1: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: %\documentclass[twocolumn,showpacs,amsmath,amssymb]{revtex4}
3: \documentclass[preprint,amsmath,amssymb]{revtex4}
4:
5: \usepackage{graphicx}% Include figure files
6: %\usepackage{dcolumn}% Align table columns on decimal point
7:
8: \begin{document}
9: \title{Cooperative response of Pb(ZrTi)O$_3$ nanoparticles
10: to curled electric fields}
11: \author{Ivan I. Naumov and Huaxiang Fu}
12: \affiliation{Department of Physics, University of Arkansas,
13: Fayetteville, AR 72701, USA}
14: \date{\today}
15:
16: \begin{abstract}
17: Using first-principles based effective Hamiltonian and finite
18: temperature Monte Carlo simulations we investigate cooperative
19: responses, as well as microscopic mechanism for vortex switching,
20: in zero-dimensional Pb(Zr$_{0.5}$Ti$_{0.5}$)O$_3$ nanoparticles
21: under curled electric fields. We find that the generally accepted
22: domain coexistence mechanism is not valid for toroid switching.
23: Instead dipoles are shown to display unusual collective behaviors
24: by forming a new vortex with perpendicular (but not opposite)
25: toroid moment. The strong correlation between the new and original
26: vortices is revealed to be critical for reversing toroid moment.
27: Microscopic insight for the puzzling collective response is
28: discussed. Based on our finding, we further describe a technological
29: approach that is able to drastically reduce the magnitude of the
30: curled electric field needed for vortex switching.
31: \end{abstract}
32:
33:
34: \pacs{77.80.Bh, 77.80.Fm, 64.70.Nd, 77.22.Ej}
35:
36: \maketitle
37:
38: In the study of ferroelectric physics, one field that is currently attracting
39: widespread attention is low-dimensional ferroelectrics
40: (FEs).\cite{Sai,Neaton,Weissman,Lee,Fong}
41: Among FE nanostructures that have been synthesized,
42: FE nanoparticles are particularly interesting because the strong depolarizing fields along all three
43: directions force dipoles in these particles to adopt a topological
44: vortex configuration,\cite{Naumov1} which is a new form of dipole
45: ordering\cite{Ginzburg}. This ordering opens up a different class
46: of cooperative phenomena and phase transitions in FEs of reduced dimension,
47: which is a subject that has generated long-standing interest
48: for decades but is still poorly understood \cite{kittel,mermin,thouless,anderson,bruno,Gamb,landau}.
49:
50: Another field of fundamental interest in ferroelectrics is to
51: influence dipole behaviors by, e.g., applying electric
52: fields~\cite{Lines,Garcia} or pressure\cite{Samara,Wu}. The
53: interest stems from the fact that external stimuli often alter the
54: balance among different interactions inside
55: FEs~\cite{Cohen}, thereby giving rise to new phenomena. In this
56: context, curled electric field (namely, the electric field
57: generated not by static charges, but by time-dependent magnetic
58: field via $\nabla\times\mathbf {E}_{\rm curl}=-\partial \mathbf
59: B/\partial t$) is an interesting route for modifying the behaviors
60: of dipoles since the field possesses non-vanishing curl.
61: The $\mathbf {E}_{\rm curl}$
62: fields can be generated on FE particles using focused laser beams
63: and by varying the {\bf B}-field component of the laser. But
64: effects of curled fields in FE nanoparticles have never been
65: studied before (to our knowledge). The knowledge as to
66: how the interaction between FE dipoles and curled electric fields
67: changes the phase properties remains entirely unknown. Technology
68: wise, understanding the response of FE vortex to curled electric
69: fields is of immense relevance, since use of bistable vortex
70: states promises to increase the density of nonvolatile
71: ferroelectric random access memories by five orders of
72: magnitude.\cite{Naumov1} Reversing the toroid moment, by applying
73: curled electric fields, is a key process for making this possible.
74: Recently, a study of using {\it
75: static} electric fields to influence the formation of vortex in FE
76: particles was reported\cite{LB1}, where the system is first heated
77: to high temperatures so that the original vortex disappears, and
78: then the paraelectric particle is annealed under the influence of
79: external static fields. By adopting different external fields,
80: vortices of varied orientations were obtained after annealing.
81: Obviously this is not a switching process in which a constant
82: temperature is maintained.
83:
84: Independent of both subjects given above, a third topic of
85: importance is to understand switching mechanism in FEs. In {\it
86: bulk} FEs, coexistence of -{\bf P} and {\bf P} domains has been
87: widely accepted as the mechanism for polarization
88: switching.\cite{Lines} More specifically, the switching begins by
89: nucleation of a -{\bf P} domain of reversed polarization; crystal
90: defects (domain walls, twin boundaries, etc.) often are the
91: nucleation sites.\cite{Merz} Subsequently, the -{\bf P} domain
92: grows while the original {\bf P} domain shrinks under external
93: fields, leading eventually to switching. In FE
94: nanoparticles, formation of domain walls is less likely and energetically
95: unfavorable. However, the vortex phase in FE particles
96: does have an intrinsic topological defect---the vortex core.
97: Based on the fact that a large strain energy exists near the center
98: of the vortex where dipoles turn their directions sharply,
99: vortex cores are thus less stable. When imposed under a switching
100: curled field, dipoles near the vortex core are likely to flip first
101: to form a -{\bf G} domain of reversed moment. This -{\bf G} domain
102: coexists with and gradually conquers the {\bf G} domain, leading to switching.
103:
104: Here we perform {\it ab initio} based studies to investigate for
105: the first time the collective response to curled electric fields,
106: and the vortex switching mechanism, in ferroelectric nanoparticles
107: made of Pb(Zr$_{0.5}$Ti$_{0.5}$)O$_3$ solid solution. The purpose
108: of this Letter is (1) to demonstrate that the generally accepted
109: mechanism of domain coexistence is invalid for vortex switching,
110: and in fact, vortex reversal occurs by following an interesting
111: G$_z$-G$_y$-(-G$_z$) phase transformation sequence; (2) to report
112: the unusual collective phenomena and energetics in 0D FEs in response to
113: curled electric fields. Based on the key insight obtained
114: in this study, we further propose and demonstrate
115: an alternative approach (which could be technologically very important)
116: to overcome the energy barrier of vortex switching.
117: It should be pointed out that this work is profoundly
118: different from a previous study\cite{Naumov2} where effects
119: of {\it homogeneous} electric fields in FE particles were investigated.
120: Homogeneous electric fields, possessing no curl, can only transform
121: a vortex phase into a ferroelectric phase of polarization. As a result,
122: the switching of toroid moment, which is the key subject of this study,
123: was not (and can not even be) addressed in Ref.\onlinecite{Naumov2}.
124:
125:
126: Technically, we use first-principle derived effective
127: Hamiltonian\cite{Zhong,Bellaiche} and finite temperature Monte
128: Carlo (MC) simulations. Besides
129: internal energy U, the Hamiltonian also consists of a
130: $-\sum_{i}{\bf E}({\bf r}_i)\cdot \mathbf{p}_{i}$
131: term,\cite{Garcia} describing the coupling between external
132: electric field $\mathbf{E}(\mathbf r)$ and local dipoles
133: $\mathbf{p}_{i}$. Nanoparticles of disk shape are considered, with
134: cylindrical {\it z}-axis parallel to the crystallographic [001]
135: direction. Switching simulations under curled electric fields are
136: performed at 64K where the vortex phase is stable. A fixed acceptance rate of 0.25 and
137: 10000 MC sweeps are used throughout all simulations.\cite{FN1}
138: We will mainly present results for a disk with
139: diameter d=19 and height h=14 (both quantities are measured in
140: lattice constant $a$=4{\AA} of pseudocubic bulk), since the diameter
141: (7.6nm) of this disk falls into the size range (5-50nm) of experimentally
142: fabricated nanoparticles\cite{Zhang}. In this work we use MC rather
143: than molecular dynamics (MD) to study switching dynamics,
144: based on the following considerations.
145: (i) In both methods the basic quantity governing the dynamics is the same, which is the energy.
146: The difference is that the energy is reflected in MD by
147: using forces, while in MC the energy itself is used in determination
148: of next configuration. (ii) Monte Carlo step can be quantified as a
149: ``quantum'' time, and the equivalence of two methods was rigorously
150: demonstrated in Ref.\onlinecite{Nowak}, where Monte Carlo
151: simulations and Langevin dynamics yield excellent agreement.
152: (iii) Vortex reversal in FE particles is essentially a thermally
153: activated process (which is slow). MD simulations will be time consuming. But
154: computations of MC simulations are moderate.
155: The method in Ref.\onlinecite{Zhong} is extended here to include
156: both static homogeneous {\bf E}$_{\rm h}$ electric field and
157: curled {\bf E}$_{\rm curl}$ field, i.e., $\mathbf E (\mathbf r) =
158: \mathbf E_{\rm h}+ \mathbf {E}_{\rm curl} (\mathbf r)$. The latter
159: is given as $\mathbf {E}_{\rm curl}=\frac{1}{2} \,\mathbf S \times
160: \mathbf{r}$, where $\mathbf{S}=S{\bf e}_z$
161: measures the vorticity of the field. With
162: the $\mathbf E_{\rm h}$ and $\mathbf {E}_{\rm curl}$
163: fields acting together, the interaction energy
164: $-\sum_{i}\mathbf{E(r_{i}})\cdot \mathbf{p}_{i}$ becomes
165: -N$\,(\mathbf{E_{\rm h}}\cdot \langle\mathbf{ p
166: }\rangle+\mathbf{S}\cdot \mathbf{G})$, where N is the number of
167: 5-atom cells in nanoparticles, $\langle\mathbf{ p }\rangle $ the
168: average dipole moment, $\mathbf{G}=\frac{1}{\rm 2N}\sum _i {\bf
169: r}_i\times {\bf p}_i$ the toroid moment. Vector $\mathbf S$ thus
170: is the thermodynamically conjugated field for toroid moment
171: $\mathbf G$.
172:
173: Figure \ref{FGm}a depicts the collective behaviors of the toroid moment {\bf G}, evolving
174: during simulation as a function of MC sweeps (denoted as $n$, in
175: units of 200), in the d=19 disk under a curled field of S=0.25
176: mV/{\AA}$^2$ that has an opposite vorticity with respect to the
177: initial FE vortex. One key finding in Fig.\ref{FGm}a is that the
178: vortex reversal process is predicted to consist of three evolution
179: phases in which the system displays distinct dipole behaviors.
180: These phases are, when MC sweep n is below n$_1$=30 (to be named
181: as phase I), n is between n$_1$=30 and n$_2$=43 (to be
182: referred to as phase II), and n$>$n$_2$ (phase III).
183: At the initial instant of the simulation, the vortex has only a
184: G$_z$ toroid moment, while the G$_x$ and G$_y$ components are
185: null. As the curled field is turned on, most notable results
186: occurring in Fig.\ref{FGm}a are: (1) The G$_z$ moment remains
187: remarkably stable and subjects to only marginal decrease
188: throughout phase I. However, the system does not idly wait;
189: another critical activity is taking place in phase I. That is,
190: the FE nanoparticle is building its G$_y$ moment which increases
191: appreciably and becomes very significant (with a value of 15
192: e{\AA}$^2$) in the end of phase I. (2) As the system enters phase
193: II, dramatic difference occurs. The G$_z$ moment starts to decline
194: sharply, reverses its direction, and then is stabilized at its
195: negative saturated value at n$_2$. (3) While G$_z$ switches its
196: direction in phase II, the G$_y$ moment undergoes a complex
197: evolution, by first rising (with a different slope than the
198: preceding process in phase I) and later dropping swiftly to null.
199: Precisely at the moment when G$_z$ is zero, G$_y$ reaches its
200: maximum magnitude.
201:
202: From simulations, the key toroid component during switching is
203: G$_z$, G$_y$ and -G$_z$ for the phases of n$<$n$_1$,
204: n$_1$$<$n$<$n$_2$, and n$>$n$_2$, respectively. The reverse of FE
205: vortex therefore occurs by undergoing an evolution sequence of
206: $G_z\rightarrow G_y\rightarrow (-G_z)$, in which appearance of the
207: lateral G$_y$ moment is a critical bridging process.
208: Formation of this lateral vortex is unusual (and puzzling), for
209: neither the initial vortex of the system nor the curl of the external
210: field have nonzero components in the lateral $xy$ directions
211: which are capable of causing the observed collective behaviors.
212: To verify the above switching sequence,
213: calculations with varied initial dipole configurations
214: or temperatures or slightly different {\bf E}$_{\rm curl}$ fields
215: were performed.
216: All confirm the occurrence of a lateral vortex as the bridging phase.
217: We also find that the toroid moment of the intermediate lateral vortex
218: can point at any direction within the $xy$ plane,
219: since $x$- and $y$-axis are equivalent. In simulations of Fig.\ref{FGm}a,
220: the lateral vortex happens to be mainly along the $y$-axis.
221:
222:
223: To explain why the G$_y$ toroid moment forms and how it influences
224: the reversal of the G$_z$ component, we now provide {\it
225: microscopic} insight on the field-induced dipole behaviors.
226: Fig.\ref{Fpat} shows the snapshots of local
227: dipole distributions corresponding to selected MC sweeps. More
228: specifically, Fig.\ref{Fpat}a-d describe the development of the
229: G$_y$ toroid moment, while Fig.\ref{Fpat}e-h show the evolution of
230: the inplane G$_z$ vortex. Our discussion shall follow the
231: sequential order of evolution.
232: Initial response to the curled electric field is subtle, and
233: begins with the dipoles located at the vortex core. Common
234: wisdom tells us that, under the {\bf E}$_{\rm curl}$ field, the
235: dipoles near the vortex core should flip into opposite directions
236: to form a -{\bf G} domain. Interestingly, this does not take place
237: in our simulations. We find instead that these dipoles rotate
238: towards the {\it z}-axis (Fig.\ref{Fpat}a). The dipole rotation
239: is, however, energetically less favorable, since it generates a
240: {\it z}-direction depolarization field that can not be compensated
241: by the curled field. Note that this situation is unlike the case when
242: a static homogeneous electric field is applied in the {\it z}-axis,
243: where the homogeneous field can overcome the existence
244: of the depolarization field. The delicate balance between the
245: internal depolarization field and the external curled field leads
246: to the dipole configuration in Fig.\ref{Fpat}a, where the
247: dipoles in the lower part point downwards as a result of the
248: curled field while the dipoles in the upper part point
249: horizontally to reduce the depolarization field. As evolution
250: proceeds to n=20 (Fig.\ref{Fpat}b), dipoles of the lower part
251: rotate also horizontally, but opposite to the dipole direction in
252: the upper part, to avoid lateral depolarizing field. Consequently,
253: the dipole pattern developed in Fig.\ref{Fpat}b explains how the
254: nucleation seed of lateral G$_y$ vortex is formed (which is
255: important). Another key feature of the dipole pattern in
256: Fig.\ref{Fpat}b is that the left and right arms of the G$_y$
257: vortex are largely absent. Interestingly, as the G$_y$ vortex is nucleated at
258: n=20, the inplane G$_z$ vortex is nearly intact (see the G$_z$ vortex in Fig.\ref{Fpat}e, also
259: at n=20), which is consistent with the result in Fig.\ref{FGm}a
260: where the G$_z$ moment remains robust throughout phase I.
261:
262: The system behaves differently when evolving into phase II,
263: highlighted in Fig.\ref{Fpat}f by the dramatic disappearance of
264: inplane dipole components in regions A and B. More precisely,
265: dipoles in region A rotate toward the negative {\it z}-axis, while
266: dipoles in region B toward the positive {\it z}-direction.
267: Disappearance of the inplane components for dipoles in A and B
268: regions leads to three marked effects: (i) the onset of a sharp
269: decline of the G$_z$ moment, as seen in Fig.\ref{FGm}a at the
270: beginning of phase II, (ii) the core of the G$_z$ vortex to shift
271: away from the cylindrical axis (Fig.\ref{Fpat}f), and (iii) a
272: rapid increase of the {\it z}-component magnitudes for those
273: dipoles located at the left and right arms of the lateral G$_y$
274: vortex, as revealed by contrasting Fig.\ref{Fpat}c with
275: Fig.\ref{Fpat}b. It should be recognized that nucleation of the
276: lateral G$_y$ vortex, right in the form of Fig.\ref{Fpat}b, is
277: critical for the destruction of the original G$_z$ vortex, since
278: the oppositely oriented horizontal dipoles at the upper and bottom
279: surfaces in Fig.\ref{Fpat}b produce only a small strain energy
280: when the dipoles in regions A and B rotate into the {\it z}
281: direction, and simultaneously, no significant increase in the
282: depolarizing field occurs along the {\it z}-axis.
283:
284: As the G$_z$ vortex continues to be uncurled, the G$_z$ moment
285: becomes zero at n=33 (see Fig.\ref{FGm}a). A vanishing G$_z$ moment
286: does not mean disappearance of the {\it xy} components
287: of all dipoles, however. Instead, calculation result in Fig.\ref{Fpat}g
288: reveals that the dipole components on the {\it xy} plane evolve,
289: interestingly, into a stripe pattern, which also yields zero
290: G$_z$. The stripe is formed due to the fact that the curled field
291: continues to push the G$_z$-vortex core in Fig.\ref{Fpat}f to the
292: (lower) surface. Starting with the stripe in Fig.\ref{Fpat}g,
293: dipoles in the upper part rotate and/or flip their directions,
294: forming the critical embryo of the {\it reversed} -G$_z$ vortex at n=34
295: (see Fig.\ref{Fpat}h). After the nucleation of this -G$_z$ vortex,
296: the lateral G$_y$ vortex starts to disassemble itself by rotating
297: its dipoles at both arms back to the {\it xy} plane (see
298: Fig.\ref{Fpat}d at n=40), which eventually leads to the
299: development of a full and reversed -G$_z$ vortex.
300:
301: From the above striking microscopic insight,
302: a novel mechanism that governs the switching of FE vortex is thus
303: obtained. That is, vortex reversal occurs, not by the coexistence
304: of G$_z$ and -G$_z$ domains, but by formation of a new and lateral
305: vortex. In fact, we find from the entire evolution process in Fig.\ref{Fpat} that, the G$_z$ and
306: -G$_z$ domains never coexist at any time in the simulations. Only
307: after the original G$_z$ vortex is dismantled does the nucleation
308: of the -G$_z$ vortex start to form (Fig.\ref{Fpat}h). We further
309: numerically find that the magnitude of the total toroid moment
310: $|{\bf G}|=\sqrt{G^2_x+G^2_y+G^2_z}$ is almost a constant
311: throughout the switching (see the diamond symbols in
312: Fig.\ref{FGm}a), demonstrating that vortex reversal occurs by
313: obeying the conservation of the $|{\bf G}|$ moment.
314:
315: We now study the energetics of vortex switching. Here we are
316: interested in internal energy U, rather than free energy F=U-$\sum
317: _i$ {\bf E}({\bf r}$_i$)$\cdot ${\bf p}$_i$, since the former
318: yields important knowledge concerning the energy barrier. The
319: internal energy is given in Fig.\ref{FGm}a, revealing new
320: observations that do not appear in collective behaviors. First,
321: the highest energy barrier occurs around n=30, which corresponds
322: to the instant when the G$_y$ vortex has formed and the inplane
323: G$_z$ vortex just starts to annihilate its {\it xy} components.
324: Height of the energy barrier is 2.4
325: meV per 5-atom cell. Second, and interestingly, a local energy
326: {\it valley} appears at n=33, showing that the fully developed
327: G$_y$ vortex (with G$_z$=0) is actually a stable structure even if
328: the curled field is removed. Third, the two energy barriers
329: surrounding this energy valley are asymmetric, with a lower
330: barrier of 0.8 meV at n=35 (favoring the formation of the -G$_z$
331: vortex).
332: To examine how the energy barrier may vary with temperature and
333: size, we have performed switching calculations at 160K (as compared
334: to 64K in Fig.\ref{FGm}a), for the same d=19 and h=14 dot. The energy
335: barrier is found to be 2.1 meV at 160K, close to the value of 2.4 meV
336: at 64K. This is consistent with the fact that the energy barrier is
337: determined mainly by depolarization field, not by temperature.
338: We have also performed calculations for other $d$ diameters,
339: all showing that formation of the lateral vortex is a key step
340: for switching the G$_z$ vortex.
341: We further find that depending on the $h/d$ ratio, the number of lateral
342: vortices---that occur at the switching moment (i.e., when G$_z$
343: equals zero)---differs. In the {\it d}=9 and {\it
344: h}=14 disk with $h/d \approx$2, two G$_y$ vortices and one G$_x$ vortex
345: (with dipole configuration similar to the structural B phase in
346: Ref.\onlinecite{Naumov1}) take place at the switching moment. The energy
347: barrier for the $d$=9 disk is found to be 1.7 meV at 64K.
348:
349: Interestingly, we previously found that {\it homogeneous} {\bf E}$_{\rm h}=E_h{\bf e}_z$
350: electric fields can also give rise to a lateral vortex\cite{Naumov2},
351: despite that homogeneous fields and curled fields are drastically
352: different in nature.
353: Based on the key findings that
354: (1) the {\bf E}$_{\rm h}$ and {\bf E}$_{\rm curl}$ fields both are
355: capable of inducing the G$_y$ vortex, and (2) formation of the
356: G$_y$ vortex is the major energy barrier for toroid switching (as
357: demonstrated in this study), we further propose an alternative
358: approach to switch FE vortex, by means of a combined
359: action of the {\bf E}$_{\rm h}$ and {\bf E}$_{\rm curl}$
360: fields. More precisely, an {\bf E}$_{\rm h}$ field of sufficient
361: magnitude is applied to precipitate the formation of G$_y$ vortex,
362: and an {\bf E}$_{\rm curl}$ field is then added on (while the {\bf
363: E}$_{\rm h}$ field remains) to generate the reversed G$_z$ vortex.
364: A significant advantage of this approach is, while a strong curled
365: field is hard to achieve since it requires fast oscillation of
366: alternating magnetic fields, a strong {\bf E}$_{\rm h}$ field of
367: 10$^9$V/m order can be easily generated on nanometer scale.
368: We demonstrate the approach using the same d=19 disk, by
369: placing it first under a homogeneous E$_{\rm h}$=1.9V/nm field
370: (which quickly induces a G$_y$ vortex), and then under a curled
371: E$_{\rm curl}$=0.04mV/{\AA}$^2$ field for ten thousand MC sweeps.
372: Fig.\ref{FGm}b describes the evolution of the toroid moment after
373: the curled field is turned on. Drastic differences are evident in
374: Fig.\ref{FGm}b as compared to Fig.\ref{FGm}a: (1) When the curled
375: field is switched on, the G$_y$ value in Fig.\ref{FGm}b is already
376: notably large. This G$_y$ moment is formed by the homogeneous
377: field, confirming the result of Ref.\onlinecite{Naumov2}. (2) The
378: long period of phase I---that lasts five thousand MC sweeps in
379: Fig.\ref{FGm}a---no longer exists in Fig.\ref{FGm}b. Instead,
380: G$_z$ starts to decline sharply in Fig.\ref{FGm}b as soon as the
381: curled field is imposed. (3) The magnitude of the curled field
382: used in Fig.\ref{FGm}b is considerably reduced, by a factor of
383: 600\% as compared in Fig.\ref{FGm}a. This reduction factor can be
384: further increased when a larger {\bf E}$_{\rm h}$ field is used.
385: Analysis of the internal energy (solid symbols in Fig.\ref{FGm}b)
386: shows that the energy barrier almost disappears, and the curled
387: field acts to guide the formation of the reversed vortex.
388:
389: In summary, response of FE particles to curled electric fields was
390: studied for the first time. Simulations revealed a novel mechanism
391: that governs the switching of FE toroid vortex. A curled electric
392: field, of only {\it z}-axis vorticity momentum, was shown to lead
393: to the unexpected formation of a lateral vortex, manifesting {\it
394: macroscopically} the balance of various {\it microscopic}
395: interactions. We demonstrated that the strong correlation
396: between the original vortex and the new lateral vortex plays a critical
397: role for vortex reversal. On the other hand, the -{\bf G} and {\bf
398: G} domains never coexist during switching, and the switching
399: mechanism by coexisting domains is not valid. Moreover, based on the
400: findings that formation of lateral vortex is the major energy barrier
401: for vortex switching, and that this lateral vortex can also occur
402: by use of homogeneous field, we described an effective
403: approach that is able to reduce substantially (by 600\%) the magnitude of the
404: switching curled field. Finally, we believe that the microscopic
405: insight and energetics (which are difficult to obtain in
406: experiments) will also be of immense value.
407:
408:
409: \begin{thebibliography}{1}
410: \bibitem{Sai} N. Sai, B. Meyer, and D. Vanderbilt, Phys. Rev.
411: Lett. {\bf 84}, 5636 (2000).
412: \bibitem{Neaton} J.B. Neaton and K.M. Rabe, Appl. Phys. Lett.
413: {\bf 82}, 1586 (2003).
414: \bibitem{Weissman} M.P. Warusawithana {\it et al.} Phys. Rev.
415: Lett. {\bf 90}, 036802 (2003).
416: \bibitem{Lee} H.N. Lee {\it et al.}, Nature {\bf 433}, 395 (2005).
417: \bibitem{Fong} D.D. Fong {\it et al.}, Science {\bf 304}, 1650
418: (2004).
419: \bibitem{Naumov1} I.I. Naumov, L. Bellaiche, and H. Fu, Nature
420: {\bf 432}, 737 (2004).
421: \bibitem{Ginzburg} V.L. Ginzburg {\it et al.}, Solid State Commun. {\bf 50}, 339
422: (1984).
423: \bibitem{kittel} C. Kittel, Phys. Rev. {\bf 70}, 965 (1946).
424: \bibitem{mermin} N.D. Mermin and H. Wagner, Phys. Rev. Lett.
425: {\bf 17}, 1133 (1966).
426: \bibitem{thouless} D.J. Thouless, Phys. Rev. {\bf 187}, 732 (1969).
427: \bibitem{anderson} P.W. Anderson, G. Yuval, and D.R. Hamann,
428: Phys. Rev. B {\bf 1}, 4464 (1970).
429: \bibitem{bruno} P. Bruno, Phys. Rev. Lett. {\bf 87,} 137203 (2001).
430: \bibitem{Gamb} P. Gambardella {\it et al.}, Nature {\bf
431: 416}, 301 (2002).
432: \bibitem{landau} L.D. Landau and E.M. Lifshitz,
433: {\it Statisical Physics, Part 1} (Pergamon, NY, 1980).
434: \bibitem{Lines} M.E. Lines and A.M. Galss, {\it Principles and
435: Application of Ferroelectrics and Related Materials} (Clarendon,
436: Oxford, 1979).
437: \bibitem{Garcia} A. Garcia and D. Vanderbilt, Appl. Phys. Lett. {\bf
438: 72}, 2981 (1998).
439: \bibitem{Samara} G.A. Samara, in {\it Solid State Physics} (ed. H. Ehrenreich
440: and F. Spaepen), Vol. 56, p 240, and references therein.
441: \bibitem{Wu} Z. Wu and R.E. Cohen, Phys. Rev. Lett. {\bf 95},
442: 037601 (2005).
443: \bibitem{Cohen} R.E. Cohen, Nature {\bf 358}, 136
444: (1992); R.E. Cohen and H. Krakauer, Phys. Rev. B {\bf 42}, 6416 (1990).
445: \bibitem{LB1} S. Prosandeev {\it et al.}, Phys. Rev. Lett.
446: {\bf 96}, 237601 (2006).
447: \bibitem{Merz} W.J. Merz, Phys. Rev. {\bf 95}, 690 (1954).
448: \bibitem{Naumov2} I. Naumov and H. Fu, Phys. Rev. Lett. {\bf 98}, 077603 (2007); cond-mat/0612549.
449: \bibitem{Zhong} W. Zhong, D. Vanderbilt, and K. M. Rabe,
450: Phys. Rev. Lett. {\bf 73}, 1861 (1994).
451: \bibitem{Bellaiche} L. Bellaiche, A. Garcia, and D. Vanderbilt,
452: Phys. Rev. Lett. {\bf 84}, 5427 (2000).
453: \bibitem{FN1} Using the same acceptance rate of 0.25, we find that the
454: bulk polarization in Pb(Zr$_{0.5}$Ti$_{0.5}$)O$_3$ is switched
455: within ten thousand MC sweeps.
456: \bibitem{Zhang} C. Liu, B. Zou, A.J. Rondinone, and Z.J. Zhang,
457: J. Am. Chem. Soc. {\bf 123}, 4344 (2001).
458: \bibitem{Nowak} U. Nowak, R.W. Chantrell, and E.C. Kennedy, Phys. Rev. Lett.
459: {\bf 84}, 163 (2000).
460: \end{thebibliography}
461:
462:
463: \begin{figure}
464: \centering
465: \includegraphics[scale=1.6]{Fig1.eps}
466: \caption{Evolution of toroid moment {\bf G} and its magnitude
467: $|{\bf G}|$ (empty symbols, using the left vertical axis) and
468: internal U energy per 5-atom cell (solid dots, using the right
469: vertical axis) in a d=19 nanodisk: (a) under a S=0.25mV/{\AA $^2$}
470: curled electric field; (b) under the combined action of a E$_{\rm
471: h}$=1.9V/nm homogeneous field and a S=0.04mV/{\AA $^2$} curled
472: field. For clarity of display, G$_z$ is plotted after multiplying
473: -1.} \label{FGm}
474: \end{figure}
475:
476: \begin{figure}
477: \centering
478: \includegraphics[scale=1.55]{Fig2.eps}
479: \caption{Dipole configurations in a d=19 nanodisk under a curled
480: field of S=0.25mV/{\AA $^2$} at different MC sweeps. Upper panel
481: (a-d): dipoles on the central {\it xz} cross section revealing the
482: evolution of the G$_y$ vortex, at the following MC sweeps (a)
483: n=10, (b) n=20, (c) n=30, (d) n=40. Lower panel (e-h): dipoles on
484: the central {\it xy} cross section displaying the evolution of the
485: G$_z$ vortex, at (e) n=20, (f) n=30, (g)
486: n=33, (h) n=34. All (n) numbers are in units of 200 MC sweeps.
487: Arrows show the magnitudes and directions of local dipoles,
488: projected on the corresponding plane. For convenience, (b) and (e)
489: corresponding to the same n=20 sweep are given on the same column
490: [so are (c) and (f)].} \label{Fpat}
491: \end{figure}
492:
493:
494: \end{document}
495: