1: \documentstyle[12pt]{article}
2:
3: \voffset -2cm
4: \hoffset -1.1cm
5: \textheight 240mm
6: \textwidth 150mm
7:
8: \renewcommand{\thefootnote}{\fnsymbol{footnote}}
9: \renewcommand{\arraystretch}{2}
10:
11: \begin{document}
12: \baselineskip 18pt
13:
14: \begin{center}
15: {\Large\bf Dynamic nonlinear (cubic) susceptibility
16:
17: in quantum Ising spin glass}\\
18:
19: \bigskip
20:
21: {G. Busiello$^a$\footnote{Corresponding
22: author,tel.(0)39-89-965426 -
23: fax (0)39-89-965275 - e-mail: busiello@sa.infn.it}}\\
24:
25: \medskip
26:
27: {\small\it Dipartimento di Fisica ``E.R. Caianiello",
28: Universit\`a
29: di Salerno,\\
30: 84081 Baronissi - Salerno and INFM - Unit\`a di Salerno, Salerno, Italy}\\
31: \medskip
32: {R. V. Saburova and V. G. Sushkova}\\
33: {\small\it Kazan State Power-Engineering University, Kazan, Russia}
34: \end{center}
35:
36: \bigskip
37: {\begin{abstract} Dynamic nonlinear (cubic) susceptibility in
38: quantum d-dimensional Ising spin glass with short-range
39: interactions is investigated on the basis of quantum droplet model
40: and quantum-mechanical nonlinear response theory. Nonlinear
41: response depends on the tunneling rate for a droplet which
42: regulates the strength of quantum fluctuations. It shows a strong
43: dependence on the distribution of droplet free energies and on the
44: droplet length scale average. Comparison with recent experiments
45: on quantum spin glasses like disordered dipolar quantum Ising
46: magnet is discussed .\end{abstract}}
47:
48: \bigskip \bigskip \bigskip \bigskip \bigskip
49:
50: \noindent PACS numbers: 75.40.Gb ; 75.10.Nr ; 64.70.Pf \\
51: \bigskip \bigskip
52: Keywords: A. Spin glasses ; D. Phase transition
53: \newpage
54:
55:
56: The dynamics of glassy systems is an attractive and rapidly
57: developing field of physics [1-5]. Spin glasses and quantum spin
58: glasses are very interesting systems for theoretical and
59: experimental investigation of dynamic phenomena [2-4]. Two major
60: theoretical description of spin glass phenomena have evolved over
61: the past twenty years:the mean field theory based on the Parisi
62: replica symmetry-breaking approach and the droplet model based on
63: renormalization group arguments [1-6]. The two pictures give very
64: different physical interpretation of observable spin-glass
65: phenomena. In this paper we investigate theoretically nonlinear
66: cubic dynamic susceptibility as function of frequency and
67: temperature in the Ising spin glass in a transverse field in terms
68: of quantum droplet model at very low temperatures (quantum
69: regime). The quantum phase transition is governed by quantum
70: fluctuations of the system which may tunnel from one local minimum
71: of the free energy to another; new physical effects such as
72: quantum channel of relaxation appear. There are few theoretical
73: studies on the nonlinear static response in quantum spin glasses
74: [7-10] and almost no studies on the dynamic nonlinear response
75: [3]. In [3] dynamic nonlinear response of a quantum spin glass was
76: found frequency independent and nonsingular in quantum critical
77: regime in contrast to the behavior in usual spin glass. There are
78: experimental data on the nonlinear dynamic response in classical
79: [11-13] and quantum [14] spin glasses
80: investigated by Fourier-transform technique. The
81: third-order nonlinear susceptibility is negative and diverges at
82: an ordinary spin glass transition temperature $T_{g}$ from both
83: the upper and the lower sides. But when $\chi'_{3}$ is measured by
84: a finite probing frequency the response falls out equilibrium
85: before the transition temperature and does not diverge at $T_{g}$.
86: Then $\chi'_{3}(\omega)$ shows a maximum at $T\simeq
87: T_{f}(\omega)$ where $T_{f}(\omega)$ is the freezing temperature
88: which is the upper bound on $T_{g}$ and $T_{g}=T_{f}(\omega
89: \rightarrow 0)$. Such behavior was observed, for example, for
90: classical Ising spin glass ${\rm Fe}_{0.5}{\rm Mr}_{0.5}{\rm
91: TiO}_{3}$ [13]. W. Wu et al. [14] measured nonlinear
92: susceptibility $\chi'_{3}(\omega , T)$ in quantum spin glass (the
93: diluted dipolar-coupled Ising spin glass ${\rm LiHo}_{0.167}{\rm
94: Y}_{0.833}{\rm F}_{4}$ in the transverse field) tuning transverse
95: field $\Gamma$ from the $\Gamma = 0$ classical to the $T=0$
96: quantum limit. At $mK$ temperatures they found a clear dynamic
97: signature of the spin glass to paramagnet transition whether
98: dominated by thermal or quantum fluctuations. In [14] it was shown
99: that the $\chi'_{3}$ depends on frequency for $\omega > 10 Hz$.
100: However it depends very weakly on $\omega$ for $\omega < 10Hz$.
101: There is a crossover between high $\omega$ ($\omega$ - dependent)
102: and low $\omega$ ($\omega$ - independent) behaviors. Nonlinear
103: susceptibility contains a diverging component which dominates at
104: $T=98 mK$, but disappears by $25 mK$. The $\chi'_{3}(\omega)$ does
105: not diverge but shows a maximum at $T_{f}(\omega)$.
106: $\chi'_{3}(\omega)$ measured at a higher temperature and lower
107: transverse field has a larger maximum than $\chi'_{3}(\omega)$
108: measured at a lower temperature and larger transverse field. The
109: analysis of these experimental data seems not clear [15] because
110: frequencies used in the experiments [14] are not sufficiently low
111: such as to determine the equilibrium behavior of system. Contrary
112: to the theoretical expectations, quantum transitions may be
113: qualitatively different from thermally driven transitions in real
114: spin glasses. Recently the linear dynamic susceptibility (in-phase
115: and out of phase components) at $T=0$ was investigated
116: theoretically for the Ising spin glass in a transverse field in
117: terms of quantum droplet model by M.~J.~Thill and D.~A.~Huse [6].
118: A basic assumption of the droplet picture is that the spin glass
119: dynamics is governed by large-scale excitations whose relaxation
120: time increases with length scale. In previous papers [16] we have
121: calculated the real and imaginary parts of linear dynamic
122: susceptibility in the same model, as in [6], for very low nonzero
123: temperatures using the general linear response theory of magnetic
124: dispersion and absorption phenomena for quantum systems by Kubo
125: and Tomita [17]. We note that in [6] the real part of the cubic
126: nonlinear ac susceptibility was defined as the in-phase $3\omega$
127: magnetization response $M(3\omega)$ to a small time-dependent
128: applied field $h\cos (\omega t)$
129: \begin{equation}
130: \chi'_{3}=\lim_{h\rightarrow 0}{24M(3\omega)\over h^{^3}V}
131: \end{equation}
132: where $V$ is the sample volume. The authors of [6] gave some
133: expression for $\chi'_{3}$ they only expect at zero temperature.
134:
135: The full nonlinear response theory despite its generality and
136: importance is of limited practical value because it is
137: mathematically difficult. It is necessary to make approximations,
138: as the well-known perturbation expansion of the time-evolution
139: operator, using Hamiltonian with some small parameter. Nonlinear
140: response theory was developed and described, for example, in
141: [17-22]. Here we summarize briefly the theory of higher-order
142: dynamic response. It is based on the Hamiltonian
143: \begin{equation}
144: \hat{\cal H}_{t}=\hat{\cal H}_{0}+\hat{\cal H}_{1}=\hat{\cal
145: H}_{0}-\hat{A}_{j}F_{j}(t),\,\,t \geq t_{0}
146: \end{equation}
147: where $\hat{\cal H}_{0}$ is nonperturbation Hamiltonian of system,
148: $\hat{\cal H}_{1}$ is perturbation Hamiltonian which describes the
149: interaction between the Heisenberg operators $\hat{A}_{j}$
150: (material operators) and the external perturbation $\hat{F}_{j}$.
151: At time $t>t_{0}$ one is interested in the expectation value of
152: the Heisenberg operator $\hat{B}_{i}$ which is given by
153: \begin{equation} %3
154: \langle\hat{B}_{i}(t)\rangle = Tr[\rho_{0}\hat{B}_{i}(t)] =
155: Tr[\hat{\rho}(t)\hat{B}_{i}]
156: \end{equation}
157: where the density matrix $\hat{\rho}(t)=\hat{U}(t,
158: t_{0})\hat{\rho}(t_{0})\hat{U}^{\dagger}(t,t_{0})$. The
159: time-evolution operator $\hat{U}$ satisfies the Schr\"odinger
160: equation $ih{d\hat{U} \over dt}=\hat{\cal H}\hat{U}$. It is
161: difficult to find an expression for $\hat{U}$ in closed form. It
162: was used that $\hat{\cal H}_{1}(t)$ is in some sense small. We
163: define the total response of the system at time $t$ to the
164: external force $F_{j}$ as the difference
165: \begin{equation} %5
166: \Delta B_{i}(t)\equiv \langle\hat{B}_{i}(t)\rangle-\langle\hat{B}_{i}\rangle_{0}
167: \end{equation}
168: where the subscript zero on expectation values refers to the
169: equilibrium expectations. One can then understand the behavior of
170: the system in terms of the dynamical response. Using
171: aforementioned expressions [2-3] the dynamical response can be
172: written through the third order in the perturbation $\hat{\cal
173: H}_{1}$ in the following form [20]
174: %6
175: $$
176: \bigl<\hat{B}_{i}(t)\bigr>-\bigl<\hat{B}_{i}\bigr>_{0}=\Delta B_{i}(t)\simeq
177: \int_{t_0}^tdt'\varphi_{ij}(t-t^{'})F_{j}(t^{'})+ $$
178: $$
179: \int_{t_{0}}^{t}dt_{1}
180: \int_{t_{0}}^{t_1}dt_{2}\,\varphi_{ijk}(t-t_{2},
181: t_{1}-t_{2})F_{k}(t_{1})F_{j}(t_{2}) + $$
182: \begin{equation} \int_{t_{0}}^{t}dt_1
183: \int_{t_{0}}^{t_1}dt_2\int_{t_{0}}^{t_2}dt_3\,\varphi_{ijkl}(t-t_{3}, t_{2}-t_{3},
184: t_{1}-t_{2})F_{e}(t_{1})F_{k}(t_{2})F_{j}(t_{3})+ \ldots
185: \end{equation}
186: where $\varphi_{ij}$, $\varphi_{ijk}$, $\varphi_{ijkl}$ are the
187: first-, second- and third order response functions,
188: \begin{equation} %7
189: \varphi_{ij}(t-t^{'})={1 \over ih}\langle[\hat{A_{j}},
190: B_{i}(t-t^{'})]\rangle_{0},
191: \end{equation}
192:
193: \begin{equation} %8
194: \varphi_{ijk}(t-t_{2}, t_{1}-t_{2}) = {1\over
195: (ih)^{2}}\langle[\hat{A}_{j}, [\hat{A}_{k} (t_{1}-t_{2}),
196: \hat{B}_{i}(t-t_{2})]]\rangle_{0},
197: \end{equation}
198:
199: \begin{equation} %9
200: \varphi_{ijkl}(t-t_{3}, t_{2}-t_{3}, t_{1}-t_{2}) = {1\over
201: (ih)^{3}} \langle[\hat{A}_{j}, [A_{k}(t_{1}-t_{2}), [A_{l}(t_{2}-t_{3}),
202: B_{i}(t-t_{3})]]]\rangle_{0}.
203: \end{equation}
204: Here we have employed the summation convention over repeated
205: indices and cyclic invariance of the trace. The expressions (6-8)
206: can be written in a more revealing form if we set $t_{0}=-\infty$
207: and change integration variables including an adiabatic switching
208: factor if necessary. We take (instead of $t$, $t_{1}$, $t_{2}$,
209: $t_{3}$) $\tau_{1}=t-t_{1}$, $\tau_{2}=t_{1}-t_{2}$ time
210: differences further. Ordinary linear response theory utilized only
211: $\varphi_{ij}(\tau_{1})$ and it is the simplest approximation to
212: the full theory of linear dynamic response [23]. Using new
213: variables we may write the expressions for response functions in
214: the form
215:
216: \begin{equation} %10
217: \varphi_{ij}^{(1)}=-{1 \over ih}\langle[A(\tau), B(0)]\rangle,
218: \end{equation}
219:
220: \begin{equation} %11
221: \varphi_{ijk}^{(2)}= {1\over (ih)^{2}}\langle[[A(\tau_{1}+\tau_{2}),
222: A(\tau_{2})], B]\rangle,
223: \end{equation}
224:
225: \begin{equation} %12
226: \varphi_{ijkl}^{(3)}=-{1\over (ih)^{3}}
227: \langle[[[A(\tau_{1}+\tau_{2}+\tau_{3}), A(\tau_{2}+\tau_{3})],
228: A(\tau_{3})], B]\rangle
229: \end{equation}
230: where the bracket $\langle\ldots \rangle$ denotes an expectation
231: value with respect to the equilibrium ensemble. The useful
232: interpretation is generated from eq. (5) in the case that $t_{0}=
233: -\infty$ if we consider the external force $F$ is constant and
234: vanishes for $t \geq 0$. For $t=0$ the system is in partial
235: equilibrium and starts to relax to equilibrium. It is convenient
236: to write nonlinear response for this case (initial value case
237: [21]) as
238:
239: \begin{equation} %13
240: \langle\hat{B}(t)\rangle-\langle\hat{B}\rangle_{0}= R^{(1)}(t)F+{1\over 2}R^{(2)}(t)
241: FF+{1\over 3}R^{(3)}(t)FFF+\ldots
242: \end{equation}
243: where $R^{\alpha}(t)$ are the relaxation functions,
244: \begin{equation} %14
245: R^{(1)}(t)=\int_{0}^{\infty}d\tau\,\varphi_{ij}^{(1)}(\tau_{1}),
246: \end{equation}
247: \begin{equation} %15
248: R^{(2)}(t)=\int_{0}^{\infty}d\tau_{1}
249: \int_{0}^{\infty}d\tau_{2}\,\varphi_{ijk}^{(2)}(\tau_{2},
250: \tau_{1}+\tau_{2}),
251: \end{equation}
252: \begin{equation} %16
253: R^{(3)}(t)=\int_{0}^{\infty}d\tau_{1}\int_{0}^{\infty}d\tau_{2}
254: \int_{0}^{\infty}d\tau_{3}\,\varphi_{ijkl}^{(3)}(\tau_{3},
255: \tau_{2}+\tau_{3}, \tau_{1}+\tau_{2}+\tau_{3}).
256: \end{equation}
257:
258: In this form response may describe relaxation of the system. If
259: response function $\varphi(t)^{(1)}$ vanishes as $t\rightarrow
260: \infty$, then $\varphi(t)^{(1)}=-{\delta \over \delta t
261: }R^{(1)}(t)$, so $R^{(1)}(t)$ contains much more information than
262: the response function.
263:
264: Let the $ac$ magnetic field $h_{\omega}=h\cos (\omega t)$ be
265: applied to a magnetic system. The magnetization nonlinear
266: response $M(\omega, t)=\sum_{k=1}^{n} \{ \Theta_{k}^{'}\cos
267: (k\omega t)+\Theta^{''}_{k}\sin (k\omega t) \}$ to harmonic
268: magnetic field $h$ contains only odd harmonic;
269: $\Theta_{1}^{'}\sim\chi_{1}^{'}h$, $ \Theta_{3}^{'}\sim
270: \chi_{3}^{'}h^{3}$, etc [11]. In expression for $M(\omega ,
271: t)$ the magnitudes $\Theta'_k$ and $\Theta''_k$ are
272: real and imaginary parts of harmonic amplitudes respectively. For the
273: general theory of nonlinear processes one can evaluate [11]
274: $$
275: \Theta_{1}^{'}=\chi_{1}^{'}(\omega t)h + [\chi_{3}^{'}(\omega,
276: 0, \omega)]{h^{3}\over 4}+[4\chi_{5}^{'}(\omega, 0,\omega, 0,
277: \omega)+2\chi_{5}^{'}(\omega,
278: 0, \omega, 2\omega, \omega) + $$
279: \begin{equation}
280: 2\chi_{5}^{'}(\omega, 2\omega, \omega, 0, \omega) +
281: \chi_{5}^{'}(\omega, 2\omega, \omega, 2\omega, \omega) +
282: \chi_{5}^{'}(\omega, 2\omega, 3\omega, 2\omega,
283: \omega)]{h^{5}\over 16}+ \ldots
284: \end{equation}
285: $$
286: \Theta_{3}^{'}=\chi_{3}^{'}(3\omega, 2\omega,
287: \omega){h^{3}\over 4} + [\chi_{5}^{'}(3\omega, 2\omega, 3\omega,
288: 2\omega, \omega)+\chi_{5}^{'}(3\omega, 4\omega, 3\omega, 2\omega,
289: \omega) $$
290: \begin{equation}2\chi_{5}^{'}(3\omega, 2\omega, \omega, 0, \omega) +
291: \chi_{5}^{'}(3\omega, 2\omega, \omega, 2\omega,
292: \omega)]{h^{5}\over 16}+ \ldots ,
293: \end{equation}
294: \begin{equation}
295: \Theta_{5}^{'}=\chi_{5}^{'}(5\omega, 4\omega, 3\omega, 2\omega,
296: \omega){h^{5}\over 16}+ \ldots
297: \end{equation}
298: The measurement of all the harmonic amplitudes $\Theta_{k}$ gives
299: a measurement of the susceptibilities $\chi_{a}$ in two limits: a)
300: if $\chi_{1}^{'}h\gg \chi_{3}^{'}h^{3} \gg \chi_{5}^{'}h^{5}$ the
301: back reaction is negligible and each harmonic measures the
302: susceptibility of the same order; b) in the static ($\omega
303: \rightarrow 0$) limit the solution of the linear system [15] fully
304: accounts for the back reaction. In the absence of $ac$ magnetic
305: field, the back reaction can be made small so that the dynamic
306: susceptibilities can be obtained from eq. (15). In a more compact
307: notation we may write
308: \begin{equation}
309: M(\omega, t) \sim [M_{0}+M_{\omega}+M_{3\omega}+\ldots]+(complex
310: \ \ conjugate)
311: \end{equation} % 20
312: where $M_0$ is the equilibrium magnetization in zero field;
313: $M_\omega$ is the $\omega $-magnetization response; $M_{3\omega}$
314: is the $3\omega$-magnetization response and so on.
315:
316: The expressions (9-11) may be considered as solution of the
317: corresponding quantum equations considered above. The external
318: $ac$ field we assume classical value. This field interacts with
319: quantum system and system behavior is determined by quantum laws.
320: We shall focus on the real part of the third-order nonlinear
321: dynamic susceptibility $\chi'_3(3\omega, 2\omega, \omega)$ and
322: denote it as $\chi'_3(\omega)$. In this paper we are interested in
323: the response when the $ac$ magnetic field is applied in
324: $z$-direction; $\chi_1=\chi_{zz}$ and so on. In formulas (2-11)
325: both $A_i$ and $B_i$ are the magnetic dipole moment operators.
326: Considering the initial value case [21] we suppose, like in [6],
327: that the system is in equilibrium with a small time-independent
328: field $h=h_z$ for $t\le 0$ and that the external field is turned
329: off at $t=0$, then for $t\ge 0$ the induced magnetization of the
330: sample in $z$-direction to first order of perturbation theory is
331: given by $M(t)-M_0\approx R^{(1)}(t)h$, where a relaxation
332: function
333: \begin{equation}
334: R^{(1)} (t)=\int_0^\infty \varphi^{(1)} (\tau)d\tau
335: \end{equation} % (21)
336: and the first order response function is [20]
337: \begin{equation}
338: \varphi^{(1)} _{ij} (\tau)\approx -{1\over i\hbar} \langle
339: \left[M_i(\tau), M_j\right]\rangle \ .
340: \end{equation} % (22)
341: The higher-order response functions are given by
342: \begin{eqnarray}
343: \varphi^{(2)} _{ijk} (\tau_1, \tau_2)&\approx &{1\over (i\hbar)^2}
344: \langle \left[\left[ M_i(\tau_1+\tau_2), M_j(\tau_2)\right], M_k \right]\rangle
345: \ ; \\ % (23)
346: \varphi^{(3)} _{ijkl} (\tau_1, \tau_2, \tau_3)&\approx & -{1\over
347: (i\hbar)^3} \langle \left[\left[\left[ M_i(\tau_1+\tau_2+\tau_3),
348: M_j(\tau_2+\tau_3)\right], M_k(\tau_3)\right], M_l
349: \right] \rangle. % 24
350: \end{eqnarray}
351: The linear and the nonlinear dynamic susceptibilities (admittances
352: in the spectral representation [22]) may be found through response
353: functions (21-23). In order to find the complete expression for
354: susceptibility, we should use its symmetry and causality
355: properties [22].
356:
357: The nonlinear susceptibilities may be chosen so that these
358: susceptibilities were symmetrical relative to simultaneous
359: permutation of tensor indices and corresponding to them
360: arguments, for example, the second rank tensors are $\chi_{ijk}
361: (\omega_1,\omega_2)= \chi_{ikj} (\omega_2,\omega_1)$, and the
362: fourth rank tensors are
363: \begin{equation}\chi_{ijkl}
364: (\omega_1,\omega_2,\omega_3) = \chi_{ikjl}
365: (\omega_2,\omega_1,\omega_3)= \chi_{ijlk}
366: (\omega_1,\omega_3,\omega_2)= \ldots \end{equation}
367: according to causality property
368: \begin{equation}
369: \chi_{ij}=0\ \hbox{for} \ \tau_1<\ \hbox{max}\ (\tau_2, \tau_3,
370: \ldots).
371: \end{equation} % (25)
372: Linear and nonlinear dynamic susceptibilities are given by
373: \begin{eqnarray}
374: \chi_{ij}(\omega)& =& \int_0^\infty d\tau \varphi^{(1)}_{ij}
375: (\tau) e^{i\omega\tau}\\ % (26)
376: \chi_{ijk}(\omega_1,\omega_2) &=&
377: {1\over 2!} \int_0^\infty d\tau_1 \int_0^\infty d\tau_2
378: \left\{\varphi^{(2)}_{ijk}
379: (\tau_1,\tau_2) e^{i(\omega_1+\omega_2)\tau_1+ i\omega_1\tau_1} \right\}\\ % (27)
380: \chi_{ijkl}(\omega_1,\omega_2, \omega_3) &=& {P_3\over 3!}
381: \int_0^\infty d\tau_1 \int_0^\infty d\tau_2 \int_0^\infty d\tau_3
382: \varphi^{(3)}_{ijkl} (\tau_1,\tau_2, \tau_3)\times \nonumber
383: \\ && \times \exp \left[i(\omega_1+\omega_2+\omega_3)\tau_1
384: + i\omega_2\tau_2 + i\omega_3\tau_3 \right] % (28)
385: \end{eqnarray}
386: In eq. (28) $P_3$ means the summation over all permutations of the
387: subscripts $(\omega_1j)$, $(\omega_2k)$ and $(\omega_3l)$ [22];
388: response functions $\varphi$ are given by expressions (21-23).
389:
390: In particular, the second harmonics generation process is
391: characterized by tensor $\chi_{ijk}(\omega,\omega)$. If
392: $\omega_1=\omega_2=\omega_3=\omega$, the triple frequency
393: $3\omega$ is formed; the frequency tripling is described by tensor
394: $\chi_{ijkl} (\omega,\omega,\omega)$ (we note it as
395: $\chi'_3(\omega)$).
396:
397: The droplet model of classical Ising spin glass was considered by
398: D.~S.~Fisher and D.~A.~Huse [4]. The features of this model are
399: described also, for example, in [23]. In the droplet model there
400: are only two pure thermodynamical states related to each other by
401: a global spin flip. In magnetic field there is no phase
402: transition. A droplet is an excited cluster in an ordered state
403: where all the spins are inverted. The natural scaling ansatz for
404: droplet free energy $\epsilon_L$ (which are considered to be
405: independent random variables) is $\epsilon_L \sim L^\theta, L\ge
406: \zeta (T)$; $\zeta$ is the correlation length, $L$ is the length
407: scale of droplet and $\theta$ is the zero temperature thermal
408: exponent, $\theta\le(d-1)/2$. One droplet consists of order $L^d$
409: spins. Below the lower critical dimension $d_l$, $\theta<0$; above
410: $d_l$ one has $\theta>0$.
411:
412:
413:
414: Recently M.~J.~Thill and D.~A.~Huse [6] have shown that the
415: $d$-dimensional quantum ising spin glass in a transverse field
416: with Hamiltonian
417: \begin{equation}
418: {\cal H} = -\sum_{i,j} {\cal I} _{ij} S^z_i S^z_j - \Gamma \sum_i
419: S^x_i
420: \end{equation}
421: (where $S_i$ are the Pauli matrices, $\Gamma$ is the strength of
422: the transverse field and the nearest neighbor interactions ${\cal
423: I}_{ij}$ are independent random variables of mean zero) can be
424: represented as the Hamiltonian of the independent quantum
425: two-level systems (low energy droplets) of the form
426: \begin{equation}
427: {\cal H} ={1 \over 2} \sum^{\sim}_L \sum_{D_L} \left( \epsilon
428: _{D_L} S^z_{D_L} +\Gamma_L S^x_{D_L} \right)
429: \end{equation}
430: where $S^z_{D_L} $ and $S^x_{D_L}$ are the Pauli matrices
431: representing the two states of the droplet; the sum is over all
432: droplets $D_L$ at length scale $L$ and over all length scales
433: $L$, and
434: \begin{equation} \sum^{\sim}_L\sim \int^\infty_{L_0}
435: {dL\over L}
436: \end{equation}
437: with a short-distance cutoff $L_0$. The value $\Gamma_L$ regulates
438: the strength of quantum fluctuations ($\Gamma_L\to 0$ corresponds
439: to the classical limit).
440: \begin{equation}
441: \Gamma_L=\Gamma_0 e^{-\sigma L^d}
442: \end{equation}
443: is the tunneling rate for a droplet of linear size $L,\,\Gamma_0$
444: is the microscopic tunneling rate; $\sigma$ is defined from the
445: equation $2K=\sigma L^d$ where $2K$ is the surface free energy of
446: an interface between the two droplet states, so $\sigma$ is a
447: reduced surface tension for this interface; $\sigma$ is
448: approximately the same for all droplet. We will assume $\Gamma_L$
449: is the same for all droplets of scale $L$. The droplet
450: excitations have a broad distribution of their free energies at
451: scale $L$ for large $L$ in a scaling form [4,6]
452: \begin{equation}
453: P_L (\epsilon_L) d\epsilon_L = {d\epsilon_L \over \gamma (T) L^\theta}
454: {\cal P} \left( \epsilon_L \over \gamma (T) L^\theta\right),\,
455: L\to\infty\ .
456: \end{equation}
457: It is assumed that $P_L (x\to 0) >0$, $P_L(0)-P_L(x)\sim
458: x^\phi$ at $x\to 0$. $\gamma(T)$ is a generalized temperature
459: dependent stiffness modulus which is of order of characteristic
460: exchange ${\cal I}= \overline{\left({\cal
461: I}_{ij}^2\right)}^{1\over 2} $ at $ T=0$ and vanishes for $T\ge
462: T_g$.
463:
464: There is a crossover length scale, $L^*(T)$, defined by condition
465: $\Gamma_{L^*(T)}= k_BT$ or $L^*(T)=\left({1\over \sigma} \log
466: {\Gamma_0 \over k_B T} \right)^{1/d}$. For droplets with
467: $L\ll L^*(T)$ and $\Gamma_L \gg k_BT$ the energy
468: $\sqrt{\epsilon^2_L+\Gamma^2_L}$ is always more than $k_BT$ and
469: thermal fluctuations are insignificant at temperature $T$.
470: Droplets with $L\gg L^*(T)$ have $\Gamma_L \ll k_BT$ and
471: behave classically. The large droplets ($\epsilon_L\le k_B T,\,
472: \Gamma_L \le k_BT$) are thermally active. At low $T$ only a small
473: fraction of droplets is thermally active, but many low-$T$ static
474: properties are dominated by these droplets at the crossover
475: length $L^*(T)$.
476:
477: The total magnet moment of a droplet will scale as $qL^{d\over 2}$
478: where $q$ a random number with mean zero and $\overline{q^2}\sim
479: q_{EA}$, $q_{EA}= \overline{\langle S^z_i\rangle^2}$ is the
480: Edwards-Anderson order parameter [4]. The total magnetization $M$
481: of the sample will be
482: \begin{equation} %36
483: M=\sum^{\sim}_L {V\over L^d} \sum_{D_L} \langle S^z_{D_L} \rangle
484: qL^{d\over 2}
485: \end{equation}
486: and
487: \begin{equation}
488: m={M\over V}=\sum^{\sim}_L \overline{\langle S^z_{D_L} \rangle
489: qL^{-d\over 2}}
490: \end{equation} % (36)
491: where $\overline{\langle S^z_{D_L} \rangle qL^{-d\over 2}}$ means
492: the average over the droplets energies $\epsilon_L$. The static
493: susceptibilities are defined in terms of the expansion of
494: magnetization into Taylor series approximately as
495: \begin{equation}
496: M=\chi_1 h -\chi_3 h^3 + \ldots \ .
497: \end{equation} % (37)
498: For enough small external field it is possible to be restricted
499: by few terms of this expansion. The susceptibilities are got as
500: derivatives with respect to $h$ at $h=0: \chi_1=\left.{\partial
501: m\over \partial h}\right|_{h=0}$, $\chi_3= \left.{\partial^3m\over
502: \partial h^3}\right|_{h=0}$.
503:
504: M.~J.~Thill and D.~A.~Huse have calculated static linear and
505: nonlinear susceptibilities for droplet system described by the
506: Hamiltonian (29). The static linear susceptibility diverges at
507: $T=0$ below the lower critical dimension $d_l$. The static
508: nonlinear susceptibility diverges in all dimensions $d$. The
509: static linear susceptibility appears to start away from the
510: nonzero constant $T=0$ value decreasingly versus T to lowest order
511: [6].
512:
513: Now using the quantum droplet model of the short-range Ising spin
514: glass in a transverse field and quantum-mechanical case
515: $\beta\Gamma_L \gg 1$ (quantum regime) we calculate the third
516: order nonlinear dynamic response at very low finite temperatures.
517:
518: When we consider the droplets at finite temperatures they may
519: have two characteristic rates, (a) the Rabi frequency (is of
520: order $\Gamma_L$) and (b) the rate of classical activation over
521: energy barrier B for annihilation and creation of the droplet
522: excitations [6]
523: \begin{equation}
524: t \sim \tau_0 e^{B\over k_BT }\ ,
525: \end{equation}
526: where $B\sim \Delta L^\psi$, $0\le\psi\le d-1$, $\psi$ is some
527: exponent [4], $\Delta$ is a barrier energy at $T\ll T_g$,
528: $\Delta\sim {\cal I}$; $\tau_0$ is a microscopic time. There is a
529: complicated dynamical classical-to-quantum crossover depending on
530: temperature, frequency of $ac$ external field and length scale
531: $L$. According to [6] the crossover dynamic length is determined
532: from the condition $\Gamma^{-1}_L= T$, i.e.
533: \begin{equation}
534: L^*_{dyn}(T) \sim \left({\sigma\over \Delta}k_B T\right) ^{1\over
535: \psi-d}
536: \end{equation} % (40)
537: The system behaves presumably classically or quantum mechanically
538: when the dominant length scale $L$ is above or below $L^*_{dyn}$
539: for frequency $\omega$.
540:
541: Now we consider dynamic third-order susceptibility $\chi'_3(\omega
542: ,T)$ at finite very low temperatures (quantum regime) when
543: $\beta\Gamma_L\gg 1$. We define nonlinear third order dynamic
544: susceptibility $\chi'_3(\omega,T)$ by expression (23) and (28) $$
545: \chi'_3(\omega,T) = {1\over (i\hbar )^3} {P_3\over 3!}
546: \int_0^\infty d\tau_1 \int_0^\infty d\tau_2 \int_0^\infty d\tau_3
547: \exp \left[ i \left( 3\omega\tau_1 +\omega\tau_2 +
548: \omega\tau_3\right)\right]\times$$
549: \begin{equation}
550: \langle \left[\left[\left[ M_i(\tau_1+\tau_2+\tau_3), M_j
551: (\tau_2+\tau_3)\right], M_k (\tau_3)\right], M_l\right] \rangle
552: .
553: \end{equation} %41
554: The contribution of a single droplet to the
555: real part of dynamic third-order susceptibility up to some factor
556: $\sim q^2_{EA} L^{2d}$ is proportional to
557: \begin{equation}
558: \chi'_{3D_L} \sim q^2_{EA} {\left( \sum^5_{k=0} A_k
559: (\omega,\Gamma_L) \epsilon^{2k}_L\right) \tanh \left( {1\over
560: 2}\beta \sqrt{\epsilon^2_L+\Gamma_L^2}\right) \over
561: \left(\epsilon^2_L +\Gamma_L^2\right)^{5\over 2}
562: \left(\epsilon^2_L +\Gamma_L^2-9\omega^2\right)
563: \left(\epsilon^2_L +\Gamma_L^2-\omega^2\right)^2
564: \left(\epsilon^2_L +\Gamma_L^2-{\omega^2\over 4}\right)\omega^2
565: }\ ,
566: \end{equation} % (42)
567: where $A_0=\omega^2\Gamma^{10}_L +4\omega^4 \Gamma^8_L -5 \omega^6
568: \Gamma^6_L$; $A_1=2\Gamma^{10}_L - 12,5 \omega^2 \Gamma^8_L -
569: 49,75 \omega^4 \Gamma^6_L - 23,5 \omega^6 \Gamma^4_L - 2,25
570: \omega^8 \Gamma^2_L $; $A_3=12 \Gamma^6_L - 45,5
571: \omega^2\Gamma^{4}_L +41,75 \omega^2 \Gamma^4_L$; $A_4=8
572: \Gamma^4_L - 15,5 \omega^2\Gamma^{2}_L$; $A_5= 2 \Gamma^2_L$.
573:
574: We have to average $\chi_{3D_L}$ over droplet energies
575: $\epsilon_L$ using the distribution of droplet free energies (33)
576: and changing variables from $\epsilon_L$ to $x=\beta\epsilon_L$.
577:
578: After averaging over droplet energies for cases $\Gamma_L
579: >3\omega, \Gamma_L \sim 3 \omega, \Gamma_L<3 \omega$ we receive
580: that the real part of the nonlinear susceptibility is dominated by
581: droplets of length scale
582: \begin{equation}
583: L_{dom(3\omega )} \sim \left(\frac{1}{\sigma}\left|\log
584: \left(\frac{3 \omega}{\Gamma_0}\right)\right|\right)^{\frac{1}{d}}
585: \end{equation} % (43)
586: which is determined by condition $\Gamma_L \sim 3 \omega$. Then for
587: $\Gamma_L> 3 \omega$ the result of two averages is given by the following
588: expression
589: $$
590: \chi'_3 \sim \frac{q^2_{EA}}{\gamma^{1+\phi}} \left \{ \pi
591: \sec \left[\frac{\pi \phi}{2}\right]
592: \sum_{k=-2,0,2} A_k \omega^{k-2} \Gamma_0^{\phi -k}
593: \frac{(\sigma (\phi -k))^{-\alpha}}{d} {\rm G}\left[\alpha,\left|\log
594: \left(\frac{3 \omega}{\Gamma_0}\right)\right| (\phi -k)\right]+\right.$$
595: \begin{equation}\left.
596: \frac{\left(\frac{1}{\sigma}\left|\log
597: \left(\frac{3 \omega}{\Gamma_0}\right)\right|\right)^\alpha}{\alpha d}
598: e^{-3 \beta \omega} \sum_{k=0,1} B_k \omega^{\frac{\phi -5-2k}{2}}
599: \beta^{\frac{-\phi-1-2k}{2}}
600: \right \},
601: \end{equation} % 44
602: where ${\rm G}[\alpha,z]$ is incomplete gamma-function. The
603: coefficients in this expression depend on $\phi$ only and are
604: given by
605:
606: $\alpha=\frac{d-\theta (1+\phi)}{d};$
607:
608: $A_{-2}=-\frac{1}{4}-\frac{\phi}{4},\,A_0=-\frac{6263}{1600}-\frac{23\phi}{100}
609: +\frac{6441}{6400}2^{\frac{\phi+1}{2}},\,A_2=-\frac{135}{10}+12\,2^
610: {\frac{\phi+1}{2}},$
611:
612: $B_0={\rm
613: G}\left[\frac{\phi-1}{2}\right]\frac{2^{\frac{\phi-7}{2}}\,3^
614: {\frac{\phi-1}{2}}(1435\phi^2+843\phi-1988)}{175},\, B_1={\rm
615: G}\left[\frac{\phi-1}{2}\right]\frac{2^{\frac{\phi-7}{2}}\,3^
616: {\frac{\phi-3}{2}}(791-791\phi^2)}{5}$.
617:
618: $$\chi'_3(\omega,T)\sim \frac{q_{EA}^2}{\gamma^{1+\phi}}
619: \frac{\left(\frac{1}
620: {\sigma}\left|\log\left(\frac{3\omega}{\Gamma_0}\right)\right|\right)^\alpha}
621: {\alpha
622: d}\left\{C_0\pi\sec\left[\frac{\pi\phi}{2}\right]\omega^{\phi-2}+
623: e^{-3\beta\omega}\sum_{k=1}^4C_k\omega^{\frac{\phi+1-2k}{2}}\beta^{
624: \frac{-\phi+5-2k}{2}}\right\} $$ for $\Gamma_L\sim3\omega$ where
625:
626: $C_0=\frac{243\,35^{\frac{\phi-1}{2}}2^{7-\phi}}{185}-\frac{68931\,2^{3\phi-3}}
627: {925}+\frac{3^\phi}{2}-\frac{391\,2^{\frac{\phi+5}{2}}3^\phi}{925}-\frac{27\,
628: 2^{\frac{3\phi-5}{2}}\phi}{5},\,C_1=-{\rm
629: G}\left[\frac{\phi-1}{2}\right]\frac{2^{
630: \frac{\phi-3}{2}}\,3^{\frac{\phi+3}{2}}}{5},$
631:
632: $C_2={\rm G}\left[\frac{\phi-1}{2}\right]\frac{2^{
633: \frac{\phi-3}{2}}\,3^{\frac{\phi+1}{2}}}{5}-{\rm
634: G}\left[\frac{\phi+1}{2}
635: \right]\frac{2^{\frac{\phi+1}{2}}\,3^{\frac{\phi-5}{2}}(458+945\phi)}{35},$
636:
637: $
638: C_3=-{\rm G}\left[\frac{\phi+1}{2}
639: \right]\frac{11897\,2^{\frac{\phi-7}{2}}\,3^{\frac{\phi-5}{2}}}{175}+
640: {\rm G}
641: \left[\frac{\phi+3}{2}\right]\frac{10206\,2^{\frac{\phi-7}{2}}\,3^{
642: \frac{\phi-5}{2}}}{35},$
643:
644: $
645: C_4=-{\rm G}
646: \left[\frac{\phi+3}{2}\right]\frac{359106\,2^{\frac{\phi-7}{2}}\,3^{
647: \frac{\phi-5}{2}}}{175}$.
648:
649: This expression does not diverge if $2<\phi<3$ and
650: $-1+\frac{d}{\theta}<\phi$. This expression has singularity at
651: $\omega\sim \frac{\Gamma_0}{3}$. When the frequency increases the
652: values of $\chi'_3$ are growing to infinity while
653: $\omega\to\frac{\Gamma_0}{3}$. $\chi'_3$ maintains this property
654: when $\omega$ is more than $\frac{\Gamma_0}{3}$. The temperature
655: dependence of $\chi'_3$ in this case has no extremes, we observe
656: monotonous decrease of values of $\chi'_3$ with temperature.
657:
658: Nonlinear susceptibility given by expression (42) does
659: not diverge if $2<\phi<3$ and $-1+\frac{d}{\theta}<\phi$.
660:
661: We observe that nonlinear susceptibility has strong dependence on
662: distribution function $P_L(\epsilon_L)$, i. e. on $\phi$, on
663: droplet microscopic tunneling rate $\Gamma_0$ and other
664: parameters. One can see that the real part $\chi'_3(\omega,T)$
665: varies approximately logarithmically with frequency. This
666: signalizes broad distribution of relaxation times of the system.
667:
668: Let us take for numerical calculation the following numbers:
669: $\theta=1,\Gamma_0=10^{10}s^{-1},d=3,\phi=2.5,\gamma=10^{-15}erg,
670: \sigma=10^{-15}, q=0.5$. The frequency dependence of $\chi'_3(\omega,T)$
671: is shown in Fig. 1. We give $\log_{10} f$-dependence at $\log_{10} f$
672: from 0 to 11 at fixed several temperatures:
673: $T_1=0.001,T_2=0.005,T_3=0.01,T_4=0.05$.
674:
675: The frequency interval covers some decades of frequencies. Our numerical
676: calculations show the crossover between low-$\omega$ and high-$\omega$
677: behaviors. In low-$\omega$ region the nonlinear response is found nonsingular
678: and slowly decreasing. When frequency increases the curve falls down more
679: quickly, the nonlinear response diverges at $\omega \sim \frac{\Gamma_0}{3}$,
680: then the curve rises to some value. In low-$\omega$ region we have a
681: qualitative agreement with experimental data for disordered dipolar magnet
682: ${\rm LiHo}_x{\rm Y}_{1-x}{\rm F}_4$. At different low fixed temperatures
683: the behavior of $\chi'_3(\omega,T)$ is the same and the values of
684: $\chi'_3(\omega,T)$ are approximately the same. Therefore we give only one
685: curve for all fixed temperatures.
686:
687: In Fig. 2 we give the temperature dependence of
688: $\chi'_3(\omega,T)$ at temperatures from 0 to $10^{-2}$K at fixed
689: several frequencies: $f_1=10^7Hz, f_2=2.5\times 10^7Hz,f_3=5
690: \times 10^7Hz,f_4=7.5\times 10^7Hz,f_5=10^8Hz$ of $ac$ field ($
691: f=\frac{\omega}{2\pi}$). The behavior of $\chi'_3(\omega,T)$
692: indicates the following glassy-like features. The curves of the
693: temperature dependence of $\chi'_3(\omega,T)$ have maxima
694: depending on fixed frequency. The temperature of $\chi'_3$-maximum
695: $T_f(\omega)$ depends on frequency. The nonlinear susceptibility
696: magnitudes at different fixed frequencies are remarkably
697: distinguishable. The temperatures of maximum values are different.
698: When the frequency increases the temperature of $\chi'_3$-maximum
699: shifts towards high temperatures. The similar curve of temperature
700: variation of $\chi'_3$ was observed in spin glasses at more high
701: temperatures [2,13]. If we consider only T-dependent part of
702: $\chi'_3$ we see that the $\chi'_3$-maxima are sharp (Fig. 3).
703:
704: The cubic dynamic susceptibility $\chi'_3(3\omega)$ is
705: analitically and numerically calculated in quantum spin glass in
706: terms of quantum droplet model on the basis of general dynamic
707: nonlinear quantum-mechanical response theory. We have carefully
708: analyzed the susceptibility temperature-frequency behavior to
709: study the properties of the low temperature magnetic state and to
710: determine whether or not a conventional spin glass state exists
711: below $T_f$. Comparing with the case of a true spin glass
712: transition we see that our data indicate that the magnetic state
713: below $T_f$ does not correspond to a conventional spin glass state
714: below $T_f$. We find a glassy type slow dynamics. Similar
715: frequency dependence was observed by W. Wu et al.[14].
716:
717: Our calculations at $T=0$ coincide with $T=0$ result of
718: M.~J.~Thill and D.~A.~Huse [6]. For finite temperatures we find
719: some features which have been recently observed [13]. We suppose
720: that at some very low temperature $T_f$ (temperature of maximum of
721: $\chi'_3(\omega,T)$) there is a phase transition. If $\theta>0$
722: and $d=3$ we suppose a true phase transition at very low
723: temperature $T_f\sim 10^{-4}\div8.5\times 10^{-4}K$ for $f=10^7
724: \div10^8Hz$ respectively (Fig. 3).
725:
726: Besides frequency and temperature dependence the shape of
727: $\chi'_3(3\omega)$ depends crucially on the probability
728: distribution of droplet free energies, on the tunneling rate for a
729: droplet of linear size L, on the material parameters. In
730: consequence of this dependence there is divergence (or
731: convergence) of $\chi'_3(3\omega)$. We need to take into account
732: (in future paper) the dipole-dipole interaction between droplets
733: and also droplet-lattice interaction.
734:
735: Applying our results to the reported experimental data on the
736: nonlinear dynamic susceptibility of ${\rm LiHo}_x{\rm Y}_{1-x}{\rm
737: F}_4$ we observe that a fairly good agreement may be achieved.
738:
739:
740: \newpage
741: \centerline{\Large References}
742:
743: \begin{enumerate}
744: \item Glassy dynamics and optimization (Ed.: Y.L. van Hemmen, M.
745: Morgenstern) Springer Verlag, Berlin Heidelberg, 1987.
746:
747: \item M.Mezard, G.Parisi and M.A.Virasoro "Spin Glass Theory and
748: Beyond",World Scientific, Singapore, (1987);
749: A.P. Young (Ed.) "Spin-glasses and random fields", World Scientific, Singapore, (1997).
750:
751: \item T. Kopec, Phys. Rev. Lett. {\bf 79}, 4266 (1997)
752:
753: \item D. S. Fisher and D. A. Huse, Phys. Rev. Lett. {\bf 56}, 1601 (1986); Phys. Rev. B
754: {\bf 36}, 8937 (1987); Phys. Rev. B {\bf 38}, 373,386 (1988);
755: D.S. Fisher, J. Appl. Phys. {\bf 61}, 3672 (1987).
756:
757: \item A. Barrat and L. Berthier, preprint cond-mat/0102151
758: 8 Feb. 2001;
759:
760: Y.G. Joh and R. Orbach. Phys. Rev. Lett. {\bf 77}, 4648 (1996).
761:
762: \item M.J. Thill and D.A. Huse, Physica {\bf A241} , 321 (1995)
763:
764: \item H. Ishii and T. Yamamoto, J. Phys. {\bf C18}, 6225 (1985);
765:
766: \item N. Read, S. Sachdev and J. Ye, Phys. Rev. B{\bf 52}, 384
767: (1995).
768:
769: \item H. Rieger and A.P. Young, Phys. Rev. Lett. {\bf 72}, 4141 (1994).
770:
771: \item L.P. L\'evi and A.T. Ogielski, Phys. Rev. Lett {\bf 26},
772: 3288 (1986)
773:
774: \item L.P. L\'evi, Phys. Rev. Lett B{\bf 38},4963 (1988)
775:
776: \item T. Jonsson, K. Jonason, P. J\"onsson and P. Nordblad, Phys. Rev. B{\bf 59}, 8770 (1999)
777:
778: \item K. Gunnarsson et al. Phys. Rev. B. {43}, 8199 (1991);
779:
780: M. Hagiwara et al. J MMM, {\bf 177-181}, 175 (1998).
781:
782: \item W. Wu, D. Bitko, T. F. Rosenbaum and G. Aeppli, Phys. Rev. Lett.
783: {\bf 71}, 1919 (1993);
784:
785: D. Bitko, T. F. Rosenbaum and G. Aeppli, Phys. Rev. Lett.
786: {\bf 75}, 1679 (1995).
787:
788: \item J. Mattsson, Phys. Rev. Lett. {\bf 75}, 1678 (1995).
789:
790: \item G. Busiello and R. V. Saburova, Int. J. Mod. Phys. {\bf B14}, 1843 (2000);
791:
792: R. Saburova, G.P. Chugunova and G. Busiello. The physics of metals
793: and metallography, {\bf 87}, 509-515 (1999);G. Busiello, R. V.
794: Saburova and V.G. Sushkova, Sol.State Comm.{\bf 119},545 (2001).
795:
796: \item R. Kubo and K. Tomita, J. Phys. Soc. Japan {\bf 9}, 888 (1954)
797:
798: R. Kubo , J. Phys. Soc. Japan {\bf 12}, 570 (1957)
799:
800: \item W. Bernard and H.B. Callen, Rev. Mod. Phys. {\bf 31}, 1017
801: (1959).
802:
803: R.L. Peterson, Rev. Mod. Physics {\bf 39}, 69 (1967)
804:
805: \item W.T. Grandy, Jr. "Foundation of statistical mechanics" D.
806: Reidel Publishing Company, Dordrecht, Holland, 1988.
807:
808: \item W. Brenig "Statistical theory of heat" Springer-Verlag,
809: 1989.
810:
811: \item R.L. Stratonovich "Nonlinear nonequilibrium thermodynamics
812: I" Springer-Verlag, 1992;
813:
814: V.M. Fain. Kvantovaya Radiophizika. Izdatelstvo Sovietskoe Radio.
815: Moskva, 1972 (Russian).
816:
817: \item J.A. Mydosh "Spin Glasses: an experimental introduction.
818: Taylor \& Francis, London, 1993.
819:
820: \item P. Esquinazi (Ed.) Tunneling systems in amorphous and
821: crystalline solids. Springer-Verlag-Heidelberg, 1998.
822:
823: \end{enumerate}
824: \newpage
825: \centerline{\Large Figure captions}
826:
827: Fig.1 - The frequency dependence of the real part of the nonlinear
828: dynamical susceptibility at fixed temperature.
829:
830: Fig.2 - The temperature dependence of the real part of the
831: nonlinear dynamical susceptibility at various frequencies.
832:
833: Fig.3 - The temperature dependence of the T-dependent part of the
834: nonlinear dynamical susceptibility at various frequencies f.
835:
836:
837: \end{document}
838: