1: \documentclass[aps,pra]{revtex4}
2:
3: \usepackage{latexsym}
4: \usepackage{amsmath}
5: \usepackage{graphicx}
6:
7: \newcommand{\be}{\begin{equation}}
8: \newcommand{\ee}{\end{equation}}
9: \newcommand{\Tr}{ {}{\rm Tr}{} }
10: %
11: % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12: % %
13: % % JOURNALS MACRO%
14: \def\JPA{{J.\ Phys.\ A\ }}
15: \def\ADP{{Adv.\ Phys.\ }}
16: \def\JPC{{J.\ Phys.\ C\ }}
17: \def\NPB{{Nucl.\ Phys.\ B\ }}
18: \def\PRB{{Phys.\ Rev.\ B\ }}
19: \def\PRL{{Phys.\ Rev.\ Lett.\ }}
20: \def\RMP{{Rev.\ Mod.\ Phys.\ }}
21: \def\ZPB{{Z.\ Phys.\ B\ }}
22: \def\PLA{{Phys.\ Lett.\ A}}
23:
24: % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
25:
26:
27: \begin{document}
28:
29: \title{One-step replica symmetry breaking solution of the quadrupolar glass
30: model}
31:
32:
33: \author{F. P. Mancini}
34: \address{Dipartimento di Fisica and
35: Sezione I.N.F.N., Universit\`a di
36: Perugia, Via A. Pascoli, Perugia, 06123, Italy}
37: %\ead{FrancescoPaolo.Mancini@pg.infn.it}
38: \author{D. Sherrington}
39: \address{Rudolf Peierls Centre for Theoretical Physics, University
40: of Oxford,1 Keble Road, Oxford, OX1 3NP, UK}
41: % \ead{D.Sherrington1@physics.oxford.ac.uk}
42:
43:
44:
45: \begin{abstract}
46: We consider the quadrupolar glass model with infinite-range random
47: interaction. Introducing a simple one-step replica symmetry
48: breaking ansatz we investigate the para-glass continuous
49: (discontinuous) transition which occurs below (above) a critical
50: value of the quadrupole dimension $m^*$. By using a mean-field
51: approximation we study the stability of the one-step replica
52: symmetry breaking solution and show that for $m>m^{*}$ there are
53: two transitions. The thermodynamic transition at a temperature
54: $T_{D}$ is discontinuous but there is no latent heat. At a higher
55: temperature we find the dynamical or glass transition temperature
56: $T_G$ and the corresponding discontinuous jump $q_G$ of the order
57: parameter.
58: \end{abstract}
59:
60: %PACS numbers: 75.10.Nr, 05.50.+q
61:
62:
63: \maketitle
64:
65:
66: \section{INTRODUCTION}
67: In the last decades quadrupolar glasses have found widespread
68: experimental and theoretical interest \cite{hochli_90}. Disordered
69: quadrupolar glasses are produced by random dilution of
70: (quadrupolar) molecular crystals with atoms which have no
71: quadrupole moments; well-known examples of such systems are
72: $K(CN)_{x}Br_{1-x}$ or $Na(CN)_{x}Cl_{1-x}$, or $N_{2} \: Ar$,
73: $CuCN$, or solid hydrogen (see Ref. \cite{binder_92} for a
74: review). The success of the Sherrington-Kirkpatrick (SK) model
75: \cite{sk} in providing a good theory to describe systems of
76: interacting magnetic or electric dipole moments, suggests to
77: extend to quadrupolar glasses the same kind of analysis. However,
78: there are differences between standard spin glass systems and
79: quadrupolar glasses; the latter do not have the global inversion
80: symmetry $S_i \to - S_i$ for all spins. For several systems
81: without reflection symmetry, and close to the transition
82: temperature, only one step in the Parisi replica symmetry breaking
83: scheme is sufficient to describe the transition para-glass above a
84: lower critical dimension \cite{gold_elder_85}. Indeed, the
85: one-step replica symmetry breaking ($1RSB$) scheme has proven to
86: provide stable solutions for the Potts glass model
87: \cite{gks,ck,desantis}, the spherical $p$-spin model
88: \cite{crisanti_92} and the $p$-spin Ising spin glass model
89: \cite{gardner_85}. It is the purpose of the present paper to show
90: that the $1RSB$ scheme can be applied also to the quadrupolar
91: glass model, and indeed it provides a stable solution in certain
92: regimes.
93:
94: %%%%%%%%%%%%%%
95:
96: In the present paper we shall consider a perturbative evaluation
97: of the free energy by means of a Taylor expansion up to fourth
98: order in the order parameter. It is obvious that the perturbative
99: approach is most reliable near the transition temperature. We
100: shall show that the transition from the replica symmetric ($RS$)
101: state to the $1RSB$ occurs either discontinuously or continuously,
102: depending on the value of the quadrupole dimension $m$. A similar
103: dependence of the $RS$ to $1RSB$ transition, though on the value
104: of an external field, is exhibited by the spherical $p$-spin model
105: \cite{crisanti_92} and the Ising $p$-spin model
106: \cite{gardner_85}. For any $p>2$, the transition is discontinuous
107: (continuous) for fields weaker (larger) than a critical value of
108: the external field $h_c$, which depends on $p$.
109:
110: The plan of the paper is as follows: in Sec. II
111: we shall use a rather pedagogical approach mainly to review the results obtained
112: in the mean-field analysis of the quadrupolar glass model in the
113: framework of the replica symmetry ansatz \cite{gs}. Section III is devoted to
114: the study of the $1RSB$ solutions of the
115: saddle-point equations, assuming all the transitions to be continuous or at worst
116: weakly discontinuous. In Sec. IV we shall perform the de Almeida-Thouless (AT)
117: stability analysis while the dynamical transition is discussed in Sec. V. The
118: concluding remarks are given in Sec. VI.
119:
120:
121:
122: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
123:
124:
125: \section{Uniaxial Quadrupolar Glass}\label{section_UQG}
126: \noindent
127: The infinite-ranged quadrupolar glass model has been first introduced by
128: Goldbart and Sherrington (GS) \cite{gs}. The model assumes the quadrupole-quadrupole
129: interaction to be more dominant than the interactions between dipoles.
130: This appears to be the case in several experimental situations, where the
131: quadrupolar species occupy the sites of a regular lattice, but share this lattice with a
132: dilutant without quadrupole moment: argon in the case of interacting $N_{2}$, parahydrogen
133: in the case of interaction with orthohydrogen, and $K Br$ in the case of $KCN$, etc.
134: \cite{binder_92}.
135:
136: To construct the mean-field theory of a set
137: of uniaxial quadrupoles interacting through randomly quenched and
138: frustrated isotropic exchange one may adopt the Hamiltonian
139: \begin{equation}
140: H = -\sum_{(i,j)} J_{ij} \sum_{\mu \nu} S_{\mu}^i S_{\nu}^i S_{\mu}^j
141: S_{\nu}^j = - \sum_{(i,j)} J_{ij} \left( \mbox{\bf S}^i \cdot \mbox{\bf S}^j
142: \right)^2 ,
143: \label{quadrup_ham}
144: \end{equation}
145: where the spin vector ${\bf S}^i$ is defined via the component $f^{i}_{\mu
146: \nu}= \left(S^{i}_{\mu}S^{i}_{\nu}-\delta_{\mu \nu}/m \right)$ of the electric
147: quadrupole moment tensor \cite{gs}. The summation $(i,j)$ runs over
148: all the distinct pairs. Each ${\bf {S}}^i$ has $m$ components $S^i_{\mu}$ ($\mu
149: =1, \ldots ,m$) and, for convenience, is assumed to be a vector with fixed
150: length $\vert {\bf {S}}^i \vert=m$. Of course, taking general $m$ does not
151: describe the experimental quadrupolar glasses. Rather it is a natural model to
152: consider theoretically for classification. By analogy with SK, the spins are taken
153: to interact via independent random interactions $J_{ij}$ which are assumed
154: Gaussian distributed:
155: \begin{equation}
156: P \left( J_{ij} \right) = \left( \frac{N}{2 \pi J^2} \right)^{1/2} \exp
157: \left[ - \frac{(J_{ij}- J_0 /N)^2}{2 J^{2}/N} \right] .
158: \end{equation}
159: The mean $J_0$ and the variance $J$ of the
160: distribution depend on the total number of quadrupoles $N$ to ensure a
161: meaningful thermodynamic limit $(N \rightarrow \infty)$ with an extensive
162: energy: $J_0 = \tilde{J}_0/N$ and $J=\tilde{J}/N^{1/2}$.
163: The Hamiltonian (\ref{quadrup_ham}) can be seen as the Hamiltonian of an
164: infinite-range model for $N$ classical vector spins ${\bf {S}}^i$ and
165: zero external field is assumed.
166:
167: In Ref. \cite{gs} it has been shown that in terms of the order parameters
168: \be
169: \begin{split}
170: Q^{ab}_{\mu \nu \lambda \rho} & = \frac{ \mbox{tr } \left[ S_{\mu}^{a}
171: S_{\nu}^{a} S_{\lambda}^{b} S_{\rho}^{b} \: \exp L \right]} { \mbox{tr }
172: \left[ \exp L \right]} = \langle S_{\mu}^{a} S_{\nu}^{a} S_{\lambda}^{b}
173: S_{\rho}^{b} \rangle \\
174: M^{a}_{\mu \nu} &= \frac{ \mbox{tr } \left[S_{\mu}^{a} S_{\nu}^{a}\: \exp
175: L \right] } { \mbox{tr } \left[ \exp L \right] } = \langle S_{\mu}^{a}
176: S_{\nu}^{a} \rangle ,
177: \label{tensors}
178: \end{split}
179: \ee
180: the free energy per spin $f$ - by means of the replica trick - is given by
181: \be
182: \begin{split}
183: - \beta f &= \frac{1}{N} \lim_{n \rightarrow 0} \frac{1}{n} \left(
184: \overline{Z^n} - 1 \right)
185: = \lim_{n \rightarrow 0} \frac{1}{n} \left[ -\frac{\beta J_0}{2}
186: \sum_{a} \sum_{\mu \nu} \left( M^{a}_{\mu \nu} \right)^2 \right.
187: \\
188: &- \left. \left( \frac{\beta J}{2} \right)^2 \sum_{ab} \sum_{\mu \nu
189: \lambda \rho} \left( Q^{ab}_{\mu \nu \lambda \rho} \right)^2 + \mbox{log
190: tr } \exp L \right],
191: \end{split}
192: \label{fee}
193: \ee
194: where $n$ is the number of replicas, $\beta=1/k_{B}T$, and $L$ is
195: \begin{equation}
196: L= \beta J_0 \sum_{a} \sum_{\mu \nu} M^{a}_{\mu \nu} S^{a}_{\mu}S^{a}_{\nu}+
197: \frac{ ( \beta J)^2}{2} \sum_{ab} \sum_{\mu \nu \lambda \rho} S_{\mu}^{a}
198: S_{\nu}^{a} S_{\lambda}^{b} S_{\rho}^{b} Q^{ab}_{\mu \nu \lambda\rho} .
199: \label{elle}
200: \end{equation}
201: The elements of the order parameters $Q^{ab}_{\mu \nu \lambda
202: \rho}$ and $M^{a}_{\mu \nu}$ are not independent quantities and
203: they can be parameterized in terms of five sets of independent
204: parameters $A^a$, $B^{ab}$, $C^{ab}$, $D^{ab}$, and $E^{ab}$. The
205: non-zero extremal values of the above sets describe possible glass
206: ordering \cite{gs}.
207:
208: Upon decreasing the temperature, when one of the parameters becomes
209: different from zero, the high-temperature disordered phase becomes
210: unstable. By assuming continuous transition in the
211: replica symmetric ansatz and provided that the
212: average interaction is not too positive,
213: \begin{equation}
214: \frac{J_{0}}{J}<
215: \frac{-m^{2}+m+8}{m+4},
216: \label{ferro}
217: \end{equation}
218: there is a transition to an isotropic glass state occurring at the temperature $T_{RS}=(2m
219: J)/[k_{B}(m+2)]$. This highest temperature phase transition is associated with the order
220: parameter $B^{ab}$ acquiring a non-zero value, hinting at the presence of
221: isotropic quadrupolar order \cite{gs}.
222: As GS showed, below $T_{RS}$ the
223: replica symmetric solution isotropic glass phase is unstable with respect to
224: fluctuations in the space of broken replica symmetric isotropic
225: glass order parameters, the instability being stronger than in conventional
226: systems.
227:
228: It is worth noting that Eq. (\ref{ferro}) is the equivalent of the
229: condition $J_{0}/J<1$ in the {\it SK} model to ensure a transition from
230: the paramagnet state to spin glass state. Furthermore, when the
231: numerator of the right hand side of Eq. (\ref{ferro}) becomes negative,
232: it is necessary to introduce a negative value for $J_{0}$ in order to
233: obtain a glass phase at low temperatures; this happens for $m>3.37
234: \ldots$. Increasing the negativity of $J_{0}$ reduces the
235: temperature of ferromagnetism onset, though it cannot be stopped.
236:
237:
238:
239: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
240:
241:
242:
243:
244: \section{One-step replica symmetry breaking Theory}
245: \noindent
246: The further analysis of GS leads to the conclusion that a replica
247: symmetric ansatz cannot give a stable, and hence physical, solution of
248: the quadrupolar glass model. Thus, one has to resort to a
249: replica symmetry breaking ansatz.
250: Close to the isotropic quadrupolar glass transition, i.e. confining
251: attention to regions of parameter space $(J_0, J)$ in which the highest
252: transition temperature does correspond to a phase transition in the
253: order parameter $B^{ab}$, it is sufficient to consider only this
254: parameter different from zero in the free energy (see Ref.
255: \cite{gs}). Thus, one requires $J_0$ to satisfy the inequality
256: (\ref{ferro}) for $T \lesssim T_{RS}$, i.e. in the neighbourhood of the
257: transition temperature. The free energy is then given by
258: \begin{equation}
259: \begin{split}
260: \beta f &=
261: \lim_{n \rightarrow 0}
262: \frac{1}{n}\left\{ \frac{\left( \beta J\right)
263: ^{2}}{2}(m-1)(m+2){\sum_{ab}}^{\prime} \left( B^{ab}\right)^2 +m\left( \beta
264: J\right)^2 {\sum_{ab}}^{\prime}B^{ab}\right.
265: \\
266: &- \left.\log \Tr \: \exp
267: \left[ \left( \beta J\right)^2 {\sum_{ab}} ^{\prime} \sum_{\mu \nu}
268: B^{ab}S^{a}_{\mu} S^{a}_{\nu}S^{b}_{\mu} S^{b}_{\nu} \right]\right\} .
269: \end{split}
270: \label{fe_sa}
271: \ee
272: The symbol $\sum_{ab}^{\prime }$ stands for a sum which excludes terms
273: with any equal indices; i.e. $a \neq b$. The paramagnetic
274: contribution $\beta f_{PM}=-(\beta \: J)^2 m(m-3)/2$ has been subtracted as it does not
275: depend on the order parameter. One then looks for a
276: replica symmetry breaking ansatz for $B^{ab}$. Here we consider this to the first
277: level in the standard Parisi procedure.
278:
279: Using the standard procedure of the replica symmetry breaking
280: method, one groups the $n$ replicas in blocks of $x$, where $x$ is
281: a parameter (between $1$ and $n$) to be located by the saddle
282: points equations. Each block contains $x$ replicas. Thus, one has
283: \begin{equation}
284: B^{ab}=
285: \begin{cases}
286: q & \text{if $I(a/x)=I(b/x)$} \\
287: 0 & \text{otherwise}
288: \end{cases}
289: \label{rsba}
290: \end{equation}
291: where $I(y)$ is an integer valued function: its value is the smallest
292: integer which is greater than or equal to $y$.
293: Upon substituting in Eq. (\ref{fe_sa}), one has
294: \begin{equation}
295: \begin{split}
296: \beta f &= \frac{\left( \beta J\right) ^{2}}{2}(m-1)(m+2)(x-1)q^2 +\left(
297: \beta J\right) ^{2}(x-1)m q +\left( \beta J\right) ^{2}m^2 q
298: \\
299: &- \lim_{n \rightarrow 0} \frac{1}{n}\log \Tr \exp \left[ \left( \beta
300: J\right)^2 q \sum_{\mu \nu} \sum_{k}^{n/x} \left( \sum_{a \subset
301: block(k)}S^{a}_{\mu} S^{a}_{\nu} \right)^2 \right] ,
302: \end{split}
303: \label{fe_rsb}
304: \ee
305: where there are $n/x$ blocks labelled by $k=1, \ldots, n/x$. The
306: index $a$ belongs to block $k$ if: $I(a/x)=k$.
307: We shall now perturbatively compute the free energy (\ref{fe_rsb}) by
308: Taylor-expanding the free energy up to fourth order in $q$. This
309: approximation implies that one is assuming the transitions to be
310: continuous or at worst weakly discontinuous.
311: After some lengthy but straightforward algebra,
312: one obtains the following expression for the free energy:
313: \begin{equation}
314: \begin{split}
315: \beta f (q,x) &=-\frac{\alpha_{2}}{2}\:(x-1)\:t \:
316: q^{2}-\frac{\overline{\alpha}_{3}}{3}\:(x-1) \:
317: q^{3}-\frac{\hat{\alpha}_{3}}{3}\:(x-1)(x-2) \: q^{3} \\
318: &-\frac{\beta_{1}}{12}\:(x-1)q^{4}-\frac{\beta_{3}}{12}\:(x-1)(x-2)q^{4}-
319: \frac{\beta_{5}}{12}\:(x-1)(x-2)(x-3)q^{4}+ O[q^5],
320: \end{split}
321: \label{functional}
322: \end{equation}
323: where the coefficients in the free energy expansion are given by
324: \be
325: \begin{split}
326: \alpha_2 &= \frac{(m-1)(m+2)^3}{4m^2 \left(1-t\right)^2} \\
327: %%
328: \overline{\alpha}_3 &=\frac{(m-1)(m-2)(m+2)^5}
329: {4m^3(m+4)\left(1-t\right)^3} \\
330: %%
331: \hat{\alpha}_3 &= \frac{(m-1)(m+2)^4}{4m^3\left(1-t\right)^3} \\
332: %%
333: \beta_1 &= \frac{3(m-1)(m-4)(m+2)^6(m^2 +m -3)}
334: {4m^4(m+4)(m+6)\left(1-t\right)^4} \\
335: %%
336: \beta_3 &= \frac{3(m-1)(m-2)(m+2)^6}{m^4(m+4)\left(1-t\right)^4} \\
337: %%
338: \beta_5 &= \frac{3(m-1)(m+2)^5}{4m^4\left(1-t\right)^4} ,
339: \end{split}
340: \ee and $t$ is the reduced temperature, defined as
341: $t=1-\left(T^2/T_{C}^2\right)$, where
342: \be
343: T_{C}=\frac{2m
344: \,J}{(m+2)k_B} .
345: \ee
346: For $m$ less than a critical value $m^*$, discussed below, the transition is
347: continuous, with the same $T_{C}$ as predicted by $RS$ theory \cite{gs}. In this case
348: one may use the saddle point method to evaluate the extremal values of the
349: parameters $q$ and $x$. These saddle point equations are
350: %%%%%%%%
351: \begin{equation}
352: \frac{\partial f}{\partial q}=\frac{\partial f}{\partial x}=0,
353: \nonumber
354: \end{equation}
355: %%%%%%%%%%%%%%%%%
356: from which one has
357: \be
358: \begin{split} 0&= 3t\alpha _{2}+3q(\overline{\alpha}_{3}+(x-2)
359: \hat{\alpha } _{3})+q^{2}\left[ \beta _{1}+(x-2)(\beta_{3}+(x-3)
360: \beta _{5}) \right]
361: \\
362: 0 &= 6t\alpha _{2}+4q(\overline{\alpha}_{3}+(2x-3)
363: \hat{\alpha } _{3})+q^{2}\big[\beta _{1}+(2x-3)\beta_{3}
364: +(3x^{2}-12x+11)\beta _{5}\big].
365: \end{split}
366: \label{sad_point}
367: \ee
368: According to the value of the quadrupole dimension $m$, the transition from the
369: higher temperature $RS$ phase can occur either continuously or discontinuously.
370: The transition is continuous in $q$ for $m<m^*\simeq 3.37$, i.e. when the coefficient
371: $\hat{\alpha}_{3}$ becomes larger than $ \overline{\alpha}_{3}$. $q$ and $x$ satisfy the
372: following equations that express the extremum of the free energy functional
373: (\ref{functional})
374: \begin{subequations}
375: \begin{eqnarray}
376: t<0 &:&
377: \begin{cases}
378: q &=0 \\
379: x & \text{undetermined}
380: \end{cases}
381: \label{s_p_sol_1}
382: \\
383: t>0 &:&
384: \begin{cases}
385: q & =-\frac{t\alpha _{2}}{2(\overline{\alpha }_{3}-\hat{\alpha }_{3})}
386: +
387: \frac{t^{2}\alpha _{2}}{48\left(\overline{\alpha }_{3}-\hat{\alpha}_{3}\right)^{3}\:
388: \hat{\alpha}_{3}^{2}}
389: \left\{ \left( \overline{\alpha }_{3}^{2}+4\overline{
390: \alpha }_{3}\hat{\alpha }_{3}-15\hat{\alpha }_{3}^{2}\right) \beta _{5}
391: \right.
392: \\
393: & + \left. \hat{
394: \alpha }_{3}\left[ -5\beta _{1}\hat{\alpha }_{3}+\beta _{3}\left( 7\hat{
395: \alpha }_{3}-2\overline{\alpha}_{3}\right) \right] \right\}
396: +O \left[ t^{3}\right],
397: \\
398: x & =\frac{\overline{\alpha }_{3}}{\hat{\alpha }_{3}}+\frac{t\alpha _{2}
399: \left[ \hat{\alpha }_{3}^{2}(-\beta _{1}+\beta _{3})+\beta _{5}(\overline{
400: \alpha }_{3}^{2}-2\overline{\alpha }_{3}\hat{\alpha }_{3}-\hat{\alpha }
401: _{3}^{2})\right] }{4(\overline{\alpha }_{3}-\hat{\alpha }_{3})\hat{\alpha }
402: _{3}^{3}}+O[t^{2}].
403: \end{cases}
404: \label{s_p_sol_2}
405: \end{eqnarray}
406: \label{s_p_sol_tutti}
407: \end{subequations}
408: %%%%%%%%%%%%%%%%%%%%%%%
409: Close to the transition temperature $T_C$, i.e. when $t\gtrsim 0$, the
410: solution(\ref{s_p_sol_2}) is valid only for $2<m\le m^{*}\simeq 3.37$ within $1RSB$
411: subspace. At $m^*$ the cubic term in the free energy functional
412: (\ref{functional}) changes sign. This coincides with $x=1$. Thus, above
413: $m^*$, the transition is not anymore continuous. The parameters $q$ and
414: $x$ are plotted for $m=3$ within this approximation in Fig.
415: \ref{quadru_q_x_T}.
416: %%%%%%%%%%%%%%%%%%
417: \begin{figure}
418: \begin{center}
419: \includegraphics[scale=.35]{fig1.eps}
420: \caption{\label{quadru_q_x_T} (a) Plot of the one-step parameter $q$ as
421: a function of the temperature $T$ (in units of $k_B/J$); (b) Plot of the one-step breaking
422: parameter $x$ as a function of the temperature $T$ (in units of $k_B/J$).
423: Both plots are drawn for $m=3$ and within the quartic approximation for
424: the free energy (\ref{functional}).}
425: \end{center}
426: \vspace{8mm}
427: \end{figure}
428: %%%%%%%%%%%%%%%%%%%%%%%%%%%%
429: By substituting Eq. (\ref{s_p_sol_2}) in Eq. (\ref{functional}) one finds
430: the free energy of the glass phase close to the transition, to be
431: \be
432: \beta f = \frac{\alpha_2^3 \: t^3}{24 \hat{\alpha }_{3}\left(\hat{\alpha
433: }_{3}- \overline{\alpha }_{3}\right)} + O[t^{4}]=
434: -\frac{(m-1)(m+2)(m+4)\: t^3}{96(m^2-m-8)} + O[t^{4}] .
435: \ee
436: Below the transition the free energy is larger than that of
437: the paramagnetic phase.
438:
439:
440: On the other hand, when $m>m^{*}$ there is still a glass solution to the saddle
441: point equations but with a discontinuous onset of $q$ from the higher
442: temperature $RS$ phase. The
443: transition may be found with the additional requirement that
444: the free energy in the paramagnetic phase is equal to the one in
445: the glass phase with break point $x$ equal to $1$. Denoting this
446: transition temperature by $T_{D}$, to the quartic order in $q$
447: for the free energy, one finds that
448: \be
449: t_{D}=\frac{2(\overline{\alpha }_{3}-\hat{\alpha}_{3})^{2}}{3\alpha _{2} (\beta _{1}-\beta
450: _{3}+2\beta _{5})}=\frac{2(m+6)(m^2-m-8)^2}{9(m+4)(168+ 34 m - 35 m^2 - 5 m^3 +m^4)} ,
451: \label{t_disc}
452: \ee
453: where $t_{D}=1-\left(T_{D}/T_{C}\right)^2$. There is a discontinuous
454: jump in the order parameter at $T_{D}$ from zero to
455: \be
456: q_{D}=-\frac{2(\overline{\alpha }_{3}-\hat{\alpha }_{3})}{\beta
457: _{1}-\beta _{3}+2\beta _{5}}
458: =
459: -\frac{2m(m+6)(m^2-m-8)(1-t_{D})}
460: {3(m+2)(168+ 34 m - 35 m^2 - 5 m^3 +m^4)}.
461: \label{q_disc}
462: \ee
463: where $q_{D}=q(T=T_{D}^-)$. In the neighbourhood of the
464: transition temperature $T_{D}$ one finds that $(x-1) \propto
465: (t-t_{D})$ and that the free energy of the quadrupolar glass phase is
466: given by
467: %%%%%%%%%%%%%%%%%
468: \be
469: \beta f=q_{D} \left(t-t_{D}\right)^{2}+O\left[
470: \left(t-t_{D}\right)^{3} \right] .
471: \nonumber
472: \ee
473: Even though the transition is discontinuous in the order parameter $q$, there
474: is no discontinuity in any thermodynamic quantities. Moreover, there is no
475: latent heat at the transition. This behaviour is qualitatively common to a whole
476: class of mean-field models of spin glasses e.g. the $p>2$ spin model beneath a
477: critical field \cite{gardner_85} and the Potts glass model above the critical
478: Potts dimension $p=4$ \cite{gks,kirkwoly_87,ck}.
479:
480: Since the order parameter has a discontinuous
481: jump at the transition temperature when $m>m^{*}$,
482: the perturbative approach should not be valid anymore.
483: However, one may control the approximation by setting $
484: m=m^{*}+\varepsilon$, where $\varepsilon \ll 1$. Thus, one obtains a quadrupolar
485: glass phase with broken replica symmetry appearing below $t_{D} \propto
486: {\varepsilon}^{2}$, with $q_{D} \propto \varepsilon$ and $x(T \to
487: T_{D}^{-}) \to 1$. Explicitly, to leading order, one has
488: \begin{subequations}
489: \be
490: q_{D}(m=m^*+\varepsilon) =\frac{2m^{*} g(m^*) (m^{*}+4)(1-2m^{*}) }
491: {m^{*}+2}\,\varepsilon +O[\varepsilon^{2}],
492: \label{q_RSB}
493: \ee
494: \be
495: t_{D}(m=m^*+\varepsilon) =4 g(m^*)(m^{*2}-m^{*}-5/2)
496: \,\varepsilon^{2}+O[\varepsilon ^{3}],
497: \label{t_RSB}
498: \ee
499: \end{subequations}
500: where
501: $g(m^*)=(m^*+6)/\left[3(m^*+4)(m^{*4}-5m^{*3}-35m^{*2}+34m^{*}+168)\right]<0$. In
502: the next section, we shall investigate the stability of the $1RSB$ solution found
503: against small further $RSB$ fluctuations.
504:
505:
506: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
507:
508:
509:
510: \section{Stability analysis}
511: \noindent In order to study the stability of the $1RSB$ ansatz one
512: introduces $1RSB$-breaking fluctuations
513: %
514: \be
515: B^{ab}=q \, \delta_{G_{a}G_{b}}+\eta ^{ab} ,
516: \label{fluct}
517: \ee
518: and expands the free energy to second order in the fluctuations
519: $\eta^{ab}$ \cite{at}. The group Kronecker delta $
520: \delta_{G_{a}G_{b}}$ is unity if $a$ and $b$ belong to the same
521: group and zero otherwise \cite{ck}. One has to compute the second
522: derivatives of the free energy (\ref{fe_sa}) with respect to
523: $\{B^{ab}\}$ at the $1RSB$ solution. The Taylor expansion of the
524: free energy around the $1RSB$ solution is
525: %
526: \be
527: f=f(q_D\, \delta_{G_{a}G_{b}})+\frac{\partial f}{\partial B^{ab}}
528: \; \eta^{ab}+ \frac{\partial^2 f}{\partial B^{ab}\partial
529: B^{cd}} \; \eta^{ab} \eta^{cd}+...
530: \ee
531: The quadratic form
532: \be
533: \Delta =\frac{\partial^{2}f}{\partial B^{ab}\partial B^{cd}}
534: \, \eta^{ab}\eta^{cd}
535: \label{Delta}
536: \ee
537: should be positive definite for a stable solution of the problem. It is easy to
538: see that the only non-vanishing terms are the ones where $a$, $b$, $c$ and $d$
539: all belong to the same group and the ones where $a$ and $c$ belong to group $k$,
540: and $b$ and $d$ belong to group $k'$. The
541: Hessian matrix $S$ associated with the quadratic form (\ref{Delta}) factorizes
542: in $n/x$ identical submatrices of dimension $x(x-1)/2\times x(x-1)/2$ which
543: couple intragroup fluctuations and $(n/x) \: \left[ (n/x) -1\right]/2$
544: identical submatrices of dimension $x^{2}\times x^{2}$ which couple
545: intergroup fluctuations. We shall give details only on the intragroup
546: matrices, because lengthy computations show that the intergroup matrices
547: always have positive eigenvalues in the range of validity of the
548: solution (see also Ref. \cite{ck}).
549:
550:
551: If the $1RSB$ is stable thermodynamically, all of the eigenvalues of the stability matrix
552: should be positive. The intragroup matrices $S^{(ab)(cd)}$ have three different types of
553: matrix element:
554: \begin{eqnarray}
555: S^{(ab)(ab)} &=&-2\alpha
556: _{2}t-4\overline{\alpha }_{3}B^{ab}-2\beta _{1}(B^{ab})^{2}-\frac{1}{3}\beta_3
557: \sum_{k\neq a\neq b}B^{bk}B^{ka} \nonumber
558: \\
559: S^{(ab)(ac)} &=&-2\hat{\alpha }_{3}B^{bc}-\frac{1}{6}\beta _{3} \left[
560: (B^{bc})^2 +2 B^{ab} B^{bc}+2 B^{bc}B^{ca} \right] \nonumber
561: \\
562: &-& \frac{2}{3}\beta_5
563: \sum_{k\neq a\neq b \neq c} B^{kb}B^{ck} \nonumber
564: \\
565: S^{(ab)(cd)} &=& - \frac{2}{3} \beta_5 \left( B^{bc} B^{da}+ B^{ac}B^{db}
566: \right), \nonumber
567: \end{eqnarray}
568: where $a\neq b\neq c\neq d$. Since $a$, $b$, $c$ and $d$ belong to same
569: group
570: \begin{eqnarray}
571: S^{(ab)(ab)} &=&-2\alpha _{2}t-4\overline{\alpha }_{3}q- 2\beta _{1}q^{2}-
572: \frac{1}{3}\beta_3 (x-2)q^2 \nonumber
573: \\
574: S^{(ab)(ac)} &=& -2\hat{\alpha}_{3}q-\frac{5}{6}\beta_{3} q^2 -\frac{2}{3}
575: \beta_5 (x-3) q^2 \nonumber
576: \\
577: S^{(ab)(cd)} &=& - \frac{4}{3} \beta_5 q^2 . \nonumber
578: \end{eqnarray}
579: The eigenvalues of the intragroup matrices, to order $q^2$, are given
580: by
581: \be
582: \begin{split}
583: \lambda_{1} &=-2t{{\alpha}_2} - 4q \left[ \overline{\alpha }_{3}+
584: (x-2)\hat{\alpha}_3 \right] -
585: 2q^2\big[\beta_1 + (x -2){{\beta}_3} + (x^2-5x+6){{\beta}_5} \big]
586: \\
587: \lambda_{2} &=-2t \,\alpha_2 - 2q \left[2\overline{\alpha }_{3} + (x
588: -4){\hat{\alpha}_3} \right] - \frac{1}{6}\, q^2\left[12{{\beta}_1}
589: -\big(24 - 7x \big)\beta_3 + 4\big(x^2-9x+18\big)\beta_5
590: \right] \\
591: \lambda_{3} &=-2 t \alpha_2 + 4q\big( \hat{\alpha}_{3}
592: -\overline{\alpha }_{3}\big) - \frac{1}{3}\, q^2 \, \left[ 6{{\beta }_1}
593: + \big( x -7\big) {{\beta }_3} - 4 (x -4) {{\beta }_5} \right].
594: \end{split}
595: \label{eigenvalues}
596: \ee
597: %%%%%%%%%%%%%%%%%%%%
598: The behaviour of the eigenvalues in the ordered phase is obtained
599: by substituting in to the above equations the values of the
600: parameters $q$ and $x$ pertinent to the continuous or
601: discontinuous transition. Close to the continuous transition
602: temperature, one finds that the first two eigenvalues are positive
603: in the range of validity of the solution $2<m<m^{*}$:
604: \be
605: \begin{split}
606: \lambda_1 &= \frac{(m-1) (m+2)^3}{2 m^2} \, t +O[t^2]
607: \\
608: \lambda_2 &=\frac{(m+2)^3 \left(m^3-3 m^2-10 m+12\right)}{4 m^2 \left(m^2-m-8\right)} \, t
609: +O[t^2]
610: .\end{split}
611: \ee
612: The last one, to order $t^2$,
613: \be
614: \lambda _{3}= \frac{(m-1) (m+2)^3 \left(2m^5 + 15m^4 - 8m^3- 104m^2- 32m
615: +96 \right)}{16m^2(m+6)(m^2 -m-8)^2}\, t^2 +O[t^3] ,
616: \ee
617: is positive only for $m>m_{2}^{*}\simeq 2.46$. Thus, one finds a $1RSB$ stable mean-field
618: theory with a continuous transition only in the range $m_{2}^{*}<m<m^{*}$. This lower limit
619: is the same as given in Ref. \cite{gold_elder_85} and obtained by means of a complementary
620: calculation based on the full replica symmetry breaking ($FRSB$) ansatz near $T_C$ with a
621: perturbation treatment. For $m>m^*$ the behaviour of the eigenvalues in the ordered phase is
622: obtained by setting $x=1$ and by substituting $q$ with the value $q_D$ obtained in Eq.
623: (\ref{q_RSB}). Upon inserting these values, one easily finds that all the fluctuations
624: around the ordered $1RSB$ phase are finite and positive:
625: \be
626: \begin{split}
627: \lambda_1 &=\lambda_3 =
628: \frac{g(m^*) \left(m^*-1\right) \left(m^*+2\right)^3
629: \left(-10 m^{*2}+10 m^* +3\right)}{m^{*2} }
630: \: \varepsilon^2 >0
631: \\
632: \lambda_2 &=\frac{ g(m^*) \left(m^*-4\right) \left(m^*-1\right)
633: \left(m^*+2\right)^3 \left(2 m^*-1\right) \left(2
634: m^*+5\right)}{ m^{*2} } \: \varepsilon>0 .
635: \end{split}
636: \ee
637: Thus, within our approximation, one finds a $1RSB$ stable
638: mean-field theory with a discontinuous transition when $m>m^*$.
639:
640: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
641:
642:
643:
644: \section{Dynamical transition}
645: Generally, disordered systems with a discontinuous transition have a
646: temperature $T_G$ where a dynamic instability appears. This temperature is
647: called the glass temperature and is higher than the transition
648: temperature $T_{D}$ where the replica symmetry breaks thermodynamically, if the
649: latter breaking is discontinuous.
650:
651: In the soft spin version of the Potts glass model \cite{kirkthiru}
652: it has been shown - by means of dynamical studies of the
653: mean-field theory - that indeed there is another transition at
654: temperature $ T_{G}>T_{D}$ as in the $p$-spin model for $p>2$
655: \cite{kirkthiru2}. Both static and dynamic transitions in the
656: Potts ($p>4)$ case, have also been found in Refs.
657: \cite{ck,desantis,kirkwoly_87}. In the study of the thermodynamics
658: of the quadrupolar glass, $T_G$ can be computed by means of
659: marginal stability \cite{marg_stabili}. By requiring the vanishing
660: of the first and second derivative of the free energy
661: (\ref{functional}) with respect to $q$ \be \left( \frac{\partial
662: f}{\partial q} \right)_{q=q_G} = 0 \quad \quad ;\quad \quad
663: \left( \frac{\partial^2 f}{\partial q^2} \right)_{q=q_G} = 0 .
664: \label{marg_stab_eqs} \ee one finds, within our approximation, the
665: dynamical transition temperature $T_G$ and the corresponding
666: discontinuous jump $q_G$ of the quadrupolar glass model \be t_G =
667: \frac{3(\overline{\alpha}_3- \hat{\alpha}_3)^2}{4\alpha_2 (
668: \beta_1- \beta_3 +2\beta_5)} = \frac{(m+6)(m^2-m-8)^2}
669: {4(m+4)(168+ 34 m - 35 m^2 - 5 m^3 +m^4)} \label{t_G} \ee and \be
670: q_G = -\frac{3(\overline{\alpha}_3-
671: \hat{\alpha}_3)}{2(\beta_1-\beta_3 +2\beta_5)}
672: = -\frac{m(m+6)(m^2-m-8)(1-t_{G})}
673: {2(m+2)(168+ 34 m - 35 m^2 - 5 m^3 +m^4)} ,
674: \ee
675: where $t_{G}=1-\left(T_{G}/T_{C}\right)^2$ and $q_{G}=q(T=T_{G}^-)$.
676: %%%%%%%%%%%%%%%%%
677: \begin{figure}
678: \begin{center}
679: \includegraphics[scale=.35]{fig2.eps}
680: \caption{\label{stat_dyn_q}The static $q_{D}=q\,(T=T_{D}^-)$ and
681: dynamic $q_{G}=q\,(T=T_{G}^-)$ order parameter as a function of
682: $1/\log(m)$. The dotted line is for the static value, the bold line is
683: for the dynamical one. }
684: \end{center}
685: \vspace{8mm}
686: \end{figure}
687: %%%%%%%%%%%%%%%%%
688: Again, by assuming the jump $q_G$ near the temperature $T_G$ to be small,
689: one can control the approximation by letting $m=m^*+\varepsilon$, $\varepsilon \ll 1$. Thus,
690: one has
691: \begin{subequations}
692: \be
693: q_{G}(m=m^*+\varepsilon) =\frac{3 m^* g(m^*) (1-2 m^* ) (m^*+4)}{
694: 2(m^*+2)}\, \varepsilon+O[\varepsilon^{2}],
695: \label{q_G_eps}
696: \ee
697: \be
698: t_{G}(m=m^*+\varepsilon) =\frac{9}{2} \,g(m^*)\, \left(m^{*2}- m^* -5/2\right) \,
699: \varepsilon ^2 +O[\varepsilon^{3}],
700: \label{t_G_eps}
701: \ee
702: \end{subequations}
703: Within the approximation used, one finds
704: \be
705: \frac{q_G}{q_{D}}=\frac{3}{4}+
706: \frac{3}{16} g(m^*) \left(-2 m^{*2}+2 m^*
707: +5\right) \, \varepsilon^2+
708: O\left[\varepsilon^3 \right].
709: \ee
710: It is worth noting that, to the leading order, exactly the same value for this ratio has
711: been obtained for the Potts glass in Refs. \cite{desantis, ck, kirkwoly_87}, suggesting a
712: sort of universality related to the same general structure of the free
713: energy for both quadrupolar and Potts glass models (see also Ref.
714: \cite{gold_elder_85}).
715: The results for $q_{D}$ and $q_G$ as a
716: function of $1 / \log( m )$ are shown in Fig. \ref{stat_dyn_q}.
717: %
718: The ratio between the two transition temperatures $T_G / T_{D}$ is very
719: close to one
720: \be
721: \frac{T_G}{T_{D}}=1+
722: \frac{1}{8}\, g(m^*) \, \left(-2 m^{*2}+2 m^* +5\right)\, \varepsilon^2
723: +O\left[\varepsilon^3 \right],
724: \ee
725: with $T_G$ always bigger than $T_{D}$, as shown in
726: Fig. \ref{quad_diag_phase}. For $m = 4$ (the smallest integer value of $m$
727: compatible with a discontinuous transition) one finds that
728: \be
729: \begin{split}
730: q_{D} &=0.056... \quad \quad ; \quad \quad T_{D}=1.342... \\
731: q_{G} &=0.042... \quad \quad ; \quad \quad T_{G}=1.343...
732: \end{split}
733: \ee
734: Of course, a naive extension of our result to include large $m$ results
735: is not possible since we have assumed all the transitions to be at worst
736: weakly discontinuous, implying the possibility to explore only the range $m$
737: close to $m^*$.
738: However, since the mean-field theory of the Potts glass is qualitatively
739: very similar to that of the quadrupolar glass, one may have an idea of the large
740: $m$ limit by considering the large $p$ limit in the Potts glass model. There, the ratio
741: $q_G/q_{D}$ stays close to $3/4$ for a large range of $p$, though it increases for very
742: large $p$, approaching unity. The ratio $T_G / T_{D}$ grows very slowly with $p$
743: \cite{desantis}.
744: This behaviour is different from the case of the $p$-spin model, where the ratio $q_G/q_{D}$
745: is not close to $3/4$ even for small values of $p$ and, in the limit $p \to \infty$, it
746: converges to unity. Moreover, the ratio $T_G / T_{D}$ grows faster with $p$ than in the
747: Potts problem \cite{desantis}.
748: %%%%%%%%%%%%%%%%%%
749: \begin{figure}[t]
750: \begin{center}
751: \includegraphics*[scale=.35]{fig3.eps}
752: \caption{\label{quad_diag_phase}
753: Phase diagram of the quadrupolar glass model in the $(T,1/m)$
754: plane. $T$ is in units of $k_B/J$. For $2<m<m_2^*$ the transition from the higher
755: temperature $RS$ phase occurs within the full replica symmetry breaking mechanism.
756: For $m<m^*$ the thermodynamic and dynamical transitions
757: coincide. For $m>m^*$ the dynamical transition, denoted by the dotted line, is
758: higher that the thermodynamic one (solid line). The plot is shown
759: within the quartic approximation for the free energy expansion in $q$. }
760: \end{center}
761: \vspace{8mm}
762: \end{figure}
763: %%%%%%%%%%%%%%%%%
764:
765: A useful and new representation of the phase diagram is obtained
766: by plotting the phase boundary line in the plane $(T,1/m)$. One
767: may identify three stable phases, $RS$ paramagnetic and both
768: $1RSB$ and $FRSB$ glasses. In
769: Fig. \ref{quad_diag_phase}, the phases are labelled by their
770: symmetry breaking and the manner of the onset from the paramagnet.
771: The $1RSB$ transition is continuous between $m_2^* \le m \le m^*$,
772: whereas it is discontinuous above $m^*$. Note that, at $m=m^*$,
773: the transition from $RS$ passes continuously from continuous $1RSB$ ($C1RSB$) to
774: discontinuous $1RSB$ ($D1RSB$) within the one-step $RSB$
775: phase. The dotted line in the figure corresponds to the $m>m^*$
776: dynamical transition temperature given in Eq. (\ref{t_G}).
777: %
778: The situation is analogous to that of the Potts glass model which shows a
779: crossover from the continuous transition to the discontinuous transition as the
780: number of Potts states increases \cite{gks,ck}.
781: The $p$$>$$2$-spin Ising and spherical spin glasses also show transitions
782: from $C1RSB$ to $D1RSB$ as an applied field is reduced but differ from
783: the present problem in that the critical field makes also a maximum in
784: the transition temperature, in contrast to the present monotonic
785: variation with $1/m$. For $m<m_2^*$ the transition is continuous to full replica symmetry
786: breaking. A phase line (not shown, but continuous) separates the one-step and full replica
787: symmetry breaking phases within the $RSB$ region.
788:
789: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
790:
791:
792: \section{Concluding remarks}
793: \noindent In this paper we have investigated the quadrupolar glass
794: model in the framework of the replica method. Upon introducing a
795: simple one-step replica symmetry breaking ansatz, one may find a
796: stable mean-field theory with a continuous or discontinuous
797: transition, according to the value of the quadrupole dimension
798: $m$. The transition is continuous to one-step replica symmetry
799: breaking in the range of the quadrupolar dimension $2.46<m<3.37$.
800: For the discontinuous transition ($m>3.37$) there are two
801: different transition temperatures. We have computed the ratio
802: $q_G/q_{D}$, where $q_G$ and $q_{D}$ are the order parameters
803: associated, respectively, with the dynamic and thermodynamic
804: transition. The ratio between the two transition temperatures
805: $T_G/T_{D}$ is also computed. Within the approximation used, the
806: values of these ratios, $q_G/q_{D}=3/4$ and $T_G/T_{D}\simeq 1$
807: (to leading order), are the same as those found in Refs. \cite{ck,
808: desantis, kirkwoly_87} for the Potts glass model.
809:
810: The results we have obtained confirm the general wisdom that the
811: properties of the quadrupolar glass, with continuous ($m<m^*$) and
812: discontinuous transitions ($m>m^*$), are similar to those of the
813: $p<4$ and $p>4$ Potts glass well studied in the literature
814: \cite{gks,ck,desantis,kirkthiru2,kirkwoly_87}. Although the
815: investigation focused on the quadrupolar glass phase, in the wider
816: $(J_0,J,T)$ space there should exist different types of
817: ferromagnet, collinear and canted, see e.g. Ref. \cite{gs}.
818:
819:
820: The full phase diagram should include also another curve which may
821: be captured by complexity arguments. In analogy with the $p$-spin
822: spin glass there should be another critical line $T^{compl.}$
823: associated with the onset of macroscopic complexity
824: \cite{cavagna99}.
825:
826:
827: %
828: \section*{Acknowledgements}
829: We thank P. M. Goldbart for the benefit of stimulating
830: discussions. We also thank the EPSRC for
831: financial support in Oxford under grants GR/R83712/01 and 97304299 and ESF
832: programme SPHINX for the opportunity to meet to continue this work.
833:
834: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
835:
836:
837: \begin{thebibliography}{30}
838:
839: \bibitem{hochli_90} H\"ochli U T, Knorr K and Loidl A 2002 \ADP {\bf 51} 589
840:
841: \bibitem{binder_92} Binder K and Reger J D 1992 \ADP {\bf 41} 547
842:
843: \bibitem{sk} Sherrington D and Kirkpatrick S 1975 \PRL {\bf 35} 1792
844:
845: \bibitem{gold_elder_85} Goldbart P and Elderfield D 1985 \JPC {\bf 18} L229
846:
847: \bibitem{gks} Gross D J, Kanter I and Sompolinsky H 1985 \PRL {\bf 55} 304
848:
849: \bibitem {ck} Cwilich G and Kirkpatrick T R 1989 \JPA {\bf 22} 4971
850:
851: \bibitem{desantis} De Santis E, Parisi G and Ritort F 1995 \JPA {\bf 28} 3025
852:
853: \bibitem{crisanti_92} Crisanti A and Sommers H-J 1992 \ZPB {\bf 87} 341
854:
855: \bibitem{gardner_85} Gardner E \NPB {\bf 257} 747 (1985)
856:
857: \bibitem{gs} Goldbart P and Sherrington D 1985 \JPC {\bf 18} 1923
858:
859: \bibitem{goldbart85} Goldbart P 1985 Ph.D. Thesis Imperial College
860:
861: \bibitem{kirkwoly_87} Kirkpatrick T R and Wolynes P G 1987 \PRB {\bf 36} 8552
862:
863: \bibitem {at} de Almeida J R L and Thouless D J 1978 \JPA {\bf 11} 983
864:
865: \bibitem{kirkthiru} Kirkpatrick T R and Thirumalai D 1988 \PRB {\bf 37} 5342
866:
867: \bibitem{kirkthiru2} Kirkpatrick T R and Thirumalai D 1987 \PRB {\bf 36} 5388
868:
869: \bibitem{marg_stabili} Sommers H-J 1983 \ZPB {\bf 50} 97; Kondor I and De
870: Dominicis C 1986 Europhys. Lett. {\bf 2} 617
871:
872: \bibitem{cavagna99} Cavagna A, Garrahan J P and Giardina I 1999 \JPA
873: {\bf 32} 711
874:
875:
876: \end{thebibliography}
877:
878: \end{document}
879: