Electron mean free path in elemental metals

Daniel Gall

Citation: J. Appl. Phys. 119, 085101 (2016); doi: 10.1063/1.4942216

View online: http://dx.doi.org/10.1063/1.4942216
View Table of Contents: http://aip.scitation.org/toc/jap/119/8
Published by the American Institute of Physics
 JOURNAL OF APPLIED PHYSICS 119, 085101 (2016)

Electron mean free path in elemental metals

Daniel Gall
Department of Materials Science and Engineering, Rensselaer Polytechnic Institute, Troy, New York 12180,
USA

(Received 23 December 2015; accepted 5 February 2016; published online 23 February 2016)
The electron mean free path k and carrier relaxation time s of the twenty most conductive
elemental metals are determined by numerical integration over the Fermi surface obtained from
first-principles, using constant k or s approximations and wave-vector dependent Fermi velocities
vf (k). The average vf deviates considerably from the free-electron prediction, even for elements
with spherical Fermi surfaces including Cu (29% deviation). The calculated product of the bulk
resistivity times k indicates that, in the limit of narrow wires, Rh, Ir, and Ni are 2.1, 1.8, and 1.6
times more conductive than Cu, while various metals including Mo, Co, and Ru approximately
match the Cu resistivity, suggesting that these metals are promising candidates to replace Cu for
narrow interconnect lines. V C 2016 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4942216]

I. INTRODUCTION is a reasonable approximation for alkali and group 11 metals.

In fact, the free electron predictions for room temperature
The electrical resistivity of metal wires increases as their
Cu and Ag of k ¼ 39 and 53 nm are in good agreement with
width decreases.1–4 This size effect represents a major chal-
values obtained from fitting the measured resistivity of epi-
lenge for nanoelectronics5 and for the further downscaling of
taxial metal layers vs their thickness using the FS
integrated circuits6,7 as the resistivity of, for example, 10-
model.14,23,24 In contrast, the limited reported results from
nm-wide Cu interconnect lines is approximately an order of
metals with non-spherical Fermi surfaces suggest differences
magnitude larger than that of bulk Cu.2,8 The primary reason
by up to an order of magnitude between the k determined
for the resistivity increase is electron scattering at external
from the free electron model and from fitting resistivity vs
surfaces and grain boundaries, which are most commonly
thickness data.25–29 These results are further complicated by
described by the classical models by Fuchs and Sondheimer
anisotropy effects30–33 and the possible breakdown of the FS
(FS)9,10 and Mayadas and Shatzkes (MS),11,12 respectively.
model for narrow conductors.25
Both models predict in their approximate form an additive
In this paper, I present calculated values of the product
resistivity contribution which is proportional to qo  k/d,
qo  k based on the bulk electronic structure predicted from
where qo and k are the bulk resistivity and mean free path
first-principles density functional calculations. Thus, these
for electron phonon scattering, and d is the relevant length
calculations correctly account for the anisotropy in the Fermi
scale, that is, the wire width or the grain size for surface and
surface as well as the variation in the electron velocity v(k)
grain boundary scattering, respectively. Therefore, in the
as a function of wave vector k; that is, they go beyond the
limiting case of thin wires and/or small grain sizes, the wire
common effective mass approximation of the FS and MS
resistivity becomes proportional to qo  k for any given fixed
models that are expected to cause inaccuracies for metals
wire dimension and grain size distribution. Thus, the metal
with anisotropic Fermi surfaces. Electron transport is
with the lowest product qo  k is expected to exhibit the
described within a semi-classical Boltzmann transport for-
highest conductivity in the limit of a small wire width. This
malism with individual scattering events that completely ran-
argument neglects variations in the surface scattering specu-
domize the electron momentum, either within a constant
larity and grain boundary reflection coefficient which are
mean free path or within a constant mean free time approxi-
expected to differ for different metals and enter the FS and
mation. That is, the electron-phonon scattering probability is
MS models as phenomenological parameters that also
approximated by a k-independent average distance or aver-
depend on the surface13–18 and grain boundary3,19–22 struc-
age time, respectively. The tabulated results suggest that
ture and chemistry. Nevertheless, minimizing the product
quite a few (5–10) metals have an effective resistivity that is
qo  k is a useful starting point in the focused search for met-
lower than that of copper in the limit of narrow wires, despite
als that form high-conductivity narrow wires.
that their bulk resistivity is up to a factor of five larger.
Values for the bulk resistivity of elemental metals are
Therefore, the decision regarding which of those metals will
well established from measurements at room temperature
replace Cu for narrow interconnect lines may ultimately be
where electron scattering is dominated by phonons and resis-
dominated by secondary factors including processing feasi-
tivity contributions due to impurities and crystalline defects
bility and electromigration performance.
are negligible for carefully prepared samples. In contrast,
there is only limited information regarding the bulk electron
II. COMPUTATIONAL APPROACH
mean free path. One approach is to calculate k from the
measured qo using the free electron model. This assumes a The electronic structure of the twenty most conductive ele-
constant Fermi velocity and a spherical Fermi surface, which mental metals is obtained from density functional calculations

0021-8979/2016/119(8)/085101/5/$30.00 119, 085101-1 C 2016 AIP Publishing LLC

V
085101-2 Daniel Gall J. Appl. Phys. 119, 085101 (2016)

using the Vienna ab initio simulation package (VASP), Transport is simulated within a semiclassical approach,
employing periodic boundary conditions, a plane wave basis such that the bulk conductivity ro is given by
set, the Perdew-Burke-Ernzerhof generalized gradient ðð
approximation (GGA) exchange correlation functional,34 2e2 X sn ðkÞv2t;n ðkÞ
ro ¼ dS: (1)
and the projector-augmented wave method.35 Computational 8p3 h n jvn ðkÞj
SnF
convergence with respect to the energy cut-off for the plane
wave expansion, the k-point grid, and the unit cell size and
Here, the sum is over bands and the integration is over
shape is tested and adjusted such that all values reported in
the Fermi surface SnF of band n. The carrier relaxation time
this letter have a computational uncertainty <1%. In addition
sn(k), the electron velocity along the transport direction
to the outer-most electron shells, the following electrons are
vt,n(k), and the electron velocity vector vn(k) are functions of
explicitly calculated; that is, they are not included in the core
the wave vector k for each band with index n. The factor two
of the pseudo potentials: p for alkali, alkaline earth, and tran-
in Eq. (1) is accounting for both spins for the case on non-
sition metals up to column VI, 4d for In, 2s and 2p for Al, magnetic metals, while this factor is removed for magnetic
and 1s for Be. Calculations are done using primitive unit materials (Co and Ni in this letter) and the contribution from
cells containing one atom for fcc, bcc, and bct and two atoms each spin is summed up separately, starting with a spin-
for hcp structures, and lattice parameters are fixed at the polarized density functional calculation.
established experimental room temperature values. The latter Using a constant relaxation time approximation, the
are typically slightly smaller (1%) than the values obtained relaxation time becomes independent of band, spin, and
when fitting calculated energy vs lattice parameters, as wave vector. That is, sn(k) ¼ s and thus
typical for the generalized gradient approximation.36
ðð
Self-consistent calculations using a C-centered 40  40 1 e2 X v2t;n ðkÞ
 40 k-point grid are used to determine the charge distribu- ¼ 3 dS: (2a)
sqo 4p h n jvn ðkÞj
tion, which is subsequently used for non-self-consistent cal- SnF
culations with a finer 200  200  200 k-point mesh. Such a
fine k-point sampling is required for good convergence of Correspondingly, using a constant mean free path approx-
the numerical integration across the Fermi surface, which is imation, the carrier relaxation time becomes sn(k) ¼ k/jvn ðkÞj
very sensitive to band crossings. Careful convergence tests and therefore
are performed for all reported elements to achieve the ðð 2 ð Þ
1 e2 X vt;n k
desired <1% numerical accuracy, requiring, for example, for ¼ 3 dS: (2b)
kqo 4p h n v2n ðkÞ
Be an increase to 400  400  400 k-points. SnF
The Fermi surface is determined from the calculated
bands, i.e., the calculated electron energy vs k curves En(k) I note here that for cubic symmetry, the integral in Eq.
where n is the band index. This is done by dividing the (2b) is exactly one third of the area of the Fermi surface Af,
Brillouin zone into irregular tetrahedra that are defined by such that the product k  qo is inversely proportional to the
four neighboring k-points. Linear interpolation between the Fermi surface area, i.e., kqo ¼ 12p3h/(e2Af) as used, for exam-
corners yields the intercepts of the Fermi surface with the ple, in Ref. 26. This fact is not explicitly used in the numerical
tetrahedron edges. Three or four intercepts per tetrahedron code, primarily to keep modularity and flexibility of the com-
define one or two triangles of the Fermi surface, respec- putational approach, which first interpolates band energies in
tively,37 with an electron velocity vn(k) ¼ 1h rk En(k) that is the reciprocal coordinate system of the primitive unit cell vec-
perpendicular to the triangle surface and is defined by the tors, second defines the Fermi surface as a list of triangles
slope of the band. More elaborate interpolation schemes with associated velocity vectors in a reciprocal Cartesian coor-
have been tested, including three-dimensional polynomial dinate system, and third determines transport quantities by
interpolation to create k-point submeshes, but do not consis- summation over all triangles according to Eqs. (2a) and (2b).
tently improve k-point convergence, which is attributed to Integration is done for all calculated metals using multiple
these approaches being more sensitive to discontinuities at transport directions corresponding to conventional [100],
band-crossings. [010], [001], [110], [1 1 0], and [111] directions. The results
I reiterate here that all calculations in this paper are are identical within <1% deviation for all directions for the
done for bulk materials; that is, quantum size effects in nar- case of cubic fcc and bcc structures, and for directions within
row wires are not accounted for during the electronic struc- the basal plane for hcp and bct. This is expected based on the
ture calculations. This approximation is justified by the small 3-dimensional isotropic resistivity for cubic materials and 2-
screening length in metals, resulting in a thin film ballistic dimensional isotropy within the basal plane of hexagonal and
conductance that matches the bulk conductance for all but tetragonal symmetries. The values reported, in this letter, are
the narrowest layers. For example, a 2-nm-thick Cu layer has the average from the [100] and [010] directions, while for hcp
a calculated conductance of 0.58  1015 X1 m2 (Ref. 17), and bct the [001] values are separately listed.
in perfect agreement with the bulk value.38 In contrast, sur-
III. RESULTS AND DISCUSSION
face roughness in narrow conductors can cause a destruction
of isotropic Fermi surface sheets which affects transport Fig. 1 shows typical Fermi surfaces, in this case, for Cu,
properties.17 Mo, Ru, and Rh. The Brillouin zones are outlined by black
085101-3 Daniel Gall J. Appl. Phys. 119, 085101 (2016)

separate pockets. The Fermi surface of Rh plotted in Fig. 1(d)

is relatively large, 4.73  1021 m2, and the average velocity
small vf ¼ 0.67  106 m/s. It exhibits multiple nestled surfaces
with both electron and hole character around the zone center
and near the zone boundary, respectively. The large Fermi
surface of Rh leads to a small product k  qo, which is the
smallest of all investigated metals, as discussed below, such
that Rh becomes a promising candidate for narrow high-
conductivity metal wires.
Table I is a list of the calculated transport data for the
twenty most conductive elemental metals. The list is sorted
according to increasing bulk resistivity at room temperature
qo,rt, as reproduced from Ref. 40. The values for the average
Fermi velocity vf and the columns s  qo and k  qo are
directly obtained from first principles calculations, using
Eqs. 2(a) and 2(b) for the constant s and k approximations,
respectively. These products are temperature-independent
because qo is proportional to the electron-phonon scattering,
while s and k are inverse proportional to the electron-phonon
FIG. 1. The Fermi surfaces of (a) Cu, (b) Mo, (c) Ru, and (d) Rh, as obtained
scattering. The listed values for the room temperature srt and
from first principles calculations. Colors indicate the Fermi velocity vf. krt are determined from the s  qo and k  qo products by
dividing by the known room temperature bulk resistivity.
lines, while the Fermi surface is colored according to the Two values are given for hexagonal and tetragonal metals,
Fermi velocity vf(k). The Fermi surface of Cu in Fig. 1(a) is reflecting transport perpendicular and parallel to the hexago-
nearly spherical, with eight necks that reach along h111i nal/tetragonal axes. That is, the first values are for conduc-
directions to touch the hexagonal faces of the zone and are tion within the 2-dimensional basal planes and have double
due to the reduction of the band energy near the zone bound- the weight in comparison to the second values for conduction
ary.39 The total Fermi surface area of 2.28  1021 m2 almost in the 3rd dimension. I note here that, for purposes of data
perfectly matches ( <2% deviation) the free-electron predic- consistency, the listed bulk resistivity values are from large-
grain randomly oriented polycrystalline samples, even for
tion of 2.32  1021 m2. In contrast, the average Fermi veloc-
anisotropic (hcp and bct) metals. Thus, while the calculated
ity of 1.11  106 m/s is 29% smaller than the free electron
data-pairs for s  qo and k  qo correctly account for the ani-
prediction of vf ¼ 1.57  106 m/s. The velocity is relatively
sotropy in Fermi surface and velocity of hcp and bct metals,
constant over most of the surface, decreasing from a maxi-
the columns for srt and krt contain two values, which is an
mum vf ¼ 1.31  106 m/s along h100i to 1.0  106 m/s along
artifact of this approach. For example, the calculated s  qo
h110i, and drops to 0.7  106 m/s for the neck at the zone
for Mg along the hexagonal axis is 15% smaller than the
boundary. The Fermi surface for Mo in Fig. 1(b) is
perpendicular value, indicating a resistivity anisotropy ratio
2.55  1021 m2 and much more complex than for Cu, as
qjj/q? ¼ 0.85 for Mg. This is in excellent agreement with a
multiple 4d bands cross the Fermi level in Mo. The Fermi reported measured anisotropy value of 0.84.41 The agree-
velocity varies by nearly an order of magnitude, from 0.2 to ment suggests, since the calculations are done with a con-
1.5  106 m/s, with an average vf ¼ 0.92  106 m/s. More spe- stant s, that the electron-phonon scattering cross-section in
cifically, vf is 0.8  106 m/s for the electron jack near the Mg is isotropic. Correspondingly, the bulk resistivity could
zone center and decreases continuously to a minimum be listed in Table I as two values: q? ¼ 4.62 lX cm and
vf ¼ 0.2  106 m/s in the electron knobs that extend along the qjj ¼ 3.93 lX cm which, in turn, would result in single value
h100i directions, vf ¼ 0.8–1.5  106 m/s for the hole octahe- for srt of 1.86 fs. However, a similar argument within the
dron and vf ¼ 0.9–1.3  106 m/s for the hole ellipsoid near the constant mean free path approximation would yield an ani-
zone boundary along h100i and h110i, respectively. The sotropy of 0.90, and therefore slightly different q? and qjj
Brillouin zone for Ru in Fig. 1(c) is wider within the basal values, which would make data presentation in Table I
plane than along the hexagonal axis, reflecting the Ru lattice confusing. Moreover, the resistivity anisotropy has not been
constants which are 2.706 and 4.282 Å perpendicular and par- reported previously for all metals, contains considerable
allel to the hexagonal axis, respectively. The Fermi surface uncertainty with deviations between different reports of up
area is 3.32  1021 m2 and the average vf ¼ 0.72  106 m/s. to 30%,41,42 and also differs considerably (0%–25%) from
The anisotropy is evident from both the shape and the velocity our calculated anisotropy ratios.41–44
distribution, with the inner “electron star” exhibiting a rela- Cu has a calculated k  qo ¼ 6.70  1016 X m2, which
tively low vf ¼ 0.3  106 m/s along [0001] but a three times is within 2% of 6.59  1016 X m2 determined from the free
larger vf ¼ 0.9  106 m/s within the hexagonal plane while the electron approximation. Correspondingly, krt ¼ 39.9 nm
“hole ring” has vf ¼ 0.3–0.8  106 m/s over most of its surface matches the 39 nm reported for room temperature Cu based
but also exhibits pockets with vf ¼ 0.9–1.2  106 m/s which is on the free electron model2,45 which also matches resistivity-
almost as high as vf ¼ 0.9–1.3  106 m/s for the additional vs-thickness scaling data.13–15 This agreement between the
085101-4 Daniel Gall J. Appl. Phys. 119, 085101 (2016)

TABLE I. List of the twenty most conductive elemental metals, sorted with increasing bulk room-temperature resistivity qo,rt from Ref. 40. The average Fermi
velocity vf and the products s  qo and k  qo are obtained from first principles calculations, while values for the room temperature carrier relaxation time srt
and mean free path krt are obtained by dividing the previous columns by qo,rt. For hexagonal and tetragonal crystal structures (hcp and bct), the two listed val-
ues are for transport perpendicular and parallel to the hexagonal/tetragonal axis.

Element Crystal structures qo,rt (lX cm) vf (105 m/s) s  qo (1022 X m s) k  qo (1016 X m2) srt (fs) krt (nm)

Silver Ag fcc 1.587 14.48 5.84 8.46 36.8 53.3

Copper Cu fcc 1.678 11.09 6.04 6.70 36.0 39.9
Gold Au fcc 2.214 13.82 6.04 8.35 27.3 37.7
Aluminum Al fcc 2.650 15.99 3.13 5.01 11.8 18.9
Calcium Ca fcc 3.36 4.80 24.7 11.9 73.6 35.4
Beryllium Be hcp 3.56 12.62 13.2/21.0 17.1/24.3 37.0/59.1 48.0/68.2
Magnesium Mg hcp 4.39 11.63 8.60/7.30 9.81/8.80 19.6/16.6 22.3/20.0
Rhodium Rh fcc 4.7 6.67 4.85 3.23 10.3 6.88
Sodium Na bcc 4.77 10.21 14.4 14.7 30.2 30.9
Iridium Ir fcc 5.2 8.54 4.32 3.69 8.30 7.09
Tungsten W bcc 5.28 9.71 8.44 8.20 16.0 15.5
Molybdenum Mo bcc 5.34 9.18 6.53 5.99 12.2 11.2
Zinc Zn hcp 5.90 15.66 6.80/4.90 10.3/8.1 11.5/8.31 17.4/13.7
Cobalt Co hcp 6.2 2.55 13.1/10.9 7.31/4.82 21.2/17.6 11.8/7.77
Nickel Ni fcc 6.93 2.34 10.0 4.07 14.5 5.87
Potassium K bcc 7.20 7.94 28.6 22.7 39.7 31.5
Cadmium Cd hcp 7.5 15.55 8.35/6.89 12.6/11.3 11.1/9.18 16.8/15.1
Ruthenium Ru hcp 7.8 7.24 6.88/5.51 5.14/3.81 8.82/7.07 6.59/4.88
Indium In bct 8.8 16.32 4.64/4.45 7.62/7.18 5.27/5.05 8.65/8.16
Osmium Os hcp 8.9 8.19 7.60/5.51 6.41/4.33 8.54/6.19 7.20/4.87

result from the calculated Fermi surface and a spherical free- of the value for Cu, despite their 3–5 times larger bulk
electron surface is attributed to a combination of (i) the cal- resistivity.
culated surface area to be nearly identical to the free-
electron area, as discussed above, and (ii) Eq. (2b) for cubic IV. CONCLUSIONS
symmetry to be only dependent on the surface area, that is,
independent of v(k). In contrast, the calculated s  qo of The calculated product k  qo provides guidance in the
6.04  1022 X m s is 44% larger than the free-electron search for metals that may be suitable for narrow intercon-
value of 4.19  1022 X m s, reflecting the importance of nect lines, because the resistance of narrow lines is expected
correctly accounting for v(k). to be proportional to k  qo. The lowest k  qo is found for
The product k  qo listed in Table I is interesting for Rh, with a value that is 2.1 times smaller than for Cu, poten-
determining the most promising metals for narrow high- tially yielding narrow lines that are twice as conductive as
conductivity lines, because, based on the FS and MS models, copper. A whole range of other metals, listed in Table I, also
the smallest k  qo value is expected to yield the lowest re- become competitive with Cu in regard to effective conduc-
sistance in the limit of thin wires with fixed dimension, grain tivity for the limiting case of narrow wires, suggesting that
size distribution, and interface scattering cross sections, as processing viability and electromigration performance will
discussed above. Various metals have a lower k  qo product ultimately determine the best metal to replace Cu for inte-
than Cu, which is the metal currently used for interconnects grated circuits.
in integrated circuits. Particularly, noteworthy are Rh, Ir, and
Ni, with values that are 2.1, 1.8, and 1.6 times smaller than ACKNOWLEDGMENTS
that for Cu, respectively. In addition, metals that have k  qo This research was funded by the Semiconductor
values comparable to that of Cu but exhibit either processing Research Corporation under Task 1292.094 and through the
and/or electromigration advantages may become competitive STARnet center FAME funded by MARCO, DARPA, and
as the wire cross-sectional area shrinks. Processing advan- SRC. The author also acknowledges the NSF under Grant
tages include metals with an established atomic layer deposi- No. 1309490. Computational resources were provided by the
tion process or metals that can be processed through Center for Computational Innovations at RPI.
annealing steps to form large grains, reducing the resistivity
contribution from grain boundaries. Electromigration advan- 1
J. J. Plombon, E. Andideh, V. M. Dubin, and J. Maiz, Appl. Phys. Lett. 89,
tages primarily refer to metals that exhibit low electromigra- 113124 (2006).
2
tion such that adhesion/liner/barrier layers may become R. L. Graham, G. B. Alers, T. Mountsier, N. Shamma, S. Dhuey, S.
unnecessary, freeing up space which effectively increases Cabrini, R. H. Geiss, D. T. Read, and S. Peddeti, Appl. Phys. Lett. 96,
042116 (2010).
the interconnect cross-sectional area and, in turn, lowers 3
K. Barmak, A. Darbal, K. J. Ganesh, P. J. Ferreira, J. M. Rickman, T. Sun,
their effective resistivity. Promising metals in that regard are B. Yao, A. P. Warren, and K. R. Coffey, J. Vac. Sci. Technol. A 32,
W, Mo, Co, and Ru, all exhibiting k  qo values within 30% 061503 (2014).
085101-5 Daniel Gall J. Appl. Phys. 119, 085101 (2016)

4 25
J. S. Chawla, F. Gstrein, K. P. O’Brien, J. S. Clarke, and D. Gall, Phys. D. Choi, X. Liu, P. K. Schelling, K. R. Coffey, and K. Barmak, J. Appl.
Rev. B 84, 235423 (2011). Phys. 115, 104308 (2014).
5 26
M. M. Maqableh, X. Huang, S.-Y. Sung, K. S. M. Reddy, G. Norby, R. H. D. Choi, C. S. Kim, D. Naveh, S. Chung, A. P. Warren, N. T. Nuhfer, M.
Victora, and B. J. H. Stadler, Nano Lett. 12, 4102 (2012). F. Toney, K. R. Coffey, and K. Barmak, Phys. Rev. B 86, 045432 (2012).
6 27
D. Josell, S. H. Brongersma, and Z. T}okei, Annu. Rev. Mater. Res. 39, 231 J. S. Chawla, X. Y. Zhang, and D. Gall, J. Appl. Phys. 113, 063704
(2009). (2013).
7 28
See http://www.itrs.net for International Roadmap Committee, P. Cotti, E. M. Fryer, and J. L. Olsen, Helv. Phys. Acta 37, 585 (1964).
29
International Technology Roadmap for Semiconductors, 2013. G. Brandli and J. L. Olsen, Mater. Sci. Eng. 4, 61 (1969).
8 30
C. Y. Pan and A. Naeemi, IEEE Electron Device Lett. 35, 250 (2014). R. Englman and E. H. Sondheimer, Proc. Phys. Soc. London, Sect. B 69,
9
K. Fuchs, Math. Proc. Cambridge Philos. Soc. 34, 100 (1938). 449 (1956).
10 31
E. H. Sondheimer, Adv. Phys. 1, 1 (1952). D. Choi, M. Moneck, X. Liu, S. J. Oh, C. R. Kagan, K. R. Coffey, and K.
11
A. F. Mayadas and M. Shatzkes, Phys. Rev. B 1, 1382 (1970). Barmak, Sci. Rep. 3, 2591 (2013).
12 32
A. F. Mayadas, M. Shatzkes, and J. F. Janak, Appl. Phys. Lett. 14, 345 P. J. Price, IBM J. Res. Dev. 4, 152 (1960).
33
(1969). E. Hashimoto and Y. Ueda, J. Phys.: Condens. Matter 10, 6727 (1998).
13 34
J. S. Chawla, F. Zahid, H. Guo, and D. Gall, Appl. Phys. Lett. 97, 132106 J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865
(2010). (1996).
14 35
J. S. Chawla and D. Gall, Appl. Phys. Lett. 94, 252101 (2009). J. P. Perdew, J. A. Chevary, S. H. Vosko, K. A. Jackson, M. R. Pederson,
15
P. Y. Zheng, R. P. Deng, and D. Gall, Appl. Phys. Lett. 105, 131603 D. J. Singh, and C. Fiolhais, Phys. Rev. B 46, 6671 (1992).
36
(2014). J. Paier, M. Marsman, K. Hummer, G. Kresse, I. C. Gerber, and J. G.
16
F. Zahid, Y. Q. Ke, D. Gall, and H. Guo, Phys. Rev. B 81, 045406 (2010). 
Angy an, J. Chem. Phys. 124, 154709 (2006).
17 37
V. Timoshevskii, Y. Q. Ke, H. Guo, and D. Gall, J. Appl. Phys. 103, E. Asier and G. G. Idoia, New J. Phys. 16, 063014 (2014).
38
113705 (2008). K. M. Schep, P. J. Kelly, and G. E. W. Bauer, Phys. Rev. B 57, 8907
18
Y. Q. Ke, F. Zahid, V. Timoshevskii, K. Xia, D. Gall, and H. Guo, Phys. (1998).
39
Rev. B 79, 155406 (2009). N. Ashcroft and D. Mermin, Solid State Physics (Thomson Learning,
19
J. M. Rickman and K. Barmak, J. Appl. Phys. 114, 133703 (2013). 1976).
20 40
T.-H. Kim, X. G. Zhang, D. M. Nicholson, B. M. Evans, N. S. Kulkarni, W. M. Haynes, CRC Handbook of Chemistry and Physics, 95th ed. (CRC
B. Radhakrishnan, E. A. Kenik, and A.-P. Li, Nano Lett. 10, 3096 Press, Boca Raton, Fla., 2014).
41
(2010). J. E. A. Alderson and C. M. Hurd, Phys. Rev. B 12, 501 (1975).
21 42
M. Cesar, D. P. Liu, D. Gall, and H. Guo, Phys. Rev. Appl. 2, 044007 P. Pecheur and G. Toussaint, J. Phys. Chem. Solids 33, 2281 (1972).
43
(2014). P. G. Tomlinson, Phys. Rev. B 19, 1893 (1979).
22 44
L. Lu, Y. Shen, X. Chen, L. Qian, and K. Lu, Science 304, 422 (2004). N. V. Volkenshtejn, V. E. Startsev, V. I. Cherepanov, V. M. Azhazha, G.
23
J. S. Chawla and D. Gall, J. Appl. Phys. 111, 043708 (2012). P. Kovtun, and V. A. Elenskij, Fiz. Met. Metalloved. 45, 1187 (1978).
24 45
J. S. Chawla, X. Y. Zhang, and D. Gall, J. Appl. Phys. 110, 043714 W. Zhang, S. H. Brongersma, Z. Li, D. Li, O. Richard, and K. Maex,
(2011). J. Appl. Phys. 101, 063703 (2007).