1: \documentstyle[prl,aps,floats,epsfig]{revtex} % single-column format
2:
3: \begin{document}
4: \draft
5: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname
6: @twocolumnfalse\endcsname
7:
8: %---------------------------------------------------------------------
9: \title{Compositional Inversion Symmetry Breaking in Ferroelectric
10: Perovskites}
11:
12: \author{Na Sai, B. Meyer, and David Vanderbilt }
13:
14: \address{Department of Physics and Astronomy,\\
15: Rutgers University, Piscataway, New Jersey 08854-8019, USA}
16: \date{March 14, 2000}
17: \maketitle
18: %------------------------------------------------------------------------------
19:
20: \begin{abstract}
21: Ternary cubic perovskite compounds of the form
22: (A$_{1/3}$A$'_{1/3}$A$''_{1/3}$)BO$_3$ and
23: A(B$_{1/3}$B$'_{1/3}$B$''_{1/3}$)O$_3$, in which the
24: differentiated cations form an alternating series of monolayers,
25: are studied using first-principles methods. Such compounds are
26: representative of a possible new class of materials in which
27: ferroelectricity is perturbed by compositional breaking of
28: inversion symmetry. For isovalent substitution on either
29: sublattice, the ferroelectric double-well potential is found to
30: persist, but becomes sufficiently asymmetric that minority
31: domains may no longer survive. The strength of the symmetry
32: breaking is enormously stronger for heterovalent substitution, so
33: that the double-well behavior is completely destroyed. Possible
34: means of tuning between these behaviors may allow for the
35: optimization of resulting materials properties.
36: \end{abstract}
37:
38: \pacs{PACS numbers: 77.80.-e, 77.84.Dy, 61.50.Ah, 81.05.Zx}
39:
40: \vskip2pc]
41: \baselineskip=11.6pt
42: % normally set to 12.0pt ; this is a cheat!
43:
44: \narrowtext
45:
46: In the last decade, the extensive use of first-principles theoretical
47: methods to study ferroelectric perovskite oxides has led to
48: a greatly expanded understanding of the ferroelectric and
49: piezoelectric properties of this important class of
50: materials. Theoretical investigations of electronic, dynamical, and
51: structural properties have been shown
52: to be in good accord with experimental observations
53: for the simple ABO$_3$ perovskites \cite{Opinion}.
54: Moreover, these studies provide microscopic insight into the
55: ferroelectric instabilities, their relation to the long-range
56: Coulomb interactions\cite{Ghosez}, and related questions about the origins
57: of the piezoelectric response\cite{Szabo,LBDVpiezo}.
58:
59: However, the materials of most interest for technological
60: applications are generally not the simple ABO$_3$ perovskites, but
61: {\it solid solutions} with stoichiometric substitutions of A or B metal atoms.
62: Examples include PZT (PbZr$_{1-x}$Ti$_x$O$_3$), currently one of the most
63: widely used ferroelectrics, and PZN (PbZn$_{1/3}$Nb$_{2/3}$O$_3$)
64: and PMN (PbMg$_{1/3}$Nb$_{2/3}$O$_3$) and their solid solutions
65: with PbTiO$_3$, which have recently been shown to have enormous
66: piezoelectric response in single-crystal form \cite{Park}.
67: In fact, this class of materials shows great promise for the
68: development of {\it new materials} having improved dielectric
69: and electromechanical properties. Not only is there an enormous
70: space of chemical compositions to explore, but it may also be
71: possible to optimize the desired material properties by tuning the
72: degree and type of compositional order for the desired
73: application \cite{Setter,Davies}.
74:
75: In particular, Eckstein \cite{Eckstein} has recently
76: suggested that the artificial atomic-layer growth of {\it
77: compositionally ordered structures that break inversion symmetry}
78: might be especially exciting and fruitful in this regard.
79: The resulting asymmetry of the ferroelectric double-well potential in
80: such a material suggests the prospect of {\it qualitatively new behavior,}
81: e.g., ``self-poling'' materials with tailored piezoelectric or
82: dielectric properties.
83:
84: In this Letter, we explore this exciting prospect by carrying
85: out {\it ab-initio} theoretical calculations for several prospective
86: model structures of this type. Specifically, we envision the
87: artificial growth of materials of overall composition
88: (A$_{1/3}$A$'_{1/3}$A$''_{1/3}$)BO$_3$ or
89: A(B$_{1/3}$B$'_{1/3}$B$''_{1/3}$)O$_3$,
90: in which the three different cations alternate layer-by-layer
91: along the ferroelectric direction
92: as illustrated in Figs.~\ref{fig:procedure}(a-b) for A-site and B-site
93: modulation respectively. We will {\it assume} that
94: such atomic-layer control will become possible \cite{Eckstein},
95: and investigate the energy landscapes and ferroelectric
96: properties of the resulting materials.
97: As will be shown below, we find that the asymmetry may easily be strong
98: enough for self-poling to occur, even in the case of isovalent
99: substitution. Moreover, we find a surprisingly strong
100: variation of the strength of the symmetry breaking with the strength of
101: the compositional perturbation, allowing for a very wide tunability of
102: ferroelectric properties. Experimentally, such tunability might be
103: exploited by growing alloy structures with a cyclic modulation of the
104: concentration variable.
105: %
106: We thus find very strong motivations for the development of such
107: materials, and hope that our work will encourage experimental efforts
108: directed toward their synthesis.
109:
110:
111: % ---
112: % Theoretical model assumptions
113: % ---
114:
115: We adopt bulk BaTiO$_3$ as a prototypical parent compound
116: and construct a series of model systems that allow us to test
117: for the effects of A-site {\it vs.}\ B-site and
118: isovalent {\it vs.}\ heterovalent substitution. Specifically,
119: we consider (Ba$_{1/3}$Sr$_{1/3}$Ca$_{1/3}$)TiO$_3$ as an example
120: of an ``A-iso'' system (isovalent substitution on the A site),
121: Ba(Sc$_{1/3}$Ti$_{1/3}$Nb$_{1/3}$)O$_3$ as an example of a
122: ``B-hetero'' system (heterovalent substitution on the B site), and
123: Ba(Ti$_{1/3}$Zr$_{1/3}$Hf$_{1/3}$)O$_3$ as an intermediate
124: ``B-iso'' case.
125: We construct 15-atom supercells by tripling the primitive unit
126: cell along the $z$ direction and cyclically alternating the identity
127: of the A or B atom, as shown in Figs.~1(a-b). We assume that any
128: ferroelectric order develops only along the $z$-direction, so that
129: the material remains tetragonal,
130: and only $z$ displacements need be considered.
131: (It may be possible to realize this situation by
132: appropriate choice of lattice-mismatched substrate for epitaxial
133: growth, but it is not our purpose here to investigate this
134: possibility.) We also assume perfect control
135: of layer-by-layer composition, resulting in the ideal stacking of
136: Figs.~1(a-b), and carry out our theoretical studies only at
137: zero temperature. Much future work clearly remains to be done in
138: relaxing these assumptions, but the study of simple prototypical
139: systems is a natural first step.
140:
141: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
142: %%Calculation Method and searching strategy%%%
143: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
144:
145: \begin{figure}
146: \centerline{\epsfig{file=fig1.ps,height=7cm}}
147: \vskip 5pt
148: \caption{Side views of
149: (a) (A$_{1/3}$A$'_{1/3}$A$''_{1/3}$)BO$_3$, and
150: (b) A(B$_{1/3}$B$'_{1/3}$B$''_{1/3}$)O$_3$ structures.
151: Energy {\it vs.} displacement along
152: (c) the cubic force direction,
153: (d) the line ${\bf r}={\bf r}_X+u_0\hat{\xi}_0$, and
154: (e) the line connecting the minima $A$ and $B$.
155: Solid (dashed) curves are for A-iso (B-hetero) systems.
156: \label{fig:procedure}}
157: %\vskip 2mm
158: \end{figure}
159:
160: The {\it ab-initio} calculations are carried out using the
161: Vanderbilt ultra-soft pseudopotential scheme\cite{David90}
162: in the local-density approximation. Details of the
163: pseudopotentials can be found in Ref.~\onlinecite{KSDV94}. Good $k$-point
164: convergence is obtained using a (6,6,2) Monkhorst-Pack mesh,
165: corresponding to the bulk (6,6,6) mesh\cite{explan-k}, and a 25-Ry plane-wave cutoff is used throughout.
166: Because experimental lattice constants $a$ and $c$ are not available
167: for the ordered compounds of interest, the calculated theoretical
168: equilibrium values were usually used. However, in some cases the
169: $c/a$ ratio was fixed and only the cell volume was optimized, as
170: detailed below.
171:
172: In the presence of the broken inversion symmetry, it is not always
173: easy to locate both local energy minima corresponding to the two
174: ferroelectric ground states. We have found that the following
175: procedure is quite reliable for finding both minima, if they exist.
176: First, we place the atoms at the ideal cubic coordinates and calculate
177: the pattern of forces, thus defining a ``cubic force direction''
178: $\hat{\xi}_{\rm cf}$ in the 15-dimensional configuration space.
179: We then use $\hat{\xi}_{\rm cf}$ as the search direction for a line
180: minimization, and denote the minimum along this line
181: ${\bf r}=u_{\rm cf}\,\hat{\xi}_{\rm cf}$ to be $X$, as
182: shown in Fig.~\ref{fig:procedure}(c). Next, starting from $X$,
183: we carry out a second line minimization along the line
184: ${\bf r}={\bf r}_X+u_0\,\hat{\xi}_0$, where
185: $\hat{\xi}_0$ corresponds to the ferroelectric mode unit vector of
186: bulk BaTiO$_3$. As shown in Fig.~\ref{fig:procedure}(d), two scenarios
187: can be identified. (i) If two minima $A'$ and $B'$ are
188: found along this line, then we carry out a steepest-descent
189: minimization from each, obtaining two distinct local minima $A$ and $B$.
190: Fig.~\ref{fig:procedure}(e) shows the perturbed double-well potential
191: plotted along a line connecting points $A$ and $B$ (direction
192: $\hat{\xi}$) in such a case.
193: (ii) If only one minimum is found, as for the dashed line of
194: Fig.~\ref{fig:procedure}(d), then the double-well potential has
195: been destroyed, and a subsequent steepest-descent minimization
196: identifies the unique minimum.
197:
198:
199: %%%%%%%%%%%%%%%%%%%%%%%%%
200: %Results and Discussions%
201: %%%%%%%%%%%%%%%%%%%%%%%%%
202:
203: \begin{figure}
204: \centerline{\epsfig{file=fig2.ps,angle=270,width=6cm}}
205: \vskip 5pt
206: \caption{Energy vs. displacement along the line
207: ${\bf r}={\bf r}_X+u_0\,\hat{\xi}_0$ in the A-iso system for
208: several cell volumes.
209: \label{fig:Aiso}}
210: \end{figure}
211:
212: We first consider the application of this procedure to the A-iso system
213: (Ba$_{1/3}$Sr$_{1/3}$Ca$_{1/3}$)TiO$_3$. The relaxation from the
214: ideal cubic structure to the point $X$ illustrated in
215: Fig.~\ref{fig:procedure}(c) is found to be dominated by ionic
216: size effects: one sees a simple shift of the atoms toward
217: smaller A-site cations and away from larger ones.
218: Figure \ref{fig:procedure}(d) then illustrates the
219: minimization along ${\bf r}={\bf r}_X+u_0\,\hat{\xi}_0$ for one
220: particular fixed cell volume (1343\,a.u.$^3$),
221: indicating a modest
222: asymmetry between minima $A'$ and $B'$. However, since it is well
223: established that the presence or absence of a ferroelectric
224: instability depends strongly on cell volume in perovskites
225: \cite{KSDV94}, we have repeated this step for a series of fixed
226: atomic volumes in Fig.~\ref{fig:Aiso}.
227: (The curves in Fig.~\ref{fig:Aiso} could thus
228: be roughly interpreted as corresponding
229: to a series of related compounds having differing tendencies
230: towards ferroelectric instability.) The results illustrate the
231: typical evolution of the double-well potential with increased
232: ferroelectric tendency, from a slightly asymmetric single-well
233: minimum ($V$=1123\,a.u.$^3$) to a slightly asymmetric double-well
234: potential ($V$=1463\,a.u.$^3$) \cite{explan-ca}.
235: It happens that the theoretical
236: equilibrium volume $V$=1193\,a.u.$^3$ is a nearly borderline case;
237: we find only a single minimum, but there are still two inflection
238: points. In such a case, we modify the procedure of the previous
239: paragraph by carrying out a steepest-descent minimization from
240: each inflection point, and by doing so we succeed in finding the
241: two distinct minima ${\bf r}_A$ and ${\bf r}_B$ in
242: Fig.~\ref{fig:procedure}(e). For this case ($V$=1193\,a.u.$^3$),
243: we find a soft-mode amplitude of 0.3\,a.u.\ and an average well depth
244: (relative to the saddle point $S$) of 0.48\,mHa per 5-atom cell,
245: fairly close to the corresponding values of 0.25\,a.u.\ and 0.43 mHa
246: \cite{KSDV94}, respectively, for tetragonal bulk BaTiO$_3$.
247: The energy difference between the twin wells is modest,
248: $\sim 15-20\%$ of the well depth.
249:
250: Applying the same approach to the B-iso system
251: Ba(Ti$_{1/3}$Zr$_{1/3}$Hf$_{1/3}$)O$_3$, we
252: observe very similar behavior: a strong volume-dependence of
253: the ferroelectric tendency, and a modest inversion asymmetry
254: for both single-well and double-well volumes. In this case,
255: the structure does not develop a double-well potential until
256: the volume is increased to 1468\,a.u.$^3$,
257: but the degree of asymmetry is similar as for the A-iso case.
258: Thus, we conclude that the choice of A vs.~B site for an isovalent
259: chemical substitution does not strongly affect the strength of
260: the asymmetry or the qualitative behavior of the system.
261:
262: While the asymmetries may appear small in the case of isovalent
263: substitution, they are large by one important measure. We define
264: an ``effective electric field'' ${\cal E}_{\rm eff}=\Delta E/\Delta P_s$,
265: where $\Delta E$ and $\Delta P_s$ denote the energy and polarization
266: differences between the two local minima. If this
267: quantity is larger than the coercive field of the material, it
268: means that the thermodynamic preference for the energetically
269: preferred minimum is strong enough to overcome the pinning of the
270: domain walls and to spontaneously switch the material into a
271: single-domain state. Using the Berry-phase approach \cite{KSDVBerry}
272: to calculate $\Delta P_s$, we find ${\cal E}_{\rm eff}$=90\,kV/cm at
273: the equilibrium volume
274: in the A-iso system. Since the typical coercive field of most
275: perovskite ferroelectrics is closer to 15\,kV/cm, we thus arrive at
276: the important conclusion that the symmetry breaking may easily be
277: strong enough to cause the material to self-pole, even in the
278: case of isovalent substitution.
279:
280: Turning now to the case of {\it heterovalent} chemical substitution,
281: we find very different behavior in this case. Specifically, we
282: consider the B-hetero system
283: Ba(Sc$_{1/3}$Ti$_{1/3}$Nb$_{1/3}$)O$_3$ in which the valence
284: charges are +3, +4, and +5 on Sc, Ti, and Nb, respectively.
285: Structural optimization results in a lattice constant of
286: $a=7.60$\,a.u. and $c/a=3.023$. The pattern of relaxation
287: $\hat{\xi}_{\rm cf}$ leading to configuration $X$ suggests that
288: the electrostatic interaction is a dominant effect.
289: Specifically, we observe bucklings of the AO and BO$_2$ planes
290: that are consistent with a picture of static electric fields
291: arising from the different B-atom valence charges.
292:
293: Most importantly, for the heterovalent case we find {\it only a
294: single minimum} when searching from $X$ along $\hat{\xi}_0$, as
295: shown by the dashed curve in Fig.~\ref{fig:procedure}(d). Various
296: alternate search strategies failed to identify a second minimum;
297: all trial structures were found to relax back to a unique
298: structural ground state. The absence of the second minimum was not
299: found to be sensitive to the cell volume, as was the case for
300: isovalent substitutions. Thus, we find that the symmetry breaking is
301: enormously stronger than for the isovalent case, and it is clear that
302: resulting behavior is of a qualitatively different type.
303:
304: \begin{figure}
305: \centerline{\epsfig{file=fig3.ps,angle=270,width=6cm}}
306: \vskip 5pt
307: \caption{Energy vs. displacement along the line ${\bf r}={\bf
308: r}_X+u_0\,\hat{\xi}_0$ in the B-hetero system with continuous
309: $\delta$.
310: \label{fig:B_het}}
311: \end{figure}
312:
313: In order to gain a better understanding of this behavior, and in
314: particular to track the disappearance of the secondary minimum,
315: we have developed a model system to study
316: the effects of ``turning on'' the heterovalent symmetry-breaking
317: perturbation gradually. We introduce a continuous variable
318: $\delta$, and construct artificial atoms with fractional nuclear
319: charges that deviate by $\pm\delta$ from those of Ti ($Z$=22).
320: Constructing a crystal out of a cyclic series of Ti$-\delta$, Ti,
321: and Ti$+\delta$ atomic layers, we can continuously tune the system
322: from ferroelectric BaTiO$_3$ ($\delta$=0) to a full-fledged heterovalent
323: system Ba(Sc$_{1/3}$Ti$_{1/3}$V$_{1/3}$)O$_3$ ($\delta$=1). The
324: energies calculated along the $\hat{\xi}_0$ direction from $X$ are
325: plotted for a series of $\delta$ values in Fig.~\ref{fig:B_het}.
326: As $\delta$ increases from 0 to 0.3, the asymmetry of the double
327: well increases and the well depth decreases. At $\delta=0.4$, the
328: curve exhibits only a single energy minimum, signaling the
329: transition from double-well to single-well behavior in response to
330: the increasingly strong symmetry-breaking perturbation.
331: To confirm the disappearance of the secondary minimum more directly,
332: we also tracked its evolution as $\delta$ was ``turned on'' in a
333: sequence of small steps, using the relaxed structure at one $\delta$
334: as a starting guess for the next.
335: At $\delta=0.34$,
336: the minimum was confirmed to disappear, and subsequent relaxation
337: led back to the principal (now global) minimum.
338:
339: To gain more insight into the structural relaxations, we have found
340: it useful to introduce a measure of the
341: ``strength of the symmetry breaking.'' While it is tempting to
342: choose a measure that is related to the energy difference between
343: the two local minima, like the ${\cal E}_{\rm eff}$ introduced
344: earlier, such a definition has the disadvantage of being ill-defined
345: in the single-minimum case. Thus, we have adopted instead the
346: following measure. For any
347: curve such as that of Figs.~1(d-e), 2, or 3, we locate the point
348: of minimum $d^2E/du^2$ (i.e., vanishing $d^3E/du^3$), and then
349: define $F_{\rm sb}$ to be $dE/du$ evaluated at that point.
350: We refer to $F_{\rm sb}$ as the ``symmetry-breaking force''
351: since it measures the strength of the symmetry breaking and has
352: units of force. It turns out to have a similar behavior
353: as the well-depth difference in the double-well case, but has
354: the advantage of remaining well-defined in the single-well case.
355:
356: Calculations of $F_{\rm sb}$ in the A-iso and B-iso
357: cases indicate that $F_{\rm sb}$ has only a modest and smooth
358: dependence on cell volume even when passing through the
359: transition from single-well to double-well behavior, confirming
360: that the ``strength of the symmetry breaking'' is not the
361: variable parameter in those cases. However, returning
362: to the virtual-atom B-hetero system, we find that $F_{\rm sb}$ is
363: an extremely sensitive function of $\delta$, with numerical fits
364: indicating a $\delta^3$ dependence. It is not hard
365: to see that $F_{\rm sb}$ must be an odd function of $\delta$, but its
366: cubic behavior may seem surprising at first sight. However, the
367: vanishing of the linear term can be deduced from simple symmetry
368: arguments. An intuitive form of the argument is to note that
369: the symmetry-breaking perturbation, which is of the form
370: ($+\delta$, $-\delta$, 0) in successive layers, can be regarded as
371: a superposition of two perturbations
372: ($2\delta/3$, $-\delta/3$, $-\delta/3$) and
373: ($\delta/3$, $-2\delta/3$, $\delta/3$) that {\it do not break}
374: inversion symmetry. Thus, the principle of superposition prevents
375: the occurrence of any symmetry-breaking response in linear order
376: in $\delta$, and in particular $F_{\rm sb}$ must vanish. (A more
377: systematic analysis may be made by considering the $C_{3v}$
378: symmetry group consisting of primitive translations
379: along $\hat{z}$ and mirrors $M_z$. The ferroelectric mode vector
380: and the perturbation $\delta$ are found to belong to the $A_2$ and
381: $E$ respresentations, repsectively, and thus
382: cannot couple at linear order.)
383:
384: Knowing the form of this extraordinarily strong cubic dependence of
385: $F_{\rm sb}$ on $\delta$, we can now understand the pronounced qualitative
386: differences that were observed for the cases of isovalent and
387: heterovalent substitution. We find that $F_{\rm sb}$ is about the same,
388: $\sim$0.25\,mHa/a.u., in the B-hetero system with $\delta=0.25$
389: as in the A-iso system. Increasing $\delta$ from 0.25 to 0.4 increases
390: $F_{\rm sb}$ by about a factor of four, enough to cause the transition to
391: single-well behavior. A further increase of $\delta$ from 0.4 to
392: 1.0 leads to a further increase of the symmetry-breaking force
393: $F_{\rm sb}$ by a {\it factor of about 15}. Thus, in a fully-developed
394: B-hetero system such as Ba(Sc$_{1/3}$Ti$_{1/3}$Nb$_{1/3}$)O$_3$,
395: the strength of the symmetry breaking is more than an order of
396: magnitude larger than needed to destroy the secondary minimum
397: completely. It is hardly surprising, then, that we observed no
398: secondary minimum in this case!
399:
400: The enormous disparity between the behavior in the isovalent and
401: heterovalent cases suggests that it may be of great interest to
402: find a way of tuning the system continuously from one behavior to
403: the other. One can imagine doing this by regarding $\delta$ not
404: as a variable atomic number, but rather as representing a
405: layer-by-layer composition variable. For example,
406: one could conceive of the epitaxial growth
407: of Ba(Sc$_{1-y}$Nb$_y$)O$_3$
408: in an alternating sequence of layers with $y=0.5(1+\delta)$, $y=0.5$,
409: and $y=0.5(1-\delta)$. By the same symmetry arguments, the effective
410: strength of the symmetry breaking must again scale as $\delta^3$. Thus, by
411: controlling the concentration variable $\delta$, one can hope to tune
412: the system over a very wide range of behavior.
413:
414: One would clearly like to use this tunability to optimize the
415: desired characteristics of the material, such as the piezoelectric
416: response. It might naively be expected that increasing
417: the strength of the inversion symmetry breaking will increase the
418: piezoelectric response. (For example, if the parent material at
419: $\delta$=0 were {\it paraelectric}, then the piezoelectric
420: response would be expected to scale as $\delta^3$.) On the other
421: hand, the materials that have the largest piezoelectric
422: coefficients are typically ferroelectrics, and since a very
423: strong symmetry breaking suppresses the ferroelectric behavior,
424: it might be counterproductive to make $\delta$ too large.
425: Clearly, further theoretical investigation is needed in order to
426: clarify these issues.
427:
428: Many other questions remain open and need to be resolved. For
429: example, the conditions under which the polarization will remain
430: oriented parallel to the growth direction, the behavior of the
431: system as a function of temperature, and the properties of
432: materials with simultaneous A-site and B-site substitution,
433: are obvious candidates for further study. The thermodynamic
434: behavior of these materials, which are technically pyroelectric
435: but interpolate to a ferroelectric limit as $\delta\rightarrow0$,
436: are also deserving of investigation. In the meantime,
437: we hope that our theoretical investigations will stimulate
438: attempts at experimental growth and characterization of novel
439: perovskites with compositionally broken inversion symmetry.
440:
441: % ---
442: % Acknowledgments
443: % ---
444:
445: We thank J.~Eckstein for suggesting the direction of this study.
446: Support for this work was provided by ONR Grant N00014-97-1-0048 and
447: NSF Grant DMR-9981193. We thank M.~Cohen and K.~Rabe for useful
448: discussions.
449:
450: %---
451: \begin{references}
452: %---
453:
454: \bibitem{Opinion}D. Vanderbilt, Curr. Opin. Mater. Sci
455: {\bf 2}, 701 (1997); and J. Korean Phys. Soc. S103 (1998).
456:
457: \bibitem{Ghosez} Ph. Ghosez, X. Gonze and J.-P. Michenaud,
458: Europhys. Letters {\bf 33}, 713 (1996); Ferroelectrics {\bf 186}, 73
459: (1996).
460:
461: \bibitem{Szabo} G. Saghi-Szabo, R.E. Cohen and H. Krakauer,
462: Phys. Rev. Lett. {\bf 80}, 4321 (1998); Phys. Rev. B {\bf 59},
463: 12771 (1999).
464:
465: \bibitem{LBDVpiezo} L. Bellaiche and David Vanderbilt, Phys. Rev.
466: Lett. {\bf 83}, 1347 (1999).
467:
468: \bibitem{Park} S.-E. Park and T.E. Shrout, J. Appl. Phys. {\bf 82},
469: 1804 (1997).
470:
471: \bibitem{Setter} N. Setter and L.E. Cross, J. Appl. Phys. {\bf 51},
472: 4356 (1980).
473:
474: \bibitem{Davies} M.A. Akbas and P.K. Davies, J. Am. Ceram. Soc.
475: {\bf 81}, 670 (1998).
476:
477: \bibitem{Eckstein} J. Eckstein, private communication.
478:
479: \bibitem{David90}D. Vanderbilt, Phys. Rev. B {\bf 41},7892 (1990).
480:
481: \bibitem{KSDV94} R.D. King-Smith and D. Vanderbilt, Phys. Rev. B
482: {\bf 49}, 5828 (1994).
483:
484: \bibitem{explan-k} A (4,4,2) $k$-point grid was found to be
485: sufficient for the B-iso case.
486:
487: \bibitem{explan-ca} These calculations were done at a fixed
488: $c/a=3.04$ derived from the bulk tetragonal phase of BaTiO$_3$.
489: See J. Padilla and D. Vanderbilt, Phys. Rev. B {\bf 56}, 1625
490: (1997).
491:
492: \bibitem{KSDVBerry}R.D. King-Smith and D. Vanderbilt, Phys. Rev. B
493: {\bf 47},1651(1993).
494:
495: \end{references}
496:
497:
498: \end{document}
499: