See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/230546072

First‐principles calculations on the four phases of BaTiO3

Article in Journal of Computational Chemistry · April 2012

DOI: 10.1002/jcc.22942

CITATIONS READS
110 1,291

2 authors:

Robert Evarestov Andrei Bandura

Saint Petersburg State University Saint Petersburg State University
328 PUBLICATIONS 6,434 CITATIONS 134 PUBLICATIONS 3,777 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Robert Evarestov on 28 August 2017.

The user has requested enhancement of the downloaded file.

 WWW.C-CHEM.ORG FULL PAPER

First-Principles Calculations on the Four Phases of BaTiO3

Robert A. Evarestov*[a] and Andrei V. Bandura[a]

The calculations based on linear combination of atomic orbitals techniques. It is concluded that the hybrid PBE0 [J. P. Perdew, K.
basis functions as implemented in CRYSTAL09 computer code Burke, M. Ernzerhof, J. Chem. Phys. 1996, 105, 9982.] functional is
have been performed for cubic, tetragonal, orthorhombic, and able to predict correctly the structural stability and phonon
rhombohedral modifications of BaTiO3 crystal. Structural and properties both for cubic and ferroelectric phases of BaTiO3. The
electronic properties as well as phonon frequencies were obtained comparative phonon symmetry analysis in BaTiO3 four phases has
using local density approximation, generalized gradient been made basing on the site symmetry and irreducible
approximation, and hybrid exchange-correlation density functional representation indexes for the first time. V
C 2012 Wiley Periodicals, Inc.

theory (DFT) functionals for four stable phases of BaTiO3. A

comparison was made between the results of different DFT DOI: 10.1002/jcc.22942

Introduction dependence can be analyzed to study the thermodynamic

properties. Moreover, the first-principles calculations of total
Barium titanate BaTiO3 (BTO) is of great scientific interest as
energies for small distortions of an ideal crystal with the cubic
the prototypical ferroelectric oxide—many properties of solids perovskite structure permit an effective Hamiltonian construc-
related to ferroelectricity have been found for the first time in tion for taking into account the anharmonic effects and treat-
BTO. Because of the number of different ferroelectric phases
ing the phase transitions in perovskites[14–16] using the statisti-
and the ease of switching between them BTO is also of great cal-mechanics methods. Based on this approach, temperature–
technological interest. In particular, BTO is used[1,2] in fields pressure dependence phase diagram of BaTiO3 was investi-
such as electronics, nonlinear optics, electrochemistry, data
gated.[15] However, concerning the applied DFT exchange-cor-
storage. Intense experimental efforts targeted toward the syn- relation (XC) functional, authors[15] noticed that the results of
thesis of ferroelectric nanostructures[3] (nanoparticles, nano- effective-Hamiltonian calculations are to be considered as reli-
wires, and nanotubes) based on BTO can be expected to fur-
able only at the qualitative level due to the well-known under-
ther expand its applicability. estimation of the equilibrium volumes given by local density
BTO crystallizes either in the perovskite crystal structure or approximation (LDA), which brings about a systematic error in
in its hexagonal modification. The crystal structure of the per- the location of zero of pressure (see also the discussion
ovskite polymorph (Fig. 1) changes from cubic (Pm3m) to tet- below).
ragonal (P4mm) at 403 K, then to orthorhombic (Amm2) at The first-principles plane wave (PW) DFT, full-potential line-
278 K, and to rhombohedral (R3m) at 183 K.[4] The cubic phase arized augmented PW (FLAPW) DFT, and linear combination of
is paraelectric and the other phases are ferroelectric. In all four atomic orbitals (LCAOs)-based hybrid density functional pho-
phases, there is one BTO formula unit in the unit cell so that non spectra calculations on BTO cubic phase (see Ref. [17] and
phase transitions to ferroelectric phases are due to the soft references therein) demonstrated its structural instability via
mode at C point of cubic BTO Brillouin zone (BZ).[5] soft modes with imaginary frequencies at the BZ symmetry
It is well known that BTO ferroelectricity properties can only points C(0, 0, 0), X(0, 0, 0.5), and M(0, 0.5, 0.5).
be explained based on the lattice dynamics study.[6,7] That is First-principles DFT calculations of the electronic, structural,
why the exact knowledge of the vibrational states of a crystal and vibrational properties of BTO for noncubic (ferroelectric)
is the key to the calculation of its thermodynamic properties phases, using LDA and generalized gradient approximations
as the function of temperature.[8] Thus, infrared, Raman, or (GGAs), are known for the most perovskite BTO phases. Thus,
neutron scattering are valuable techniques for the exploring of PW LDA,[18] PW GGA,[19,20] and FLAPW GGA[21] computations
BTO lattice dynamics. In particular, the frequency of the lowest were performed for the tetragonal phase; PW LDA,[12,18,22] PW
frequency transverse optical (TO) mode was predicted to be GGA,[20] and PW density functional perturbation theory[23]
strongly temperature dependent, and the rigid ion model was studies were published for rhombohedral phase. To the best
applied to tetragonal BTO phase to explain this dependence.[9]
Our knowledge concerning the nature of ferroelectric phase [a] R. A. Evarestov, A. V. Bandura
Department of Chemistry, Quantum Chemistry Division, St. Petersburg
transitions in BTO was rather limited until first-principles den-
University, Universitetsky Prosp. 26, 198504 St. Petersburg, Petergof, Russia
sity functional theory (DFT) calculations were used to study E-mail: re1973@re1973.spb.edu
BTO lattice dynamics.[10–13] Once the electronic (ground) state Contract/grant sponsor: Saint-Petersburg State University; Contract/
of a system of nuclei and electrons is found and the phonon grant number: 12.37.142.2011 (for fundamental research)
frequencies are calculated, the free-energy temperature V
C 2012 Wiley Periodicals, Inc.

Journal of Computational Chemistry 2012, 33, 1123–1130 1123

 FULL PAPER WWW.C-CHEM.ORG

hybrid XC functionals are used. These functionals mix a frac-

tion of HF exchange into the DFT exchange functional and
use DFT correlation part. Hybrid XC functionals are popular in
molecular quantum chemistry, as they have been shown to
provide accurate first-principles description of the bond
lengths, atomization energies, vibrational spectra for most
molecules.[25] The B3PW and B3LYP functionals contain pa-
rameters, which were optimized to fit the experimental data
for molecules and from this point of view are not strictly
nonempirical. In the past decade, these hybrid functionals
have been also increasingly applied for solids.[26,27] In some
cases, these functionals give reasonable results for the crystal
atomic structure. However, it was shown[28] that for cubic
BTO crystal B3LYP calculations give the value of 4.04 Å for lat-
tice constant, that is, by 1% larger than the experimental
value of 3.996 Å. Hybrid functional PBE0[29,30] with PBE[31] XC
part is parameter free as the percentage of HF exchange mix-
ing with PBE exchange is found from the perturbation theory.
The lattice constant in our PBE0 calculations is 3.992 Å (see
below), that is, differs from experimental value only by
Figure 1. Structure of conventional unit cells for a) cubic, b) tetragonal, c)
0.1%. Comparison of our result with those given in Table 1
orthorhombic, and d) rhombohedral BaTiO3. Legend: Ti—small violet balls,
from Ref. [7] shows that our result is mostly close to the ex-
Ba—middle green balls, O—large light-pink balls.
perimental data as different theoretical lattice constants of
the cubic BTO differ from the experimental data in the inter-
of our knowledge, the phonon calculation for orthorhombic val
1.4% to þ0.4%.
BTO phase is absent in the literature. The most hybrid functional applications to solids use LCAO
Cited above DFT calculations of BTO ferroelectric phases are basis set as it is done for molecules. This is explained by the
based, in most cases, on LDA for XC part of the density func- difficulties in the HF method formulation for PW basis set.
tional and to a lesser extent on the semilocal GGA. LDA
predicts good phonon frequencies, if the frequencies are
evaluated at the theoretically predicted equilibrium volume Table 1. Calculated bulk properties of BaTiO3 cubic phase; Q and Z are
because of sizable error compensation (LDA equilibrium vol- Mulliken and Born atomic charges, respectively.
ume is too small but LDA frequencies at the experimental vol-
Our This work
ume are too soft). GGA drastically underestimates the phonon previous
frequencies as a result of a too large equilibrium volume. To work[17]
overcome the limitation of DFT calculations in estimation of Property PBE0 LDA PBE PBE0 Experiment
structural parameters in ferroelectrics, Wu and Cohen[24] intro- Cell a (Å) 3.988 3.958 4.036 3.992 3.996[b]
duced a new flavor of GGA and obtained a better agreement Bulk modulus B (GPa) 185 197 166 189 162[c]
Cohesive energy Eat (eV) 29.9 38.45 32.70 30.77 31.6[c]
between calculated and experimentally observed lattice con-
Band gap Egap (eV) 3.7 1.9 1.9 4.0 3.2[d]
stants at 0 K for BTO. Sometimes in DFT calculations, the ex- High-frequency dielectric 5.2 6.32 6.34 5.11 5.4[e]
perimental data for lattice parameters are used and optimiza- constant, e1
tion is made only for the atomic positions in unit cell as it was QBa 1.90 1.78 1.77 1.80
ZBa 2.60 2.74 2.71 2.68 2.9[f ]
done by Hermet et al.[18] for tetragonal BTO. However, such an QTi 2.24 2.20 2.25 2.39
approach cannot be considered as a fully first-principles one. ZTi 7.06 7.30 7.36 6.98 6.7[f ]
However, use of first-principles calculations with Hartree–Fock QO
1.38
1.33
1.34
1.40
(HF) approximation (which excludes the effects of electron cor- ZO1
5.64
5.80
5.89
5.58
4.8[f ]
ZO2
2.01
2.12
2.09
2.04
2.4[f ]
relation) gives the calculated structure and phonon frequen- C4–(T1u)–TO1 281i 133i 240i 257i (263i) soft
cies in poor agreement with the experimental data. The essen- C4–(T1u)–TO2 196 193 175 191 (191) 181[g]
tial error of HF arises from the fact that in the HF model, the C5–(T2u)–silent 314 296 290 313 (305) 306[g]
C4–(T1u)–TO3 478 483 460 483 (474) 487[g]
XC hole in a two-electron bond is too delocalized. The simple
C4–(T1u)–LO1 182 187 165 178 (178) 180[g]
truth is not that LDA/GGA is particularly good but that HF is C4–(T1u)–LO2 472 461 451 478 (469) 468[g]
rather poor in the two-electron chemical bond description.[17] C4–(T1u)–LO3 677 698 647 697 (696) 717[g]
In the perovskites, such bonds arise between the transition Phonon frequencies[a] (cm
1) are given at the C-point of BZ; TO—trans-
metal and oxygen atoms, being important for the correct crys- verse, LO—longitudinal optical phonons.
tal properties description. [a] Frequencies calculated by PBE0 functional with reduced level
To escape the use of experimental data for structure and (XLGRID) of DFT integration accuracy are given in brackets. [b] Ref. [51].
[c] Ref. [52]. [d] Ref. [53]. [e] Ref. [54]. [f ] Ref. [55]. [g] Ref. [56].
improve the LDA and GGA models, the orbital-dependent

1124 Journal of Computational Chemistry 2012, 33, 1123–1130 WWW.CHEMISTRYVIEWS.COM

 WWW.C-CHEM.ORG FULL PAPER

Only recently hybrid functionals were implemented in PW In this work, we demonstrate that use of hybrid PBE0 XC
Vienna ab initio simulation package code for solids.[32] The Hamiltonian allows reproducing the structural, electronic,
hybrid XC functionals provide electronic structure of solids in dielectric, and vibrational properties of paraelectric cubic and
much better agreement with the experiment, in particular for three ferroelectric (tetragonal, rhombohedral, and orthorhom-
the band gap, usually underestimated in DFT calculations and bic) BTO phases in good agreement with the existing experi-
overestimated in HF method. However, the question arises, mental data. To the best of our knowledge, the phonon fre-
whether there is any need for hybrid functionals in the quencies obtained within the same theoretical approach in
ground-state calculations, in particular for phonon frequencies. cubic and all ferroelectric BTO phases are published at the first
After all, hybrid functionals are at least one order of magni- time.
tude more expensive than semilocal GGA functionals. As it is The outline of the article is as follows. In the second section,
stated in Refs. [17] and [32], apart from the band gaps (which we describe the computational details of our LCAO calcula-
are not ground state properties) there are a few other proper- tions. In the third section, we report the LCAO results obtained
ties that are predicted clearly better by the hybrid functionals. for the structural, electronic, and dielectric properties of cubic
It was found in PW DFT calculations[32] that for Sr(Ba)TiO3, the and three ferroelectric BTO phases and compare our results
ground state structure of the distorted antiferroelectric phase with the existing experimental data and with those from PW
is in better agreement with experiment and phonon modes DFT calculations. In the fourth section, the calculated phonon
are much more accurate for hybrid XC functionals. frequencies at C-point of BZ are presented for the three ferro-
A systematic comparison of DFT and hybrid XC functionals electric BTO phases and compared with those found in PW
for prototypical ferroelectric oxides BaTiO3 and PbTiO3 in cubic DFT calculations. Finally, our conclusions are given.
phase was made by Bilc at al.[33] It was demonstrated that the
new hybrid B1-WC functional allows the obtaining very good
Computational Details
structural, electronic, and ferroelectric properties as compared
to experimental data for cubic BaTiO3 and PbTiO3. The hybrid Perovskite properties and phonon frequencies in the LCAO ba-
B1-WC functional consists of PBE correlation part and WC sis were calculated with CRYSTAL09[37,38] computer code. The
exchange part,[24] being modified PBE exchange. The percent- Gaussian basis set and small core pseudopotentials (PPs) have
age of exact exchange is defined by parameter A, which is var- been used for metallic atoms. In this case, 5s, 5p, and 6s orbi-
ied in calculations with B1-WC hybrid functional. The best tals of barium, 3s, 3p, 3d, and 4s orbitals of titanium were
results were found for A ¼ 0.16. It means that use of hybrid treated as valence states. In our previous work,[17] the PP of Ba
B1-WC functional introduces additional parameter in the first- atom and the corresponding contracted Gaussian basis set
principles calculation. Meanwhile, hybrid PBE0 functional[29] is was taken from the PPs library of the Stuttgart/Cologne
parameter free as the exchange percentage (A ¼ 0.25) was group,[39] while the uncontracted Gaussian basis set and small-
found from the perturbation theory, that is, deduced on the core CRENBL ECP[40] ware used for Ti atom. In this work,
basis of pure theoretical grounds. This functional proved to because of increasing amount of computations for large super-
reproduce the properties of many strongly correlated systems cells, we admit Hay–Wadt small-core PP[41,42] for both Ba and
properly.[34] The verdict on which hybrid PBE0 functional is Ti atoms. The corresponding basis sets as well as the all-elec-
preferable for ABO3 perovskites is still open and requires fur- tron basis for O atom were taken from the work of Piskunov
ther theoretical investigations. et al.,[28] where they were optimized for SrTiO3 and BaTiO3
Till now, the hybrid functionals have been applied mainly to crystals.
cubic perovskites, see Refs. [17, 33, 35] and references therein. Our approach follows the traditional for LCAO calculations
In particular, in our previous work[17] the periodic LCAO PBE0 use of PPs generated for atoms from HF method, that is, not
approach was applied to calculate the ground state properties generated using the hybrid functionals used in our LCAO cal-
and phonon frequencies for the three crystals Ba(Ti,Zr,Hf)O3 culations. Conversely, DFT community working with PW basis
within the hybrid XC functional PBE0. It was shown that such has always favored an alternative approach that involves the
first-principles calculations with the basis set optimization and generation of the corresponding PPs for each XC functional. It
use of parameter-free hybrid XC PBE0 functional are able to pre- can be argued[28,33] that HF PPs constitute a reasonable
dict correctly the structural stability and phonon properties in choice. The PPs used describe the interaction of core and va-
cubic perovskites under consideration. In agreement with the lence electrons in good agreement with the results of all elec-
experimental data, the structure instability was found for the tron calculations as it was demonstrated[28,33] for BaTiO3 and
cubic phase in ferroelectric BaTiO3 crystal. For cubic phases of PbTiO3 crystals. The values of lattice constants and phonon
BaZrO3 and BaHfO3 crystals, imaginary phonon modes are not frequencies obtained from Ti all-electron calculations are very
found in agreement with the observed stability down to 2 K. close to those obtained using Ti HF PPs.
The performance of hybrid density functionals in the case of A very high accuracy is needed for the calculation of pho-
cubic and tetragonal BTO phases was investigated by Cora.[36] non frequencies as they are determined by the second deriva-
It was demonstrated that observables related to the curvature tives of the total energy over atomic displacements. By this
of the potential energy surface around the minimum (such as reason, the Monkhorst and Pack[43] scheme for 12  12  12
phonon frequencies, elastic constants, and bulk modulus) k-point mesh in the BZ was applied together with the toleran-
require the fraction A ¼ 0.2–0.4 of HF exchange. ces 8, 8, 8, 8, 16 for the one-electron, Coulomb and exchange

Journal of Computational Chemistry 2012, 33, 1123–1130 1125

 FULL PAPER WWW.C-CHEM.ORG

integrals calculations. Briefly speaking, these tolerances mean the results considerably. Calculated values are also compared
that during the direct lattice summations the one-electron with existing experimental data. As it is seen from Table 1, use
integrals and two-electron Coulomb integrals less than 10
8 of the hybrid PBE0 XC functional generally allows predicting
are estimated by the multipolar expansion and two-electron the considered cubic BTO properties with better accuracy than
exchange integrals less than 10
16 are ignored. (For details, other considered XC functionals. Especially, the superior agree-
see CRYSTAL09 User’s Manual.[37]) These tolerances are usually ment between PBE0 and experimental results can be seen for
tight enough for good DFT results. Use of tighter tolerances lattice constant a, band gap Egap, and high-frequency dielectric
might be necessary for dielectric constants calculations.[44] constant e1. The numerical values of Born charges in BaTiO3
Unfortunately due to the complicated structure of noncubic agree satisfactory with the experimental estimates[55] for all
BTO phases, our computer facilities were not enough to try presented XC functionals. Despite the presumable basis set de-
the effect of the tolerance tightening. A very large grid pendence of Mulliken charges, the distinction between our
(XXLGRID[37]) was also used for DFT integration. former[17] and present PBE0 charges is of the same order
The calculations of phonon frequencies were based on the (about 0.1 e) as that caused by use of the different XC.
following procedure. First, the equilibrium geometry was The phonon frequencies obtained for cubic phase in other
found. In the cubic phase, the lattice parameter, fully defining computations have been considered in our previous work,[17]
the structure, was optimized. In the ferroelectric phases, the where the advantage of PBE0 functional had been established.
lattice parameters and fractional displacements of atoms with The structural instability of cubic BTO via soft modes was
respect to the cubic location were optimized. Ionic and cell obtained by all considered XC functionals (see Table 1). Our
relaxations were performed with the convergence criterion for present data show that LCAO LDA calculation, while underesti-
the forces on atoms set to 0.005 eV/Å. The threshold on the mating the cell constant of cubic BTO, produces the phonon
energy change between optimization steps for the self-consist- frequencies approximately with the same deviation from ex-
ent cycles was 10
9 eV for both the structure optimization perimental values as PBE0 functional. In the case of PBE0 XC
and the phonon frequency calculations. Second, high-fre- functional, we also made calculations using the lower precision
quency dielectric tensors were estimated using the coupled on DFT integration. The results of this study (given in paren-
perturbed HF/Kohn–Sham method.[44–46] This is a perturbative thesis) show that decrease of integration accuracy level may
treatment that allows solving the Schro €dinger equation in the lead to errors in phonon frequencies of order 10 cm
1.
presence of the electric field perturbation. Furthermore, this is We know only two theoretical works,[13,20] where three BTO
a coupled method (in contrast with sum over states method) ferroelectric phases have been treated by the same DFT
that, by means of a self-consistent cycle, allows the wave func- method. However, in these two works mainly the structural
tion to relax under the effect of the bielectronic potential cor- properties of the ferroelectric phases have been considered. In
rected for the presence of the electric field. Convergence is Table 2, we compare our results using the various XC function-
then evaluated on a dielectric tensor element and the self-con- als, other theoretical values,[13,20] and experimental data avail-
sistent procedure stops when no relevant changes happen able for the considered crystals.
between two subsequent cycles. Third, the phonon frequen- Unexpectedly, LDA in average gives good agreement with ex-
cies were obtained by the frozen phonon method[47–50] within perimental structural data for BaTiO3 phases. Moreover, despite
the harmonic approximation at the optimized equilibrium lat- underestimating the length of the lattice vectors, LDA produces
tice constants. The zone-center phonon frequencies (the the atomic displacements of the same quality as PBE0. On the
eigenvalues of the dynamical matrix) are determined from nu- other side, as expected, both LDA and PBE XC functionals sub-
merical second-order derivatives of the ground state energy. stantially underestimate the band gap and as a rule, overesti-
To this purpose, the total energy is calculated for fixed distor- mate components of high-frequency dielectric tensor. On the
tions of the crystal. The supercell approach is used to obtain contrary, PBE0 slightly overestimates the band gap and under-
the phonon frequencies at the nonzero wave vectors. The con- estimates high-frequency dielectric constant. The main failure of
vergence of the phonon frequencies and dispersion curves in LDA method is incapability to reproduce the correct value of
terms of the supercell size was studied in ab initio frozen pho- the cohesive energy of metal oxides (it is well known that LDA
non calculations[50] on ionic crystal. The longitudinal–trans- usually overestimates the cohesive energy of solids), whereas
verse optical (LO-TO) splitting was found using the theoretical PBE0 method gives good results for this quantity. Moreover,
values of high-frequency dielectric constants given in Tables 1 LDA seems to be not able to predict the correct sequence of
and 2. total electronic energies for considered phases.
The definite conclusions about calculation quality of bulk
moduli cannot be done at present time because of lack of the
Properties of Cubic and Ferroelectric Phases
sufficiently precise experimental data.
We performed the first-principle calculations of four known
BTO phases with perovskite structure using LDA, PBE, and
Phonon Symmetry and Frequencies in BaTiO3
PBE0 XC functionals within the same LCAO approach. In Table
1, we compare BTO cubic phase properties obtained using the The crystalline phonon mode symmetry is defined by the cor-
different XC functionals. Our previous results,[17] which are also responding space group irreducible representations (IR)
given in Table 1, show that basis set change does not affect induced from the IR occupied by the atom site symmetry

1126 Journal of Computational Chemistry 2012, 33, 1123–1130 WWW.CHEMISTRYVIEWS.COM

 WWW.C-CHEM.ORG FULL PAPER

Table 2. Calculated bulk properties of tetragonal, orthorhombic, and rhombohedral phases of BaTiO3.

This work Other theoretical results

Phase; Property LDA PBE PBE0 LDA[b] PBE[c] Experiment [a]

Tetragonal, SG 99
Cell
a (Å) 3.952 4.007 3.968 3.994 3.976 3.997
c (Å) 3.987 4.186 4.137 4.036 4.172 4.031
zTi 0.0115 0.0193 0.0203 0.0143 0.0188 0.0203
zbO
0.0168
0.0436
0.0431
0.0307
0.0473
0.0258
zcO
0.0107
0.0246
0.0226
0.0186
0.0266
0.0123
Bulk modulus, B (GPa) 160 100 112 82.9 141[d], 134[e]
Cohesive energy, Eat (eV) 38.45 32.75 30.81 40.25
Band gap, Egap (eV) 2.0 1.9 4.1 2.3 1.6 3.4[f ]
e1
xx 6.20 5.86 4.77 5.19
zz 5.81 4.89 4.27 5.05[g]
Total energy, DE0 (eV)
0.004
0.046
0.041
0.045
Orthorhombic, SG 38
Cell
a (Å) 3.951 3.995 3.958 3.984 3.991 3.983
b (Å) 5.614 5.792 5.728 5.674 5.783 5.674
c (Å) 5.623 5.837 5.770 5.692 5.822 5.692
zTi 0.0093 0.0172 0.0175 0.0127 0.0159 0.0170
zaO
0.0085
0.0203
0.0186
0.0144
0.0206
0.0110
yeO
0.0078
0.0154
0.0139
0.0230
0.0159
0.0096
zeO
0.0130
0.0314
0.0307
0.0162
0.0309
0.0218
Bulk modulus, B (GPa) 154 96 109 87.4
Cohesive energy, Eat (eV) 38.46 32.76 30.82 40.26
Band gap, Egap (eV) 2.1 2.5 4.7
e1
xx 6.17 5.69 4.69
yy 6.02 5.34 4.59
zz 5.82 4.81 4.16
Total energy, DE0 (eV)
0.005
0.058
0.056
0.060
Rhombohedral, SG 160
Cell
a (Å) 3.966 4.073 4.029 4.001 4.073 4.004
a (deg) 89.958 89.710 89.727 89.87 89.74 89.8
zTi
0.0080
0.0150
0.0151
0.011
0.0150
0.0128
xO 0.0069 0.0143 0.0129 0.0133 0.0141 0.0109
zO 0.0109 0.0249 0.0242 0.0192 0.0245 0.0193
Bulk modulus, B (GPa) 153 101 114 94.6
Cohesive energy, Eat (eV) 38.46 32.77 30.83 38.27 40.26
Band gap, Egap (eV) 2.2 2.7 4.9 2.3
e1
xx 6.05 5.42 4.62 6.19
zz 5.82 4.80 4.13 5.88[h]
Total energy, DE0 (eV)
0.005
0.061
0.061
0.04
0.060

Fractional displacements xA, yA, zA of atom A are given in respect to the ideal cubic location (see Table 3 for parameterization of Wyckoff positions);
total energy, DE0, is calculated relative to the cubic phase; e1—high-frequency dielectric constant.
[a] Experimental lattice parameters and atomic positions.[57] [b] Ref. [13]. [c] Ref. [20]. [d] Ref. [58]. [e] Calculated from elastic constants obtained in Ref.
[59]. [f ] Ref. [53]. [g] Taken from Ref. [18]. [h] Ref. [60].

group corresponding to the atomic displacement (x, y, z).[5,27] responding point symmetry groups of a crystal. This corre-
Table 3 gives the symmetry of phonons in four BTO structures: spondence is given in the footnote of Table 3 and used in
paraelectric cubic and three ferroelectric ones (tetragonal, Tables 4–6, giving the calculated phonon frequencies.
orthorhombic, and rhombohedral). The phase transitions in In the cubic perovskite structure (Fig. 1a), Ba and Ti atoms
BTO are connected with the soft mode in C point of BZ, that occupy Wyckoff positions a and b with the site symmetry group
is, take place without increasing of the five atom primitive Oh and three oxygen atoms occupy Wyckoff position c with the
cell.[5] Therefore, the total number of optical TO modes is 12. site symmetry D4h. It is seen from Table 3 that IR active C4–(T1u)
In the footnote to Table 3, we give the symmetry of optical IR, modes are induced by all the five atoms displacements but the
Raman, silent, and acoustic modes, using the conventional silent C5–(T2u) mode only by three oxygen displacements.
space group IR labels.[61] In C point of BZ, there is one-to-one In the ferroelectric phases, the site symmetry groups of Ti
correspondence between these labels and those used for cor- and Ba atoms are C4v, C2v, and C3v, for tetragonal (Fig. 1b),

Journal of Computational Chemistry 2012, 33, 1123–1130 1127

 FULL PAPER WWW.C-CHEM.ORG

Table 3. Wyckoff positions of atoms (in reduced coordinates of primitive cells), site symmetry, phonon symmetry[a], and IR indexes[b] for cubic and
ferroelectric phases of BaTiO3.

Cubic phase Tetragonal phase Orthorhombic phase Rhombohedral phase

(T > 393K) O1h (T ¼ 393K278K) (T ¼ 278K183K) C14
2v (T < 183K) C53v
(Pm3m), SG 221 C14v (P4mm), SG 99 (Amm2), SG 38 (R3m), SG 160
Ba: 1a (0, 0, 0) Ba: 1a (0, 0, 0) Ba: 1a (0, 0, 0) Ba: 1a (0, 0, 0)
Oh C4v C2v C3v
T1u (x, y, z) ! C4
A1 (z) ! C1 A1 (z) ! C1 A1 (z) ! C1
E (x, y) ! C5 B1 (x), B2 (y) ! C3, C4 E (x, y) ! C3

Ti: 1bð12 ; 12 ; 12Þ Ti: 1b ð12 ; 12 ; 12 þ zTi Þ Ti: 1b ð12 ; 12 þ zTi ; 12 þ zTi Þ Ti: 1a ð12 þ xTi ; 12 þ xTi ; 12 þ xTi Þ
Oh C4v C2v C3v
T1u (x, y, z) ! C4
A1 (z) ! C1 A1 (z) ! C1 A1 (z) ! C1
E (x, y) ! C5 B1 (x), B2 (y) ! C3, C4 E (x, y) ! C3

O: 3c ð12 ; 12 ; 0Þ O(1): 1b ð12 ; 12 ; zbO Þ O(1): 1a ð0; 12 þ zaO ; 12 þ zaO Þ O: 3b ð12 þ xO ; 12 þ xO ; zO Þ

D4h C4v C2v Cs
A2u (z) ! C4
A1 (z) ! C1 A1 (z) ! C1 A00 (z) ! C2, C3
Eu (x, y) ! C5
E (x, y) ! C5 B1 (x), B2 (y) ! C3, C4 A0 (x) ! C1, C3
A0 (y) ! C1, C3
O(2): 2c ð12 ; 0; 12 þ zcO Þ O(2): 2e ð12 ; 12 þ yeO ; zeO Þ –

C2v Cs
A1 (z) ! C1, C2 A00 (z) ! C2, C3
B1 (x), ! C5 A0 (x) ! C1, C4
B2 (y) ! C5 A0 (y) ! C1, C4

For space groups (SG) there are given Schonflis symbols, international symbols and numbers.
[a] Phonon spectra constituents:
SG 221: 4C4
þ C5
; acoustic C4
; IR 3C4
, silent C5
.
SG 99: 4C1 þ C2 þ 5C5; acoustic C1 þ C5, Raman 3C1 þ 4C5, IR 3C1 þ 4C5; silent C2.
SG 38: 5C1 þ C2 þ 4C3 þ 5C4; acoustic C1 þ C3 þ C4, Raman 4C1 þ C2 þ 3C3 þ 4C4, IR 4C1 þ 3C3 þ 4C4.
SG 160: 4C1 þ C2 þ 5C3; acoustic C1 þ C3, Raman 3C1 þ 4C3, IR 3C1 þ 4C3, silent C2.
[b] Phonon mode symbols accepted in literature: SG 221—C4
: T1u, C5
: T2u; SG 99—C1:A1, C2:B1, C5:E; SG 38—C1:A1, C2:A2, C3:B1, C4:B2; SG 160—
C1:A1, C2:A2, C3:E.

orthorhombic (Fig. 1c), and rhombohedral (Fig. 1d) structures, The information collected in Table 3 is useful for the analysis
respectively. The three oxygen atoms Wyckoff position c in of the role of different atoms displacements in vibrational
cubic phase is split into two nonequivalent sets in tetragonal modes of different symmetry. For example, in the orthorhom-
and orthorhombic phases with the site symmetry (C4v, C2v) and bic phase silent modes are absent but in the rest three phases
(C2v, Cs), respectively. At last in the rhombohedral phase, all these modes are connected with the displacements of only ox-
three oxygens are equivalent with the site symmetry group Cs. ygen atoms. Using the induced IR of space groups[66] allows
one to extend Table 3 to the other symmetry
points of BZ for detailed interpretation of the
21 vibrational spectra, found in the neutron-scat-
Table 4. Calculated and experimental C-point phonon frequencies (cm ) for tetragonal
phase of BaTiO3 (sequence number is given before symbol for multiple modes). tering experiments. To the best of our knowl-
edge, the comparative phonon symmetry
This work Experiment Other theoretical results
analysis in BTO four phases is given here for
Mode LDA PBE PBE0 Ref. [62] Ref. [63] Ref. [18] Ref. [19] Ref. [21] the first time.
TO In Tables 4–6, we give the symmetry and
1E 79i 138i 139i soft 161i 122i 148i calculated phonon frequencies for all the
1A1 190 159 173 170 280 161 152 171
three ferroelectric BTO phases. We compare
2E 192 171 185 180 190 167 163 177
B1 (silent) 301 282 301 305 303 287 275 291 our results both with other theoretical values
3E 302 298 320 305 280 284 290 288 and with the existing experimental data for
2A1 237 362 384 270 302 363 323 the tetragonal and rhombohedral phases. For
4E 487 462 484 486 516 457 451 475
3A1 507 563 591 520 507 562 544 the orthorhombic phase, the experimental
LO data are sparse, the other theoretical data are
1E 188 165 178 180 140 162 – – absent. To discuss the influence of the XC
1A1 195 191 205 185 180 – –
potential on the calculated phonon frequen-
2E 302 297 318 305 303 284 – –
3E 462 446 474 463 640 444 – – cies, we give the results for LDA, GGA (PBE),
2A1 465 460 489 475 452 – – and hybrid PBE0 XC potentials. In agreement
4E 701 660 707 715 720 641 – – with the experimental data (the low tempera-
3A1 721 738 792 720 705 – –
ture phase is rhombohedral), we obtained the

1128 Journal of Computational Chemistry 2012, 33, 1123–1130 WWW.CHEMISTRYVIEWS.COM

 WWW.C-CHEM.ORG FULL PAPER

with the experimental data. The accordance between LDA fre-

Table 5. Calculated and experimental C-point phonon frequencies
(cm21) for orthorhombic phase of BaTiO3 (sequence number is given quencies and the experimental values is also sufficiently good,
before symbol for multiple modes). apparently due to the use of observed values for the lattice
parameters and/or cell volume. In average, our results are as
This work
Experiment good as LDA results despite nonuse of any experimental
Mode LDA PBE PBE0 Ref. [64] information.
TO
1B1 61i 78i 66i Conclusions
1B2 165 162 177
1A1 191 165 179
2B1 192 169 184 193 1. We report here the structure, electronic, and energy
2B2 191 276 297 270 properties of BaTiO3 stable phases obtained using different XC
3B2 298 300 308 functionals of DFT within LCAO approximation and compared
A2 301 294 317 320
2A1 211 299 319 them with available experimental data and previous theoreti-
3A1 305 333 356 cal calculations. We clearly show that such calculations with
3B1 486 458 482 490 the basis set optimization and use of parameter-free hybrid XC
4B2 490 483 500
PBE0 functional are able to predict correctly the structural sta-
4A1 506 550 578 532
LO bility, electronic properties, and phonon frequencies in cubic
1B1 188 165 178 and ferroelectric perovskites under consideration. The LDA
1B2 191 180 193 functional, and to slightly less extent, GGA PBE functional can
1A1 191 269 284
also provide the sufficiently precise information on atomic dis-
2B2 298 277 299
2A1 214 299 319 320 placements and vibrational frequencies in indicated crystals.
3A1 305 371 375 2. The main failure of LDA method is incapability to repro-
3B2 459 438 465 duce the correct value of the cohesive energy and correct
2B1 461 440 469
4A1 506 579 610
sequence of total electronic energies for considered phases of
3B1 701 666 711 BaTiO3.
4B2 716 717 765 720 3. On the base of the site-symmetry consideration, we give
the comparative phonon symmetry analysis in BaTiO3 four
phases, which can be useful for estimating the atomic displace-
ments contributions to the vibrational modes of different
structural instability via soft modes (imaginary frequencies) for symmetry.
BTO tetragonal and orthorhombic phases and all the three XC 4. We anticipate that the thermodynamic modeling based
potentials, see Tables 4 and 5. No imaginary frequencies were on the use of first-principles methods (suggested for our next
obtained for the rhombohedral phase (see Table 6) being sta- publication) can provide the valuable information on the low-
ble at low temperatures. Calculated frequencies agree well temperature behavior that could not be obtained easily by
experimental techniques.
5. The approach used seems to be efficient for the theoreti-
cal predictions of phase transitions in the crystalline systems
Table 6. Calculated and experimental C-point phonon frequencies being experimentally poorly studied.
(cm21) for rhombohedral phase of BaTiO3 (sequence number is given
before symbol for multiple modes).

This work Other theoretical results Acknowledgment

Experiment
Mode LDA PBE PBE0 Ref. [65] Ref. [18] Ref. [22] Ref. [23] All calculations have been made with the assistance of the Saint-
TO Petersburg State University Computer Center.
1E 145 165 180 163 164 162
1A1 191 167 181 173 167 169 167 Keywords: crystal structure  density functional theory  perovskite
2E 191 264 264 210 206 240
polymorphs  vibrational frequencies  phonon symmetry
A2 (silent) 297 279 301 277 278 274
3E 306 299 319 293 293 296
2A1 200 309 332 242 259 255 295 How to cite this article: R. A. Evarestov, A. V. Bandura, J. Comput.
4E 489 474 493 470 472 468
3A1 508 544 571 522 512 511 527 Chem. 2012, 33, 1123–1130. DOI: 10.1002/jcc.22942
LO
1E 190 176 190 174 175 176
[1] J. F. Scott, Ferroelectric Memories; Springer: Berlin, 2000.
1A1 193 182 197 187 178 179 183
[2] R. Waser, Nanoelectronics and Information Technology: Advanced Elec-
2E 306 299 319 310 293 293 296
tronic Materials and Novel Devices; Wiley-VCH: Weinheim, 2003.
3E 460 441 468 441 443 440
[3] S. S. Nonnenmann, J. E. Spanier, J. Mater. Sci. 2009, 44, 5205.
2A1 471 467 498 485 461 460 462
[4] P. Goudochnikov, A. J. Bell, J. Phys. Condens. Matter 2007, 19, 176201.
3A1 707 689 739 714 676 677 679
[5] H. T. Stokes, E. H. Kisi, D. M. Hatch, C. J. Howard, Acta Crystallogr. B
4E 713 705 750 687 687 688
2002, 58, 934.

Journal of Computational Chemistry 2012, 33, 1123–1130 1129

 FULL PAPER WWW.C-CHEM.ORG

[6] R. E. Cohen, Nature 1992, 358, 136. [39] Energy-Consistent Pseudopotentials of the Stuttgart/Cologne Group;
[7] K. Rabe, P. Ghosez, Top. Appl. Phys. 2007, 105, 117. Available at: http://www.theochem.uni-stuttgart.de/pseudopotentials/
[8] Q. Zhang, T. Cagin, W. A. Goddard,III, Proc. Natl. Acad. Sci. 2006, 103, clickpse.html; Last change 30.12.2006.
14695. [40] M. M. Hurley, F. Pacios, P. A. Christiansen, R. B. Ross, W. C. Ermler,
[9] J. D. Freire, R. S. Katiyar, Phys. Rev. B 1988, 37, 2074. J. Chem. Phys. 1986, 84, 6840.
[10] R. E. Cohen, H. Krakauer, Phys. Rev. B 1990, 42, 6416. [41] P. J. Hay, W. R. Wadt, J. Chem. Phys. 1984, 82, 270.
[11] K. M. Rabe, U. V. Waghmare, Ferroelectrics 1992, 136, 147. [42] P. J. Hay, W. R. Wadt, J. Chem. Phys. 1984, 82, 299.
[12] Ph. Ghosez, X. Gonze, J. -P. Michenaud, Ferroelectrics 1997, 194, 39. [43] H. J. Monkhorst, J. D. Pack, Phys Rev B 1976, 13, 5188.
[13] Ph. Ghosez, X. Gonze, J. -P. Michenaud, Ferroelectrics 1999, 220, 1. [44] M. Ferrero, M. R
erat, B. Kirtman, R. Dovesi, J. Chem. Phys. 2008, 129,
244110.
[14] W. Zhong, D. Vanderbilt, K. M. Rabe, Phys. Rev. B 1995, 52, 6301.
[45] M. Ferrero, M. R
erat, R. Orlando, R. Dovesi, J. Chem. Phys. 2008, 128,
niguez, D. Vanderbilt, Phys. Rev. Lett. 2002, 89, 115503.
[15] J. Í~
014100.
[16] T. Nishimatsu, M. Iwamoto, Y. Kawazoe, U. V. Waghmare, Phys. Rev. B
[46] M. Ferrero, M. R
erat, R. Orlando, R. Dovesi, J. Comput. Chem. 2008, 29,
2010, 82, 134106.
1450.
[17] R. A. Evarestov, Phys. Rev. B 2011, 83, 14105.
[47] K. Kunc, R. Martin, Phys. Rev. B 1981, 24, 2311.
[18] P. Hermet, M. Veithen, Ph. Ghosez, J. Phys. Condens. Matter 2009, 21, [48] K. Kunc, R. Martin, Phys. Rev. Lett. 1982, 48, 406.
215901. [49] K. Parlinski, Z.-Q. Li, Y. Kawazoe, Phys. Rev. Lett. 1997, 78, 4063.
[19] S. Sanna, C. Thierfelder, S. Wippermann, T. P. Sinha, W. G. Schmidt, [50] R. A. Evarestov, M. V. Losev, J. Comput. Chem. 2009, 30, 2645.
Phys. Rev. B 2011, 83, 054112. [51] J. W. Edwards, R. Speiser, H. L. Johnston, J. Am. Chem. Soc. 1951, 73, 2934.
[20] M. Uludogan, D. P. Guarin, Z. E. Gomez, T. Cagin, W. A. Goddard,III, [52] K. H. Hellwege, A. M. Hellwege, Eds. Ferroelectrics and Related Sub-
Comput. Mod. Eng. Sci. 2008, 24, 215. stances, Vol. 16; Landolt-B€ ornstein, New Series, Group III; Springer Verlag:
[21] X. Y. Deng, H. T. Zhang, T. Li, D. J. Li, J. B. Li, X. H. Wang, L. T. Li, Chin. Berlin, 1981.
Sci. Bull. 2010, 55, 2182. [53] S. H. Wemple, Phys. Rev. B 1970, 2, 2679.
[22] X. Wu, D. Vanderbilt, D. R. Hamann, Phys. Rev. B 2005, 72, 035105. [54] G. Burns, F. H. Dacol, Solid State Commun. 1982, 42, 9.
[23] N. Choudhury, E. J. Walter, A. I. Kolesnikov, C.-K. Loong, Phys. Rev. B [55] J. D. Axe, Phys. Rev. 1967, 157, 429.
2008, 77, 134111. [56] T. Nakamura, Ferroelectrics 1992, 137, 65.
[24] Z. Wu, R. E. Cohen, Phys. Rev. B 2006, 73, 235116. [57] G. H. Kwei, A. C. Lawson, S. J. L. Billinge, J. Phys. Chem. 1993, 97, 2368.
[25] K. E. Riley, K. M. Metz, J. Phys. Chem. A 2007, 111, 6044. [58] A. Schaefer, H. Schmitt, A. Dorr, Ferroelectrics 1986, 69, 253.
[26] F. Cora, M. Alfredson, G. Mallia, D. Middlemiss, W. C. Makrodt, R. [59] M. Zgonik, P. Bernasconi, M. Duelli, R. Schlesser, P. Günter, M. H. Gar-
Dovesi, R. Orlando, Struct. Bonding 2004, 113, 171. rett, D. Rytz, Y. Zhu, X. Wu, Phys. Rev. B 1994, 50, 5941.
[27] R. A. Evarestov, Quantum Chemistry of Solids; Springer: Heidelberg, [60] B. Wang, C. Sun, Appl. Opt. 2001, 40, 672.
2007. [61] A. Cracknell, B. Davies, S. Miller, V. Love, Kronecker Product Tables, I
[28] S. Piskunov, E. Heifets, R. I. Eglitis, G. Borstel, Comput. Mater. Sci. 2004, General Introduction and Tables of Irreducible Representations of
29, 165. Space Groups; IFI/Plenum: New-York, Washington, London, 1979.
[29] J. P. Perdew, K. Burke, M. Ernzerhof, J. Chem. Phys. 1996, 105, 9982. [62] U. D. Venkateswaran, V. M. Naik, R. Naik, Phys. Rev. B 1998, 58, 14256.
[30] C. Adamo, V. Barone, J. Chem. Phys. 1999, 110, 6158. [63] Z. Lazarevic, N. Romcevic, M. Vijatovic, N. Paunovic, M. Romcevic, B.
[31] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865. Stojanovic, Z. Dohcevic-Mitrovic, Acta Phys. Polonica A 2009, 115, 808.
[32] R. Wahl, D. Vogtenhuber, G. Kresse, Phys. Rev. B 2008, 78, 104116. [64] K. Labidi, M. D. Fontana, B. Jannot, Solid State Commun. 1990, 76, 765.
[65] D. A. Tenne, X. X. Xi, Y. L. Li, L. Q. Chen, A. Soukiassian, M. H. Zhu, A.
[33] D. I. Bilc, R. Orlando, R. Shaltaf, G.-M. Rignanese, J. Í~
niguez, Ph. Ghosez,
R. James, J. Lettieri, D. G. Schlom, W. Tian, X. Q. Pan, Phys. Rev. B 2004,
Phys. Rev. B 2008, 77, 165107.
69, 174101.
[34] I. D. Prodan, G. E. Scuseria, R. L. Martin, Phys. Rev. B 2007, 76, 033101.
[66] R. A. Evarestov, V. P. Smirnov, Site Symmetry in Crystals: Theory and
[35] R. A. Evarestov, E. Blokhin, D. Gryaznov, E. A. Kotomin, J. Maier, Phys.
Applications; Springer Series in Solid State Sciences, Vol. 108, 2nd ed.;
Rev. B 2011, 83, 134108.
Springer: Berlin, 1997.
[36] F. Cora, Mol. Phys. 2005, 103, 2483.
[37] R. Dovesi, V. R. Saunders, C. Roetti, R. Orlando, C. M. Zicovich-Wilson,
F. Pascale, B. Civalleri, K. Doll, N. M. Harrison, I. J. Bush, Ph. D’Arco, M.
Received: 10 November 2011
Llunell, CRYSTAL09 User’s Manual; University of Torino: Torino, 2010. Revised: 18 December 2011
[38] R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V. R. Saunders, C. M. Zico- Accepted: 8 January 2012
vich-Wilson, Comput. Crystallogr. 2005, 220, 571. Published online on 27 February 2012

1130 Journal of Computational Chemistry 2012, 33, 1123–1130 WWW.CHEMISTRYVIEWS.COM

View publication stats