1: \documentclass[12pt]{article}
2: \usepackage{graphicx}
3: \sloppy
4: \vbadness 10000
5: \tolerance 10000
6: \hbadness 10000
7: \setcounter{equation}{0}
8: \baselineskip4ex
9: \topmargin=-1.5cm
10: \textheight=23.0cm
11: \textwidth=16.2cm
12: \oddsidemargin=0cm
13: \begin{document}
14: \baselineskip=5.5mm
15: \newcommand{\be} {\begin{equation}}
16: \newcommand{\ee} {\end{equation}}
17: \newcommand{\Be} {\begin{eqnarray}}
18: \newcommand{\Ee} {\end{eqnarray}}
19: \def\a{\alpha}
20: \def\b{\beta}
21: \def\g{\gamma}
22: \def\G{\Gamma}
23: \def\d{\delta}
24: \def\D{\Delta}
25: \def\e{\epsilon}
26: \def\k{\kappa}
27: \def\l{\lambda}
28: \def\s{\sigma}
29: \def\t{\tau}
30: \def\om{\omega}
31: \def\Om{\Omega}
32: \def\lg{\langle}
33: \def\rg{\rangle}
34:
35: \noindent
36: \begin{center}
37: {\Large
38: {\bf
39: Dynamic heterogeneities in the out-of-equilibrium dynamics of simple
40: spherical spin models\\
41: }}
42: \vspace{1cm}
43: \noindent
44: {\bf Gregor Diezemann} \\
45: {\it
46: Institut f\"ur physikalische Chemie, Universit\"at Mainz,
47: Welderweg 11, 55099 Mainz, FRG
48: \\}
49: revised version\\
50: \end{center}
51: \vspace{2cm}
52: \noindent
53: {\it
54: The response of spherical two-spin interaction models, the spherical
55: ferromagnet (s-FM) and the spherical Sherrington-Kirkpatrick (s-SK) model, is
56: calculated for the protocol of the so-called nonresonant hole burning
57: experiment (NHB) for temperatures below the respective critical temperatures.
58: It is shown that it is possible to select dynamic features in the
59: out-of-equilibrium dynamics of both models, one of the hallmarks of
60: dynamic heterogeneities.
61: The behavior of the s-SK model and the s-FM model in three dimensions is very
62: similar, showing dynamic heterogeneities in the long time behavior, i.e. in
63: the aging regime.
64: The appearence of dynamic heterogeneities in the s-SK model explicitly
65: demonstrates that these are not necessarily related to {\it spatial}
66: heterogeneities.
67: For the s-FM it is shown that the nature of the dynamic heterogeneities changes
68: as a function of dimensionality. With incresing dimension the frequency
69: selectivity of the NHB diminishes and the dynamics in the
70: mean-field limit of the s-FM model becomes homogeneous.
71: }
72:
73: \vspace{1cm}
74: \noindent
75: PACS Numbers: 64.70 Pf,05.40.+j,61.20.Lc
76: \vspace{1.5cm}
77: \section*{I. Introduction}
78: Non-exponential relaxation behavior is found to be rather common when dealing
79: with disordered materials like glasses, spin-glasses, disordered crystals
80: or proteins\cite{a.rel}.
81: In the last decade particular attention has been payed to the question
82: to which extent the relaxation is to be viewed as dynamic
83: heterogeneous\cite{het.dyn}.
84: Different experimental techniques have been invented in order to investigate
85: the detailed nature of the relaxation particularly of amorphous
86: systems\cite{SRS91, CE95, ViRI00, DvB01}.
87: These techniques allow to specifically select a slow sub-ensemble and
88: afterwards monitor its relaxation.
89: Various applications of these techniques have shown that the primary response
90: in amorphous polymers and supercooled liquids is to be viewed as
91: heterogeneous in the sense that it is possible to select slow sub-ensembles
92: relaxing at smaller rates than the average.
93: Throughout this paper I use the definition given in ref.\cite{het.def}
94: according to which a system will be called dynamic heterogeneous if it is
95: possible to select dynamically distinguishable (slow and/or fast)
96: contributions to a relaxation.
97:
98: In addition it has been found that after a certain re-equilibration time the
99: relaxation properties of the selected sub-ensemble return to those of the
100: bulk\cite{andreas.4d, roland.4d}.
101: Therefore, these experiments indicate that the response can be described
102: as a superposition of exponentially decaying entities with different
103: relaxation rates. The various relaxation rates, however, are not static
104: quantities but apparently fluctuate in time. Different interpretations
105: have been provided for this behavior\cite{heuer97, dieze97}.
106: The NMR-techniques\cite{andreas.4d, roland.4d} have the advantage of a simple
107: interpretation in terms of equilibrium 4-time correlation functions but are
108: restricted to a rather narrow temperature regime and to certain materials.
109: The optical deep bleach technique\cite{CE95} has the advantage that it can be
110: applied in a wider temperature range with the shortcoming that up to now it has
111: not been interpreted in terms of equilibrium correlation functions.
112: At this point it is important to note that these experimental methods monitor
113: molecular reorientations and therefore are {\it not} able to address the
114: question of {\it spatial aspects} of the dynamic heterogeneities.
115: The only exception is provided by a new variant of the 4d-NMR
116: technique\cite{4dcp}, which allows to extract a length scale via monitoring
117: spin-diffusion. Originally, this technique has been applied to a polymeric
118: liquid and later on the length scale of the dynamic heterogeneities has
119: been extracted for low-molecular glass-forming systems\cite{4dcp.gly,4dcp.vgl}.
120: The corresponding length scales have been found to be on the order of
121: $1\cdots 4$nm.
122:
123: Another experimental technique allowing to monitor dynamic heterogeneities
124: is provided by the nonresonant spectral hole burning (NHB)
125: experiment\cite{SBLC96}.
126: This method is based on a pump-wait-probe field sequence with a large pump
127: field amplitude beyond the linear response regime.
128: Though originally applied to supercooled liquids\cite{SCDB97}, NHB in the
129: meantime has been used to investigate the relaxation of several materials,
130: including disordered crystals (relaxor ferroelectrics)\cite{KSB98},
131: ion-conducting glasses\cite{RB99} and spin glasses\cite{ralph99}.
132: Also the application of NHB to a solvable glass-model has been
133: presented\cite{leticia01}.
134: The interpretation of the obtained results has mainly been guided by the fact
135: that via the application of a large amplitude ac field of frequency $\Om$
136: the sample absorbs energy of an amount proportional to the imaginary part
137: of the susceptibility evaluated at the pump frequency $\Om$\cite{kubo}.
138: In case of a homogeneously broadened response one does not expect that it
139: is possible to modify the response in a frequency selective way. By contrast
140: such a goal could be achieved if the response is given by a superposition of
141: differently fast relaxing entities (heterogeneous scenario). This is because
142: in this case energy absorbtion should be largest for those sub-ensembles with
143: a relaxation time on the scale of the inverse pump frequency.
144: This intuitive picture is confirmed in the framework of a response-theory for
145: NHB for the particular case of stochastic dipole reorientations, which I have
146: developed recently\cite{dieze01}.
147:
148: In the quoted experiments it always has been found that a frequency selective
149: modification of the response indeed is possible. Regarding the
150: re-equilibration of this modification, however, the results differ not only
151: with respect to the frequency dependence but also regarding the time scale
152: of the recovery. For the latter a time scale longer than the inverse
153: burn frequency has been observed in the case of the relaxor
154: ferroelectrics\cite{KSB98}.
155:
156: As long as one is concerned with supercooled liquids, one can safely consider
157: the system to be in (metastable) equilibrium prior to the NHB field sequence.
158: This, however, is not necessarily true for the relaxor materials or the spin
159: glasses.
160: In particular, it was argued in ref.\cite{DB01} that the results obtained for
161: the solvable p-spin-glass model\cite{leticia01} are mainly to be interpreted
162: in terms of out-of-equilibrium effects.
163: In equilibrium, the $p\!=\!3$-model studied shows an exponential relaxation
164: at long times. Therefore, according to what was said above one does not expect
165: to be able to select a sub-ensemble in a frequency dependent way in the
166: equilibrated version of the model.
167: It has to be mentioned here that dynamic heterogeneities as monitored by
168: NHB have also been observed in Monte Carlo simulations on an equilibrated
169: Sherrington-Kirkpatrick mean-field spin-glass model\cite{let.reply}.
170: These calculations explicitly demonstrate that from the observation of
171: dynamic heterogeneities one cannot conclude on the existence of spatial
172: heterogeneities.
173: Also, this finding appears to be independent of whether the system was in
174: thermal equilibrium before the application of the pump-field.
175:
176: In this paper I consider the application of the NHB field sequence to the
177: spherical Sherrington-Kirkpatrick (s-SK) model, i.e. the p-spin model with
178: $p=2$ and the spherical ferromagnet (s-FM) in arbitrary dimension.
179: The Langevin dynamics for these models has been solved analytically\cite{CD95}.
180: Additionally, the s-FM model is equivalent to the O(N)-model in the limit of
181: large N and therefore is a typical model for domain coarsening
182: processes\cite{bray94}.
183:
184: I will solely consider a thermal history protocol in which the system is
185: quenched to a temperature below the critical temperature $T_c$ from infinite
186: temperature prior to the experiment. The behavior at and above $T_c$ will be
187: investigated in a forthcoming publication.
188: Therefore, all effects observed are intimately related to the
189: out-of-equilibrium dynamics of the model.
190: This is because the system never reaches equilibrium in this temperature
191: regime.
192: The outline of the paper is as follows. In the next Section I will briefly
193: recall the dynamic features of the models and calculate the response to
194: the NHB field sequence in second order regarding the pump-field amplitude
195: and linearly in the small step-field.
196: In Section III the results of the calculations are presented and discussed.
197: The paper closes with some conclusions in Section IV.
198: \section*{II. NHB in spherical two-spin interaction models}
199: The spherical models under consideration are defined by the Hamiltonian
200: \be\label{Ham.def}
201: {\bf H}=-\frac{1}{2}\sum_{i\neq k}J_{ik} s_i s_k-\sum_i h_i s_i
202: \ee
203: where in case of the s-SK model the $J_{ik}$ are chosen at random from a
204: Gaussian probability distribution with zero mean and variance $\s=1/N$
205: and are restricted to a ferromagnetic coupling $J$ (to be set to unity in the
206: following) for the s-FM model on a simple hypercubic lattice in d dimensions.
207: In addition the spin-variables are subject to the spherical constraint
208: $\sum_i s_i^2=N$.
209: The Langevin equations governing the dynamics of the model read as
210: \be\label{Langevin}
211: {\dot s}_i(t)=\sum_k J_{ik}s_k(t)+h_i-z(t)s_i(t)+\xi_i(t)
212: \ee
213: where $z(t)$ is the Lagrange multiplier enforcing the spherical constraint
214: and $\xi_i(t)$ is a $\d$-correlated Gaussian white noise.
215: These equations have been solved analytically by Cugliandolo and Dean
216: (CD)\cite{CD95} in two papers.
217: The dynamical properties of the s-FM model are discussed in ref.\cite{GL00}.
218: Further information regarding the correspondence between the two models can be
219: found in ref.\cite{ZKH00}. Here, I briefly summarize the results relevant in
220: the present context.
221:
222: Of particular importance are the violations of the fluctuation-dissipation
223: theorem (FDT), which relates the response function to the time-derivative of
224: the two-time correlation function,
225: \be\label{FDT}
226: R(\t)=-{1\over T}{dC(\t)\over d\t}
227: \ee
228: In particular, it has proven extremely useful in out-of -equilibrium
229: situations to define the so-called fluctuation-dissipation ratio $X(t,t_w)$
230: via\cite{FDT_rev98},
231: \be\label{X_FDT}
232: R(t,t_w)={ X(t,t_w)\over T}{\partial C(t,t_w)\over\partial t_w}
233: \ee
234: the limiting long-time behavior, $X_\infty$, of which is known to vanish for
235: domain coarsening models\cite{X_infty}. Here, $t_w$ denotes the time that has
236: elapsed after a quench to the working temperature prior to the measurement.
237: Also for the models considered in the present paper one has
238: $X_\infty=0$\cite{GL00}.
239: Thus, concerning this measure of typical distances
240: from equilibrium, the domain coarsening models in finite dimension do not show
241: any differences to the mean-field s-SK model.
242:
243: The response of the system, $R_h(t,t')=\sum_i^N\lg s_i(t)\xi_i(t')\rg/(2NT)$,
244: in the presesence of a field $h(t)$ can be obtained in the same way as
245: calculated by CD for the zero field case.
246: As shown by Berthier et al.\cite{BCI01} this yields:
247: \be\label{R.t.tw}
248: R_h(t,t')=\theta(t-t')\frac{W_h(t')}{W_h(t)}g\!\left(\frac{t-t'}{2}\right)
249: \ee
250: where the function $g(t)$ is defined by
251: \Be\label{g.t.def}
252: &&g(t)=\left[\exp{(-4t)}I_0(4t)\right]^{d}\quad \mbox{s-FM}\nonumber\\
253: &&g(t)=\exp{(-4t)}\frac{I_1(4t)}{2t}\quad \mbox{s-SK}
254: \Ee
255: with $I_n(x)$ denoting the generalized Bessel function and $W_h(t)$ is the
256: solution of
257: \Be\label{Wh.def}
258: W_h(t)^2=
259: &&\hspace{-0.6cm}
260: g(t)+2T\int_0^t\!d\t W_h(\t)^2g(t-\t)
261: \nonumber\\
262: &&\hspace{-0.4cm}
263: +\int_0^t\!dt_1\int_0^{t}\!dt_2h(t_1)h(t_2)W_h(t_1)W_h(t_2)
264: g\!\left(t-\frac{t_1+t_2}{2}\right)
265: \Ee
266: which has its origin in the normalization of the equal-time correlation,
267: $C(t,t)=1$, i.e. the spherical constraint.
268:
269: \subsection*{Zero field response}
270:
271: Before I turn to the calculation of the response following the NHB pulse
272: sequence, it is appropriate to summarize the known results for the reponse
273: and the correlation in zero field, for a more detailed discussion see
274: refs.\cite{CD95,GL00,ZKH00}.
275: In zero field, eq.(\ref{Wh.def}) simplifies to the following Volterra equation:
276: \be\label{W.t}
277: W(t)^2=g(t)+2T\int_0^t\!d\t W(\t)^2g(t-\t)
278: \ee
279: For the s-SK model, this equation has been solved by CD for random initial
280: conditions (i.e. a quench from $T=\infty$ at $t=0$).
281: No simple analytical solution exists for the s-FM model.
282: However, for $t\gg 1$ it can be shown\cite{CD95, GL00, ZKH00} that for
283: $T<T_c$
284: \be\label{W.as}
285: W(t)^2=\frac{1}{(1-T/T_c)^2}g_{as}(t)
286: \ee
287: Here, the asymptotic behavior of the $g(t)$ are given by
288: \Be\label{g.as.def}
289: &&g_{as}(t)=(8\pi t)^{-d/2}\hspace{1.3cm}\mbox{s-FM}\nonumber\\
290: &&g_{as}(t)=(32\pi)^{-1/2}t^{-3/2}\quad\mbox{s-SK}
291: \Ee
292: The critical temperatures are given by $T_c=1$ for the s-SK model and
293: $T_c$ depends on the spatial dimension in case of the s-FM model,
294: with $T_c(d=3)\simeq 3.9568$\cite{BK52}. For the other dimensions used in the
295: present paper I find $T_c(d=5)\simeq 8.6482$, $T_c(d=7)\simeq 12.7982$ and
296: $T_c(d=9)\simeq 16.8579$.
297:
298: If $t_w$ denotes the time elapsed after a quench from $T=\infty$ to $T<T_c$,
299: the following behavior is found for $R(t+t_w,t_w)$ from
300: eqns.(\ref{R.t.tw},\ref{W.as},\ref{g.as.def}) in the limit of long $t_w$:
301: \Be\label{R.as}
302: &&R(\t+t_w,t_w)
303: =\l^{-\e/4}g(\t/2)
304: \quad\mbox{with}\quad
305: \l=\frac{t_w}{t_w+\t} \quad ; \quad t_w \gg 1
306: \nonumber\\
307: &&\hspace{1cm} \e=d\quad\mbox{(s-FM)}\quad ; \quad \e=3\quad\mbox{(s-SK)}
308: \Ee
309: From this expression it is evident already that the dynamic properties of the
310: s-SK model are very similar to those of the s-FM in $d=3$.
311: Therefore, the following expressions are given for the s-FM model. The only
312: differences between the s-FM model in $d=3$ and the s-SK model stem from the
313: different prefactors in eq.(\ref{g.as.def}).
314: In particular, it is evident from eq.(\ref{R.as}) that the temperature is an
315: irrelevant variable in the whole low-temperature phase.
316:
317: Two time-sectors are to be distinguished:
318: \begin{itemize}
319: \item
320: For short times $\t$ such that $\t\ll t_w$ - the so-called stationary regime -
321: one has $\l\simeq 1$ and accordingly:
322: \be\label{R.stat}
323: R(\t+t_w,t_w) = R(\t) = g(\t/2)\quad \t\ll t_w
324: \ee
325: In particular, in this stationary regime the FDT, eq.(\ref{FDT}), holds.
326: In the long time limit, $1\ll \t$, eq.(\ref{R.stat}) shows that $R(\t)$
327: decays according to $R(\t)\sim \t^{-d/2}$.
328: \item
329: In the so-called aging regime $1\ll \t\sim t_w$, one finds from eq.(\ref{R.as}):
330: \be\label{R.aging}
331: R(\t+t_w,t_w)=\l^{-d/4}(4\pi \t)^{-d/2} \quad 1\ll \t\sim t_w
332: \ee
333: In this regime the FDT is strongly violated.
334: If in addition $\t\gg t_w$, the response behaves as $R(\t,t_w)\sim \t^{-d/4}$.
335: \end{itemize}
336:
337: The overall behavior is shown for various values of the waiting time in Fig.1.
338: From this plot it is seen that for all dimensions shown, $d=3,5,7$, there is a
339: crossover from the $\t^{-d/2}$ to the $\t^{-d/4}$ behavior. Also the explicit
340: dependence on the waiting time in the aging regime is evident.
341: It should be pointed out that the crossover from the stationary regime to the
342: aging (domain growth) regime takes place around $\t\sim t_w$ independent of
343: spatial dimension.
344: Remember, that only in the stationary regime the FDT holds.
345: In particular, the behavior of the response (and also the two-time correlation
346: function) does not change qualitatively around $d\!=\!4$, above which the
347: model behaves mean-field like concerning the statics and the exponents.
348: (The same holds for the dynamic fluctuations, as will be shown elsewhere.)
349:
350: The above discussion shows that the relaxation is extremely non-exponential.
351: Therefore, the question as to what extent the response can be viewed as
352: dynamic heterogeneous naturally arises.
353: \subsection*{\bf Nonresonant hole burning}
354: In the following, the response will be calculated for the NHB-field-sequence,
355: cf. Fig.2.
356: At $t=0$ the system is quenched from $T=\infty$ to the working temperature $T$.
357: One (or more) cycles of the pump-field $h_p(t)=h_p\sin{[\Om(t-t_q)]}$ are
358: applied after a time $t_q$ has elapsed.
359: Following a waiting time $t_w$ the response $R^*(\hat{t}+\t,\hat{t}+\t')$ is
360: measured, where I defined
361: \[\hat{t}=t_q+t_p+t_w\quad\mbox{and}\quad t_p=2N\pi/\Om\]
362: for brevity. Here, $N$ denotes the number of cycles of the sinusoidal
363: pump-field.
364: In the following calculations of $R^*$ the $W_h(t)$ are needed in second
365: order with respect to the pump-field amplitude $h_p$.
366: From eq.(\ref{Wh.def}) it is evident that a perturbation expansion follows
367: from
368: \be\label{W.hp}
369: W_{h_p}(t)^2=W(t)^2+h_p^2 \D(t)^2+{\cal O}(h_p^4)
370: \ee
371: Inserting this expression into eq.(\ref{Wh.def}) yields eq.(\ref{W.t}) for
372: the zero'th order term and, assuming a time-dependent field of the form
373: $h(t)=h_p\sin{[\Om(t-t_q)]}\theta(t-t_q)$ according to Fig.2:
374: \Be\label{del.W}
375: &&\hspace{-1.0cm}\D(t_q+\t)^2=
376: 2T\int_0^\t\!ds \D(t_q+s)^2g(\t-s) + \D_0(t_q+\t)
377: \nonumber\\
378: &&\hspace{-1.0cm}
379: \D_0(t_q+\t)=\int_0^{t_m}\!dt_1\int_0^{t_m}\!dt_2\sin{(\Om t_1)}\sin{(\Om t_2)}
380: W(t_q+t_1)W(t_q+t_2)
381: g\!\left(\t-\frac{t_1+t_2}{2}\right)
382: \Ee
383: where $t_m=$Min$(\t,t_p)$.
384: Using eq.(\ref{W.hp}) the response function in ${\cal O}(h_p^2)$ then is found
385: to be given by ($\t>\t'$):
386: \Be\label{R.DR.allg}
387: &&\hspace{-0.8cm}R^*(\hat{t}+\t,\hat{t}+\t') = R(\hat{t}+\t,\hat{t}+\t')
388: +\D R(\hat{t}+\t,\hat{t}+\t')
389: \nonumber\\
390: &&\hspace{-0.8cm}R(\hat{t}+\t,\hat{t}+\t')
391: =\frac{W(\hat{t}+\t')}{W(\hat{t}+\t)}g\!\left(\frac{\t-\t'}{2}\right)\\
392: &&\hspace{-0.8cm}\D R(\hat{t}+\t,\hat{t}+\t')=
393: -\frac{h_p^2}{2}
394: \left[\frac{\D(\hat{t}+\t)^2}{W(\hat{t}+\t)^2}-
395: \frac{\D(\hat{t}+\t')^2}{W(\hat{t}+\t')^2}\right]
396: R(\hat{t}+\t,\hat{t}+\t')\nonumber
397: \Ee
398: In the NHB-protocol of Fig.2, however, the response to a small step field is
399: recorded, i.e. the integrated response (the thermoremanent magnetization):
400: \be\label{Chi.def}
401: \chi^*(\hat{t},\t)=\chi(\hat{t},\t)+\D\chi(\hat{t},\t)
402: =\int_0^\t\!ds R^*(\hat{t}+\t,\hat{t}+s)
403: \ee
404: according to eq.(\ref{R.DR.allg}) with the zero-field integrated response
405: $\chi(\hat{t},\t)=\int_0^\t\!ds R(\hat{t}+\t,\hat{t}+s)$.
406: Eqns.(\ref{del.W})$\cdots$(\ref{Chi.def}) allow the calculation of the results
407: of a NHB experiment at any desired temperature.
408: \section*{III. Results and discussion}
409: Many of the general features of the modification $\D\chi(\hat{t},\t)$ can
410: already be seen for $T=0$, which is the simplest case.
411: Afterwards finite temperatures will be discussed as well as the dependence
412: of the observed features on spatial dimension. Finally, a direct comparison
413: between the three-dimensional s-FM model and the s-SK model will by carried
414: out.
415:
416: Throughout the remainder of the paper the dependence on $\hat{t}$ will be
417: skipped whenever no confusion can occur, i.e. the shorthand notation
418: $\chi(\t)\equiv\chi(\hat{t},\t)$ and $\D\chi(\t)\equiv\D\chi(\hat{t},\t)$ will
419: be used. Times and frequencies will be given in dimensionless units.
420: \subsection*{A. Spherical ferromagnet}
421: In this subsection the details of the results for calculations of the
422: response following the NHB field sequence are discussed.
423: The discussion is kept general with regard to spatial dimension but the
424: actual calculations are carried out for $d\!=\!3$, cf. Figs.3$\cdots$6.
425: The dependence of the results on spatial dimension will be presented in the
426: next subsection.
427: \subsubsection*{T=0:}
428: For $T\!=\!0$, eq.(\ref{del.W}) can be solved trivially and the modified
429: response is easily calculated. From eq.(\ref{W.t}) one explicitly has
430: \be\label{W.del.T0}
431: W(t)^2=g(t)\quad\mbox{and}\quad \D(t)^2=\D_0 (t)\quad ; \quad T=0
432: \ee
433: The corresponding expressions for $\chi(\t)$ and $\D\chi(\t)$ are easily
434: obtained from eqns.(\ref{R.DR.allg},\ref{Chi.def}).
435:
436: From the discussion of the zero field response in the last section it is
437: evident that the time $t_q$ elapsed after the quench and before application of
438: the pump-field is a very important parameter.
439: For small $t_q$ some transient features are expected due to the interplay
440: of the approach of the aging-regime and the additional non-equilibrium features
441: induced by the application of the pump-field.
442: Of course, the cross-over to a $\Om$ independent behavior will depend crucially
443: on the pump-frequency $\Om$ as this determines the time $t_p$ of the imposed
444: non-equilibrium situation.
445: This is demonstrated in Fig.3a, where $\D\chi(\t)$ is plotted for $d\!=\!3$,
446: $T\!=\!0$, $t_w\!=\!0$, $\Om=0.1, 10$ and several values of $t_q$.
447: From this figure two features are evident immediately. First of all it is seen
448: that $\D\chi(\t)$ is non-zero only in a limited time interval and the time of
449: the maximum modification depends on the burn frequency $\Om$, thus
450: demonstrating dynamic heterogeneous behavior.
451: Additionally, the curves for $t_q\!=\!0$ differ from the others in that they
452: change from positive to negative values in a limited time range. This transient
453: behavior also depends on $\Om$.
454: However, it is always possible to choose $t_q$ in a way that the mentioned
455: interplay between the two sources of transient features can be neglected.
456: Thus, it is interesting to consider the asymptotic regime determined by
457: \[
458: t_q\gg 1\quad\mbox{where}\quad t_q=t_q(\Om).
459: \]
460: The detailed dependence of $t_q$ on $\Om$ has to be found empirically in the
461: sense that no transient effects should show up in the modified response for
462: a given pump frequency $\Om$.
463: In order to further demonstrate the relative independence of the results from
464: the chosen value of $t_q$, in Fig.3b I have plotted the value of the maximum
465: modification, $\D\chi_{max}=\D\chi(\t_{max})$ vs. $t_q$ for
466: $\Om=0.01, 0.1, 10$ in a scaled way.
467: This plot demonstrates the features already mentioned above.
468: For small $t_q$ there is some time interval in which $\D\chi(\t)$ is negative
469: and therefore the maximum is reduced. The small hump around
470: $t_q\sim 2\pi/\Om$ in the curves is roughly located at those values of $t_q$
471: where also the negative part vanishes.
472: Finally for long $t_q$ a plateau is observed and the results are independent
473: of $t_q$.
474: Obviously, the minimum value of $t_q$ satisfying the constraint that
475: $\D\chi(\t)$ is independent of $t_q$ depends on $\Om$.
476: In the whole range of $\Om$ considered in the present paper it turned out that
477: a value of $t_q\!=\!10^6$ is sufficient. Therefore all further calculations
478: are performed for $t_q\!=\!10^6$ unless stated otherwise.
479: (I have checked via explicit calculations that
480: $\chi(t_q+t_p,\t)-\chi(t_q,\t)\simeq 0$ independent of $\Om$. Otherwise, the
481: interpretation of the $\Om$-dependence of $\D\chi(\hat{t},\t)$ in terms of dynamic
482: heterogeneities would be meaningless.)
483: From eqns.(\ref{g.as.def}) and (\ref{del.W}) it is seen that in this regime
484: one has for $\D_0(t_q+\t)$ for arbitrary dimension ($t_m=$Min$(\t,t_p)$):
485: \be\label{del.as}
486: \D_0(t_q+\t)=(8\pi)^{-d/2}\int_0^{t_m}\!dt_1\int_0^{t_m}\!dt_2
487: \sin{(\Om t_1)}\sin{(\Om t_2)}
488: { g\!\left(\t-\frac{t_1+t_2}{2}\right)\over
489: [(t_q+t_1)(t_q+t_2)]^{d/4}}
490: \quad t_q\gg 1
491: \ee
492: This expression along with eq.(\ref{R.DR.allg}) for the modification of the
493: response also explains the observed relatively weak dependence of
494: $\D\chi(\t)$ on $t_q$. In lowest order
495: $\D_0(\hat{t}+\t')$ behaves as $t_q^{-d/2}$ and according to eq.(\ref{W.as})
496: the same holds for $W(\hat{t}+\t')$. Therefore, $R$ and $\D R$ are independent
497: of $t_q$ in this order and the same holds for $\D\chi$. An explicit
498: $t_q$-dependence enters only in higher order.
499:
500: For $T=0$, the above expression is only needed for $t_m=t_p=2N\pi/\Om$, where
501: $N$ denotes the number of cycles of the sinusoidal field.
502: The reason for the fact that smaller values of $t_m$ are irrelevant in this
503: case is that according to eq.(\ref{W.del.T0}) $\D(t)^2=\D_0 (t)$ and that the
504: response is measured only after the pump-field is switched off.
505: Therefore, the question as to which extent the results depend on the number of
506: cycles naturally occurs.
507: Fig.3c shows $\D\chi(\t)$ as a function of $\t$ for $N=1, 5, 10$ for
508: $\Om=0.1$.
509: The curves for $N=5$ and $N=10$ are hardly distinguishable in this plot.
510: The saturation of the maximum amplitude is demonstrated in Fig.3d, where
511: $\D\chi_{max}$ is plotted vs. $N$ for $\Om=0.1$ and $10$ ($t_q=10^6$).
512: For $N$ values larger than roughly 6, $\D\chi_{max}$ become $N$-independent.
513: In all of these calculations one has $t_q\gg t_p$, whereas the opposite limit
514: is met at small $t_q$ as discussed above, although for $N\!=\!1$.
515: This demonstrates that the field sequence of NHB is unable to drive the system
516: much further away from equilibrium than it already is due to the quench
517: at $t=0$.
518:
519: After this consideration of the influence of the parameters $t_q$ and $N$ now
520: the more important issue of the $\Om$-dependence of $\D\chi(\t)$ will be
521: discussed.
522: Before presenting the results of model calculations it is instructive to
523: investigate analytically the limiting behavior of $\D_0(t_q+\t)$ according to
524: eq.(\ref{del.as}) which in turn determines the behavior of $\D\chi(\t)$.
525: One finds that $\D_0$ vanishes in the limits of large and small $\Om$
526: according to
527: \Be\label{del.limits}
528: &&\Om\to 0:\quad\D_0(\hat{t}+\t)\sim\Om^{d/2-2}\quad(\Om t_q\gg 1)\nonumber\\
529: &&\hspace{1.9cm}\D_0(\hat{t}+\t)\sim\Om^{d-2}\hspace{0.7cm}(\Om t_q\ll 1) \\
530: &&\hspace{-0.25cm}\Om\to\infty:\quad\D_0(\hat{t}+\t)\sim\Om^{-4}
531: \quad\hspace{0.6cm}\mbox{$\forall$ d}\nonumber
532: \Ee
533: demonstrating that the modification induced by the NHB pulse sequence in
534: principle depends on $\Om$.
535: The universal $\Om^{-4}$-dependence for large $\Om$ can easily be understood
536: by slightly rewriting eq.(\ref{del.W}) in the form ($M\!\equiv\!\Om t_m$)
537: \[
538: \D_0(t_q+\t)=\Om^{-2}\int_0^{M}\!dt_1\int_0^{M}\!dt_2\sin{t_1}\sin{t_2}
539: W\left(t_q+{t_1\over\Om}\right) W\left(t_q+{t_2\over\Om}\right)
540: g\left(\t-{t_1+t_2\over 2\Om}\right)
541: \]
542: The behavior for large $\Om$ is obtained via second-order expansions of
543: $W(t_q+\Om^{-1}x)$ and of $g(\t-\Om^{-1}x)$. In a simple calculation one then
544: finds the quoted $\Om^{-4}$ behavior.
545: Therefore, this represents a universal result which is valid for {\it any}
546: model obeying eq.(\ref{del.W}). As the $\Om$-dependence of $\D_0(t_q+t)$
547: uniquely determines the one of $\D(t_q+t)^2$, this result additionally holds
548: for all temperatures, including the disordered paramagnetic phase.
549:
550: In Fig.4a the dependence of $\D\chi(\t)$ on $\Om$ is demonstrated in detail
551: for $d\!=\!3$.
552: It is clearly seen that $\D\chi(\t)$ shows a very pronounced $\Om$-dependence
553: for $\Om\!<\!0.1$ which, however, diminishes with increasing $\Om$.
554: Additionally, it is evident that the spectral modifications become 'broader',
555: i.e. are non-vanishing in a larger time interval, as $\Om$ decreases.
556:
557: In Fig.4b I have plotted the time of the maximum modification, $\t_{max}$ vs.
558: $\Om$. From this plot one can see that $\t_{max}$ varies as $\Om^{-1}$ for
559: small $\Om$.
560: This is the behavior typical for an extremely broad distribution of relaxation
561: times.
562: Additionally, $\t_{max}$ becomes independent of $\Om$ for $\Om\sim 10$.
563: Thus, for the s-FM model in $d\!=\!3$, the response is dynamic homogeneous
564: in the short time regime, whereas it is dynamic heterogeneous for long times.
565: Therefore, the aging dynamics in this model is dynamic heterogeneous.
566: Also, it is seen from Fig.4b that the behavior is independent of the time
567: $t_q$ elapsed after the quench.
568:
569: Another important question regards the time scale of the 'recovery' of the
570: modification, i.e. the waiting time dependence. In ref.\cite{dieze01} I have
571: found that in a model of reorientational dynamics there is no extra time scale
572: for the recovery. However, in the experiments on the relaxor materials a very
573: long recovery time scale has been found\cite{KSB98}. In order to investigate
574: this question, in Fig.5a I have plotted the maximum modification (normalized to
575: the value at $t_w\!=\!0$), $\D\chi(\t_{max})_{norm.}$, vs. scaled waiting
576: time $\Om t_w$ for a variety of burn frequencies $\Om$. It is evident that
577: the life-time strongly depends on $\Om$ for small $\Om$ and that this
578: dependence diminishes for larger $\Om$.
579: It has to be mentioned at this point that the form of the modifications
580: hardly change for different waiting times.
581: In order to have a simple measure for the recovery times, I have fitted the
582: curves of Fig.5a to a Kohlrausch function, $\exp{\{-(t_w/\t_K)^{\b_K}\}}$,
583: and plotted the resulting time scales $\t_K$ vs. $\Om$ in Fig.5b.
584: I included the time of the maximum modification for $t_w\!=\!0$ (cf. Fig.4b)
585: in this plot (dot-dashed line).
586: As the behavior of these two time scales is almost identical to within a
587: factor of two, the conclusion to be drawn from these calculations is that
588: there does not exist an extra time scale for the recovery in the s-FM model.
589: This appears plausible on intuitive grounds because the only time scale set
590: by the pump is roughly $1/(2\pi\Om)$ in the domain growth regime and
591: just unity ($1/(2\pi))$ in the short-time regime.
592: This is qualitatively different from a complex domain-structured system like
593: the relaxor materials, where pinning effects play a dominant role.
594: \subsubsection*{T$>$0:}
595: So far, I have considered $T\!=\!0$ solely.
596: As already mentioned in the last section, the response in the ferromagnetic
597: phase is independent of temperature. For the NHB field sequence, however,
598: eqns.(\ref{del.W}) and (\ref{R.DR.allg}) show that $\D R$ and correspondingly
599: also $\D\chi(\t)$ do depend on temperature.
600: The physical reason for this dependence is quite clear.
601: Though the thermal noise is irrelevant for the linear response, this changes
602: under NHB conditions due to the aligning effect of the pump-field.
603: Here, thermal fluctuations tend to counterbalance the induced alignment.
604:
605: In order to investigate the importance of this effect, eq.(\ref{del.W})
606: is solved numerically for finite temperatures for $t_q\!\gg\!1$, i.e. using
607: eq.(\ref{W.as}) in the expression for $\D_0(t)$. (To ensure that the physical
608: solution of the equation is met, I first performed calculations for very small
609: T, in which case eq.(\ref{del.W}) can be solved in terms of an expansion in
610: $T/T_c$.) The results of such temperature dependent calculations
611: are shown in Fig.6a, where $\D\chi(\t)$ is plotted for various $\Om$ and for
612: temperatures up to $0.5T_c$ using $t_q\!=\!10^{6}$ and $t_w\!=\!0$.
613: Two features are evident by inspection of that plot.
614: First, the position of the maximum modification hardly changes as a function
615: of temperature. It should be mentioned that also the shape of the $\D\chi(\t)$
616: does not change as a function of temperature. This means that no effect
617: of 'motional narrowing' is observable.
618:
619: The most prominent feature is the increasing intensity of $\D\chi(\t)$. In
620: order to demonstrate this behavior in more detail, the intensity at the time
621: of the maximum modification, $\D\chi(\t_{max})$, is shown as a function of
622: reduced temperature $1-T/T_c$ in Fig.6b for $\Om\!=\!0.01, 100$.
623: This plot demonstrates a scaling behavior of these quantities.
624: The exponents, however, are different for various frequencies and do not seem
625: to have an obvious explanation.
626: It would be interesting to further analyze the $\Om$-dependence of
627: $\D\chi(\t_{max})$, which, however, is beyond the scope of the present study.
628:
629: The conclusion to be drawn from these calculations is that the effect of
630: temperature is seen mainly in the amplitude of the modification, but narrowing
631: effects are absent. This means that the nature of the dynamic heterogeneities
632: are not affected by temperature-effects. Therefore, in the following, I will
633: concentrate on $T=0$ for simplicity.
634: \subsection*{B. Spherical-ferromagnet: NHB for varying dimension}
635: In contrast to the s-SK model the s-FM model offers the opportunity to study
636: the behavior of the observed dynamic heterogeneities as a function of spatial
637: dimension.
638: As already mentioned in Sect.II the fluctuation-dissipation ratio $X_\infty$
639: vanishes for the s-FM model independent of spatial dimension. Additionally, it
640: is known\cite{GL00} that the usual static critical exponents take on their
641: mean-field values for $d\!>\!4$ while some dynamical exponents depend on
642: dimension for arbitrary $d$.
643: The only hint for a change in the dynamic properties of coarsening models
644: for $d\!>\!4$ stems from the large N-model, where it has been shown that the
645: aging contribution to the integrated response changes qualitatively for
646: $d\!>\!4$\cite{CLZ02}.
647: The two-time quantities in the spherical models considered here, however, do
648: not show any signature of a change in behavior around $d\!=\!4$.
649: Therefore, in this subsection, the dependence of the behavior of $\D\chi(\t)$
650: and thus of the dynamic heterogeneities on spatial dimension will be
651: investigated for $T\!=\!0$.
652:
653: Fig.7a shows a plot of $\D\chi(\t)$ versus $\t$ for $d\!=\!5$ (upper panel) and
654: $d\!=\!7$ (lower panel), which is to be compared to Fig.4a.
655: (I do not consider even spatial dimensions like $d\!=\!6$, because the
656: calculations are much more involved for technical reasons.)
657: In both cases the $\Om$-dependence is much weaker than in $d\!=\!3$.
658: Whereas there is some weak $\Om$-dependence for $d\!=\!5$, such a dependence
659: is hardly visible for $d\!=\!7$.
660: To quantify this diminishing $\Om$-dependence, in Fig.7b I have plotted the
661: time, $\t_{max}$, of the maximum modification as a function of $\Om$.
662: Whereas one has a $\Om^{-1}$ dependence in the long-time regime for
663: $d\!=\!3$, this is weaker, roughly $\Om^{-1/2}$, for $d\!=\!5$ and finally in
664: $d\!=\!7$ and $d\!=\!9$ there is no visible $\Om$-dependence. This means that
665: the dynamics in $d\!=\!7$ an $d\!=\!9$ is {\it dynamic homogeneous} in the
666: short-time {\it and} the long-time regime.
667: Thus, also the aging dynamics becomes dynamic homogeneous for higher spatial
668: dimension.
669:
670: The conclusion from these calculations is that the dynamics of the s-FM model
671: becomes dynamic homogeneous in the mean-field limit, $d\!\gg\!1$.
672: However, as pointed out earlier, the mean-field limit holds for all $d\!>\!4$
673: when concerned with static properties. For $d\!=\!5$, Fig.7 reveals that the
674: aging dynamics still is heterogeneous, although the shape of an effective
675: distribution of relaxation times is changed relative to $d\!=\!3$.
676:
677: \subsection*{C. NHB in the spherical Sherrington-Kirkpatrick model}
678: As already noted in the preceeding section the dynamic behavior of the s-FM
679: model in $d\!=\!3$ and the s-SK model is very similar in the low temperature
680: phase.
681: Concerning the modified response $\chi^*(\t)$ the same holds true, in
682: particular in the limit of large $t_q$.
683: The only difference stems from the functions $g(t)$ occuring in the expressions
684: for $\D_0(t_q+\t)$ and $R(\hat t+\t,\hat t +\t')$,
685: eqns.(\ref{R.DR.allg},\ref{del.as}), for the two models.
686: Plotting the two functions $g_{FM}(t)$ and $g_{SK}(t)$ reveals that
687: $g_{FM}(t)\!\simeq\! g_{SK}(t/3)$ in a good approximation for moderate $t$ and
688: asymptotically one has
689: $g_{SK}(t)\simeq[(8\pi)^{3/2}/(32\pi)^{1/2}]g_{FM}(t)\!\simeq\!12.57g_{FM}(t)$,
690: cf. eq.(\ref{g.as.def}).
691: For large $\Om$ small times $\t$ are relevant and therefore one expects
692: $\D\chi_{SK}(\t)\simeq\D\chi_{FM}(\t)$ to hold with the difference that
693: the maximum modification occurs at $\t_{max}^{SK}\!\simeq\!3\t_{max}^{FM}$.
694: On the other hand, for small $\Om$ long times are most important and all
695: functions $g(t)$ can be replaced by their asymptotic values. Thus, one
696: roughly has $\D\chi_{SK}(\t)\simeq[(8\pi)^{3}/(32\pi)]\D\chi_{FM}(\t)$
697: and $\t_{max}^{SK}\!\simeq\!\t_{max}^{FM}$.
698: In Fig.8a I have plotted $\D\chi(\t)$ for both models for $\Om=10^{-4}$
699: (upper panel) and $\Om=100$ (lower panel).
700: Apart from a factor of two in the amplitude for the larger frequency the
701: behavior just discussed is recovered.
702:
703: Fig.8b shows the time of the maximum modification, $\t_{max}$ versus
704: frequency $\Om$.
705: The full line is for the 3d s-FM, cf. Fig.4b, and the dashed line is for the
706: s-SK model.
707: Also included is $(\t_{max}^{SK}/3)$ for $\Om>1$ (dot-dashed line).
708: From the above discussion it is clear that one just finds the expected
709: behavior: for very small $\Om$ one has $\t_{max}^{SK}\!\simeq\!\t_{max}^{FM}$
710: whereas $\t_{max}^{SK}\!\simeq\!3\t_{max}^{FM}$ holds for large $\Om$.
711:
712: These considerations show that the qualitative behavior of the s-FM model for
713: $d\!=\!3$ and the s-SK model are extremely similar in the low temperature
714: phase, including the dynamic heterogeneities.
715: The similarity between the 3d s-FM and the s-SK model is well known and it has
716: been argued earlier that the s-SK model does not generically behave like a
717: spin glas, concerning both, the statics and the dynamics\cite{ZKH00}.
718:
719: The fact that there are dynamic heterogeneities observable in the s-SK model
720: might appear somewhat astonishing in view of the fact that the s-SK
721: model is a mean-field model.
722: However, this result is not unique.
723: The first observation of dynamic heterogeneous relaxation in a mean-field
724: model was reported in ref.\cite{leticia01}.
725: Also the Monte Carlo simulations on the Sherrington-Kirkpatrick model
726: mentioned above\cite{let.reply} show dynamic heterogeneous behavior.
727: The present results therefore confirm the fact that an identification of
728: {\it dynamic} and {\it spatial} heterogeneities is not possible in general.
729: \section*{IV. Conclusions}
730: In the present paper I presented calculations of the response of simple
731: spherical spin models to the field sequence of the nonresonant hole burning
732: experiment, a technique allowing the detection of dynamic heterogeneities if
733: they exist.
734: The calculations were restricted to the generic situation of aging in the
735: low temperature phase. It was assumed that the system is quenched from
736: infinite temperature to $T\!<\!T_c$ at the beginning, correlations in the
737: initial conditions thus being neglected completely. The equilibrium dynamics
738: will be considered in a forthcoming publication.
739:
740: The main result of the present paper is the fact that the non-equilibrium
741: dynamics of the s-FM model in $d\!=\!3$ is dynamic heterogeneous in the
742: aging regime whereas it appears homogeneous in the short-time regime.
743: In higher dimensions, $d\!\geq\!7$, the response becomes dynamic homogeneous
744: also in the aging regime.
745: Therefore, the nature of the aging dynamics changes and in the mean-field
746: limit, $d\!\gg\!1$, the model displays homogeneous dynamics.
747: This is what one would expect as in this limit each spin interacts with an
748: infinite number of neighboring spins.
749: However, for the s-SK model, a simple mean-field spin glass model, a behavior
750: very similar to that of the s-FM model in $d\!=\!3$ is found.
751: This explicitly demonstrates that the existence of dynamic heterogenities does
752: not tell us anything about spatial heterogeneities in the system.
753: Furthermore, a strongly heterogeneous dynamics has also been observed in the
754: aging behavior of a short range spin glass model\cite{CCCK02} showing strong
755: relations between local responses and correlations.
756: The mere existence of dynamic heterogeneities in a short range spin glass
757: is of course to be expected and here also the spatial aspects of these
758: heterogeneities are of importance.
759:
760: When considering the spherical two-spin interaction models, it is tempting to
761: associate some disorder with the spherical constraint which forces the equal
762: time correlation function to unity at all times.
763: The lengths of the spins are not static quantities and are random to some
764: extent.
765: This might be viewed as a kind of 'dynamic' disorder and the similarities
766: between the s-FM model in $d\!=\!3$ and the s-SK model hint towards the
767: irrelevance of the quenched disorder in the latter when concerned with
768: dynamic quantities.
769: On a speculative level the fact that the dynamics in the s-FM model becomes
770: homogeneous for $d\!\gg\!1$ can be understood from the following argument.
771: Assuming the existence of an effective distribution of relaxation rates at a
772: given instant of time, i.e. for a given distribution of $s_i(t)^2$
773: the width of this distribution is expected to decrease with an increasing
774: number of nearest neighbor interactions and consequently the dynamics
775: becomes more homogeneous. This of course does not mean that the response
776: decays in an exponential way for large $d$ because the considered distribution
777: is not a static quantity.
778: Notice that this argument also implies that the life-time of the
779: heterogeneities is finite. Unfortunately, NHB does not allow to determine
780: this life-time.
781: It is, however, important to point out that the argument concerning the
782: increasing number of neighbors, though appealing with respect to the
783: s-FM model, has important drawbacks. Taking it serious in case of the
784: s-SK model would predict homogeneous dynamics in contrast to what is
785: observed.
786:
787: Although the s-FM model is a typical model for phase-ordering kinetics,
788: there is no obvious relationship between the observed dynamic heterogeneities
789: and the domain size distribution in a coarsening system.
790: A possible way to investigate such a relationship could be to perform
791: calculations on the Ising model using spatially varying magnetic fields.
792: It would also be interesting to perform an analysis along the lines of
793: refs.\cite{CCCK02, MT03} in order to see whether the dynamic heterogeneities
794: in a coarsening system behave similar to those observed in glassy systems.
795: Furthermore, such calculations would allow to investigate the dependence on
796: spatial dimension and therefore to check whether the one observed in the
797: present paper for the s-FM model also is found for other domain-coarsening
798: models.
799:
800: In summary, I have shown that heterogeneous aging can be observed in the
801: low-temperature phase of the spherical model of a ferromagnet. The aging
802: dynamics becomes homogeneous on increase of the spatial dimension despite the
803: fact that no qualitative change in the two-time quantities like the correlation
804: function or the response is observed.
805: Quenched disorder does not play any significant role with respect to
806: heterogeneities in spherical models.
807: \subsection*{Acknowledgements}
808: I thank R. B\"ohmer, G. Hinze, U. H\"aberle and R. Schilling for fruitful
809: discussions.
810: This work was supported by the DFG under contract Di693/1-1.
811: \newpage
812: \begin{thebibliography}{XX}
813: \bibitem{a.rel}
814: for reviews see:
815: K. Binder and A.P. Young; Rev. Mod. Phys. {\bf 58} 801 (1986)\\
816: C.A. Angell; J. Phys. Chem. Solids {\bf 8} 863 (1988)\\
817: U. Mohanty; Adv. Chem. Phys. {\bf 89} 89 (1995)\\
818: M.D. Ediger, C.A. Angell and S.R. Nagel; J. Phys. Chem. {\bf 100} 13200
819: (1996)
820: \bibitem{het.dyn}
821: M.D. Ediger; Annu. Rev. Phys. Chem. {\bf 51} 99 (2000)\\
822: H.Sillescu; J. Non-Cryst. Solids {\bf 243} 81 (1999)\\
823: R. B\"ohmer; Current Opinion in Solid State and Material Science
824: {\bf 3} 378 (1998)\\
825: R. B\"ohmer, R.V. Chamberlin, G. Diezemann, B. Geil, A. Heuer, G. Hinze,
826: S.C. Kuebler, R. Richert, B. Schiener, H. Sillescu, H.W. Spiess, U. Tracht
827: and M. Wilhelm; J. Non-Cryst. Solids {\bf 235-237} 1 (1998)
828: \bibitem{SRS91}
829: K. Schmidt-Rohr and H.W. Spiess; Phys. Rev. Lett. {\bf 66} 3020 (1991)
830: \bibitem{CE95}
831: M.T. Cicerone and M.D. Ediger; J. Chem. Phys. {\bf 103} 5684 (1995)
832: \bibitem{ViRI00}
833: E. Vidal- Russell and N.E. Israeloff; Nature {\bf 408} 695 (1999)
834: \bibitem{DvB01}
835: L.A. Deschenes and D. Vanden Bout; Science {\bf 292} 255 (2001)
836: \bibitem{het.def}
837: H. Sillescu, R. B\"ohmer, G. Hinze and G. Diezemann;
838: J. Non-Cryst. Solids {\bf 307-310} 16 (2002)
839: \bibitem{andreas.4d}
840: A. Heuer, M. Wilhelm, H. Zimmermann and H.W. Spiess; Phys. Rev. Lett.
841: {\bf 75} 2851 (1995)
842: \bibitem{roland.4d}
843: R. B\"ohmer, G. Hinze, G. Diezemann, B. Geil and H. Sillescu; Europhys. Lett.
844: {\bf 36} 55 (1996)
845: \bibitem{heuer97}
846: A. Heuer; Phys. Rev. {\bf E 56} 730 (1997)
847: \bibitem{dieze97}
848: G. Diezemann; J. Chem. Phys. {\bf 107} 10112 (1997)
849: \bibitem{4dcp}
850: U. Tracht, M. Wilhelm, A. Heuer, H. Feng, K. Schmidt-Rohr and H.W. Spiess;
851: Phys. Rev. Lett. {\bf 81} 2727 (1998)
852: \bibitem{4dcp.gly}
853: S.A. Reinsberg, X.H. Qiu, M. Wilhelm, H.W. Spiess and M.D. Ediger;
854: J. Chem. Phys. {\bf 114} 7299 (2001)
855: \bibitem{4dcp.vgl}
856: S.A. Reinsberg, A. Heuer, B. Doliwa, H. Zimmermann and H.W. Spiess;
857: J. Non-Cryst. Solids {\bf 307-310} 208 (2002)
858: \bibitem{SBLC96}
859: B. Schiener, R.B\"ohmer, A. Loidl and R.V. Chamberlin; Science {\bf 274} 752
860: (1996)
861: \bibitem{SCDB97}
862: B. Schiener, R.V. Chamberlin, G. Diezemann and R. B\"ohmer;
863: J. Chem. Phys. {\bf 107} 7746 (1997)
864: \bibitem{KSB98}
865: O. Kircher, B. Schiener and R. B\"ohmer; Phys. Rev. Lett. {\bf 81} 4520 (1998)
866: \bibitem{RB99}
867: R. Richert and R. B\"ohmer; Phys. Rev. Lett. {\bf 83} 4337 (1999)
868: \bibitem{ralph99}
869: R.V. Chamberlin; Phys. Rev. Lett. {\bf 83} 5134 (1999)
870: \bibitem{leticia01}
871: L.F. Cugliandolo and J.L. Iguain; Phys. Rev. Lett. {\bf 85} 3448 (2000)
872: \bibitem{kubo}
873: R. Kubo, M. Toda and N. Hashitsume {\it Statistical Physics II} Springer,
874: Berlin-Heidelberg-New York (1992)
875: \bibitem{dieze01}
876: G. Diezemann; Europhys. Lett. {\bf 53} 604 (2001)
877: \bibitem{DB01}
878: G. Diezemann and R. B\"ohmer; Phys. Rev. Lett. {\bf 87} 129602 (2001)
879: \bibitem{let.reply}
880: L.F. Cugliandolo and J.L. Iguain; Phys. Rev. Lett. {\bf 87} 129603 (2001)
881: \bibitem{CD95}
882: L.F. Cugliandolo and D.S. Dean; J. Phys. {\bf A 28} 4213 (1995)\\
883: L.F. Cugliandolo and D.S. Dean; J. Phys. {\bf A 28} L453 (1995)
884: \bibitem{bray94}
885: A.J. Bray; Adv. Phys. {\bf 43} 357 (1994)
886: \bibitem{GL00}
887: C. Godreche and J.M. Luck; J. Phys. {\bf A 33} 9141 (2000)
888: \bibitem{ZKH00}
889: W. Zippold, R. K\"uhn and H. Horner; Eur. Phys. J. {\bf B 13} 531 (2000)
890: \bibitem{FDT_rev98}
891: J.P. Bouchaud, L.F. Cugliandolo, J. Kurchan and M. Mezard; in
892: {\it Spin glasses and random fields}, Ed. A.P. Young, World Scientific,
893: Singapore (1998)
894: \bibitem{X_infty}
895: A. Barrat; Phys. Rev. {\bf E 57} 3629 (1998)\\
896: L. Berthier, J.L. Barrat and J. Kurchan; Eur. Phys. J. {\bf B 11} 635 (1999)
897: \bibitem{BCI01}
898: L. Berthier, L.F. Cugliandolo and J.L. Iguain;
899: Phys. Rev. {\bf E 63} 051302 (2001)
900: \bibitem{BK52}
901: T.H. Berlin and M. Kac; Phys. Rev. {\bf 86} 821 (1952)
902: \bibitem{CLZ02}
903: F. Corberi, E. Lippiello and M. Zannetti; Phys. Rev. {\bf E 65} 046136 (2002)
904: \bibitem{CCCK02}
905: H.E. Castillo, C. Chamon, L.F. Cugliandolo and M.P. Kennett;
906: Phys. Rev. Lett. {\bf 88} 237201 (2002)
907: \bibitem{MT03}
908: A. Montanari and F. Ricci-Tersenghi; Phys. Rev. Lett. {\bf 90} 017203 (2003)
909: \end{thebibliography}
910: \newpage
911: \section*{Figure captions}
912: \begin{description}
913: \item[Fig.1 : ] The response function $R(\t+t_w,t_w)$ for the s-FM model
914: in the ferromagnetic phase, $T<T_c$, for $t_w=10^2,10^3,10^4,10^5$ (from upper
915: to lowest line). Upper panel: $d=3$, middle panel: $d=5$, lower panel: $d=7$.
916: \item[Fig.2 : ] The field sequence for the nonresonant hole burning (NHB)
917: experiment: A time $t_q$ after a quench from $T=\infty$ one or more cycles
918: of a strong sinusoidal field $h_p(t)=h_p\sin{(\Om(t-t_q))}$ are applied.
919: After a waiting time $t_w$ the response to an infinitesimally small field is
920: monitored.
921: \item[Fig.3 : ]
922: {\bf a:} The modification of the response, $\D\chi(\t)$ vs. $\t$ for
923: the s-FM, $d\!=\!3$. Upper panel: $\Om\!=\! 0.1$; lower panel:
924: $\Om\!=\! 10$.
925: The time after the quench from infinite temperature, $t_q$, is chosen as
926: $t_q\!=\! 0, 10, 10^6$.\\
927: {\bf b:} $\D\chi_{max, sc.}=\D\chi_{max}(t_q)/\D\chi_{max}(t_q\!=\!10^{-6})$
928: versus $t_q$ for $\Om=0.01, 0.1, 10$.\\
929: {\bf c:} $\D\chi(\t)$ versus $\t$ for $\Om=0.1$ and $t_q=10^6$ for
930: $N=1$ (full line), $N=5$ (dashed line) and $N=10$ (dotted line).
931: The curves for $N=5$ and $N=10$ are hardly to distinguish.\\
932: {\bf d:} $\D\chi_{max, sc.}=\D\chi_{max}(N)/\D\chi_{max}(N\!=\!1)$ as a
933: function of the number of cycles of the sinusoidal field, $N$ ($t_q=10^6$).
934: \item[Fig.4 : ]
935: {\bf a:} $\D\chi(\t)$ vs. $\t$ for $t_q\!=\!10^6$, $t_w\!=\!0$,
936: $d\!=\!3$ and $\log{(\Om)}=1, 0, -1, -2, -3, -4$ (from left to right).
937: \\
938: {\bf b:} Full line: $\t_{max}$ vs. $\Om$, dashed line: $\Om\t_{max}$ vs. $\Om$
939: for $t_q\!=\!10^6$, $t_w\!=\!0$. Additionally shown as the dot-dashed line is
940: $\t_{max}$ vs. $\Om$ for $t_q\!=\!1$ and $t_w\!=\!0$, demonstrating that the
941: behavior does not depende on $t_q$.
942: \item[Fig.5 : ]
943: {\bf a:} $\D\chi(\t_{max})_{norm.}$ vs. the scaled waiting time $\Om t_w$
944: for $\Om=10^{-4}, 1, 5, 10, 25, 50, 10^2$ (from left to right).\\
945: {\bf b:} Characteristic decay time $\t_K$ of a Kohlrausch-fit of the from
946: $\D\chi(\t_{max})_{norm.}=\exp{\{-(t_w/\t_K)^{\b_K}\}}$ ($t_q\!=\!10^6$).
947: The stretching parameter is approximately constant, $\b_K\!\simeq\!0.9$.
948: The dot-dashed line represents $\t_{max}$ for $t_w=0$ for comparison, cf.
949: Fig.4b.
950: \item[Fig.6 : ]
951: {\bf a:}
952: $\D\chi(\t)$ vs. temperature for $T=0,0.1,0.2,0.3,0.4,0.5$ for the burn
953: frequencies given in the respective panels. The remaining parameters are
954: $t_q\!=\!10^6$, $t_w\!=\!0$ and $d\!=\!3$.\\
955: {\bf b:}
956: The value of the maximum modification, $\D\chi(\t_{max})$,
957: vs. $1-T/T_c$. The other parameters are the same as in {\bf a}.
958: \item[Fig.7 : ]
959: {\bf a:}
960: $\D\chi(\t)$ vs. $\t$ for burn frequencies $\Om=1,10,20,50,100$ for
961: $d\!=\!5$ (upper panel) and $d\!=\!7$ (lower panel).
962: The remaining parameters are $t_q\!=\!10^6$ and $t_w\!=\!0$.\\
963: {\bf b:}
964: The time of the maximum modification, $\t_{max}$, for $d\!=\!3$, $d\!=\!5$
965: and $d\!=\!7$ vs. $\Om$. In the upper panel $\t_{max}$ is shown and in the
966: lower one the product $\Om\t_{max}$.
967: Full lines are for $t_q=10^{6}$ and the dot-dashed lines for $t_q=0$.
968: \item[Fig.8 : ]
969: {\bf a:}
970: $\D\chi(\t)$ vs. $\t$ for the 3-dimensional s-FM (full line) and the s-SK
971: model (dashed line). Upper panel: $\Om\!=\!10^{-4}$ and the factors
972: muliplying $\D\chi(\t)$ are $N(SK)=32\pi$ and $N(FM)=(8\pi)^3$. Lower panel:
973: $\Om\!=\!10^2$.
974: The remaining parameters are $t_q\!=\!10^6$ and $t_w\!=\!0$.\\
975: {\bf b:}
976: The time of the maximum modification, $\t_{max}$, for the 3d s-FM model (c.f.
977: Fig.4b, full line) and the s-SK model (dashed line) vs. $\Om$ for the same
978: parameters as in {\bf a}.
979: Also included is $\t_{max}^{SK}/3$ (dot-dashed line) for $\Om\!>\!1$.
980: \end{description}
981: \newpage
982: \begin{figure}
983: \includegraphics[width=15cm]{fig1.eps}
984: \end{figure}
985: \newpage
986: \begin{figure}
987: \includegraphics[width=15cm]{fig2.eps}
988: \end{figure}
989: \newpage
990: \begin{figure}
991: \includegraphics[width=15cm]{fig3a.eps}
992: \end{figure}
993: \newpage
994: \begin{figure}
995: \includegraphics[width=15cm]{fig3b.eps}
996: \end{figure}
997: \newpage
998: \begin{figure}
999: \includegraphics[width=15cm]{fig3c.eps}
1000: \end{figure}
1001: \newpage
1002: \begin{figure}
1003: \includegraphics[width=15cm]{fig3d.eps}
1004: \end{figure}
1005: \newpage
1006: \begin{figure}
1007: \includegraphics[width=15cm]{fig4a.eps}
1008: \end{figure}
1009: \newpage
1010: \begin{figure}
1011: \includegraphics[width=15cm]{fig4b.eps}
1012: \end{figure}
1013: \newpage
1014: \begin{figure}
1015: \includegraphics[width=15cm]{fig5a.eps}
1016: \end{figure}
1017: \newpage
1018: \begin{figure}
1019: \includegraphics[width=15cm]{fig5b.eps}
1020: \end{figure}
1021: \newpage
1022: \begin{figure}
1023: \includegraphics[width=15cm]{fig6a.eps}
1024: \end{figure}
1025: \newpage
1026: \begin{figure}
1027: \includegraphics[width=15cm]{fig6b.eps}
1028: \end{figure}
1029: \newpage
1030: \begin{figure}
1031: \includegraphics[width=15cm]{fig7a.eps}
1032: \end{figure}
1033: \newpage
1034: \begin{figure}
1035: \includegraphics[width=15cm]{fig7b.eps}
1036: \end{figure}
1037: \newpage
1038: \begin{figure}
1039: \includegraphics[width=15cm]{fig8a.eps}
1040: \end{figure}
1041: \newpage
1042: \begin{figure}
1043: \includegraphics[width=15cm]{fig8b.eps}
1044: \end{figure}
1045: \end{document}