Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Journal of ELECTRONIC MATERIALS, Vol. 37, No.

7, 2008

Regular Issue Paper

DOI: 10.1007/s11664-008-0453-0
 2008 TMS

Determination of the Elastic Properties of Au5Sn and AuSn


from Ab Initio Calculations
RONG AN,1,2 CHUNQING WANG,1 and YANHONG TIAN1
1.Microjoining Laboratory, School of Materials Science and Engineering, Harbin Institute of
Technology, Harbin 150001, China. 2.e-mail: ronganhit@yahoo.com

Elastic constants of single-crystal Au5Sn and AuSn were determined through


ab initio calculations to characterize their polycrystalline elastic behavior and
elastic anisotropy. The ideal bulk (K = 127 GPa), shear (G = 28 GPa), and
Youngs modulus (E = 78 GPa), as well as Poissons ratio (v = 0.398), of Au5Sn
were determined using the VoigtReussHill method and were very close to
the range of experimental results; the values of AuSn were 88 GPa, 15 GPa,
42 GPa, and 0.420, but its ideal Youngs modulus was much smaller than the
experimentally obtained values. This unusual discrepancy in the Youngs
modulus of AuSn is probably attributed to its extraordinarily high anisotropy
in Youngs modulus, with a 95 GPa difference between its maximum and
minimum values. Au5Sn exhibits relatively low anisotropy in Youngs modulus with a maximum-minimum difference of 38 GPa.
Key words: Ab initio calculation, intermetallic compounds, elastic property,
anisotropy

INTRODUCTION
In electronic/optoelectronic packaging, Au-Sn
solder is widely used as preforms, coatings, or
bumps for flip-chip bonding because of its environmental friendliness, superior creep resistance, high
thermal and electrical conductivities, reduced oxide
formation, and good corrosion resistance.1,2 Two
intermetallic compounds, Au5Sn (f) and AuSn (d),
constitute eutectic Au-Sn solder (29 at.% Sn), which
has a comparatively low melting point of 280C.2
Although the vast majority of research15 has
focused on the microstructure of the solder and on
its evolution during interfacial reaction between the
solder and a pad to bridge the gap between the
microstructure and mechanical property of the solder joints, surprisingly few studies of the mechanical properties of these Au-Sn compounds themselves
exist in the literature.68 This is attributable mainly
to the difficulty in preparing the single-phase samples and the limitations of the available experimental techniques.
(Received December 16, 2007; accepted March 19, 2008;
published online April 23, 2008)

968

The elasticity of the Au-Sn compounds is the most


fundamental mechanical property for modeling and
prediction of the behavior of the solder joint,
because the elastic stiffness is important to understand how these intermetallics respond to applied
mechanical perturbations. Nevertheless, there is no
good agreement in the experimentally obtained
Youngs modulus values,68 as shown in Table I; for
example, the modulus of AuSn measured using the
ultrasonic technique with cast samples after
annealing is 71 GPa,6 whereas the value determined by microindentation testing on electrochemically deposited intermetallics is up to 101 GPa.7
Moreover, the miniaturization of microelectronic
packages requires the use of finer solder joints; as
the solder joints become increasingly small (less
than 100 lm in diameter) and contain only a few
grains, their mechanical properties, however, cannot be determined from conventional mechanical
tests as with bulk samples, and there may be a
considerable variation in mechanical behavior from
joint to joint because of the anisotropy of mechanical
properties.3,4 Thus, in addition to characterizing the
morphology and crystallographic orientation of the
grains in joints, there is a strong need to determine

Determination of the Elastic Properties of Au5Sn and AuSn from Ab Initio Calculations

969

Table I. Experimental Youngs Moduli of Au5Sn


and AuSn
Material

Specimen/Technique

E (GPa)

Au5Sn (f)

Bulk, resonance
Diffusion couples,
nanoindentation
Bulk, resonance
Thick film, microindentation
Bulk, nanoindentation

0.4

626
768

0.3

716
1017
878

AuSn (d)

the elastic constants of these intermetallics to represent their elastic anisotropy. To the best of our
knowledge, however, no research on this particular
aspect has been found in the literature.
Ab initio calculation is being widely utilized for
the prediction of material properties by virtue of the
increase of computing power and the development of
the density functional theory (DFT). Mehl et al.
performed the first DFT calculations of the elastic
constants of various intermetallic compounds in a
series of studies.911 Recently, Lee et al.12,13 provided evidence on the applicability of DFT calculations with the pseudopotential method for Sn-based
intermetallic compounds. It can be seen from these
two examples that the experimental data on singlecrystal elastic constants can be enriched using the
VoigtReussHill method, and bounds can be placed
on polycrystalline elastic moduli, so that they can be
compared with existing polycrystalline data. In our
study, we computed the full set of elastic constants
by performing the first-principles pseudopotential
total energy calculations on both Au5Sn (f) and
AuSn (d). Our aims are (i) to determine the ideal
elastic properties of the f and d phases and (ii) to
investigate their elastic anisotropy.
METHODOLOGY
Au5Sn (f) crystallizes in the space group R3 (no.
146) belonging to the trigonal crystal structure
(Fig. 1a), which can be described by two cell
parameters a and c.14 The Sn atom occupies the 3a
site (0, 0, 0), and there are three Au sites: Au1
located at 3a positions (0, 0, 0.3307), Au2 located at
3a positions (0, 0, 0.6693), and Au3 resides at 9b
positions (0.3333, 0.3403, 0.1667). The crystal
structure of AuSn (d) is depicted in Fig. 1b. The
compound crystallizes in a hexagonal lattice with
the P63/mmc space group (no. 194).15 It has four
atoms per unit cell; there is one position of Au at 2a
site and one position of Sn at 2c site. We used their
conventional cells for all the ab initio calculations.
The CASTEP code16,17 was used in the present
calculations, wherein the Vanderbilt-type ultrasoft
pseudopotential,18 the PW91 form19 of the generalized gradient approximation (GGA), and a
plane-wave basis set were employed to describe
electronion interactions, to take into account

Fig. 1. The crystal structures of (a) Au5Sn and (b) AuSn.

exchangecorrelation effects, and to represent


electronic wavefunctions, respectively. The cut-off
energy of plane-wave basis sets was 540 eV for all
the calculations on either Au5Sn or AuSn. An
8 8 8 MonkhorstPack20 k-point mesh was
employed to approximate the definite integral over
the Brillouin zone of Au5Sn, and a 9 9 6 k-point
mesh was used for AuSn. For Au5Sn, increasing the
plane-wave cut-off energy to 640 eV and the k-point
mesh to 10 10 10 changed the total energy by
less than 0.01 eV/atom and lattice constants by less
than 0.2%; and for AuSn, increasing the cut-off
energy to 640 eV and the mesh to 11 11 8
changed the total energy by less than 0.007 eV/atom
and lattice constants by less than 0.1%. Therefore,
the present computations were precise enough to
reproduce the ground-state properties.
The unit cell was fully optimized in order to
obtain the equilibrium crystal structure. Lattice
parameters and internal atomic coordinates were
independently modified to minimize the total energy
and interatomic forces. The BroydenFletcher
GoldfarbShanno (BFGS) minimization scheme21

970

was used for geometry optimization. The tolerances


for geometry optimization were selected as follows:
difference in total energy within 5 10-6 eV/atom,
maximum ionic HellmannFeynman force within
, maximum ionic displacement within
0.01 eV/A
-4
5 10 A, and maximum stress within 0.02 GPa.
We set the strain of the optimized equilibrium
cells to a finite value by applying a given homogeneous deformation to optimize the internal atomic
coordinates and calculate the resulting stress; each
of the second-order elastic constants was determined by means of a least-squares linear fit of stress
against strain. This technique has been successfully
used for many compounds, such as BeO,22 Cu6Sn5,12
and Ag3Sn.13 To calculate the elastic coefficients
precisely, stricter criteria for convergences were
selected to optimize the internal atomic coordinates
as follows: difference in total energy within 1 10-6
eV/atom, ionic HellmannFeynman forces within
, and maximum ionic displacement
0.002 eV/A
. For the Au5Sn compound with R3
within 1 10-4 A
symmetry the independent elastic constants are c11,
c33, c44, c12, c13, c14, and c15; for AuSn with P63/mmc
symmetry the independent elastic constants are c11,
c33, c44, c12, and c13. Two strain patterns (the first
with nonzero xx component, and the second with
nonzero zz and yz components) were used to generate the stresses related to all seven independent
elastic constants of Au5Sn; similarly, two other
strain patterns (the first with nonzero zz component, and the second with nonzero xx and yz components) were employed to yield all five elastic
coefficients of AuSn that are independent of each
other. Three positive and three negative amplitudes
were applied for each strain component with a
maximum strain value of 0.6% in these computations. We demonstrated the convergence of the calculations by increasing the energy cutoff to 640 eV
and the k-point mesh to 10 10 10 for Au5Sn and
by increasing the energy cutoff to 640 eV and the
k-point mesh to 11 11 8 for AuSn, after which
there was no difference in the calculated elastic
stiffness constants of Au5Sn within 8 GPa and in
the calculated elastic stiffness constants of AuSn
within 6 GPa (Table III). The compliance tensor S
was calculated as the inverse of the stiffness tensor,
S = C-1. The computations of Au5Sn and AuSn were
performed, respectively, over an equivalent of 5
months and about an equivalent of 2 months, on a
dedicated personal computer with 1 GB of random
access memory and a central processing unit clock
speed of 2.0 GHz. Polycrystalline elastic parameters, such as bulk modulus and Youngs modulus,
were estimated from the compliance tensor components using the Voigt method and the Reuss method,23 and the HashinShtrikman (HS) bounds24
were also used to place tighter bounds within the
Voigt and Reuss bounds. In addition, the Voigt
ReussHill (VRH) average23 was employed to
determine the theoretical polycrystalline elastic
property.

An, Wang, and Tian

RESULTS AND DISCUSSION


The theoretical lattice parameters of Au5Sn and
AuSn are listed in Table II, together with the
experimental data for comparison. The computed
lattice constants a and c are consistent with the
reported experimental data, that is, for Au5Sn,
Da/a < 1.5%, Dc/c < 0.4%; for AuSn, Da/a < 1.8%,
Dc/c < 2.5%. In addition, the z fractional coordinates
of Au5Sn deviate from the experimental value by
only 0.25%. Therefore, the present first-principles
computation is sufficiently reliable to reproduce the
equilibrium crystal structures of these two intermetallics.
The elastic stiffness determines the response of a
crystal to an imposed strain (or stress) and provides
information about bonding characteristics near the
equilibrium state. Investigating the elastic stiffness
is essentially the first step to understanding the
mechanical properties of a solid. Table III includes
the full set of theoretical second-order elastic constants of Au5Sn and AuSn. Since their second-order
elastic constants have not been reported in the literature, it is impossible to compare the present
theoretical elastic constants with others; it is however noted that Au5Sn has larger elastic stiffness
constants relative to AuSn. For instance, the elastic
constants representing stiffness against uniaxial
strains, c11 and c33, of Au5Sn, are 160 GPa and
181 GPa, respectively, which are slightly higher
than those of AuSn (103 GPa and 164 GPa, respectively). The c44 of Au5Sn, which corresponds to the
resistance to shear deformation, is 30 GPa, and is
over twice that of AuSn (14 GPa). These results
agree with the hardness of these two Au-Sn intermetallic compounds reported in Ref. 8 that the
hardness value of Au5Sn (2.5 GPa) is much larger
than that of AuSn (1.1 GPa).
Table IV lists the isotropic bulk (K) and shear (G)
moduli calculated from the corresponding singlecrystal data using the VRH approximation. The
Youngs modulus E and Poissons ratio v were

Table II. Optimized Equilibrium and Experimental


Structure Parameters of Au5Sn and AuSn

Au5Sn

AuSn

Method

)
a (A

)
c (A

Calc.

5.167

14.388

Expt.14

5.092

14.333

Calc.

4.245

5.659

Expt.15

4.322

5.523

Fractional z
Coordinate (z/c)
z(Sn) = 0
z(Au1) = 0.3315
z(Au2) = 0.6685
z(Au3) = 0.1667
z(Sn) = 0
z(Au1) = 0.3307
z(Au2) = 0.6693
z(Au3) = 0.1667
z(Sn) = 0.2500
z(Au) = 0
z(Sn) = 0.2500
z(Au) = 0

Determination of the Elastic Properties of Au5Sn and AuSn from Ab Initio Calculations

971

Table III. Calculated Elastic Stiffness (Cij) and Compliance (Sij) Constants of Single-Crystal Au5Sn and AuSn

Au5Sn

AuSn

540 eV, 8 8 8 k points

PW91

640 eV, 10 10 10 k points

PW91

540 eV, 9 9 6 k points

PW91
PBE
LDA

640 eV, 11 11 8 k points

PW91

Cij (GPa)
Sij (10-3/GPa)
Cij (GPa)
Sij (10-3/GPa)
Cij (GPa)
Sij (10-3/GPa)
Cij (GPa)
Sij (10-3/GPa)
Cij (GPa)
Sij (10-3/GPa)
Cij (GPa)
Sij (10-3/GPa)

11

33

44

12

13

14

15

160
15
165
15
103
36
115
21
126
25
108
30

181
9.5
173
10
164
8.1
165
7.8
205
6.3
158
8.8

30
33
34
30
14
69
15
65
24
42
17
59

117
-8.4
121
-8.3
88
-29
86
-14
103
-19
89
-23

102
-3.5
104
-3.7
62
-2.7
61
-2.4
72
-2.0
66
-2.9

-1.0
0.7
-1.6
1.1

0.6
-0.4
1.4
-0.9

11, 22, etc. are the tensor subscripts.

Table IV. Bounds on the Elastic Properties


of Polycrystalline Au5Sn and AuSn
Bound
Au5Sn

AuSn

Voigt
HS upper
HS lower
Reuss
Voigt
HS upper
HS lower
Reuss

K (GPa) G (GPa) E (GPa)


127.21
127.19
127.19
127.17
88.34
87.86
87.51
87.31

28.48
27.97
27.81
27.16
17.80
15.59
13.50
11.87

79.51
78.18
77.76
76.07
50.04
44.17
38.51
34.06

v
0.3958
0.3976
0.3981
0.4003
0.4056
0.4162
0.4266
0.4350

calculated from K and G using the interrelationship


of these four elastic parameters based on the isotropic elasticity of the materials, and the results are
also listed in the table. The VRH Youngs modulus
of Au5Sn is 78 GPa, which is larger than that
(62 GPa) obtained using resonance6 and very close
to that (76 GPa) determined by Chromik et al. using
nanoindentation8 (Table I). The VRH average on
the Poissons ratio of Au5Sn (0.398) is fairly consistent with the reported experimental value (0.4). Lee
et al.12,13 reported similar results about Cu6Sn5,
Ag3Sn, and Ni3Sn4 that the Youngs moduli determined through ab initio calculations are close to
those obtained using nanoindentation, but larger
than those values measured with bulk samples
using resonance or tensile loading techniques. This
difference can be attributed to the following possible
sources: first, the DFT computation typically
deduces the elastic constants and the mechanical
parameters at absolute zero, and Youngs modulus
decreases with increasing temperature; second,
defects in the material are not taken into account in
the overall calculation. In contrast with Au5Sn and
other Sn-based intermetallics, however, the VRH
average of AuSn on Youngs modulus (42 GPa) is
surprisingly far lower than any of the values

determined through resonance, microindentation,


or nanoindentation. To clarify that the elastic constants calculated with the PW91 are not in error,25
we examined the predictions from other GGA formulations, i.e., PerdewBurkeEnzerhof (PBE), and
also the local density approximation (LDA); the
calculation results are listed in Table III. For the
GGA-PBE, no difference is observed in the calculated elastic constants within 10%, and the VRH
Youngs modulus of 54 GPa is still much lower than
the experimental results (71 GPa to 101 GPa). The
elastic constants calculated with the LDA are typically larger than that predicted from the PW91 or
PBE, but the VRH Youngs modulus of 62 GPa is
also lower than the experimental results. Thus, this
unusual negative deviation from the experimental
values can be derived not from the calculation error
but from the elastic anisotropy, which will be discussed later in this paper.
Au5Sn exhibits slight enhancements of mechanical properties compared with AuSn. For instance,
Au5Sn yields a bulk modulus of 127 GPa and a shear
modulus of 28 GPa, which are larger than those of
AuSn by 39 GPa and 13 GPa, respectively. Moreover, the Youngs modulus of Au5Sn is 1.9 times that
of AuSn. Besides, the two compounds possess a low
shear deformation resistance, which is demonstrated by their low shear moduli. Pugh26 introduced
the quotient of shear modulus to bulk modulus, G/K,
as an indication of the extent of fracture range in
metals. A low value of G/K is associated with ductility and a high value with brittleness. There are
several examples that involve some representation
based on the quotient G/K; these include the evaluation of brittleness for trialuminide alloys,27 the
determination of a transition from plasticity to
brittleness in a bulk amorphous steel,28 the discovery of the ceramics with unusual plastic behavior,29
and the explanation for the ductility of a ternary
Y-Si-O silicate.30 Theoretical c44/K and G/K values
are 0.24 and 0.22, for Au5Sn; the c44/K and G/K
values of AuSn are 0.16 and 0.17. Although the

972

An, Wang, and Tian

attribution of the extent of fracture range solely to


the G/K (or c44/K) ratio is highly simplistic, it nevertheless indicates the tendency of ductility for the
two Au-Sn intermetallic compounds.
Both Au5Sn and AuSn are noncubic crystals; their
elastic anisotropy can be measured using three
dimensionless quantities AG, AK, and AE, defined as
AG = (GV - GR)/(GV + GR), AK = (KV - KR)/(KV + KR),
and AE = (EV - ER)/(EV + ER),31,32 respectively. The
subscripts V and R here designate the Voigt and the
Reuss averaging schemes and they represent the
maximum and minimum limits of the true polycrystalline elastic moduli. A gives the relative magnitude
of the elastic anisotropy present in crystals. It is
always positive and is zero for crystals which are
elastically isotropic. For Au5Sn, the anisotropy factors, viz., AG, AK, and AE (in percent), are 2.4%,
0.016%, and 2.2%; and for AuSn, these factors are up to
20%, 0.59%, and 19%, respectively. The results indicate that AuSn is much more anisotropic than Au5Sn;
these are supported by the fact that the residual indents of AuSn in nanoindentation testing exhibited
significant asymmetric pileup different from Au5Sn,
which is due to the plastic anisotropy of AuSn.8 Furthermore, these results are also in accordance with the
data on the orientation-dependent Youngs moduli in
these crystals. A three-dimensional surface representation of elastic anisotropy is an illustrative way of
showing the variation of elastic modulus with crystallographic direction. The directional dependence of Youngs modulus for trigonal crystals can be
given by33
h
2


E 1  l23 s11 l43 s33 l23 1  l23 2s13 s44



 i1
2l2 l3 3l21  l22 s14 2l1 l3 3l22  l21 s15 ;
(1)
and for hexagonal crystals,
h
i1
2


E 1  l23 s11 l43 s33 l23 1  l23 2s13 s44 ;
(2)

where sij are the single-crystal elastic compliance


constants in the two suffix notation, and l1, l2, and l3
are the direction cosines to the a, b, and c axes,
respectively. In this representation, an isotropic
system would have a spherical shape, and so the
degree of deviation of the geometry from a sphere
indicates the degree of anisotropy in a specific
property of a system. The representations on the
directional dependence of the Youngs moduli of
both Au5Sn and AuSn show clear deviations from a
spherical shape as shown in Figs. 2 and 3, but
greater deviation occurs in AuSn. Therefore, the two
intermetallics have a high degree of anisotropy in
Youngs modulus, but the degree in AuSn is much
higher than that in Au5Sn.
The anisotropic Youngs moduli of both the compounds in the (100), (010), and (001) planes are
depicted in Figs. 2b and 3b. For either Au5Sn or
AuSn, significant in-plane elastic anisotropy
appears in the (100) and (010) planes, but no elastic
anisotropy in the (001) plane, which is consistent
with the atomic arrangements in itself. The maximum Youngs modulus of Au5Sn is 106 GPa, which
is in the [001] direction; the minimum is 68 GPa,
 10, 3] direction (Fig. 2c). For
which is along the [5,
AuSn, the maximum modulus of 123 GPa is reached
in the [001] crystallographic orientation; the minimum of 28 GPa is obtained along any orientation in
the (001) plane, as shown in Fig. 3b. The difference
between the maximum and minimum moduli of
Au5Sn is 38 GPa, which is 49% of the VRH Youngs
modulus; this large difference implies that the distinct disagreement in its experimentally obtained
Youngs modulus can be attributed to the strong
elastic anisotropy plus the differences induced during sample preparation and the restriction imposed
by limited experimental techniques available. This
explanation is partially supported by making the
comparisons between the maximumminimum difference of Youngs modulus and the bound of the
reported experimental results. Examples include
the cases of Cu6Sn5 and Ag3Sn; these two Sn-based
intermetallics have the orthorhombic crystal structure. The difference between the maximum and

Fig. 2. (a) Directional dependence of Youngs modulus in Au5Sn. Plane projections of the directional dependence of Youngs modulus on (b) the
(100), (010), and (001), and (c) (120) planes.

Determination of the Elastic Properties of Au5Sn and AuSn from Ab Initio Calculations

973

obvious that the experimentally obtained Youngs


moduli will be much larger than the VRH Youngs
modulus. So far it has been impossible to quantitatively compare the modulus values obtained with
samples that possess different textures because the
textures of samples are not readily accessible.
However, because of such distinct anisotropy of
elasticity, it is clear that, the fewer grains there are
in the intermetallic compounds of the increasingly
small solder joints, the greater the difference
between the Youngs modulus value of the bulk
samples and that measured with the intermetallics
in the solder joints.
CONCLUSIONS

Fig. 3. (a) Directional dependence of Youngs modulus in AuSn. (b)


Plane projections of the directional dependence of Youngs modulus
on the (100), (010), and (001) planes.

The seven elastic constants of single-crystal


Au5Sn and the five elastic constants of single-crystal
AuSn have been determined from ab initio calculations using the pseudopotential plane-wave method.
The ideal bulk, shear, and Youngs moduli, as well
as Poissons ratios, of these two polycrystalline
intermetallic compounds were calculated using the
VRH and HS methods. Au5Sn exhibits quite distinct
anisotropy in Youngs modulus, with a difference of
38 GPa between its maximum and minimum modulus; AuSn shows a much greater degree of anisotropy than Au5Sn, that is, the maximumminimum
difference of the Youngs moduli of AuSn is 95 GPa.
In addition to the metallurgical imperfections of test
samples and the restrictions imposed by the limited
experimental techniques available, high elastic
anisotropy is one of the causes of the discrepancy in
experimentally and theoretically obtained elastic
modulus values. It is believed that with time the
elastic anisotropy of these intermetallics will become a more important consideration in the accurate characterization of the mechanical behavior of
small solder joints.
ACKNOWLEDGEMENTS

minimum moduli of Cu6Sn5 (calculated using the


elastic constants in Ref. 12) is more than 20 GPa,
and the experimental Youngs moduli summarized
in Ref. 12 show a wide range from 85 GPa to
125 GPa. Here again, the calculated difference of
Ag3Sn (using the data in Ref. 13) is 34 GPa, and the
reported experimental values are scattered within
the large range of 70 GPa to 94 GPa. For AuSn, the
difference of Youngs modulus between the maximum and minimum is up to an unusally large value
of 95 GPa, which is 2.3 times the VRH Youngs
modulus. This is illustrated graphically in the surface representation of its elastic anisotropy. The
representation of AuSn has a needle-like shape,
meaning that a slight degree of deviation of crystallographic orientation from the [001] direction in
AuSn will produce a significant decrease in Youngs
modulus, as depicted in Fig. 3a. Thus, when the test
samples primarily have the [001] texture, it is

The authors thank the Natural Science Foundation of China (No. 50675047) for its financial support; J. Cheng and J.C. Zhu, Harbin Institute of
Technology, for their valuable help with the execution of CASTEP code; and Shanghai Supercomputer
Center (SSC) for supercomputing resources provided.
REFERENCES
1. G.S. Matijasevic, C.C. Lee, and C.Y. Wang, Thin Solid Films
223, 276 (1993). doi:10.1016/0040-6090(93)90533-U.
2. D.G. Ivey, Micron 29, 281 (1998). doi:10.1016/S09684328(97)00057-7.
3. H. Song, J. Morris, and M. McCormack, J. Electron. Mater.
29, 1038 (2000). doi:10.1007/s11664-000-0170-9.
4. H. Song, J. Ahn, and J. Morris, J. Electron. Mater. 30, 1083
(2001). doi:10.1007/s11664-001-0133-9.
5. J.-W. Yoon, H.-S. Chun, and S.-B. Jung, J. Mater. Res. 22,
1219 (2007). doi:10.1557/jmr.2007.0145.
6. F.G. Yost, M.M. Karnowsky, W.D. Drotning, and J.H. Gieske,
Metall. Mater. Trans. A 21, 1885 (1990). doi:10.1007/
BF02647236.

974
7. A. Vicenzo, M. Rea, L. Vonella, M. Bestetti, and P.L.
Cavallotti, J. Solid State Electrochem. 8, 159 (2004).
doi:10.1007/s10008-003-0428-2.
8. R.R. Chromik, N. Wang, A. Shugar, L. Limata, M.R. Notis,
and R.P. Vinci, J. Mater. Res. 20, 2161 (2005). doi:10.1557/
JMR.2005.0269.
9. M.J. Mehl, J.E. Osburn, D.A. Papaconstantopoulos, and
B.M. Klein, Phys. Rev. B 41, 10311 (1990). doi:10.1103/
PhysRevB.41.10311.
10. J.E. Osburn, M.J. Mehl, and B.M. Klein, Phys. Rev. B 43,
1805 (1991). doi:10.1103/PhysRevB.43.1805.
11. M.J. Mehl, D.J. Singh, and D.A. Papaconstantopoulos,
Mater. Sci. Eng. A 170, 49 (1993). doi:10.1016/0921-5093
(93)90368-O.
12. N.T.S. Lee, V.B.C. Tan, and K.M. Lim, Appl. Phys. Lett. 88,
031913 (2006). doi:10.1063/1.2165280.
13. N.T.S. Lee, V.B.C. Tan, and K.M. Lim, Appl. Phys. Lett. 89,
141908 (2006). doi:10.1063/1.2358832.
14. K. Osada, S. Yamaguchi, and M. Hirabayashi, Trans. Jpn.
Inst. Met. 15, 256 (1974).
15. J.-P. Jan, W.B. Pearson, A. Kjekshus, and S.B. Woods, Can.
J. Phys. 41, 2252 (1963).
16. M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, and J.D.
Joannopoulos, Rev. Mod. Phys. 64, 1045 (1992). doi:10.1103/
RevModPhys.64.1045.
17. M.D. Segall, P.J.D. Lindan, M.J. Probert, C.J. Pickard,
P.J. Hasnip, S.J. Clark, and M.C. Payne, J. Phys.: Condens.
Matter 14, 2717 (2002). doi:10.1088/0953-8984/14/11/301.
18. D. Vanderbilt, Phys. Rev. B 41, 7892 (1990). doi:10.1103/
PhysRevB.41.7892.

An, Wang, and Tian


19. J.P. Perdew and Y. Wang, Phys. Rev. B 45, 13244 (1992).
doi:10.1103/PhysRevB.45.13244.
20. H.J. Monkhorst and J.D. Pack, Phys. Rev. B 13, 5188 (1976).
doi:10.1103/PhysRevB.13.5188.
21. T.H. Fischer and J. Almlof, J. Phys. Chem. 96, 9768 (1992).
doi:10.1021/j100203a036.
22. V. Milman and M.C. Warren, J. Phys.: Condens. Matter 13,
241 (2001). doi:10.1088/0953-8984/13/2/302.
23. R. Hill, Proc. Phys. Soc. A 65, 349 (1952). doi:
10.1088/0370-1298/65/5/307.
24. J.P. Watt and L. Peselnick, J. Appl. Phys. 51, 1525 (1980).
doi:10.1063/1.327804.
25. A.E. Mattsson, P.A. Schultz, M.P. Desjarlais, T.R. Mattsson,
and K. Leung, Modell. Simul. Mater. Sci. Eng. 13, R1 (2005).
doi:10.1088/0965-0393/13/1/R01.
26. S.F. Pugh, Philos. Mag. 45, 823 (1954).
27. C.L. Fu, J. Mater. Res. 5, 979 (1990). doi:10.1557/
JMR.1990.0971.
28. X.J. Gu, A.G. McDermott, S.J. Poon, and G.J. Shiflet, Appl.
Phys. Lett. 88, 211905 (2006). doi:10.1063/1.2206149.
29. D. Music and J.M. Schneider, Appl. Phys. Lett. 88, 031914
(2006). doi:10.1063/1.2165285.
30. J.Y. Wang, Y.C. Zhou, and Z.J. Lin, Acta Mater. 55, 6019
(2007). doi:10.1016/j.actamat.2007.07.010.
31. D.H. Chung and W.R. Buessem, J. Appl. Phys. 38, 2010
(1967). doi:10.1063/1.1709819.
32. D.H. Chung and W.R. Buessem, J. Appl. Phys. 39, 2777
(1968). doi:10.1063/1.1656672.
33. J.F. Nye, Physical Properties of Crystals (Oxford: Oxford
University Press, 1985).

You might also like