You are on page 1of 6

Article

pubs.acs.org/cm

Ab initio Calculations of Intrinsic Point Defects in ZnSb


Lasse Bjerg, Georg K. H. Madsen,*, and Bo B. Iversen

Department of Chemistry & iNANO, Aarhus University, Denmark


Department of Atomistic Modelling and Simulation, ICAMS, Ruhr-Universitt Bochum, Germany

S Supporting Information
*

ABSTRACT: Several ecient thermoelectric materials have been found among the ternary
Zintl antimonides. If the band structure is highly asymmetric around the band gap, the eciency
as either n- or p-type may dier signicantly. The Zintl antimonides have generally been found
to be p-type. Surprisingly, this also holds true for the narrow band gap binary ZnSb and Zn4Sb3.
Using ab initio calculations, we investigate intrinsic point defects in ZnSb as a possible origin of
the p-type conductivity. Only Zn vacancies are found to be present in signicant amounts at
room temperature. The low formation energy of negatively charged Zn defects pins the
electronic chemical potential to the lower part of the band gap leading to intrinsic ZnSb being ptype. We discuss this nding as a general explanation of p-type conductivity in Zintl
antimonides, and how to overcome the doping limits in these materials.
KEYWORDS: thermoelectric materials, n-type, p-type, zintl, semiconductor, antimonides, doping bottleneck

INTRODUCTION
Thermoelectric materials can interconvert temperature dierences and electric voltages, and the most ecient are based on
heavily doped semiconductors.1,2 The electronic transport
properties are intricately related to the carrier concentration.
The thermoelectric conversion eciency of a material as either
n- or p-type may dier signicantly if the band structure is
highly asymmetric around the band gap. Furthermore,
thermoelectric devices consist of coupled n- and p-type legs,
which makes it advantageous if a material can be grown with
both carriers types. This makes it important to be able to
understand and control the defect chemistry of a material for its
application as a thermoelectric material.
One problem in thermoelectric applications is that most high
performance materials are based on rare (e.g., Te) or toxic (e.g.,
Pb) elements. As Zn and Sb are both relatively cheap,
abundant, and nontoxic, their compounds are highly attractive
for technical applications. Several ZnSb compounds have
large thermoelectric gures of merit, with the largest values
found for Zn4Sb3 (zT = 1.3 at 700 K)37 and YbCd1.6Zn0.4Sb2
(zT = 1.2 at 700 K).8
One problem of ZnSb based compounds is that they
generally have only been found to be p-type conductors. This
doping limit seems to hold for antimonides in general, Figure 1,
which is somewhat surprising as several of the compounds have
narrow band gaps which should favor symmetric doping.9
Attempts have been made at n-doping ZnSb by substituting Ga
or In for Zn,1012 and Te for Sb.12,13 However, when a negative
Seebeck coecient was found, it is reported as lost with time.12
and no evidence of phase purity is reported. Recent work shows
that the p-type conductivity is also present on single crystalline
CaZn2Sb2, EuZn2Sb2, and YbZn2Sb2, which disfavors grain
boundaries as the origin.14 Considering a ternary antimonide,
the bonding can be understood in terms of the Zintl concept,
2012 American Chemical Society

where a cation donates electrons to a covalent framework with


only small intraatomic charge transfer.15 Taking CaZn2Sb2 as an
example, one can thus write the formula as Ca2+(ZnSb)2 or,

more generally, Am+


n (ZnSb )mn. One possible explanation of the
observed p-type conductivity could be vacancies on the electron
donating Am+ cation site. However, Figure 1 shows that also the
binary ZnSb compounds come out p-type.11,16 As ZnSb and
Zn4Sb3 have no electron donating cations and narrow band
gaps, this makes the doping asymmetry especially surprising.
Zn4Sb3 has been found experimentally, and predicted by
calculations to be Zn decient compared to the Zintl balanced
stoichiometry Zn39Sb30.4,1720 It is speculated that the lack of
Zn is the cause of the observed p-type behavior. However,
Bader charge calculations on ZnSb show a charge transfer from
Zn to Sb of only 0.27 electrons.15,21 Furthermore, little is
known about the thermodynamics of charged impurities in
Zinc-Antimonides in general. Therefore, a specic investigation
of charged defects is necessary to determine if Zn deciency
leads to p-type behavior.
The hypothesis of the present work is that if we understand
the doping properties of ZnSb, it will contribute to the
understanding of the entire range of Zintl antimonides. The
doping properties are investigated by calculating the formation
energies of intrinsic point defects in ZnSb as a possible origin of
the p-type conductivity using density functional theory (DFT).
Such calculations have been used to investigate both intrinsic
and extrinsic point defects for a range of wide-band gap
semiconductors, in particular IV,24 IIIV,2527 IIVI,2831 and
CuIIIVI232,33 compounds. For thermoelectric materials,
which are typically narrow-band-gap semiconductors, a few
Received: February 27, 2012
Revised: May 14, 2012
Published: May 16, 2012
2111

dx.doi.org/10.1021/cm300642t | Chem. Mater. 2012, 24, 21112116

Chemistry of Materials

Article

Figure 1. Experimentally measured Seebeck coecients at room temperature for a range of binary and ternary Zintl antimonides.3,8,22,23 At elevated
temperatures, Ca11Sb10 also has a positive Seebeck coecient, and it thus behaves as a p-type material. The superscript letter refers to the reference
subindex.

compounds have been investigated. These include PbTe,34


Mg2Si,35 CoSb3,36 NaCoO2,37 Bi2Te3,38 and CuCrO2.39 CoSb3
is atypical as a Zintl antimonide, as it is found as both ntype4042 and p-type.4044 On the basis of ab initio calculations,
the n-type conductivity has been attributed to Co interstitial
pairs formed under Co rich conditions, whereas the p-type
conductivity is due to Co vacancies formed under Sb rich
conditions.36
The methods used for our calculations are described in the
next section. This is followed by a presentation and discussion
of the calculated formation energies. Only Zn vacancies were
found to be abundant in intrinsic ZnSb. Moreover, these were
predicted to be negatively charged, explaining the observed ptype behavior.

and if Zn or Sb are not to precipitate out, it is limited to


Hf Hf

where Hf is the formation enthalpy of ZnSb. = Hf mimics a


Zn-rich environment, whereas = Hf mimics an Sb-rich
environment. qd is the charge of the defected supercell, VBM is the
energy of the valence band maximum (VBM), and e is the chemical
potential of the electrons relative to the VBM.
Computational Details. The Ed energies and the ground-state
reference energies, bulk
i , have been calculated using the ab initio total
energy and molecular dynamics program VASP47,48 with the PBE
functional,49 and PAW basis sets50,51 with a plane wave cuto energy
of 553 eV.
The ZnSb unit cell contains 16 atoms. Calculations have been
performed for single defects in increasingly larger supercells: 2 2 2
face centered (32 atoms); 2 2 2 body centered (64 atoms); 2 2
2 (128 atoms); and 4 4 4 face centered (256 atoms). -pointcentered k grids of sizes 6 6 6, 4 5 5, 4 4 4, and 3 3
3, respectively, were used. Convergence was achieved in the sense that
doubling the k-grid density changed the energy by less than 10 meV
for the neutral Zn vacancy calculations. The unit cell parameters were
held xed at the pure ZnSb calculated equilibrium values (a = 6.284 ,
b = 7.826 , c = 8.229 compared to experimentally measured values
a = 6.204 , b = 7.741 , c = 8.098 52) to mimic the isolated defect.
All the ions were allowed to relax into their equilibrium positions.
Energy Corrections. Ideally, we would be interested in Ed for an
innitely large supercell with one defect, Ed,. When limited to nite
size supercells, electrostatic self-interaction and strain eects cause Ed
to dier from Ed,. According to Makov and Payne, the electrostatic
self-interaction is given by53

METHODS

Defect Thermodynamics. The equilibrium concentration of a


given defect in a structure, cd, is given by a Boltzmann distribution
cd = c0eGf / kBT

(1)

where Gf is the formation Gibbs free energy of the defect, c0 is the


concentration of possible defect sites, kB is the Boltzmann constant,
and T is the temperature. If only bulk phases are involved in the
formation process, and if the pressure is moderate, we can approximate
Gf by the formation energy of the defect, Ed.45,46
For a binary compound, such as ZnSb, the formation energy of a
defected structure in equilibrium with its surroundings is given
by25,45,46

1
bulk
(nZn + nSb)ZnSb
2
1
bulk
(nZn nSb)(Zn
Sbbulk )
2
1
(nZn nSb) + qd(VBM + e ).
2

Ed = Ed

Ed, = Ed +

qd2
2L

2qdQ
3L3

+ 6(L5)

(5)

where is the Magdelung constant of the supercell Bravais lattice, is


the relative static permittivity of the material, L is the distance between
defect centers, and Q is the quadrupole moment. In the current paper,
was calculated theoretically for pure ZnSb to be 36.8, using density
functional perturbation theory as implemented in VASP.54,55 This
value was used to calculate the term proportional to L1. The hereby
obtained values for dierent supercell sizes were tted to obtain a
correction proportional to L3. This correction can be expected to also
contain strain eects, which are of order L3.56,57 Terms of 6(L5)
and higher were neglected, since they were considered to contribute
minimally for the larger supercells. Generally, the calculations showed
good convergence at supercell sizes of 64 atoms after correction.

(2)

Here, Ed is the calculated energy of a defected supercell containing nZn


Zn atoms and nSb Sb atoms. bulk
is the chemical potential of the
i
compound i in the ground state bulk structure. If entropy and pressure
is set equal to the calculated
contributions are again neglected, bulk
i
energy of the structure. is dened as
bulk
= Zn Sb (Zn
Sbbulk )

(4)

(3)
2112

dx.doi.org/10.1021/cm300642t | Chem. Mater. 2012, 24, 21112116

Chemistry of Materials

Article

Larger divergences were found for the 32 atoms cell size. The reason
for this can be attributed to 6(L5) corrections being more important
at small cell sizes, and the possibility of defect wave function overlap
between cells. Therefore, only cell sizes of 64 atoms and larger were

the shifts decreased with increasing supercell size and were roughly
proportional to L3. We, therefore, did not introduce explicit shifts in
our calculations, but relied on the L3 t to compensate for a major
part of the shifts.
PBE is generally considered to underestimate the band gap of
semiconductors. Band gaps in better agreement with experiment can
be obtained using hybrid functionals with exact exchange or GW
calculations to get shifts of the VBM and CBM. These calculations are
much more costly, however, and the methods have not been
considered in our case. PBE gives a band gap in ZnSb of only 0.05
eV, whereas it has been observed experimentally to be around 0.5
eV.15,23,5860

RESULTS AND DISCUSSION


The ZnSb structure contains 16 atoms in the unit cell, but only
one crystallographically distinct Zn site and one distinct Sb site.
We considered vacancies on the Zn, VZn, and Sb, VSb, positions
and antisite defects with Sb on a Zn position, SbZn, and vice
versa, ZnSb. ZnSb is 17% more densely packed than ZnTe in
the zincblende structure, and there are no clear interstitial sites.
Therefore, interstitial defects were not considered. Charged
states from 3 through +1 were calculated for all defects. The
calculated defect formation energies are listed in Table 1. Most
striking in Table 1 is the low formation energy of both the
neutral and negatively charged Zn vacancies.
Table 1. Defect Formation Energies in eV for e = = 0
qd

VZn

VSb

SbZn

ZnSb

3
2
1
0
+1

0.51
0.30
0.32
0.52
0.88

2.47
2.05
1.83
1.79
1.70

1.83
1.50
1.33
1.34
1.39

1.33
1.26
1.28
1.41
1.67

We furthermore considered two defect complexes involving


Zn vacancies: A Zn vacancy and a Zn on a neighboring Sb
position, VZn+ZnSb, and two neighboring Zn vacancies,
VZn+VZn. The VZn+ZnSb complex was found to have an energy
only slightly lower than the sum of the individual defects. The
VZn+VZn complex was found to have a formation energy almost
twice as high as the sum of the single vacancies. Such complexes
are therefore unfavorable, and are not expected to play a
signicant role.
Depending on the position of the electron chemical
potential, dierent charge states will be the most stable for a
given defect, eq 2. In Figure 3, the lowest energy charge state is
plotted for each defect as a function of e. The slope of the line
corresponds to the charge state. It is clearly seen that the Zn
vacancy is the dominant defect throughout the band gap.
In the dilute limit of defect concentrations, the electronic
band structure is considered to be that of the defect free bulk
structure. Discrete defect states are considered to accept
electrons from or donate electrons to the bands if they are
negatively or positively charged. Conservation of electrons then
gives the equation

Figure 2. Defect formation energies as-calculated, monopole


corrected, and L3-tted for (a) Zn vacancy with q = 3 and (b)
Sb antisite with q = 3.
used for the L3 ts. Figure 2 shows the as-calculated and corrected
energies for the defects with the best and worst convergence. Plots for
the remaining defects can be found in the Supporting Information.
The MakovPayne correction is expected to be an upper limit, only
valid for localized charge states.27,33 For completely delocalized
charges, the multipole correction should be zero. We have, therefore,
also investigated corrections based solely on L3 ts. For q = 3
charge states, we nd that the formation energies are about 0.2 eV
smaller using this correction method. The remaining charge states
have dierences smaller than 0.1 eV. We will discuss this further in the
Results section.
The band energies of a defected supercell are shifted relative to the
band energies of the perfect unit cell.27,33 Hence, the VBM in eq 2
should be shifted accordingly relative to the VBM energy of the defectfree structure, 0VBM. The energy shift is normally found in one of two
ways. Either by aligning the average potential of the defected supercell
far from the defect with the average potential of the perfect cell, or by
aligning localized states at atoms far from the defect.27,33 For ZnSb, the
Zn d-states are fairly localized, and expected to be similar to the dstates of perfect ZnSb far from the defect. For our calculations, we
found that the shift was small, ranging from 40 to +30 meV for the
largest supercells and most distant Zn atoms. Moreover, we found that

VBM

n()d =

n()f (; e)d qdcd


d

(6)

where n() is the electronic density of states of the perfect


crystal, and f(;e) is the Fermi distribution. In the picture of
holes and electrons, this can be rewritten as
2113

dx.doi.org/10.1021/cm300642t | Chem. Mater. 2012, 24, 21112116

Chemistry of Materials

Article

be a positive number, indicating p-type behavior. The top of the


gure shows the obtained value of e. As the temperature rises,
e moves closer to the VBM and even crosses it because of the
increasing number of negatively charged defects. At 300 K we
found nh ne = 8.8 1017 cm3, and a total Zn vacancy
concentration of 5.1 1017 cm3, corresponding to 0.003% of
the Zn sites being vacant. At 700 K nh ne = 3.6 1019 cm3,
and the Zn vacancy concentration was 3.3 1019 cm3,
corresponding to 0.2%, or one Zn vacancy in a 1024 atoms
supercell. This is a rather large concentration, and we stress that
the dilute-limit approximations lose validity. However, the
qualitative results of an intrinsic p-type material with Zn
vacancies remain. Experimentally, the Hall carrier concentration
is found to vary between 4 1017 and 1 1018 cm3 at room
temperature for undoped samples.23 If Zn diusion is expected
to occur at a reasonable rate at room temperature, these
numbers agree very well with our calculations.
The nearest atoms around the Zn vacancy were found to
move closer to the vacancy. Figure 5 shows the coordination

Figure 3. Calculated defect energies for varying chemical potential.


The slope of the line corresponds to the charge state.

nh ne +

qdcd = 0.
d

(7)

nh and ne are the numbers of holes and electrons, respectively,


and given by
VBM

nh =

ne =

n()[1 f ( ; e )]d

(8)

CBM

n()f ( ; e )dE

(9)

CBM is the energy of the conduction band minimum (CBM).


Because cd for each defect depends on e through eqs 1 and 2
and nh and ne also depend on e, eq 7 can be solved iteratively
for any given temperature, and thus it is possible to predict the
sign of the majority carriers, as well as the equilibrium defect
concentrations.
For n() in eq 7, the DFT density of states, with the band
gap scissored to 0.5 eV, was used. Equation 7 was solved
iteratively at a range of temperatures which gave the plot in
Figure 4. Only Zn vacancy concentrations are shown for the
dierent charge states. The 2, 1, and 0 states were found to
dominate. The remaining defects have concentrations orders of
magnitude lower for all temperatures, and they thus fall out of
the range of the plot. Also shown is nh ne. This was found to

Figure 5. Coordination of the Zn (gray) and Sb (blue) atoms around


the Zn vacancy location (a) in pure ZnSb and (b) for a Zn vacancy
with q = 2.

around the vacancy position before and after introduction of


the vacancy. The neighboring Zn atom moved 0.4 closer to
the vacancy. This changed its coordination to be a more regular
tetrahedron. We speculate that the two neighboring Zn atoms
in pure ZnSb hinder each other from obtaining their desired
coordination. This can be part of the reason why the vacancy
formation energy is so low. The low ionization energy of the
vacancy, we then attribute to the willingness of Sb to take up
extra charge.
The high degree of delocalization of the charge states
suggests that a Makov-Payne correction scheme will overcompensate for the charge self-interaction, as mentioned in the
Methods section. We, therefore, also used a simpler correction
scheme based only on an L3 t. Even though the energies
changed by up to 0.2 eV, the overall conclusion remained the
same. At 300 K we found nh ne = 2.7 1018 cm3, and a total
Zn vacancy concentration of 1.4 1018 cm3. At 700 K nh ne
= 5.0 1019 cm3, and the Zn vacancy concentration was 3.9
1019 cm3.
Finally, we considered whether it was possible to eliminate
Zn vacancies by growing crystals in a Zn rich environment. Hf
for ZnSb has been calculated to be 0.068 eV. For vacancies
|nZn nSb| = 1, whereas for antisites it is 2. A Zn rich or Sb rich
environment would thus only have a small eect on the
formation energies, eqs 2 and 4, and hence on the size of nh
ne. Therefore, it is not possible to eliminate Zn vacancies and

Figure 4. Top: Position of electron chemical potential, e. Bottom:


Equilibrium concentrations of Zn vacancy defects in dierent charge
states, and the total number of holes minus electrons. Experimental
Hall carrier concentrations23 are plotted for comparison.
2114

dx.doi.org/10.1021/cm300642t | Chem. Mater. 2012, 24, 21112116

Chemistry of Materials

p-type behavior by simply providing a Zn rich environment, as


long as the ZnSb formation happens under equilibrium
conditions. Zn rich conditions will correspond to multiplying
cd by exp ((0.0034 eV)/(kBT)). At 300 K, this is 0.88, at 700
K it is 0.94; the order of magnitude is not changed.

CONCLUSION
Calculations of intrinsic defects in ZnSb showed that only Zn
vacancies are expected to be present in signicant amounts at
room temperature and that the negative charge of these defects
gives rise to intrinsic ZnSb being p-type. Our results suggest
that it will not be possible to n-dope ZnSb. The low formation
energy of Zn defects pins the electronic chemical potential to
the lower part of the band gap. As the chemical potential is
shifted up in the band gap, it becomes more and more favorable
for Zn vacancies to form. Since these vacancies will be
negatively charged, they will lead to p-type behavior. Crystallographic studies have shown that ZnSb becomes Zn decient at
temperatures above 673 K under dynamic vacuum.52 This
qualitatively conrms our predictions that Zn vacancy
formation is energetically accessible at moderate temperatures.
We chose ZnSb as an example for the study of defect stability
because its small band gap should favor symmetric doping. This
was found not to be the case, and we therefore believe that the
mechanism proposed, the formation of negatively charged Zn
vacancies, can be general for all compounds based on a ZnSb
framework. We, furthermore, found that the low formation
energy of ZnSb limits the usefulness of growing ZnSb in a Zn
rich environment. Our results suggest two routes to n-doped
ZnSb framework compounds. First of all one should look for
systems where both the A and Zn vacancy formation energy are
large, and where the overall stability of the compound is high.
Furthermore, no positions were found for interstitial defects
which have otherwise led to n-type conductivity in CoSb3,
suggesting that more open systems could be favorable.
ASSOCIATED CONTENT

* Supporting Information
S

As-calculated and corrected energies for all defects and charge


states; discussion of the density of states of the Zn vacancy.
This material is available free of charge via the Internet at
http://pubs.acs.org/.

REFERENCES

(1) Snyder, G. J.; Toberer, E. S. Nat. Mater. 2008, 7, 105114.


(2) Sootsman, J. R.; Chung, D. Y.; Kanatzidis, M. G. Angew. Chem.,
Int. Ed. 2009, 48, 86168639.
(3) Caillat, T.; Fleurial, J.; Borshchevsky, A. J. Phys. Chem. Solids
1997, 58, 11191125.
(4) Snyder, G.; Christensen, M.; Nishibori, E.; Caillat, T.; Iversen, B.
Nat. Mater. 2004, 3, 458463.
(5) Cargnoni, F.; Nishibori, E.; Rabiller, P.; Bertini, L.; Snyder, G.;
Christensen, M.; Gatti, C.; Iversen, B. Chem.Eur. J. 2004, 10, 3861
3870.
(6) Pedersen, B. L.; Yin, H.; Birkedal, H.; Nygren, M.; Iversen, B. B.
Chem. Mater. 2010, 22, 23752383.
(7) Iversen, B. B. J. Mater. Chem. 2010, 20, 1077810787.
(8) Wang, X.-J.; Tang, M.-B.; Chen, H.-H.; Yang, X.-X.; Zhao, J.-T.;
Burkhardt, U.; Grin, Y. Appl. Phys. Lett. 2009, 94, 092106.
(9) Zunger, A. Appl. Phys. Lett. 2003, 83, 5759.
(10) Justi, E.; Rasch, W.; Schneider, G. Adv. Energy Convers. 1964, 4,
2738.
(11) Schneider, G. Phys. Status Solidi 1969, 33, K133K136.
(12) Abou-Zeid, A.; Schneider, G. Z. Naturforsch. A 1975, 30, 381
382.
(13) Abou-Zeid, A.; Schneider, G. Phys. Status Solidi A 1971, 6,
K101K103.
(14) May, A. F.; McGuire, M. A.; Ma, J.; Delaire, O.; Huq, A.;
Custelcean, R. J. Appl. Phys. 2012, 111, 033708.
(15) Bjerg, L.; Madsen, G. K. H.; Iversen, B. B. Chem. Mater. 2011,
23, 39073914.
(16) Ioe, A. F. Semiconductor Thermoelements and Thermoelectric
Cooling; Infosearch Ltd., 1957.
(17) Nylen, J.; Lidin, S.; Andersson, M.; Liu, H.; Newman, N.;
Haussermann, U. J. Solid State Chem. 2007, 180, 26032615.
(18) Toberer, E. S.; Rauwel, P.; Gariel, S.; Tafto, J.; Snyder, G. J. J.
Mater. Chem. 2010, 20, 98779885.
(19) Pomrehn, G. S.; Toberer, E. S.; Snyder, G. J.; van deWalle, A.
Phys. Rev. B 2011, 83, 094106.
(20) Pomrehn, G. S.; Toberer, E. S.; Snyder, G. J.; van de Walle, A. J.
Am. Chem. Soc. 2011, 133, 1125511261.
(21) Benson, D.; Sankey, O. F.; Haeussermann, U. Phys. Rev. B 2011,
84, 125211.
(22) (a) Toberer, E. S.; May, A. F.; Scanlon, C. J.; Snyder, G. J. J.
Appl. Phys. 2009, 105, 063701. (b) Kawano, K.; Kurosaki, K.; Muta,
H.; Yamanaka, S. J. Appl. Phys. 2008, 104, 013714. (c) Gascoin, F.;
Ottensmann, S.; Stark, D.; Haile, S.; Snyder, G. Adv. Funct. Mater.
2005, 15, 18601864. (d) Zhang, H.; Zhao, J.-T.; Grin, Y.; Wang, X.J.; Tang, M.-B.; Man, Z.-Y.; Chen, H.-H.; Yang, X.-X. J. Chem. Phys.
2008, 129, 164713. (e) Zhang, H.; Fang, L.; Tang, M.-B.; Chen, H.-H.;
Yang, X.-X.; Guo, X.; Zhao, J.-T.; Grin, Y. Intermetallics 2010, 18,
193198. (f) May, A. F.; Toberer, E. S.; Snyder, G. J. J. Appl. Phys.
2009, 106, 013706. (g) Wang, H. F.; Cai, K. F.; Li, H.; Wang, L.;
Zhou, C. W. J. Alloys Compd. 2009, 477, 519522. (h) Wang, X.-J.;
Tang, M.-B.; Zhao, J.-T.; Chen, H.-H.; Yang, X.-X. Appl. Phys. Lett.
2007, 90, 232107. (i) Saparov, B.; Saito, M.; Bobev, S. J. Solid State
Chem. 2011, 184, 432440. (j) Kim, S.; Kanatzidis, M. Inorg. Chem.
2001, 40, 37813785. (k) Saparov, B.; He, H.; Zhang, X.; Greene, R.;
Bobev, S. Dalton Trans. 2010, 39, 10631070. (l) Zevalkink, A.;
Toberer, E. S.; Zeier, W. G.; Flage-Larsen, E.; Snyder, G. J. Energy
Environ. Sci. 2011, 4, 510518. (m) Zevalkink, A.; Toberer, E. S.;
Bleith, T.; Flage-Larsen, E.; Snyder, G. J. J. Appl. Phys. 2011, 110,
013721. (n) Park, S.; Choi, E.; Kang, W.; Kim, S. J. Mater. Chem. 2002,
12, 18391843. (o) Kim, S.; Ireland, J.; Kannewurf, C.; Kanatzidis, M.
J. Solid State Chem. 2000, 155, 5561. (p) Kim, S.; Hu, S.; Uher, C.;
Kanatzidis, M. Chem. Mater. 1999, 11, 31543159. (q) Zheng, W.-Z.;
Wang, P.; Wu, L.-M.; Liu, Y.; Chen, L. Inorg. Chem. 2010, 49, 5890
5896. (r) Brown, S.; Kauzlarich, S.; Gascoin, F.; Snyder, G. Chem.
Mater. 2006, 18, 18731877. (s) Yi, T.; Cox, C. A.; Toberer, E. S.;
Snyder, G. J.; Kauzlarich, S. M. Chem. Mater. 2010, 22, 935941.
(t) Brown, S. R.; Kauzlarich, S. M.; Gascoin, F.; Snyder, G. J. J. Solid
State Chem. 2007, 180, 14141420. (u) Ponnambalam, V.; Gao, X.;

Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: georg.madsen@rub.de.
Notes

The authors declare no competing nancial interest.

ACKNOWLEDGMENTS
This work was supported by the Danish National Research
Foundation (Center for Materials Crystallography), the Danish
Strategic Research Council (Centre for Energy Materials), and
the Danish Center for Scientic Computing. G.K.H.M.
acknowledges nancial support through ThyssenKrupp AG,
Bayer MaterialScience AG, Salzgitter Mannesmann Forschung
GmbH, Robert Bosch GmbH, Benteler Stahl/Rohr GmbH,
Bayer Technology Services GmbH, and the state of NorthRhine Westphalia, as well as the European Commission in the
framework of the European Regional Development Fund
(ERDF).
2115

dx.doi.org/10.1021/cm300642t | Chem. Mater. 2012, 24, 21112116

Chemistry of Materials

Article

Lindsey, S.; Alboni, P.; Su, Z.; Zhang, B.; Drymiotis, F.; Daw, M. S.;
Tritt, T. M. J. Alloys Compd. 2009, 484, 8085. (v) Candol, C.;
Lenoir, B.; Chubilleau, C.; Dauscher, A.; Guilmeau, E. J. Phys.-Condens.
Mater. 2010, 22, 025801. (w) Condron, C. L.; Kauzlarich, S. M.;
Gascoin, F.; Snyder, G. J. J. Solid State Chem. 2006, 179, 22522257.
(23) Bottger, P. H. M.; Pomrehn, G. S.; Snyder, G. J.; Finstad, T. G.
Phys. Status Solidi A 2011, 208, 27532759.
(24) Goss, J. P.; Jones, R.; Heggie, M. I.; Ewels, C. P.; Briddon, P. R.;
O berg, S. Phys. Rev. B 2002, 65, 115207.
(25) Zhang, S. B.; Northrup, J. E. Phys. Rev. Lett. 1991, 67, 2339
2342.
(26) Neugebauer, J.; Van de Walle, C. G. Phys. Rev. Lett. 1995, 75,
44524455.
(27) Van de Walle, C.; Neugebauer, J. J. Appl. Phys. 2004, 95, 3851
3879.
(28) Kohan, A. F.; Ceder, G.; Morgan, D.; Van de Walle, C. G. Phys.
Rev. B 2000, 61, 1501915027.
(29) Zhang, S. B.; Wei, S.-H.; Zunger, A. Phys. Rev. B 2001, 63,
075205.
(30) Park, C. H.; Zhang, S. B.; Wei, S.-H. Phys. Rev. B 2002, 66,
073202.
(31) Lany, S.; Zunger, A. Phys. Rev. Lett. 2007, 98, 045501.
(32) Zhao, Y.; Persson, C.; Lany, S.; Zunger, A. Appl. Phys. Lett.
2004, 85, 58605862.
(33) Persson, C.; Zhao, Y.-J.; Lany, S.; Zunger, A. Phys. Rev. B 2005,
72, 035211.
(34) Xiong, K.; Lee, G.; Gupta, R. P.; Wang, W.; Gnade, B. E.; Cho,
K. J. Phys. D Appl. Phys. 2010, 43, 405403.
(35) Kato, A.; Yagi, T.; Fukusako, N. J. Phys.Condens. Mater. 2009,
21, 205801.
(36) Park, C.-H.; Kim, Y.-S. Phys. Rev. B 2010, 81, 085206.
(37) Yoshiya, M.; Okabayashi, T.; Tada, M.; Fisher, C. J. Electron.
Mater. 2010, 39, 16811686 10.1007/s116640101237-x..
(38) Hashibon, A.; Elsasser, C. Phys. Rev. B 2011, 84, 144117.
(39) Scanlon, D. O.; Watson, G. W. J. Mater. Chem. 2011, 21, 3655
3663.
(40) Sharp, J. W.; Jones, E. C.; Williams, R. K.; Martin, P. M.; Sales,
B. C. J. Appl. Phys. 1995, 78, 10131018.
(41) Liu, W.-S.; Zhang, B.-P.; Li, J.-F.; Zhao, L.-D. J. Phys. D: Appl.
Phys. 2007, 40, 566.
(42) Liu, W.-S.; Zhang, B.-P.; Li, J.-F.; Zhao, L.-D. J. Phys. D: Appl.
Phys. 2007, 40, 6784.
(43) Morelli, D. T.; Caillat, T.; Fleurial, J. P.; Borschevsky, A.;
Vandersande, J.; Chen, B.; Uher, C. Phys. Rev. B 1995, 51, 96229628.
(44) Caillat, T.; Borshchevsky, A.; Fleurial, J. J. Appl. Phys. 1996, 80,
44424449.
(45) Qian, G.-X.; Martin, R. M.; Chadi, D. J. Phys. Rev. B 1988, 38,
76497663.
(46) berg, D.; Erhart, P.; Williamson, A. J.; Lordi, V. Phys. Rev. B
2008, 77, 165206.
(47) Kresse, G.; Hafner. J. Phys. Rev. B 1993, 47, 558561.
(48) Kresse, G.; Furthmuller, J. Phys. Rev. B 1996, 54, 1116911186.
(49) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,
38653868.
(50) Blochl, P. E. Phys. Rev. B 1994, 50, 1795317979.
(51) Kresse, G.; Joubert, D. Phys. Rev. B 1999, 59, 17581775.
(52) Mozharivskyj, Y.; Pecharsky, A.; Budko, S.; Miller, G. Chem.
Mater. 2004, 16, 15801589.
(53) Makov, G.; Payne, M. C. Phys. Rev. B 1995, 51, 40144022.
(54) Gajdos, M.; Hummer, K.; Kresse, G.; Furthmuller, J.; Bechstedt,
F. Phys. Rev. B 2006, 73, 045112.
(55) Baroni, S.; Resta, R. Phys. Rev. B 1986, 33, 70177021.
(56) Grecu, D.; Dederichs, P. H. Phys. Lett. A 1971, 36, 135136.
(57) Dederichs, P. H.; Pollmann, J. Z. Phys. 1972, 255, 315324.
(58) Turner, W. J.; Fischler, A. S.; Reese, W. E. Phys. Rev. 1961, 121,
759767.
(59) Zavetova, M. Phys. Status Solidi 1964, 5, K19K21.
(60) Yamada, Y. Phys. Status Solidi B 1978, 85, 723732.

2116

dx.doi.org/10.1021/cm300642t | Chem. Mater. 2012, 24, 21112116

You might also like