1: \documentclass[showpacs,aps,prl,twocolumn]{revtex4}
2: \usepackage{epsfig}
3: \usepackage{amsmath}
4:
5: \begin{document}
6:
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8:
9: \title{
10: {\em Ab initio} Study of the Phase Diagram of Epitaxial BaTiO$_3$
11: }
12:
13: \author{Oswaldo Di\'eguez}
14: \author{Silvia Tinte}
15: \author{A.~Antons}
16: \author{Claudia Bungaro}
17: \author{J.~B.~Neaton}
18: \altaffiliation{Present address: The Molecular Foundry, Materials Sciences
19: Division, Lawrence Berkeley National Laboratory,
20: Berkeley, CA 94720, USA.}
21: \author{Karin M. Rabe}
22: \author{David Vanderbilt}
23:
24: \affiliation{Department of Physics and Astronomy, Rutgers University,
25: Piscataway, New Jersey 08854-8019, USA}
26:
27: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
28:
29: \begin{abstract}
30: Using a combination of first-principles and effective-Hamiltonian
31: approaches, we map out the structure of BaTiO$_3$ under
32: epitaxial constraints applicable to growth on perovskite
33: substrates.
34: We obtain a phase diagram in temperature and misfit strain that is
35: qualitatively different from that reported by Pertsev {\em et al.}
36: [Phys.~Rev.~Lett. {\bf 80}, 1988 (1998)], who based their results on an
37: empirical thermodynamic potential with parameters fitted at temperatures in the
38: vicinity of the bulk phase transitions.
39: In particular, we find a region of `{\em r} phase' at low temperature where
40: Pertsev {\em et al.} have reported an `{\em ac} phase'.
41: We expect our results to be relevant to thin epitaxial films of BaTiO$_3$ at
42: low temperatures and experimentally-achievable strains.
43: \end{abstract}
44:
45: \date{February 3, 2004}
46:
47: \pacs{
48: % In 77. DIELECTRICS, PIEZOELECTRICS, AND FERROELECTRICS AND THEIR PROPERTIES:
49: 77.55.+f, % Dielectric thin films
50: 77.80.Bh, % Phase transitions and Curie point
51: 77.84.Dy, % Niobates, titanates, tantalates, PZT ceramics, etc.
52: % In 81. MATERIALS SCIENCE:
53: 81.05.Zx % New materials: theory, design, and fabrication
54: }
55:
56: \maketitle
57:
58: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
59:
60: The perovskite oxide barium titanate (BaTiO$_3$) is a prototypical
61: ferroelectric, an insulating solid whose macroscopic polarization can be
62: reoriented by the application of an electric field \cite{Lines.book.1977}.
63: In the perovskite ferroelectrics, it is well known both experimentally and
64: theoretically that the polarization is also strongly coupled to strain
65: \cite{Cohen.Nature.1992}, and thus that properties such as the ferroelectric
66: transition temperature and polarization magnitude are quite sensitive to
67: external stress.
68:
69: Experimentally, the properties of ferroelectrics in thin film form generally
70: differ significantly from those in the bulk \cite{Ahn.S.2004}.
71: While many factors are expected to contribute to these differences, it has been
72: shown that the properties of perovskite thin films are strongly influenced by
73: the magnitude of the epitaxial strain resulting from lattice-matching the film
74: to the substrate.
75: For example, Yoneda {\em et al.}\ \cite{Yoneda.APL.1998} used molecular-beam
76: epitaxy (MBE) to grow BaTiO$_3$ (lattice constant of 4.00~\AA) on
77: (001)-oriented SrTiO$_3$ (lattice constant of 3.91~\AA); they found that the
78: ferroelectric transition temperature exceeds 600~$^\circ$C, to be compared
79: to the bulk Curie temperature of $T_{\rm C} =$ 130~$^\circ$C.
80: Other studies have shown that the amount of strain in BaTiO$_3$/SrTiO$_3$
81: superlattices on SrTiO$_3$ substrates strongly influences properties including
82: the observed polarization, phase transition temperature, and dielectric
83: constant \cite{Chang.JAP.2000,Li.APL.2001,Shimuta.JAP.2002,Neaton.APL.2003}.
84:
85: In a seminal paper, Pertsev, Zembilgotov and Tagantsev \cite{Pertsev.PRL.1998}
86: introduced the concept of mapping the equilibrium structure of a ferroelectric
87: perovskite material versus temperature and misfit strain, thus producing a
88: ``Pertsev phase diagram'' (or Pertsev diagram) of the
89: observable epitaxial phases.
90: The effect of epitaxial strain is isolated from other aspects of thin-film
91: geometry by computing the structure of the {\em bulk} material with homogeneous
92: strain tensor constrained to match a given substrate with square
93: surface symmetry \cite{note:cubic}.
94: In addition, short-circuit electrical boundary conditions are imposed,
95: equivalent to ideal electrodes above and beneath the film
96: \cite{Pertsev.PRL.1998}.
97: Given the recognized importance of strain in determining the properties of
98: thin-film ferroelectrics, Pertsev diagrams have proven to be of enormous
99: interest to experimentalists seeking to interpret the results of experiments
100: on epitaxial thin films and heterostructures.
101:
102: In \cite{Pertsev.PRL.1998}, the mapping was carried out with a phenomenological
103: Landau-Devonshire model taken from the literature.
104: This should give excellent results in the temperature/strain regime in which
105: the model parameters were fitted, but will generally be less accurate when
106: extrapolated to other regimes.
107: In Fig.~\ref{fig:diagramsPERTSEV}, we compare two Pertsev diagrams for
108: BaTiO$_3$ computed using two different sets of Landau-Devonshire parameters,
109: used by Pertsev and coworkers in \cite{Pertsev.PRL.1998} and
110: \cite{Koukhar.PRB.2001}, respectively.
111: While both give the same behavior near the bulk $T_{\rm C}$ and small misfit
112: strains, they predict completely different low-temperature phase behavior.
113:
114: With first-principles methods, it is possible not only to resolve such
115: discrepancies arising in phenomenogical theories, but also to generate a wealth
116: of microscopic information about the structure and properties of epitaxial
117: phases at various temperatures and substrate lattice constants.
118: In this Letter, using parameter-free total-energy methods
119: based on density functional theory (DFT),
120: we map out the equilibrium structure of BaTiO$_3$ as a function of epitaxial
121: constraints at zero temperature, and then extend the results to finite
122: temperature via an effective-Hamiltonian approach.
123: The Pertsev diagram obtained in this way has a similar global topology as that
124: of Fig.~\ref{fig:diagramsPERTSEV}(b) (but not to the one in
125: \cite{Pertsev.PRL.1998}).
126: This allows us to predict the impact of misfit strain on the magnitude
127: and orientation of the polarization and Curie temperature of BaTiO$_3$.
128: Our results should thus be of considerable importance for understanding
129: experimental growth of high-quality, coherent epitaxial thin films of BaTiO$_3$
130: on perovskite substrates, as well as more generally illustrating the utility
131: of first-principles Pertsev diagrams.
132:
133: \begin{figure}
134: \centerline{\epsfig{file=figure1.ps,width=3.00in}}
135: \caption{Phase diagrams of epitaxial BaTiO$_3$ as predicted by the theory
136: of Pertsev {\em et al.}\ \protect\cite{Pertsev.PRL.1998}.
137: (a) Using the parameters quoted in \protect\cite{Pertsev.PRL.1998}.
138: (b) Using the parameters quoted in \protect\cite{Koukhar.PRB.2001}.
139: The second- and first-order phase transitions are represented by
140: thin and thick lines, respectively.}
141: \label{fig:diagramsPERTSEV}
142: \end{figure}
143:
144: The first-principles DFT calculations are carried out in the
145: Kohn-Sham framework \cite{Hohenberg.PR.1964+Kohn.PR.1965} using the
146: VASP software package \cite{Kresse.PRB.1993+Kresse.PRB.1996}. The
147: electron-ion interaction is described by the projector augmented
148: wave method \cite{Blochl.PRB.1994+Kresse.PRB.1999}; semicore
149: electrons are included in the case of Ba ($ 5s^2 5p^6 6s^2 $) and
150: Ti ($ 3s^2 3p^6 4s^2 3d^2 $). The calculations
151: employ the Ceperley-Alder \cite{Ceperley.PRB.1980} form of the
152: local-density approximation (LDA)
153: exchange-correlation functional \cite{note:gga}, a 700~eV plane-wave
154: cutoff, and a $6 \times 6 \times 6$ Monkhorst-Pack sampling
155: of the Brillouin zone \cite{Monkhorst.PRB.1979}.
156:
157: We begin by systematically performing optimizations of the five-atom unit
158: cell of the cubic perovskite structure (space group $Pm\bar{3}m$) in the
159: six possible phases considered by Pertsev {\em et al.}\ in
160: \cite{Pertsev.PRL.1998}.
161: A description of these phases is given in Table \ref{table:structure}.
162: Starting from a structure in which the symmetry is established by displacing
163: the Ti and O atoms, we relax the atomic positions and the out-of-plane
164: cell vector until the value of the Hellmann-Feynman forces and $zz$, $yz$ and
165: $zx$ stress tensor components fall below some given thresholds
166: (0.001 eV/\AA~and 0.005 eV, respectively).
167:
168: In Fig.~\ref{fig:energiesDFT} we present the computed energy for each phase
169: as a function of the misfit strain $s = a/a_0 - 1$, where
170: $a_0$ is our DFT lattice constant for free cubic BaTiO$_3$
171: (3.955~\AA).
172: For large compressive strains, the lowest energy corresponds to the {\em c}
173: phase; for large tensile strains, the {\em aa} phase is favored.
174: At a misfit strain of $s_{\rm max}(c) = -6.4 \times 10^{-3}$ ($a = 3.930$ \AA),
175: there is a second-order transition
176: from the {\em c} phase to the {\em r} phase, with the polarization in the
177: {\em r} phase continuously rotating away from the {\em z} direction as the
178: misfit strain increases.
179: At misfit strain $s_{\rm min}(aa) = 6.5 \times 10^{-3}$ ($a = 3.981$ \AA), the
180: {\em r} phase polarization completes
181: its rotation into the {\em xy} plane, resulting in a continuous transition to the
182: {\em aa} phase.
183: The minimum energy {\em r} phase is at misfit strain of $2.2 \times 10^{-3}$
184: ($a = 3.964$ \AA);
185: lattice matching to the substrate would be optimal at this point.
186: At the misfit strain of the {\em c}$\rightarrow${\em r} transition, the
187: polarization could also begin a continuous rotation into the (010) plane,
188: corresponding to the {\em ac} phase.
189: However, it is clear from the figure that the energy of the {\em ac} phase is
190: always higher than that of the {\em r} phase, which makes sense given that
191: the {\em r} phase is an epitaxial disortion of the ground-state rhombohedral
192: phase of bulk BaTiO$_3$, while the {\em ac} phase is related
193: to the higher-energy bulk orthorhombic phase.
194: We conclude that the phase sequence at low temperatures is not
195: {\em c}$\rightarrow${\em ac}$\rightarrow${\em aa} as given in
196: \cite{Pertsev.PRL.1998}, but {\em c}$\rightarrow${\em r}$\rightarrow${\em aa}.
197:
198: \begin{table}
199: \caption{Summary of possible epitaxial BaTiO$_{3}$ phases.
200: In-plane cell vectors are fixed at ${\bf a}_1 = a \hat x$,
201: ${\bf a}_2 = a \hat y$. Columns list, respectively:
202: phase; space group; out-of-plane lattice vector; number of
203: free internal displacement coordinates; and form of the
204: polarization vector.}
205: \begin{tabular}{ccccc}
206: \hline
207: \hline
208: Phase & SG & ${\bf a}_3$ & $N_{\rm{p}}$
209: & Polarization \\
210: \hline
211: {\em p} & $P4mmm$ & $c \hat z$ & 0
212: & 0 \\
213: {\em c} & $P4mm$ & $c \hat z$ & 3
214: & $P_z\hat z$ \\
215: {\em aa} & $Amm2$ & $c \hat z$ & 4
216: & $P(\hat x+\hat y)$ \\
217: {\em a} & $Pmm2$ & $c \hat z$ & 4
218: & $P\hat x$ \\
219: {\em ac} & $Pm$ & $c_\alpha \hat x + c \hat z$ & 8
220: & $P \hat x + P_z \hat z$ \\
221: {\em r} & $Cm$ & $c_\alpha (\hat x + \hat y) + c \hat z$ & 7
222: & $P (\hat x+\hat y ) + P_z \hat z$ \\
223: \hline
224: \hline
225: \end{tabular}
226: \label{table:structure}
227: \end{table}
228:
229: Figure \ref{fig:displacementsDFT} shows the computed behavior of the atomic
230: displacements for the lowest-energy phase with increasing misfit strain.
231: For large compressive strains, the pattern of displacements corresponds to the
232: $c$ phase, and atoms relax only along the [001] direction.
233: As the in-plane strain increases, we observe a second-order phase
234: transition ({\em c}$\rightarrow${\em r}), and while the magnitude of the
235: atomic displacements continues to diminish along [001], the displacements in
236: the {\em xy} plane begin to grow.
237: With increasing tensile strain, the displacements along [001] vanish at the
238: {\em r}$\rightarrow${\em aa} transition, while the displacements
239: in the {\em xy} plane continue to grow
240: smoothly.
241: Similar results are found when we analyze the
242: {\em c}$\rightarrow${\em ac}$\rightarrow${\em a} sequence (not shown), where
243: what was said for the {\em xy} plane applies now to the [100] direction.
244: The clear change in character of the displacement pattern within the
245: {\em r} phase witnessed here illustrates the quantitative limitations
246: of using a single misfit-strain-independent local mode to model the
247: phase diagram.
248:
249: \begin{figure}
250: \centerline{\epsfig{file=figure2.ps,width=3.00in}}
251: \caption{Energies of the possible epitaxial BaTiO$_3$ phases for different
252: misfit strains, as obtained from the full {\em ab initio}
253: calculations. The vertical lines denote the phase transition
254: points given by the stability analysis.}
255: \label{fig:energiesDFT}
256: \end{figure}
257:
258: A stability analysis provides the precise
259: limits of phase stability shown in Figs.~\ref{fig:energiesDFT}
260: and \ref{fig:displacementsDFT}. At each value of misfit in the {\em c}
261: phase, for example, we carry out finite-difference calculations of
262: $x$ forces and $xz$ stress as the atomic $x$ coordinates and
263: $xz$ strain are varied. The zero crossing of the lowest eigenvalue
264: of the resulting $6 \times 6$ Hessian matrix identifies
265: the critical misfit. A similar analysis is used to consider $z$
266: displacements and shear strains in the {\em a} and {\em aa} phases.
267: By properly considering zone-center phonons, elastic
268: shear, and linear cross-coupling between them, this analysis allows
269: us to locate the second-order phase boundaries much more precisely
270: than is possible through direct comparison of total energies \cite{note:ifc}.
271:
272: Having established the first principles zero-temperature phase diagram,
273: we now extend our study of epitaxial BaTiO$_3$ to finite temperatures using
274: the effective Hamiltonian approach of Zhong, Vanderbilt, and
275: Rabe~\cite{Zhong.PRL.1994+Zhong.PRB.1995}.
276: In this method, the full Hamiltonian is mapped onto a
277: statistical mechanical model by a subspace projection, and parameterized
278: through {\em ab initio} calculations of small distortions of bulk BaTiO$_3$ in
279: the cubic perovskite structure.
280: The reduced subspace is composed of a set of relevant degrees of freedom
281: identified for ferroelectric perovskites as the unit cell distortions
282: corresponding to local polarization, expressed in the form of local modes.
283: This subspace is augmented by the inclusion of the homogeneous strain.
284:
285: It is straightforward to impose the constraint of fixed in-plane strain by
286: fixing three of the six tensor strain components during the Monte Carlo
287: (MC) simulations. For each value of in-plane strain, MC
288: thermal averages are obtained for the unconstrained components of the
289: homogeneous strain and the average polarization \cite{note:mcdetails},
290: and phase transitions are identified by monitoring the symmetry of these
291: quantities.
292: Following \cite{Zhong.PRL.1994+Zhong.PRB.1995}, all the simulations were
293: performed at the same negative external pressure of $P=-4.8$~GPa.
294: Misfit is defined relative to $a_0=3.998$~\AA, the lattice
295: constant at the bulk cubic-to-tetragonal transition as computed
296: with this approach \cite{Zhong.PRL.1994+Zhong.PRB.1995}.
297: The resulting phase diagram appears in Fig.~\ref{fig:diagramHEFF}, where all
298: phase lines represent second-order transitions.
299:
300: The Pertsev diagrams of Figs.~\ref{fig:diagramsPERTSEV}(a),
301: \ref{fig:diagramsPERTSEV}(b), and \ref{fig:diagramHEFF} share the
302: same topology above and just below $T_{\rm C}$: {\em p} at high
303: temperature, {\em c} at large compressive misfit, {\em aa} at
304: large tensile misfit, and
305: a 4-phase point connecting these phases with the {\em r} phase at
306: $T_{\rm C}$. At lower temperature, there is a drastic difference
307: between Figs.~\ref{fig:diagramsPERTSEV}(a) and
308: \ref{fig:diagramsPERTSEV}(b), with our theory supporting the
309: latter. While our theory underestimates the temperature of
310: the 4-phase crossing point in Fig.~\ref{fig:diagramHEFF} by about
311: 100~$^\circ$C, this is the price we pay for insisting on a
312: first-principles approach; indeed, this effective Hamiltonian
313: underestimates the temperature of the bulk cubic-to-tetragonal
314: transition by about the same amount.
315:
316: \begin{figure}
317: \centerline{\epsfig{file=figure3.ps,width=3.00in}}
318: \caption{Displacements of the atoms from the cubic perovskite cell positions,
319: for the most energetically favorable configuration at a given misfit
320: strain.
321: The vertical lines denote the phase transition points obtained
322: from the stability analysis.
323: $\Delta_3\text{(Ti)}$ labels the displacement of the Ti atom in units
324: the third lattice vector, etc.}
325: \label{fig:displacementsDFT}
326: \end{figure}
327:
328: At low temperature, our Pertsev diagram shows the sequence of second-order phase
329: transitions {\em c}$\rightarrow${\em r}$\rightarrow${\em aa}.
330: The {\em r} phase is predicted to exist in a range that is more than twice
331: as broad as that shown in Fig.~\ref{fig:energiesDFT}.
332: We have found that this range is reduced to about 1.5 times that of Fig.~2
333: when the negative-pressure correction is not included.
334: The remaining discrepancy is related to technical differences between the DFT
335: calculations used in \cite{King-Smith.PRB.1994} to obtain the parameters for
336: the effective Hamiltonian method and the DFT calculations we report here
337: \cite{note:technical}.
338: We should also mention that the effective Hamiltonian used here does not
339: include the physics related to the zero-point motion of the ions.
340: This quantum effect should alter the shape of the lines of the diagram at very
341: low temperatures, and it would result in those lines approaching the
342: misfit-strain axis with infinite derivative (see, for example,
343: \cite{Iniguez.PRL.2002}).
344: In any case, at zero temperature, the phase sequence is quite unambiguously
345: established by the first-principles results.
346: This clearly indicates that the low-temperature extrapolation of the
347: Landau-Devonshire parameters fitted near $T_{\rm C}$ can give rise to spurious
348: results, such as the stability of the {\em ac} phase obtained in
349: Ref. \cite{Pertsev.PRL.1998}.
350:
351: \begin{figure}
352: \centerline{\epsfig{file=figure4.ps,width=3.00in}}
353: \caption{Phase diagram of epitaxial BaTiO$_3$ obtained using the effective
354: Hamiltonian of Zhong, Vanderbilt and Rabe
355: \protect\cite{Zhong.PRL.1994+Zhong.PRB.1995}.}
356: \label{fig:diagramHEFF}
357: \end{figure}
358:
359: Finally, we comment on the effect of the assumptions made in the construction
360: of this first-principles Pertsev diagram.
361: In principle, we should consider the possibility of equilibrium structures with
362: larger unit cells, particularly those with cell-doubling octahedral rotations,
363: which have been shown to be important in SrTiO$_3$, and could condense in
364: BaTiO$_3$ under sufficiently large misfit strains.
365: As an example, we have checked that the paraelectric phase of the film is
366: stable with respect to octahedral rotations about the [001] direction (with
367: M$_3$ symmetry) up to an epitaxial compressive strain of
368: $-70.9 \times 10^{-3}$ ($a = 3.675$ \AA), far larger than those likely to be
369: experimentally relevant.
370: In addition, while we have studied only the effects of epitaxial strain,
371: other physical effects may also be relevant to the structure and properties of
372: thin films, such as atomic rearrangements at the film-substrate interface and
373: free surface, and the instability to formation of multiple domain
374: structures \cite{Li.APL.2003}.
375:
376: To summarize, we have performed density-functional theory calculations in order
377: to obtain the ``Pertsev diagram'' of epitaxial BaTiO$_3$ at zero temperature.
378: The results we obtain differ from those computed previously
379: \cite{Pertsev.PRL.1998} using a Landau-Devonshire theory where the
380: parameters needed were obtained from experimental information about bulk
381: BaTiO$_3$ at the phase transitions temperatures.
382: Alternatively, the use of a similar theory where the constants of the model
383: are computed using an {\em ab initio} method is consistent with both the
384: first principles results at zero temperature, and with the work of Pertsev
385: {\em et al.} \cite{Pertsev.PRL.1998} at high temperature.
386:
387: It is a pleasure to thank
388: Jorge \'I\~niguez,
389: Javier Junquera,
390: and Jos\'e Juan Blanco-Pillado
391: for useful discussions.
392: This work was supported by ONR Grants N0014-97-1-0048, N00014-00-1-0261, and
393: N00014-01-1-0365, and DOE Grant DE-FG02-01ER45937.
394:
395: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
396:
397: \begin{thebibliography}{0}
398:
399: \bibitem{Lines.book.1977}
400: M.~E.~Lines and A.~M.~Glass,
401: {\em Principles and Applications of Ferroelectrics and Related Materials}
402: (Clarendon Press, Oxford, 1977).
403:
404: \bibitem{Cohen.Nature.1992}
405: R.~E.~Cohen,
406: Nature {\bf 358}, 136 (1992).
407:
408: \bibitem{Ahn.S.2004}
409: C.~H.~Ahn, K.~M.~Rabe, J.-M.~Triscone,
410: Science {\bf 303}, 488 (2004).
411:
412: \bibitem{Yoneda.APL.1998}
413: Y.~Yoneda, T.~Okabe, K.~Sakaue, H.~Terauchi, H.~Kasatani, and K.~Deguchi,
414: J. Appl. Phys. {\bf 83}, 2458 (1998).
415:
416: \bibitem{Chang.JAP.2000}
417: W.~Chang, C.~M.~Gilmore, W.-J.~Kim, J.~M.~Pond, S.~W.~Kirchoefer, S.~B.~Qadri,
418: D.~B.~Chrisey, and J.~S.~Horwitz,
419: J. Appl. Phys. {\bf 87}, 3044 (2000).
420:
421: \bibitem{Li.APL.2001}
422: H.~Li, A.~L.~Roytburd, S.~P.~Alpay, T.~D.~Tran, L.~Salamanca-Riba, and
423: R.~Ramesh,
424: Appl. Phys. Lett. {\bf 78}, 2354 (2001).
425:
426: \bibitem{Shimuta.JAP.2002}
427: T.~Shimuta, O.~Nakagawara, T.~Makino, S.~Arai, H.~Tabata, and T.~Kawai,
428: J. Appl. Phys. {\bf 91}, 2290 (2002).
429:
430: \bibitem{Neaton.APL.2003}
431: J.~B.~Neaton and K.~M.~Rabe, Appl. Phys. Lett. 82, 1586 (2003).
432:
433: \bibitem{Pertsev.PRL.1998}
434: N.~A.~Pertsev, A.~G.~Zembilgotov, and A.~K.~Tagantsev,
435: Phys. Rev. Lett. {\bf 80}, 1988 (1998).
436:
437: \bibitem{note:cubic}
438: The applicability of our study is thus not limited to cubic substrates; it
439: also applies, e.g., to tetragonal perovskite substrates.
440:
441: \bibitem{Koukhar.PRB.2001}
442: V.~G.~Koukhar, N.~A.~Pertsev, and R.~Waser,
443: Phys. Rev. B {\bf 64}, 214103 (2001).
444:
445: \bibitem{Hohenberg.PR.1964+Kohn.PR.1965}
446: P.~Hohenberg and W.~Kohn,
447: Phys. Rev. {\bf 136}, B864 (1964);
448: W.~Kohn and L.~J.~Sham,
449: Phys. Rev. {\bf 140}, A1133 (1965).
450:
451: \bibitem{Kresse.PRB.1993+Kresse.PRB.1996}
452: G.~Kresse and J.~Hafner,
453: Phys. Rev. B {\bf 47}, 558 (1993);
454: G.~Kresse and J.~Furthm\"uller,
455: Phys. Rev. B {\bf 54}, 11169 (1996).
456:
457: \bibitem{Blochl.PRB.1994+Kresse.PRB.1999}
458: P.~E.~Bl\"ochl,
459: Phys. Rev. B {\bf 50}, 17953 (1994);
460: G.~Kresse and D.~Joubert,
461: Phys. Rev. B {\bf 59}, 1758 (1999).
462:
463: \bibitem{Ceperley.PRB.1980}
464: D.~M.~Ceperley and B.~J.~Alder,
465: Phys. Rev. Lett. {\bf 45}, 566 (1980).
466:
467: \bibitem{note:gga}
468: Using the generalized gradient approximation (GGA) instead of the LDA does
469: not lead to substantial improvements in the case of BaTiO$_3$; see
470: D. J. Singh,
471: Ferroelectrics {\bf 164}, 143 (1995).
472:
473: \bibitem{Monkhorst.PRB.1979}
474: H.~J.~Monkhorst and J.~D.~Pack,
475: Phys. Rev. B {\bf 13}, 5188 (1976).
476:
477: \bibitem{note:ifc}
478: Similarly, the {\em a} phase is stable against the {\em ac} phase down
479: to $s_{\rm min}(a) = 10.3 \times 10^{-3}$ ($a = 3.993$ \AA).
480: Omitting the strain and working only with the
481: $5 \times 5$ Hessian matrix matrix results in very little error:
482: $s_{\rm max}(c) = -6.2 \times 10^{-3}$ ($a = 3.931$ \AA),
483: $s_{\rm min}(aa) = 6.1 \times 10^{-3}$ ($a = 3.980$ \AA), and
484: $s_{\rm min}(a) = 9.3 \times 10^{-3}$ ($a = 3.992$ \AA).
485:
486: \bibitem{Zhong.PRL.1994+Zhong.PRB.1995}
487: W.~Zhong, D.~Vanderbilt, and K.~M.~Rabe,
488: Phys. Rev. Lett. {\bf 73}, 1861 (1994);
489: W.~Zhong, D.~Vanderbilt, and K.~M.~Rabe,
490: Phys. Rev. B {\bf 52}, 6301 (1995).
491:
492: \bibitem{note:mcdetails}
493: MC simulations were performed using a 12$\times$12$\times$12 supercell.
494: Typically 30,000~MC sweeps were used to equilibrate the system, and
495: an additional 50,000 to obtain averages of local-mode variables with a
496: statistical error below $10\%$. The temperature was increased in steps
497: of 5~K.
498:
499: \bibitem{King-Smith.PRB.1994}
500: R.~D.~King-Smith and D.~Vanderbilt,
501: Phys. Rev. B {\bf 49}, 5828 (1994).
502:
503: \bibitem{note:technical}
504: In particular, the soft-mode eigenvalue has been found to be very sensitive to
505: the fineness of the grid used to evaluate the Fourier transforms.
506: This discrepancy vanishes almost completely if instead of working with the
507: soft-mode eigenvalue calculated in \protect\cite{King-Smith.PRB.1994}
508: (0.0350 a.u.) we use the one given by our new DFT calculations
509: (0.0223 a.u.) that was calculated using a finer grid than
510: in \protect\cite{King-Smith.PRB.1994}.
511:
512: \bibitem{Iniguez.PRL.2002}
513: J.~\'I\~niguez and D.~Vanderbilt,
514: Phys. Rev. Lett. {\bf 89}, 115503 (2002).
515:
516: \bibitem{Li.APL.2003}
517: Y.~L.~Li, S.~Choudhury, Z.~K.~Liu, and L.~Q.~Chen,
518: Appl. Phys. Lett. {\bf 83}, 1608 (2003).
519:
520: \end{thebibliography}
521:
522: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
523:
524: \end{document}
525: