cond-mat0205219/text
1: %file  Tadic_Revtex file
2: %\documentstyle[aps,preprint,epsf]{revtex}
3: \documentstyle[aps,prl,multicol,epsf]{revtex}
4: %\documentstyle[aps,prl,twocolumn,epsf]{revtex}
5: %\tighten
6: 
7: \begin{document}
8: 
9: \title{Switching current noise and relaxation of ferroelectric domains }
10: \author{Bosiljka Tadi\'c }
11: 
12: \address{Jo\v{z}ef Stefan Institute,
13: P.O. Box 3000, 1001 Ljubljana, Slovenia  }
14: 
15: %\date\today
16: 
17: \maketitle
18: \begin{abstract}
19: \newline
20: We simulate field-induced  nucleation and switching of domains in a
21:  three-dimensional  model of ferroelectrics with quenched  disorder
22: and varying domain sizes. We study (1) bursts of the switching current
23: at  slow driving  along the  hysteresis loop  (electrical Barkhausen noise)
24:  and (2) the  polarization reversal when a strong electric field was
25:  applied and  back-switching after the field was removed.
26: We show how these processes  are related to the  underlying structure of
27: domain walls, which in turn is  controlled by the  pinning at quenched
28: local electric fields.
29:  When the depolarization fields of bound charges are properly screened
30: we find that the fractal switching current noise may appear with two
31: distinct universal behaviors. The critical depinning of plane domain walls
32: determines the universality class in the case of  weak random fields,
33: whereas for large randomness the  massive nucleation of
34: domains in the bulk leads to different scaling properties.
35:  In both cases the scaling exponents decay logarithmically  when the
36: driving frequency is increased.
37: The polarization reverses in the applied  field as a power-law, while
38: its relaxation in zero field  is a stretch exponential function of time.
39: The stretching  exponent depends  on the strength of pinning. The results
40: may be applicable for uniaxial relaxor ferroelectrics, such as doped SBN:Ce.
41: 
42: 
43: \end{abstract}
44: \pacs{PACS numbers: 77.80.Dj, 77.80.Fm, 05.40.Ca}
45: 
46: %\narrowtext
47: \begin{multicols}{2}
48: 
49: 
50: 
51: \newpage
52: 
53: \section{Introduction}
54: 
55: 
56: Polarization reversal and  properties of the accompanying  switching
57: current  are relevant  for various applications of ferroelectric materials.
58: In particular, it is a key process that takes place in electronic devices
59: based on ferroelectrics thin films, where a reproducible switching process is
60: required.
61: Apart from the extensive experimental and theoretical work
62: \cite{Rudyak,book,Shur_rev}, a complete understanding of the domain-wall
63: motion and predictability of the switching processes remains elusive in
64: real ferroelectrics, especially  when several physical parameters,
65: determining either structure or dynamics, are varied.
66: Domain switching and relaxation (back-switching) in ferroelectrics can be
67: studied by hysteresis loop, switching current, and switching current
68: bursts (electrical Barkhausen noise).
69: 
70: 
71: 
72: The development of domain structure and motion of domain walls  during
73: switching can be regarded in the scope of nonlinear dynamic  processes
74: occurring in driven disordered systems, which are studied extensively in
75: the past decade.
76: Particularly, the measurements on ferromagnets (a short review is given in
77: \cite{BT_summary}) show that Barkhausen noise exhibits interesting fractal
78: properties, that can be characterized by a full set of scaling exponents
79: of the sizes, durations and energies of the bursts and of the power spectrum.
80: An example with a  complete set of measured scaling exponents and
81: corresponding fractal dimensions with  derivation of the associated
82: scaling relations can be found in Ref.\ \cite{Djole}.
83: A possible grouping of the scaling exponents into different
84: universality classes of scaling behavior  offers a new and
85: simple link between the  measured Barkhausen  noise
86: and the domain structure. For example, in  a recent study  \cite{packing}
87: such a  relation was established  between the universal scaling exponents of
88: acoustic emission noise and the packing properties near the first-order
89: structural transformation.
90: 
91: 
92:  The electrical Barkhausen noise  has been observed experimentally in
93: ${\mathrm{BaTiO_3}}$ \cite{Rudyak,book,BTiO3} and
94: ${\mathrm{Gd_2(MoO_4)_3}}$ and ${\mathrm{LiTaO_3}}$
95: \cite{Shur,Shur_et_al}  single crystals and
96: in  ${\mathrm{9.4/65/35 (Pb,La)(Zr,Ti)O_3}}$ (PLZT) ceramics
97: \cite{ceramics} and  ${\mathrm{PbMg_{1/3}Nb_{2/3}O_3}}$ (PMN)
98: relaxor ferroelectric \cite{Weissmann}.
99: Physically, the polarization reversal and domain wall motions in
100: ferroelectrics is accompanied with the local change of sign of bound  charges,
101: which influences values of the electric field in the sample. The
102: depolarization field of bound charges can be partially screened by an
103: adequate addition of charges on the surface and by
104: other methods (see \cite{Shur_rev} and references therein).
105: Hence at a given point $x$ in the sample at time $t$ un-compensated
106: depolarization field is given by
107: \cite{Shur_rev}
108: \begin{equation}
109: \delta E_{dep}(x,t)=
110: E_{dep}(x,t)-\sum E_{sreen}(x,t)\ ,
111: \label{eff_field}
112: \end{equation}
113: where $\sum E_{sreen}(x,t)$ is a total field due to screening (surface and
114: bulk)  charges at that site. Incomplete screening of the depolarization field
115: tends to change  the electric field inside the sample, thus
116: influencing the course of switching process itself. For instance, there is
117: a finite probability that inverse jumps of polarization (negative
118: Barkhausen pulses) \cite{Rudyak,Shur_rev} occur, that slow down the switching.
119: Recently a series of sophisticated measurements  were done
120: \cite{Shur_et_al,FTT} that reveal the effects of the depolarization field on
121: a  created plane domain wall in ${\mathrm{Gd_2(MoO_4)_3}}$ and on a
122: multi-domain structure in ${\mathrm{LiTaO_3}}$.
123: The obtained  distributions of lengths, sizes and rest time of Barkhausen
124: jumps obey a  power-law behavior with an exponential cut off, both for
125: switching and back-switching processes.
126: Using a suitable experimental set-up  the power-spectrum in PMN relaxors
127: was  measured, which was
128: found to  decay with frequency $\omega $ as  $S(\omega) \sim \omega^{-2}$
129: \cite{Weissmann}.
130: 
131: 
132: 
133: Recently studied uniaxial relaxor ferroelectrics
134: ${\mathrm{Sr_{0.61}Ba_{0.39}Nb_2O_6}}$  (SBN61) and
135: ${\mathrm{Pb_{0.61}Ba_{0.39}Nb_2O_6}}$ are regarded as
136: systems with the Ising symmetry of the order parameter,  having the tetragonal
137: structure  in the low-temperature phase with the spontaneous polarization
138: along c-axis  \cite{Dui2,Peer}.
139:  In solid solutions ${\mathrm{SBN61:X}}$ doping with ${\mathrm{X=Ce^{3+},
140: Cr^{3+},  Co^{2+}}}$ reduces the transition temperature and induces
141:  local electric random fields, which give rise to local polar clusters
142: \cite{Peer}.
143: Recently a comprehensive series of measurements in SBN:Ce and similar
144: systems \cite{Dui2,Peer,Kleemann}  gave a strong support  to
145: the presence of random electric fields in this class of relaxor
146: ferroelectrics.
147: In general, more common sources of uncorrelated random fields in
148: ferroelectrics are neutral defects that can locally break the polar symmetry,
149: as for instance in TGS (see detailed discussion in Ref.\ \cite{book_AL},
150: page 149).
151: [Note that, in contrast to ferroelectrics,  in ferromagnets local
152: random fields are not
153: allowed by the symmetry, but are theoretically constructed as being induced
154: by coarse graining due to other types of disorder (see also \cite{BT_letter}).]
155: 
156: 
157: 
158: 
159: 
160: 
161: 
162: In this work we simulate  temporal behavior of polarization driven by the
163: electric field  in three-dimensional systems of Ising spins subject to
164: {\it quenched} local random  fields and  positive nearst-neighbor
165: interaction, which is
166: compatible with the occurrence of a ferroelectric long-range order.
167:  We consider two types of driving.  First, by stepwise increase
168: (decrease) of the external field along a
169: hysteresis loop (see below) we study  bursts in the switching current
170: $j(t) = dP(t)/dt$
171: (electrical Barkhausen noise). We discuss effects of incomplete screening
172: of depolarization fields on the coarse of switching and hysteresis loop
173: properties. When the screening is complete, we study  how the
174: statistical properties of the noise depend on the strength of
175: pinning and on the frequency  with which the driving field is  changed.
176: Second,  by imposing a high opposite value of the external field in the
177: homogeneous state we let the polarization to reach its saturation at
178: that field, and then remove the field. We determine
179: reversal and spontaneous back-switching of the polarization with time at
180: different strength of disorder.
181: This type of switching is  more familiar in the experiments
182: on ferroelectrics \cite{Kleemann}.
183: 
184: 
185: 
186: 
187: 
188: 
189: 
190: Apart from allowing negative Barkhausen pulses as described below,
191: the present study is complementary compared to earlier simulations of
192: Barkhausen pulses in 3-dimensional Ising systems in the presence of disorder
193: \cite{BT_ferroel,BCN,Cornel,Zheng} in the following sense: In the earlier
194: work  \cite{BT_ferroel} we considered only strong disorder that is
195: compatible with
196: nucleation of small domains under driving  field, as for instance in
197: relaxor ceramics, and  analyzed the signal from only central part
198: of the hysteresis loop, as it is usually done on experiments.
199: In the present work  the study is extended to entire range of disorder
200: including the weak pinning region where extended domain walls may occur
201: due to lack of pinning centers \cite{comment_planarDW}.
202: In this case a potential depinning of domain walls at a critical driving
203: field contributes significantly to the scaling behavior of Barkhausen
204: pulses, as recent theoretical results on Bethe lattice suggest \cite{DD}.
205: Therefore, we include integration over entire hysteresis loop
206: in order to incorporate the contribution of critical depinning to the
207: universal scaling properties of the  induced electrical
208: noise in the case of  weak pinning \cite{comment_integration}.
209: Furthermore, here we employ the {\it simulation algorithm at finite driving
210: rates}, that mimics  closely the experimental situation with stepwise
211: increase (decrease) of the electric field.
212: We examine in detail the effects of driving rates in both strong and
213: weak pinning case by varying height of the steps in the external field.
214: The earlier simulations  in Refs.\ \cite{BCN,Cornel} were done in the
215: limit of zero driving rate, which appears physically inaccessible to
216: real experiments.  Whereas an entirely different
217: aspect concentrating on the  dynamic scaling during the growth of {\it
218: individual} avalanches was  considered in \cite{Zheng}.
219: It should be stressed that in this paper we implement strictly {\it quenched
220: disorder}, as described in Section II. Note that some speed algorithms,
221: as for instance used in  \cite{Cornel,Zheng}, allow certain  dynamic
222: variations of pinning  while the avalanche grows, which  can be  relevant
223: to  systems with annealed  rather than quenched disorder.
224: Theoretically, having unfixed pinning centers during the avalanche growth
225: may alter the long-range correlations in the avalanche-triggering fields
226: and in distances between initial points of avalanches,  which were shown
227: to be closely associated with the appearance of the fractal noise in
228: two-dimensional systems  \cite{correlations}.
229: 
230: 
231: 
232: 
233: The paper is organized as follows. In Sec.\ II we introduce the model and
234: discuss all relevant parameters that influence the hysteresis loop properties.
235: In Sec.\ III we study in detail scaling behavior of the switching current
236: noise under varying disorder and driving conditions.  Sec.\ IV contains the
237: results  of switching in a high field and
238: spontaneous back switching  of polarization when the pinning is varied.
239: The paper contains  a short summary of the results and discussion in Sec.\ V.
240: 
241: 
242: \section{Model and relevant parameters }
243: 
244: We consider a model with Ising-type  spins situated on three-dimensional
245: simple cubic lattice with positive interactions between neighbor spins and
246: local random field at each lattice site \cite{BT_letter}:
247: \begin{equation}
248: {\cal{H}} = -\sum _{x,y}J_{xy}S(x)S(y) - \sum _xh_xS(x)  -E\sum _xS(x)\ .
249: \label{Hamiltonian}
250: \end{equation}
251: We assume that the random exchange interactions $J_{x,y}$ are
252: distributed around a
253: positive value $J$ with a narrow distribution. The random fields $h_x$
254: are given by Gaussian distribution  with  zero mean and the width $f$
255: (measured in units of $J^2$). As mentioned above, the source of local
256: random fields can be  neutral defects that break polar
257: symmetry \cite{book_AL} or charge disorder due to
258: doping \cite{Peer},  which cause occurrence of polar regions in
259: relaxor ferroelectrics.
260: In addition, a random part of the interaction between these
261: polar regions  contributes to the observed glassy behavior of the
262: dielectric response \cite{Rasa}. Since the random
263: fields locally break the symmetry of the order parameter, we believe
264: that the dominant disorder effects on the
265: dynamics that we consider are due to random fields. Therefore,  here we
266: neglect the randomness in the spin-spin interaction.
267: 
268: 
269: 
270: The spin dynamics that we are considering is field-assisted and takes part
271: deeply in the ordered phase. Hence, we neglect possible temperature
272: effects, both for the reasons of keeping the number of parameters finite
273: and assuming that they play a secondary role in these processes.
274: The dynamics consists of spin alignment along a locally dominant field.
275: Thus, when the local field at site $x$ and time $t$, $h_{loc}(x,t)$ exceeds
276: zero, the spin $S(x,t)$ at that site flips and thus aligns along the local
277: field. The local field  $h_{loc}(x,t)$ consists of the interaction and pinning
278: part  $h_{ip}(x,t)= \sum_{y}J_{xy}S(y,t) +h(x)$  and
279: the driving field  $E(x,t)$. The driving field itself is  given by
280: $E(x,t) = E_{ext}(t) - \delta E_{dep}(x,t) $, where  $E_{ext}(t)$ is
281: the external field  and $\delta E_{dep}(x,t) $ a non-compensated
282: part of the depolarization field at that site \cite{Shur_rev}.
283: We employ a slow stepwise increase of the  external field
284: $E_{ext}(t_i)=-E_{sat}+ t_i\Delta E$ at discrete time intervals $\{t_i\}$,
285: while assuming that a non-compensated
286: part of the depolarization field at site $x$ may assist a probabilistic
287: (probability parameter $b$)  back-flip of an already aligned spin at that
288: site.
289: It should be stressed that the parameter $b$ represents the probability
290: that at a site $x$ in time $t$ the depolarization field opposes the effect
291: of external filed.
292: (The situation where the  depolarization field is parallel to the external
293: field is less interesting within the present numeric implementation of
294: the dynamics where driving rate is not fixed, as described below.)
295: Having in mind that the depolarization field at a given site depends on
296: the actual domain structure and its kinetics, the uniform distribution of
297: the stochastic variable $b$ can be regarded only as a first step towards
298: modeling of this complex phenomena.
299: 
300: 
301: 
302: 
303: 
304: 
305: Implementation of the dynamics is as follows:
306: Quenched random fields of Gaussian distribution (in double precision)
307: are generated and stored. They are {\it kept constant until entire
308:  hysteresis loop is completed} by the driving field. The external field is
309: increased by a small amount $\Delta E$ and the set of local fields
310: $\{h^{loc}(x,t)\}$ computed at each lattice site  and stored.
311: The system is updated {\it in parallel}, i.e., with respect to stored
312: set of local fields, and the spin flipped when
313: $h^{ip}(x,t) + E_{ext}(x,t)$ exceeds zero by a small amount $ 10^{-10}$.
314: When it is assumed that the depolarization  field is not entirely compensated,
315: the spin flip is executed with a reduced probability $1-b$.
316: This completes one time step of avalanche evolution.
317: Then a new set of local fields at affected sites (neighbors of just flipped
318: spins) is computed and stored. The conditions for spin flips at those sites
319: examined and spins flipped according to the above rule.
320: The process continues  until no more spins are found which satisfy the
321: conditions for flip. This completes one
322: Barkhausen avalanche. Then the external field is increased again by
323: the amount $\Delta E$.
324: When the  whole loop is completed, we set another distribution of quenched
325: random fields and repeat the driving. The results are averaged over the entire
326: ensemble (up to 200) of different random-field realizations.
327: 
328: In this model the following relevant parameters appear: Width of the
329: random-field distribution $f$, that determines strength of pinning and/or
330: domain size; jump $\Delta E$ of the external field, which  defines the
331: driving rate; probability of back-flips $b$,  which is related to the
332:  unscreened  part of the depolarization  field. In practice,
333:  usually an additional parameter is due to the section of
334: the hysteresis loop at which the signal is monitored, since the properties of
335: the signal vary along the loop. As mentioned above, here we integrate the
336: statistics over the entire hysteresis loop in order to capture the
337: contribution of the depinning of large domain walls to the scaling behavior
338: of noise by  weak pinning.
339: 
340: 
341: 
342: 
343: 
344: 
345: 
346: The hysteresis loop dependence on pinning by random fields and
347: on driving rates  is shown in Fig.\ 1. By varying the strength of
348: pinning $f$  the form of the hysteresis loop changes (Fig.\ 1a) from the
349: rectangular loop at low disordered 3-d sample  to the
350: slim loop, when the pinning is strong (large $f$). In the case of low
351: disorder, i.e., at weak pinning of domain walls, large domains may form
352: due to lack of pinning centers. In real ferroelectrics occurrence of
353: large domains are usually accompanied by extended plane domain walls
354: \cite{BTiO3,Shur,comment_planarDW}.
355:  On the other side, when the pinning is strong,
356: such as in the case of pinning at grain boundaries of small grain
357: ceramics \cite{ceramics}, many small domains can be nucleated at spatially
358: distributed centers with weak pinning, and their walls can propagate
359: along few active sites with weakest pinning, as the driving field increases.
360: This makes  effectively low-dimensional   walls moving in a
361: three-dimensional sample. In addition, a massive
362: nucleation of domains in the central part of hysteresis imposes  a spatial
363: restriction on the growth of later domains. Hence we expect a different
364: scaling behavior of sizes and durations of induced switched domains in
365: the case of weak and strong disorder. Our simulations confirm this
366: phenomenological picture.
367: In ferroelectrics a continuous variations  of the size of domains can be
368: achieved by increasing temperature or by applying  stress, as for
369: instance described
370: in  8/65/35 PLZT ceramics \cite{PLZT-HL}. It was suggested in Ref.\
371: \cite{Kleemann} that the appropriate illumination of samples can
372: notably reduce the strength of pinning in doped relaxor
373: ${\mathrm{SBN61:Ce^{3+}}}$ ferroelectrics.
374: 
375: 
376: We would like to stress that in the algorithm  keeping the external
377: field constant during the
378: avalanche evolution and increasing it only  {\it after} the avalanche
379: has stopped does not allow to study rest periods between avalanches.
380: However, this type of field update leads to a better resolution of the noise,
381: while it changes the time dependence of the driving field (cf. Fig.\ 1c).
382: The driving rate can be better defined as the average derivative
383: $r\equiv <dE(t)/dt>$ along the loop.
384: In the simulations we find that  when back flips occur with a finite
385: probability  $b>0$ they effectively reduce increase of the driving field
386: and prolong the duration of a jump, as well as inducing inverse jumps,
387: as shown in  Fig.\ 1 c,d.
388: 
389: 
390: Assuming that the back flips occur very rarely ($b\to 0$),
391: we obtain the switching current noise which is shown in Fig.\ 2.
392: The train of bursts of  the switching current exhibits a long-range
393: correlations---Fourier spectrum $j(\omega )$ of the
394: signal is shown in Fig.\ 2 (top).  It is related to  the power-spectrum
395: via $S(\omega )=|j(\omega )|^2$, that can be directly measured
396: (see \cite{Weissmann}).
397: Hence, the exponent of the power spectrum in the present model is
398: given by $2\times \phi =1.48\pm 0.02$ for the case of moderate pinning
399: $f=3.6$. The power-law behavior
400: of the power spectrum suggests that the switching current noise exhibits
401: fractal character when  certain conditions are met, in particular, when the
402: depolarization fields are successfully screened.
403: 
404: 
405: 
406: 
407: 
408: 
409: \section{Fractal properties of the switching-current noise }
410: 
411: The fractal noise of the switching current have a range of universal
412: scaling properties, that can be determined more clearly if the field
413: inside the sample is fully controlled. Therefore, in this section we
414: put $b=0$ in order to study in detail these scaling
415:  properties and to determine how they depend on two relevant
416: parameters---the domain size and the driving rate.
417: 
418: 
419: \subsection{Varying strength of pinning}
420: 
421: First we fix the driving rate to a small value $\Delta E/J=1\times10^{-3}$
422: and consider how the domain size or strength of pinning influence the
423: properties of the switching current.
424: In Fig.\ 3 we show the cumulative  distributions of size (area) $s$ of
425: switched domains  and duration $t$ of switching following a single
426: field update, for several values of  pinning strength $f$.
427: As the Fig.\ 3 shows, both switched areas and durations of their switching
428: show a wide range of values that are described by  power-law
429: distributions with finite cut-offs ($X$ stands for  $X\equiv s,t$):
430: \begin{equation}
431: P(X,L) = X^{-(\tau_X-1)}{\cal{P}}(XL^{-D_X}) \ .
432: \label{DX}
433: \end{equation}
434: The scaling exponents $\tau _s-1$ and $\tau _t-1$ determine
435: the respective slopes of the cumulative distributions of  switched area
436:  and duration of switching (see Fig.\ 3).
437:  The fractal dimension $D_s$ of size of switched domains
438: and the  dynamic exponent  $D_t\equiv z$ in Eq.\ (\ref{DX})
439: determine how the  corresponding cut-offs scale with the characteristic
440: length (here the system size $L$),  $X_0 \sim L^{D_X}$.
441: 
442: 
443: Both the slopes $\tau_X$ and the cut-offs, related to $D_X$, of the
444: distributions depend on the strength of pinning.
445: First, for weak pinning (small values of $f$, bottom curve in Fig.\ 3) we
446: notice that extended domains, bounded by  presumably plane  domain walls,
447:  may occur.
448: By increasing the field these  domain walls move through a weakly
449: random medium and eventually
450: depin at a critical field $H_d(f)$. Depinning of large domain walls leads to a
451: peak in the histogram at large values of $s$, or a  plateau in the
452: cumulative distribution as shown
453:  in Fig.\ 3. Before the depinning is reached a series of small avalanches
454: occurs that are power-law distributed.
455: The plateau disappears at a critical value of pinning strength
456: $f_c(r,L)\approx 2.4$,
457: where a pure power-law distribution with virtually infinite cut-off (controlled
458: only by the system size) occurs. Here the measured slopes give $\tau_s=1.74$
459: and $\tau_t=2.14$ for the fixed driving rate $\Delta E/J=1\times 10^{-3}$
460: (see  Fig.\ 3).
461: 
462: 
463: 
464: 
465: At strong  pinning (large $f$) many small domains are being nucleated at
466: isolated points in the bulk and grow slowly along a few active sites per time
467: step.  The maximum sizes of domains are finite
468: and decrease with stronger pinning  (cf. top two curves in
469: Fig.\ 3). Here the measured slopes
470: give $\tau_s=1.56$ and $\tau_t=1.94$ for size and duration,
471: respectively, suggesting another universality class  of the scaling
472: behavior, compared to the case of weak pinning.
473: The transition to the strong pinning regime for $f > f_c$ is marked
474: by disappearance of the critical depinning of large domain walls at all
475: fields and domination of the nucleation processes.
476: Numerical data in Fig.\ 3 suggest that the critical value  $f_c$ is close to
477:  2.4 for the used  system size and driving rate.
478: 
479: Qualitatively similar behavior with two universality classes but with
480: different exponents, was found in the case of two-dimensional system
481: where an extended one-dimensional domain wall was initially prepared
482: \cite{TN}. For comparison, in Table\ 1 we summarize exponents both for
483: 3-dimensional and for 2-dimensional case. (Notice that the listed
484:  exponents for three-dimensional system are obtained for a small finite
485: driving rate $\Delta E/J=4\times  10^{-4}$, whereas the simulations at zero
486: driving rate were  employed in two dimensions.)
487: The results for  2-dimensional case are expected to be
488: relevant for domains on the surfaces of the bulk samples (see Ref.
489: \cite{Peer}) and for  thin films in which the width of the domain wall
490: exceed the film thickness. An interesting physical realization  of
491: varying strength of pinning was suggested recently \cite{Co_film} in
492: ferromagnetic ${\mathrm C_o/C_oO}$ film by varying temperature through the
493: transition temperature at which the substrate orders
494: {\it anti-}ferromagnetically.
495: As already mentioned,  in ferroelectrics the local random fields are physical
496: \cite{Dui2,Peer,Kleemann,book_AL}. Therefore varying the strength of
497: pinning due to random fields in
498: ferroelectric thin films and bulk materials  can be achieved in a more
499: direct way, offering  diverse possibilities to detect the change in the
500: scaling behavior of the switching current noise and  critical properties
501: on the hysteresis.
502: 
503: 
504: 
505: 
506: 
507: 
508: 
509: \subsection{Varying driving rate}
510: 
511:  In the following we examine how the scaling properties of the switching
512: current noise change when the driving  rate is increased.
513:  Very low frequencies of the driving field are usually not accessible on
514: experiments, both for physical and technical reasons. A finite driving
515: rate, which is defined as $r\equiv <dE(t)/dt>$, can be approximated with
516: $\Delta E/J$ in the case of stepwise increase of the external field in the
517: absence of back flips $b=0$. For finite rate $r>0$ the switching may start
518: at different parts of the sample and the measured signal is a superposition
519: of these separate events. In contrast to the mathematical limit of
520: infinitely slow driving, by  finite field jump $\Delta E$  the domain wall
521:  may get enough energy to overcome several  pinning centers at the same time.
522: Thus either different small domains are initiated simultaneously, or
523: different parts of a large domain wall advance
524: in parallel. In both cases the statistical properties of the signal are
525: changed. Merging  of many  individual jumps makes the occurrence of large
526: events  more probable.
527: 
528: For a range of values of driving rates we still find a power-law behavior
529: with reduced values of the  scaling exponents and enlarged cut-offs.
530: In Fig.\ 4 we show how the distributions for sizes and durations of pulses
531: vary when the driving rate $\Delta E/J$ is increased, while the strength of
532: pinning was kept constant in the strong pinning regime. In Fig.\ 5 the
533: same type of analysis was done for the case of weak  pinning.
534: In the strong pinning regime a dramatic increase of the cut-offs
535: in the distribution of switched area occurs due to merging of many small
536: signals. Whereas, in the weak pinning the relative enlargement of the cut-off
537: is smaller, that can be understood as an effect of parallel advancement of
538: different sections of the same extended wall, rather than merging of
539: different domains.
540: Both in strong and weak pinning case the increased incidence of large events
541: leads to decrease of the scaling exponents. We find {\it logarithmic}
542: dependence
543: of the exponents for the size  and duration of
544: events on  $\Delta E/J$ both in weak and strong pinning regimes.
545: The results are shown in Fig.\ 6 and the fit lines are given by the
546: general form
547: \begin{equation}
548: \tau_X = A_X - B_X\ln (\Delta E/J) \ .
549: \label{logE}
550: \end{equation}
551: For the strong pinning both area and duration exponents decrease
552: equally, i.e.,  $B_s\approx B_t \approx 0.08$, with the constants
553: $A_s\approx 1.0$ and $A_t\approx 1.4$.
554: In the weak pinning, however, the duration exponent decreases with nearly
555: double rate compared to the size exponent. We find $B_t \approx 2B_s
556: \approx 0.11$, (see Fig.\ 6) and $A_t\approx 1.37$, $A_s \approx 1.30$
557: within numerical error bars.
558: 
559: It is important to note that the measuring conditions depend on the
560: experimental set up and they can vary from one experiment to the other.
561: The effective driving rate depends on the steps in the external field and
562: the degree of compensation of depolarization fields inside the sample.
563: In real experiments  the driving rate is always finite, resulting
564: in the scaling exponents that are always {\it smaller} (and the fractal
565: dimensions larger) compared to those computed in
566: the  theoretical limit of zero driving rate $r \to 0$.
567: Therefore, in order to compare measured and predicted scaling exponents
568: it is advisable to make several  measurements at  different driving rates and
569: find out if the results, for instance for the area exponent $\tau_s$, belong
570: to a line in the range 1.66---1.40, which is compatible with occurrence of
571: small domains, or in the range 1.78---1.60, suggesting that the sample has
572: extended plane domain walls.
573: When the driving rate is too large (cf. top line on both panels in
574: Fig.\ 5), the scaling behavior of the distributions is lost. The cumulative
575: probability distribution shows curvature both due to the lack of small
576: events and the excess of large events.
577: 
578: 
579: 
580: 
581: 
582: 
583: 
584: 
585: \section{ Polarization reversal and back-switching}
586: 
587: 
588: In the previous section we have shown how different kinds of domain walls
589: in the sample determine the scaling properties of  switching
590: current  noise, reflecting the nature of domain-wall motion when the electric
591: field is slowly ramped. Here we consider another type of driving.  First,
592: in the initially polarized   state (all spins down) we impose a large opposite
593: field $E_s$ and keep it until polarization reverses and reaches saturation
594: in that field.  Then we switch off the field to zero and monitor  how the
595:  polarization decays.  Both processes depend on the  type of domains in the
596: sample, and thus, on the strength of pinning.
597: The results for polarization as a function of time are shown in Fig.\ 7 for
598: various strengths of pinning $f$.
599: 
600: 
601: 
602: For weak pinning  spontaneous back switching of the polarization is
603: exponentially fast, reaching the value $P_r$ of remanent polarization.
604: Whereas, the relaxation becomes increasingly slow  as the strength of
605: pinning $f$ increases. The limiting  values of $P_r$ are lower, which
606: correspond to increasingly slimmer loops for strong
607: pinning  (cf. form of the hysteresis loop in Fig.\ 1a).  In general,
608: polarization decay in Fig.\ 7 can be approximated with a stretch
609: exponential law $P(t) = A+B\exp[-(x/\tau )^\sigma]$,
610: with $\sigma =\sigma (f)$  decreasing with the pinning strength. We have
611: $\sigma \approx $1, 0.92 and 0.80 for $f=$ 2.4, 3.4, and 4.4, respectively.
612: Qualitatively similar back-switching
613: was recently observed in doped  ${\mathrm{SBN:Ce^{3+}}}$ relaxor ferroelectric
614:  with two different stretching exponents corresponding respectively  to
615: illuminated and non-illuminated sample \cite{Kleemann}.
616: 
617: 
618: The polarization reversal in the strong constant field shows slow increase
619: in the presence of small domains, but with another type of time dependence
620: (see Fig.\ 7).  The positive part of the polarization
621: curve   is fitted by the power-law increase $P(t) = 1-C(t-a)^{-\nu}$,
622: where $C\approx 0.8$, $a$ is the respective shift (see Fig.\ 7) and the
623: exponent $\nu \approx$ 2, 1.75, and 1.68, for $f=$ 2.4, 3.4, and 4.4,
624: respectively. Again, the characteristic exponent $\nu$ decreases when the
625: pinning becomes stronger, thus contributing to the slow dynamics but
626: within a power-law, rather than a stretch-exponential dependence.
627: 
628: 
629: 
630: 
631: \section{Conclusions and discussion}
632: 
633: 
634: We have shown that, in addition to the  measurements of the hysteresis
635: loop properties and smooth component of the switching current, the domain
636: structure in a ferroelectric
637: sample can be indirectly determined by the non-invasive dynamic
638: measurements of the switching current noise and back-switching processes.
639: Fractal character of the dynamic response indicates that a nontrivial
640: domain structure can occur under slow field driving.
641: The measured properties in these processes can be expressed
642: in terms of universal  exponents, which turn to be related in a unique way
643: to the underlying structure and motion of domain walls.
644: We have shown how the universal features of the switching current noise
645: are quantified through two distinct sets of scaling exponents.  The
646: universal scaling behavior is associated  with a dominant scenario of the
647: domain wall dynamics: either the motion and depinning of large  domain
648: walls, in the case of weak pinning, or  bulk nucleation and growth of many
649: small 3-dimensional domains, in the case of strong pinning.
650: The  pinning in our model is provided by the local electric random fields.
651: 
652: Similarly, the polarization reversal in a strong applied field,
653: and back-switching of the domains when the field is set to zero, are quick
654: when large domains occurs, whereas they are slow when  many small domains
655: are nucleated. In the latter case the polarization relaxes according to
656: a stretch-exponential law (the stretching exponent $\sigma <1$) whereas it
657: reverses with a fractal power of time (the characteristic exponent $\nu <2$)
658:  when a large opposite field is switched on.
659: In the case of weak pinning, on the contrary, the relaxation is exponential
660: ($\sigma \approx 1$) and the reversal  algebraic in time with $\nu =2$.
661: Dependence of the stretching exponent $\sigma (f)$ on the pinning strength
662: is in qualitative agreement with measurements of back-switching
663: in doped ${\mathrm{SBN:Ce^{3+}}}$ relaxor ferroelectrics \cite{Kleemann},
664: again suggesting that the  local electric fields  contribute
665: significantly to slow the dynamics in this class of relaxors \cite{Kleemann}.
666: (Note that the characteristic time appears unrealistically small in our
667: lattice model.) We are not aware of measurements of temporal  behavior of
668: the polarization reversal and of the switching current noise in these systems.
669: 
670: The observed  scaling behavior of the switching current depends on
671: the  rate at which the driving field changes. We have demonstrated that
672: in both universality classes the scaling exponents decrease {\it
673: logarithmically} with increased driving rate $\Delta E/J$. This is in
674: sharp contrast to the cases where the dominant type of disorder is due to
675: random vacancies in spin-systems \cite{BT_summary} and other avalanche-type
676: dynamic systems driven at a finite rate \cite{TP}, where the scaling
677: exponents decrease and fractal dimensions increase {\it linearly} with the
678: driving rate.
679: Therefore, occurrence of the logarithmic dependences  of the scaling
680: exponents on the driving frequency may serve as an additional test of
681: the presence of the real random fields in  ferroelectrics.
682: 
683: 
684: 
685: 
686: Throughout this work we used the simulation algorithm at finite driving
687:  rates $\Delta E >0$, that mimics the  realistic situation in experiments.
688: The obtained set of scaling exponents, together with their
689: dependences on the driving rate, belongs to one (of the two) universality
690: classes, which implies  a definite domain structure of the system, and hence
691: the expected  character of the switching processes. Our results suggest that
692: a crossover from one to the other universal behavior both in the switching
693: current noise and in the polarization reversal and relaxation can be achieved
694: by varying  the strength of pinning due to random electric fields.
695: As mentioned above, in practice there are various ways
696: to accomplished the crossover by reducing the pinning in ferroelectrics
697: \cite{Shur_et_al,Kleemann,PLZT-HL}.
698: Here we have not discussed the properties of the dynamic phase transition
699: separating these two scaling behaviors at a critical value $f_c$, nor the
700: critical exponents describing the depinning transition along the critical
701: field $H_d(f)$ for $f<f_c$.
702: This questions require an additional study and proper finite size scaling
703: analysis, for instance along the lines as done for the two-dimensional
704: case in Ref.\cite{TN}.
705: Finally, by allowing a frequent appearance of the unscreened depolarization
706:  fields  a  finite probability  of negative pulses in the switching current
707: builds up ($b\neq 0$). The presence of negative pulses slows the switching
708: process, interferes with the effects of finite driving rates, and alters
709: the scaling properties of noise. A more complete analysis of the switching
710: processes for non-vanishing probability $b\neq 0$ is left for a future
711: study.
712: 
713: Finally, it should be stressed that, in spite of its simplicity, the proposed
714: model seems reasonable for dynamic properties of three-dimensional uniaxial
715: relaxor ferroelectrics.
716: Whereas, in the single-crystalline normal ferroelectrics a  more realistic
717: modeling of the effects of interactions and depolarization fields
718: on domain structure is required. Nevertheless, we believe that, when
719: restricted only to  the {\it universal} scaling behavior of the noisy
720: component of the switching current, the presented results for the case of
721: large domains may not depend significantly on the details of interactions.
722: Further research on the subject is necessary.
723: 
724: 
725: 
726: \acknowledgments
727: Work is supported by the Ministry
728: of Education, Science and Sports of the Republic of Slovenia.
729: 
730: \begin{references}
731: 
732: \bibitem[*]{*}Electronic address: Bosiljka.Tadic@ijs.si
733: 
734: \bibitem{Rudyak}V. M. Rudyak, Usp. Fiz. Nauk {\bf 101}, 429 (1970).
735: \bibitem{book}E. Fatuzzo and W. J. Merz, {\it Ferroelectricity},
736: North-Holland, Amsterdam (1967).
737: \bibitem{Shur_rev}V. Ya. Shur, Phase Transitions, {\bf 65},
738: 49 (1998).
739: \bibitem{BT_summary}B. Tadi\'c,  Physica A {\bf 270}, 125 (1999).
740: \bibitem{Djole}Dj. Spasojevi\'c {\it et al.}, Phys. Rev. E
741: {\bf 54}, 2531 (1996).
742: 
743: \bibitem{packing}L. Carillo {\it et al.}, Phys. Rev. Lett. {\bf 81}, 1889
744: (1998).
745: 
746: \bibitem{BTiO3}R.C. Miller, Phys. Rev. B {\bf 111}, 736 (1958); A.G. Chynoweth,
747: J. Appl. Phys. {\bf 30}, 280 (1959).
748: 
749: \bibitem{Shur}V. Ya. Shur,  E.V. Nikolaeva, E.L. Rumyantsev, {\it et al.},
750: Ferroelectrics {\bf 222}, 323 (1999).
751: \bibitem{Shur_et_al}V. Ya. Shur, E. L. Rumyantsev, D. V. Pelegov, {\it et al.},
752: Ferroelectrics (proceedings of IMF10).
753: \bibitem{ceramics}G. Pleyber {\it et al.}, Ferroelectrics {\bf 141}, 125
754: (1993).
755: \bibitem{Weissmann}E. V. Colla, L. K. Chao and M. B. Weissmann, preprint
756: (2001).
757: \bibitem{FTT}V. Ya. Shur,  E.L. Rumyantsev, V.P. Kuminov {\it et al.},
758: Phys. Solid State {\bf 41}, 269 (1999); V. Ya. Shur, V. L. Kozhevnikov,
759: D. V. Pelegov, {\it et al.}, {\it ibid.} {\bf 43}, 1128 (2001).
760: 
761: \bibitem{Dui2}W. Kleemann, J. Dec, P. Lehnen, {\it et al.},
762: Europhys. Lett. {\bf 57}, 14 (2002).
763: 
764: \bibitem{Peer}P. Lehnen,  W. Kleemann, Th. Woike, and R. Pankrath,
765: Eur. Phys. J. B {\bf 14}, 633 (2000); Ferroelectrics {\bf 253}, 11 (2001);
766: P. Lehnen, W. Kleemann, Th. Woike, and R. Pankrath, Phys. Rev. B
767: (in press);
768: \bibitem{Kleemann}T. Granzow, U. D\"orfler, Th. Woike, {\it et al.},
769: Europhys. Lett. {\bf 57}, 597 (2002).
770: 
771: \bibitem{book_AL}A. P. Levanyuk and A. S. Sigov, {\it Defects and Structural
772: Phase Transitions
773: (Ferroelectricity and Related Phenomena)}, Taylor\& Frances, (London) 1988.
774: 
775: \bibitem{BT_letter}B. Tadi\'c, Phys. Rev. Lett. {\bf 77}, 3843 (1996).
776: 
777: 
778: 
779: \bibitem{BT_ferroel}B. Tadi\'c, Ferroelectrics {\bf 259}, 3 (2001).
780: \bibitem{BCN}E. Vives {\it et al.}, Phys. Rev. Lett. {\bf 72}, 1694 (1994).
781: \bibitem{Cornel}O. Perkovi\'c {\it et al.}, Phys. Rev. B {\bf 59}, 6106 (1999).
782: \bibitem{Zheng}G.-P. Zheng and M. Li, Phys. Rev. B {\bf 63}, 36122 (2001).
783: 
784: \bibitem{comment_planarDW}Apart from weak pinning, more realistic
785: interactions, e.g.,  including elastic
786: coupling are responsible for occurrence of the plane domain walls in
787:  normal ferroelectrics such as  ${\mathrm{BaTiO_3}}$.
788: \bibitem{DD}S. Sabhapandit, P. Shukla, and D. Dhar, J. Stat. Phys. {\bf 98},
789: 103 (2000).
790: \bibitem{comment_integration}
791: Notice that the properties of noise vary with the applied field, see for
792: instance Refs.\ \cite{Shur,FTT}.
793: 
794: 
795: \bibitem{correlations}B. Tadi\'c, Physica A {\bf 282} 362 (2000).
796: \bibitem{Rasa}R. Pirc and R. Blinc, Phys. Rev. B {\bf 60}, 13470 (1999);
797: R. Blinc {\it et al.}, Phys. Rev. Lett. {\bf 83}, 424 (1999).
798: 
799: \bibitem{PLZT-HL}C. S. Lynch {\it et al.}, Ferroelectrics {\bf 166}, 11 (1995)
800: 
801: 
802: \bibitem{TN}B. Tadi\'c and U. Nowak, Phys. Rev. E {\bf 61}, 4610 (2000).
803: \bibitem{Co_film} A. Berger {\it et al.}, Phys. Rev. Lett. {\bf 85}, 4176
804: (2000).
805: \bibitem{TP}B. Tadi\'c and V. Priezzhev, Phys. Rev. E {\bf 62}, 3266 (2000).
806: 
807: 
808: 
809: \end{references}
810: 
811: 
812: 
813: 
814:  \narrowtext
815: \begin{table}
816:   \begin{center}
817:     \begin{tabular}{|c|c|c|c|}
818:                                 %\hline\hline
819:       $f$ (pinning) &Exponent& 3D & 2D\\
820: \hline \hline$f\approx f_c$& $\tau _s$& 1.78 & 1.54\\
821: (weak)& $\tau _t$& 2.20 & 1.83\\
822: \hline$f> f_c$& $\tau _s$& 1.66 & 1.30\\
823: (strong)& $\tau _t$& 2.03 & 1.47\\
824:                                 %\hline\hline
825:     \end{tabular}
826:   \caption{Scaling exponents for weak pinning ($f\approx f_c$) and for strong
827: pinning ($f> f_c$) for 3-dimensional (3D) system at driving rate $4\times
828: 10^{-4}$,  results of this work, and for
829: 2-dimensional (2D) system  at zero driving rate,  from Ref.\ [25].
830:     Estimated error bars  are within $\pm 0.03$.}
831:   \end{center}
832:   \label{table1}
833: %\end{table}
834: 
835: 
836: \narrowtext
837: \begin{figure}[b!] % fig 1
838: \epsfxsize=80mm\epsffile[46 72 513 488]{sw_fig1.ps}
839: \caption{(a) Hysteresis loop for driving rate $r\equiv \Delta E/J=$0.01 and
840: varying strength of disorder $f=$4, 3.2, and 2 (inside out).
841: (b) Sections of the first two loops from (a) for two different
842: driving  rates $r=$0.01 (dark) and $r=$0.08 (clear symbols).
843: (c) Variation of the driving field with time and (d) the electrical Barkhausen
844: noise (switching current) for finite probability of back flips $b=0.3$ .
845: Fig.\ 1 from Ref.\ [18].}
846: \label{fig1}
847: \end{figure}
848: 
849: \begin{figure}
850: \epsfxsize=80mm\epsffile[58 63 509 639]{sw_fig2.ps}
851: \caption{Part of the switching current signal $j(t)$ against cumulative
852: time $t$ (lower panel) and the Fourier spectrum of the signal (top panel)
853: for moderate pinning $f=3.6$ . Slope of the fitted area
854: $\phi = 0.738\pm 0.007$ . Only nonzero values of current are shown (see text).}
855: \label{fig2}
856: \end{figure}
857: 
858: \begin{figure}
859: \epsfxsize=80mm\epsffile[45 72 507 670]{sw_fig3.ps}
860: \caption{ Cumulative distributions of switched domain size $P(s,f)$ vs. size
861:  $s$ (lower panel) and duration of switching $P(t,f)$ vs. duration $t$
862: (top panel) for
863: varying strength of pinning $f$ =2.2, 2.4, 2.8 and 3.2 (bottom to top, both
864: panels). Driving rate is fixed $\Delta E/J$ = 0.001. Fitted values of three
865: slopes, corresponding to full lines, are (bottom to top): $\tau_s-1$=1.029(4),
866: 0.747(2), 0.568(5), on the lower panel, and $\tau_t-1$= 1.482(6), 1.145(7),
867: and 0.94(2), on top panel.
868: In total $4.32\times 10^6$ spins divided into 20 different samples was
869: used in simulations. Distributions are normalized by the actual
870: number of avalanches during the full cycle and the number of samples.
871: Data are log-binned with ratio 1.1  .}\label{fig3}
872: \end{figure}
873: 
874: \begin{figure}
875: \epsfxsize=82mm\epsffile[44 72 507 670]{sw_fig4.ps}
876: \caption{Cumulative distributions of switched domain size $P(s,\Delta E/J)$
877: vs. $s$ (lower panel) and duration of switching $P(t,\Delta E/J)$ vs. $t$
878: (top panel) for fixed pinning $f=3.2$ and varying driving rate $\Delta E/J$
879: = $4\times10^{-4}$,
880: $1\times10^{-3}$, $2\times10^{-3}$, $4\times10^{-3}$ and  $8\times10^{-3}$
881: (bottom to top on both panels). Normalization  to the actual
882: number of avalanches results in the vertical shift of the curves.
883: Scaling region indicated by the section of the straight line.
884:  Log-binning ratio 1.1 .}\label{fig4}
885: \end{figure}
886: 
887: \begin{figure}
888: \epsfxsize=82mm\epsffile[41 72 507 670]{sw_fig5.ps}
889: \caption{Same as Fig.\ 4 but for weak pinning regime $f=2.4$. Driving rates
890: are (bottom to top)  $\Delta E/J$ = $4\times10^{-4}$,
891: $1\times10^{-3}$, $2\times10^{-3}$, $4\times10^{-3}$, $8\times10^{-3}$
892: and $1.2\times10^{-2}$. Fitted area on each curve is shown by solid line. }
893: \label{fig5}
894: \end{figure}
895: 
896: 
897: \begin{figure}
898: \epsfxsize=82mm\epsffile[70 93 324 385]{sw_fig6.ps}
899: \caption{Scaling exponents for the distribution of switched area (squares)
900: and duration of pulses (circles) against driving rate $\Delta E/J$.
901: Filled and open symbols correspond to strong  and weak pinning, respectively.
902:  Fit lines:  $\tau _X= A_X -B_X\ln (\Delta E/J)$, see text.}
903: \label{fig6}
904: \end{figure}
905: 
906: 
907: \begin{figure}
908: \epsfxsize=80mm\epsffile[53 72 508 412]{sw_fig7.ps}
909: \caption{Polarization reversal at field $E/J=3$ (left part) and relaxation
910: after field was set to zero (right part) against time for three values of
911: pinning strength $f$, as indicated. Latter two lines were shifted to the right
912: for better display.  Fits according to $P=1-C(t-a)^{-\nu}$ (dotted lines),
913: and $P=A+B\exp[-(t/2-11)^\sigma]$ (full lines), see text. }
914: \label{fig7}
915: \end{figure}
916: 
917: 
918: 
919: \end{table}
920: \end{multicols}
921: \end{document}
922: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
923: 
924: 
925: #!/bin/csh -f
926: # this uuencoded Z-compressed .tar file created by csh script  uufiles
927: # for more information, see e.g. http://xxx.lanl.gov/faq/uufaq.html
928: # if you are on a unix machine this file will unpack itself:
929: # strip off any mail header and call resulting file, e.g., swfigures.uu
930: # (uudecode ignores these header lines and starts at begin line below)
931: # then say        csh swfigures.uu
932: # or explicitly execute the commands (generally more secure):
933: #    uudecode swfigures.uu ;   uncompress swfigures.tar.Z ;
934: #    tar -xvf swfigures.tar
935: # on some non-unix (e.g. VAX/VMS), first use an editor to change the
936: # filename in "begin" line below to swfigures.tar_Z , then execute
937: #    uudecode swfigures.uu
938: #    compress -d swfigures.tar_Z
939: #    tar -xvf swfigures.tar
940: #
941: uudecode $0
942: chmod 644 swfigures.tar.Z
943: zcat swfigures.tar.Z | tar -xvf -
944: rm $0 swfigures.tar.Z
945: exit
946: 
947: