. Plendl, and L. C.
Mansur, Solid State Commun. 6, 523 ~1968!.
27
J. A. Sanjurjo, E. Lopez-Cruz, P. Vogl, and M. Cardona, Phys. Rev. B 28,
4579 ~1983!.
28
K. Hayashi, K. Itoh, S. Sawaki, and I. Akasaki, Solid State Commun. 77,
115 ~1991!.
29
H. Sobotta, H. Neumann, R. Franzheld, and W. Seifert, Phys. Status Solidi
B 174, K57 ~1992!.
30
J. Manchon, D. D. J. Barker, A. S. P. J. Dean, and R. B. Zetterstrom, Solid
State Commun. 8, 1227 ~1970!.
31
P. Perlin, C. Jauberthie-Carillon, J. P. Itie, A. San Miguel, I. Grzegory,
and A. Polian, Phys. Rev. B 45, 83 ~1992!.
32
A. S. Barker, Jr. and M. Ilegems, Phys. Rev. B 7, 743 ~1973!.
33
T. Inushima, A. Nagase, A. Iso, T. Yaguchi, and T. Shiraishi, in 6th
International Conference on Silicon Carbide and Related Materials ~IOP,
Kyoto, Japan, 1995!.

A. F. Wright

Downloaded 30 Apr 2012 to 139.184.30.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

2839