1: \documentclass[aps,twocolumn,amsmath,letterpaper]{revtex4}
2: \usepackage{amssymb}
3: \usepackage{graphicx}
4: \usepackage{array}
5: \usepackage{hhline}
6: \usepackage{longtable}
7: %\usepackage{dcolumn}
8: \cleardoublepage
9: \newcommand{\und}[2]{\underset{\underset{#1}{}}{#2}}
10: \newcommand{\Mvariable}[1]{#1}
11: \newcommand{\multsp}{}
12: \renewcommand{\thetable}{\Roman{table}} \thetable
13: \newcommand{\<}[1]{\hspace{-0.33333em}#1\hspace{-0.33333em}}
14: \newcommand{\rsp}[1]{\hspace{-0.15em}#1\hspace{-0.15em}}
15: \newcommand{\vetor}[1]{\mbox{\boldmath ${#1}$}}
16:
17: %\allowdisplaybreaks[4]
18: %\topmargin 0.in
19:
20: \begin{document}
21:
22: \title{The Critical Properties of Two-dimensional Oscillator Arrays}
23:
24: \author{Gabriele Migliorini}
25: \affiliation{ The Neural Computing Research Group\\ School of
26: Engineering and Applied Sciences, Aston University\\ Birmingham B4
27: 7ET, United Kingdom.}
28:
29: \begin{abstract}
30:
31: We present a renormalization group study of two dimensional arrays
32: of oscillators, with dissipative, short range interactions. We
33: consider the case of non-identical oscillators, with distributed
34: intrinsic frequencies within the array and study the
35: steady-state properties of the system.
36: In two dimensions no macroscopic mutual entrainment is found but,
37: for identical oscillators, critical behavior of the
38: Berezinskii-Kosterlitz-Thouless type is shown to be present.
39: We then discuss the stability of (BKT) order in the physical case of
40: distributed quenched random frequencies.
41: In order to do that, we show how the steady-state dynamical properties
42: of the two dimensional array of non-identical oscillators are related
43: to the equilibrium properties of the $XY$ model with quenched randomness,
44: that has been already studied in the past. We propose
45: a novel set of recursion relations to study this system within
46: the Migdal Kadanoff renormalization group scheme, by mean of the discrete
47: clock-state formulation. We compute the phase diagram in the presence
48: of random dissipative coupling, at finite values of the clock state
49: parameter.
50: Possible experimental applications in two dimensional arrays of
51: microelectromechanical oscillators are briefly suggested. \\
52:
53: PACS numbers: 64.60A-64.60ae,64.60.Bd, 64.60.Ht\\
54:
55: \end{abstract}
56: \maketitle
57: \def\s{\rule{0in}{0.28in}}
58: %\narrowtext
59: \section{Introduction}
60:
61: The study of the dynamical properties of large arrays of
62: self-sustained oscillators
63: with distributed intrinsic frequencies is an interesting
64: problem bridging non-equilibrium statistical physics and non-linear
65: science. Large populations of interacting, self-sustained oscillators
66: are known to model a great variety of biological systems and several
67: progresses have been made in the last decades \cite{zero,zerob,zeroc}.
68: If dense arrays of coupled self-oscillators, close to an oscillatory
69: instability of the Hopf type, are expected to present interesting
70: critical behavior, the analysis of such systems in the
71: low dimensional case, i.e. when the range of the interactions is short,
72: has been quite a challenging one and very little attempts have been
73: made in this direction.
74: The mean-field Kuramoto model of phase oscillators represents a very
75: special case where one can predict in a clear way the macroscopic mutual
76: entrainment (MME) properties of the system, but attempts to generalize such
77: results to the low dimensional case are usually difficult and the
78: majority of the studies of oscillator arrays have been devoted to
79: the $all~to~all$ case, and to networks with relatively high
80: degree of connectivity. Meanwhile, the case of short range interactions
81: is a very interesting one and have a lot of potential applications,
82: in particular considering the case of dissipatively and/or reactively
83: coupled arrays of mechanical oscillators at micro and nano scales.
84:
85: The oscillator lattice problem, namely the finite-dimensional, nearest
86: neighbour version of the Kuramoto model, has been attacked in the past
87: by several authors \cite{one,two}.
88: The main conclusions are that the synchronization
89: properties one normally finds in the mean-field case are drastically
90: changed by the short range nature of the connectivity, and that no
91: synchronization is expected in two dimensions \cite{one,two,twob}.
92: An interesting analysis
93: based on real space renormalization group theory \cite{three} suggests
94: an explicit expression for the lower critical dimension, below which one
95: do not expect macroscopic mutual entrainment, the latter being the dynamical
96: analogue of ferromagnetic order.
97: A related situation occurs within the equilibrium statistical physics of
98: spin systems with continuous degree of freedom of the $XY$ type,
99: as already suggested in \cite{one}.
100: In the statistical mechanics of continuous spin models of the mean-field
101: type the system has a spontaneous magnetization and the out-of-equilibrium
102: behavior is a very interesting and open problem \cite{threeb}. These
103: models relate closely to the Kuramoto model, where the additional difficulty
104: of the intrinsic, quenched frequency distribution has to be taken into
105: account, and where the system does not reach equilibrium but is self-driven
106: in an out-of-equilibrium steady state regime.
107:
108: Both the dynamical and equilibrium properties of continuous spin models
109: change importantly when short ranged, nearest neighbour interactions
110: are considered. Rather than ordinary symmetry breaking, in the
111: two-dimensional $XY$ model algebraic order (AO) occurs.
112: Namely the system has infinite correlation length, but no
113: magnetization at finite temperature \cite{threec,threed,threee}.
114:
115: When considering the case of low dimensional arrays of coupled
116: oscillators, the problem is further complicated by the fact that
117: the system is driven out-of-equilibrium, so that the standard tools
118: of equilibrium statistical mechanics cannot apparently be exploited.
119: On the other side, if the analogy between arrays of identical oscillators and the
120: statistical physics of $XY$ type of models with continuous degrees of
121: freedom is a well known fact and attempts to relate the dynamical
122: properties of low-dimensional arrays of oscillators to the classical
123: theory of dynamical critical phenomena \cite{four,five} have been made,
124: it is interesting to see if the great
125: deal of results obtained in the past for the finite dimensional
126: $XY$ model in the presence of disorder, could possibly relate to the
127: problem of arrays of dissipatively coupled, short
128: ranged oscillators with a natural intrinsic frequency spread.
129: We refer to this last problem as the finite dimensional Kuramoto
130: model or lattice oscillator problem \cite{one}.
131: Obviously the simplest case one
132: may consider is the strictly two dimensional case, where the
133: $XY$ type of models are widely studied and where real space renormalization
134: group theory has been capable of predicting a great deal of results in the
135: past \cite{threee}.
136:
137: We will show that the formal analogy between identical,
138: dissipatively coupled arrays of oscillators and model A of
139: non-equilibrium critical phenomena \cite{fiveb} can be extended, in the two
140: dimensional case, to the case of non-identical
141: oscillators, so that the steady state non-equilibrium properties in the
142: system may be studied as a function of the quenched intrinsic frequency
143: distribution. This will allow us to describe the steady state out of
144: equilibrium properties of two dimensional oscillator arrays by mean of
145: classical statistical mechanics. The possibility to consider
146: non-identical oscillators is essential, in that a finite width of the
147: intrinsic frequency distribution is the fundamental parameter one
148: consider in all Kuramoto type of models. We will also include a physical
149: temperature corresponding to white noise in the array and consider
150: the case of bimodal frequency distributions, options also
151: commonly considered in the context of the mean-field Kuramoto model \cite{zeroc}.
152:
153: In this respect we will show that the steady state
154: properties of a two dimensional array of dissipatively coupled
155: non-identical oscillators can be related to the equilibrium
156: properties of an effective $XY$ model in the presence of random
157: Dzyaloshinskii-Moriya (DM) interactions.
158: The random (DM) interaction effectively induces phase canting between
159: neighbouring oscillators, and can be related, as we will show below, to
160: the quenched distribution of the native intrinsic frequencies.
161: It is then argued that
162: the steady state non-equilibrium properties of two dimensional arrays of dissipatively
163: coupled non-identical oscillators (in the phase approximation) are related to the
164: equilibrium properties of the two dimensional $XY$ model in the presence of
165: disorder, where again a great deal of real space renormalization group theory
166: results are available, even though the effect of quenched
167: random interactions on the classical properties of the two dimensional
168: $XY$ model is still a matter of debate within the condensed matter
169: physics community.
170: A similar approach to describe the non-equilibrium
171: properties of two dimensional superconducting arrays with external currents
172: by mean of the equilibrium properties of an effective $XY$ model has
173: been considered in the past \cite{six}.
174:
175: Before we begin our analysis, it is important to stress that
176: the properties of the Kuramoto model in low dimensions have already been discussed
177: in the past \cite{one,two,twob,three}. In particular one could
178: argue at this point that in two dimensions no macroscopic mutual
179: entrainment (MME) is expected, so that collective synchronization
180: is simply not present. This argument reminds a similar one about the very
181: nature of the critical properties of the two dimensional $XY$ model
182: \cite{threee,threed}. If magnetization is known not to be expected at finite temperatures,
183: the two dimensional $XY$ model is known however to present critical properties of a special
184: type (AO). We will show in the following that the dynamical analog of algebraic
185: order is present in dissipatively coupled two dimensional arrays of
186: oscillators (and we might refer to it as algebraic synchronization).
187: Dissipatively coupled oscillators in two dimensions are characterized by a
188: phase transition of the Berezinskii-Kosterlitz-Thouless (BKT) type, as
189: we will show in this article, even for non-identical oscillators, namely
190: when one consider the non trivial case of distributed intrinsic
191: frequencies. The main question is how algebraic order (AO) reacts to the
192: disruptive effect of the intrinsic frequency distribution, and if
193: a steady state algebraic synchronization regime exists in two
194: dimensional arrays of oscillators.
195:
196: We will address these questions by evaluating the renormalization group
197: recursion relations, within the position space Migdal-Kadanoff (MK)
198: approximation, for the case of an $XY$ model with quenched random
199: interactions induced by effective Dzyaloshinskii-Moriya (DM)
200: interactions. We consider the standard formulation of the $XY$
201: model within the (MK) approach \cite{sixb}, adopting the
202: discretization scheme introduced in reference \cite{seven} and
203: proposing a novel set of recursion relations that will allow us to
204: include both random exchange and random (DM) interactions,
205: consistently with the usual methods of real space
206: (MK) renormalization group of disordered spin systems \cite{eight}.
207: We will be considering a renormalization group rescaling length $b=3$, to treat
208: exchange and random (DM) interactions of random opposite signs on equal
209: footing \cite{nine}, effectively generalizing the position space
210: renormalization group (PSRG) recursion relations of the two dimensional
211: XY model to the case of random exchange and (DM) interactions.
212: Together with the novel set of (PSRG) recursion relations, we present
213: preliminary results obtained by means of an algorithmic
214: method capable to implement the recursion relations and evaluate
215: the corresponding phase diagram, for different model systems,
216: corresponding to different forms of disorder, by fixing the proper
217: choice of the initial conditions in the renormalization group
218: transformation. Choices we will consider ranges
219: from the $XY$ spin glass (XYSG), where randomness is in the exchange
220: interaction and no (DM) interactions are present, the
221: two dimensional ferromagnet with Dzyaloshinskii-Moriya interactions
222: (XYDM) problem, where randomness is in the (DM) term and no random exchange
223: interaction term is present, and ultimately the random gauge glass (RGG)
224: problem, as an important general case where both type of randomness are
225: present simultaneously, so that gauge invariance is restored.
226: Our MK renormalization
227: group approach will be general enough to describe any of the models
228: above, by a proper choice of the initial conditions in the
229: renormalization group flows.
230:
231: The proper initial condition of primary interest, i.e. the one we
232: wish to consider to be able to predict the type of behavior one expects in two
233: dimensional arrays of dissipatively coupled arrays of non-identical
234: oscillators, is, as we now show, of the (XYDM) type,
235: even though randomness is the dissipative/exchange coupling term is
236: also relevant when discussing two dimensional oscillator arrays.
237:
238: Our results do not restrain to the oscillator array problem only, but
239: are general enough and pertain to different type
240: of systems that have been considered in the past within the condensed
241: matter theory community, and where it is relatively common to consider
242: the properties of the $XY$ model in the presence of quenched random
243: interactions of various types.
244:
245: The shape of the phase diagram of the two dimensional $XY$ model in
246: the presence of quenched random (DM) impurities is the subject of a
247: long standing debate \cite{nineteen,twentytwo,nineb,twentyfour} we here
248: address and show to relate also to the study of dissipatively
249: coupled, two dimensional oscillator arrays.
250:
251: \section{The Model}
252:
253: The dynamical behavior of oscillator arrays in the vicinity of a
254: supercritical Hopf bifurcation can be described in terms of complex
255: amplitude equations, related to the amplitude and phase of each
256: oscillator. In the general case of reactive and dissipative coupling,
257: as well as considering the presence of stiffening non-linearities
258: of the Duffing type, and in the case of an intrinsic frequency distribution
259: spread within the oscillators, one can write the following equations for the complex
260: amplitude \cite{ten,eleven}, where, as mentioned above, we add a finite
261: temperature due to brownian fluctuations within the array so that $ \langle
262: \eta_k(t) \eta_{k'}(t') \rangle = 2T \delta_{k,k'} \delta(t-t')$.
263: We restrict the range of the interactions to neighbouring
264: oscillators on the square lattice.
265: \begin{eqnarray}
266: \frac{dA_k}{dt}&=& ( \mu + i \omega_k)A_k-( \gamma+i
267: \alpha)|A_k|^2A_k \nonumber \\
268: &+&( J+i R) \sum_{j \in { \cal L}_k}(A_j-A_k)+\eta_k
269: \label{one}
270: \end{eqnarray}
271:
272: We will assume that the oscillators are not identical so that quenched
273: intrinsic frequencies $\omega_k$ are considered, with a given
274: (e.g. bimodal) distribution $ P ( \omega)$. The two parameters $ \mu$
275: and $ \gamma $ are related to the self-sustaining mechanism that drives
276: each oscillator; for oscillators of the Van der Pol type we can
277: choose $ \mu=\gamma$. The coupling $J$ is the dissipative coupling and
278: we here assume to be uniform and short ranged. In what follows we will
279: also consider the possibility of a random dissipative/exchange coupling.
280: Finally the reactive term $R$, which is related to a mechanical coupling
281: between the oscillators, is also short ranged and eventually distributed
282: around an average value $R_o$.
283: We note at this point that the interesting case of reactively coupled
284: oscillators has been considered in reference \cite{fourteen}, in the
285: mean-field approximation. We neglect this last term, and rewrite
286: the above equation as
287:
288: \begin{eqnarray}
289: \frac{dr_k}{dt}&=&(\mu-\gamma r_k^2)r_k+J \sum_{j \in { \cal L}_k}
290: r_j \big ( 1-\cos(\theta_j-\theta_k) \big )+\eta^{r}_k \nonumber \\
291: \frac{d \theta_k}{dt}&=&(\omega_k - \alpha_k r^2_k)+J \sum_{j \in { \cal L}_k}
292: r_j/r_k \sin(\theta_j-\theta_k)+\eta^{ \theta}_k
293: \end{eqnarray}
294:
295: Including a reactive term means one should include $ R \sum_{j \in { \cal L}_k}
296: r_k \sin(\theta_j-\theta_k) $ to the right end side of the first
297: equation and $R \sum_{j \in { \cal L}_k} r_j/r_k ( \cos(\theta_j-\theta_k) -1) $
298: to the right end side of the second.
299: We consider in what follows the isochronous case, assuming that
300: the Duffing parameter $ \alpha_k$ can be neglected, as in the original
301: formulation of \cite{zerob}.
302: The magnitude of the complex amplitude crosses over to one as soon
303: as the value of $ \mu$,
304: related to the Van
305: der Pol strength of the autonomous oscillators, is properly tuned.
306: Assuming the width of the intrinsic frequency distribution to be relatively
307: high peaked around the average value $ \omega_o$, the above equation reduces
308: to the phase
309: approximation, \cite{zerob} and one writes
310: \begin{equation}
311: \frac{d \theta_k}{dt}= \omega_k +J \sum_{j \in { \cal L}_k}
312: \sin(\theta_j-\theta_k)+ \eta^{ \theta}_k
313: \label{KU}
314: \end{equation}
315: where ${ \cal L}_k$ are the neighbouring sites to $k$.
316: In the simple case of identical oscillators one recover the Langevin equations
317: for the $XY$ model. Differently, we are back to the oscillator lattice
318: problem.
319: In other terms, as mentioned in the introduction, the dynamical behavior of identical oscillators,
320: reduces, up to a global redefinition of the complex amplitudes, to the
321: coarsening dynamics of the $XY$ model, so that the above equation
322: (\ref{one}) reduces to model A of critical dynamical phenomena
323: \cite{fiveb,twelve,thirteen} and the formal analogy between a critical Hopf
324: bifurcation and phase transition theory can easily be drawn in this
325: case. The dynamical behavior of non-identical oscillators is certainly
326: harder to solve, as usually happens when dealing with models in the
327: presence of quenched randomness. We are effectively studying
328: the low-dimensional version of the Kuramoto model, and the first
329: question is how the random frequency term affects the onset
330: of algebraic order (AO) we know to be present in the case of identical oscillators.
331: We will address these questions in the rest of the paper, showing how the
332: non-equilibrium steady state properties of the problem can be related to the
333: equilibrium properties of an effective $XY$ model with random quenched
334: interactions. Before doing so, it is interesting to review how (and if)
335: numerical simulations might be useful to answer, at least in part, this
336: question before we return to a real space, renormalization group calculation
337: in the rest of the paper.
338: We clearly expect that disorder, in combination to the finite
339: dimensional character of the connectivity, might result in numerical
340: simulations to be converging very slowly, a situation we faced and
341: we will discuss in the following section.
342:
343: \section{Numerical simulations}
344:
345: In the current assumption of dissipative coupling and within the phase
346: approximation we discussed above, we shown that the amplitude equations for the
347: two-dimensional array reduce to the equations of motion of the $XY$
348: model, with the inclusion of a quenched random frequency term.
349: How are the critical properties of the $XY$ model affected by the random
350: frequency term? We firstly answer this question numerically, by
351: means of direct integration methods. We considered a
352: gaussian form for the intrinsic frequency distribution and solve the
353: Langevin dynamics explicitly, via Runge-Kutta methods for a
354: system of size $N=L \times L$, following
355: the standard methods discussed in the literature \cite{fifteen,sixteen}.
356: \begin{figure}
357: \includegraphics*[scale=0.8,angle=0]{fig1.eps}
358: \caption{Helicity Modulus computed for system sizes $L=5,10,1,20,25,50$
359: in the two dimensional $XY$ model without disorder, by mean of direct
360: Runge-Kutta integration methods. The Helicity modulus jump becomes
361: steeper for increasing system sizes. We could extrapolate a critical
362: transition value of $T_{KT} \simeq 0.89$, in agreement with previous,
363: accurate estimates \cite{eighteen}.}
364: \label{Helix}
365: \end{figure}
366: \begin{figure}
367: \includegraphics*[scale=0.6,angle=270]{fig2.eps}
368: \caption{Critical exponent $ \eta$. We computed explicitly the
369: correlation function at increasing values of the temperature $T<T_KT$,
370: observing as expected, algebraic decay. The estimated value of the
371: exponent is consistent with the value $ \eta=1/4$, expected at
372: $T=T_{KT}$ \cite{threee}.}
373: \label{Eta}
374: \end{figure}
375: We also checked that the
376: same algorithm, in the case of mean-field interactions and when random
377: frequency terms are included, reproduces the results of the Kuramoto model.
378: We also check that we were able to obtain the expected results in the
379: case of identical oscillators in two dimensions.
380: Even though the system does not order at finite temperature, phase
381: correlation effects are present and are quantified by finite values of the
382: helicity modulus \cite{seventeen}, as predicted by the Kosterlitz
383: Thouless renormalization group theory. The numerical results we here
384: show are consistent with the expected (BKT) transition
385: \cite{threee,eighteen}. The algebraic decay of the correlation function
386: can be measured and the weak violation of universality critical
387: phenomena is seen, as expected, with the critical exponent being
388: close to the expected value $ \eta=1/4$ at the transition point (see
389: Fig. \ref{Eta}).
390: In the presence of disorder, as usually occurs in low dimensional
391: systems, things become difficult from the perspective of numerics.
392: We clearly observed that for each given quenched realization of the intrinsic
393: frequency distribution, the system reaches a steady state, i.e., for any
394: realization of the disorder, we observed that the Runge-Kutta algorithm
395: reaches a time independent state. We computed the average helicity
396: modulus over several realization of disorder, for a given system size
397: ranging, as in the pure case, from $L=5$ to $L=50$, for two dimensional
398: arrays of size $L \times L$. We observed however that the results
399: we obtained from numerics in the presence of disorder are not that
400: clear, the sample-to-sample fluctuations
401: becoming very strong as the width of the intrinsic frequency distribution
402: grows. For very small system sizes we observe the helicity modulus to
403: be finite below the critical point, that decreases with the disorder
404: strength, as in Fig. \ref{helix_d}. However for systems of size as small as
405: $N=L \times L=16^2$ it becomes very difficult to have sample-to-sample
406: fluctuations under control, so no conclusive results were obtained
407: for these system sizes, even after a run of several weeks on a standard
408: workstation.
409: \begin{figure}
410: \includegraphics*[scale=0.8,angle=0]{fig3.eps}
411: \caption{Helicity Modulus computed for system size $N=16 \times 16$.
412: For this small system size we were able to control sample-to-sample
413: fluctuations and estimate the helicity modulus for increasing values
414: of the disorder strength, $\Delta=0.05,0.1,0.2,0.3$, defined as the
415: width of the gaussian distribution of the intrinsic frequency
416: distribution $P(\omega)$.}
417: \label{helix_d}
418: \end{figure}
419: Differently, in the mean field case, where each oscillator is coupled to
420: all others with equal strength, we were able to control
421: sample-to-sample fluctuations and an average value of the Kuramoto order
422: parameter was easily obtained, consistently with theoretical predictions, up to
423: relatively large system sizes $N =10^4$.
424: To conclude this section on the direct integration approach of
425: (\ref{KU}), we want to comment further the results we obtained in
426: the presence of disorder. Our findings seems
427: to indicate that the location of the (BKT) transition decreases for increasing
428: values of the disorder strength, as in the inset. On the other side, in the low
429: temperature region and for system sizes $L \ge 32$, we do not see the
430: helicity modulus to reach a constant finite value as for the small
431: system $L=16$ analyzed in Fig. \ref{helix_d}. In other terms, if algebraic order
432: is seen to occur at finite temperatures, for finite, small values of the
433: intrinsic frequency width distribution, numerics cannot state in a
434: clear manner whether such transition extends all the way down to zero
435: temperatures. The helicity modulus sample-to-sample fluctuations diverge
436: in the low temperature region, so unique conclusions on the form of the
437: phase diagram, in the space of temperature and disorder strength, cannot
438: be reached by mean of direct integration methods. We believe that the
439: difficulties we just discussed do not really depend on the specific
440: method we used, but simply, as often occurs in low dimensional systems,
441: the presence of disorder makes numerical simulations a very hard task.
442:
443: In this section we have seen how direct integration methods cannot state
444: in a clear manner how algebraic order is affected by the presence of
445: random quenched intrinsic frequencies. For this reason we will address
446: the same question within real space renormalization group theory in what
447: follows.
448:
449: \section{Real Space Renormalization Group Theory}
450:
451: The effect of quenched random interactions on the Kosterlitz-Thouless
452: type transition is an old and interesting problem that applies to a
453: variety of different physical problems. If one normally
454: expects that infinitesimal bond randomness do not affect the transition, it has not
455: been clarified in a conclusive way how other types of quenched random interactions, namely site
456: disorder and/or random interactions of the (DM) type affects algebraic
457: order. Real space renormalization group considerations suggests that at
458: finite temperature the (BKT) transition is not destroyed \cite{nineteen}.
459: However the effect of quenched random (DM) impurities on the low temperature
460: behavior of the system is usually expected to destroy (BKT) order, and
461: a reentrant phase diagram has been predicted long ago,
462: \cite{nineteen,twenty,twentyfourb}. On the other side, the validity of such
463: results have been questioned in the recent past
464: \cite{nineb,twentyeight}, and a Migdal-Kadanoff set of recursion
465: relations have been considered \cite{seven},
466: possibly suggesting that algebraic order is stable against disorder
467: at finite, small temperatures so that no reentrance would be present.
468: Before returning to this point, we want to stress why this question is
469: also relevant when studying the effect of a random intrinsic frequency
470: spread in the two dimensional array of dissipatively coupled
471: oscillators we consider in this work.
472: In the first part of this section we show that disorder in the intrinsic
473: frequency distribution of the two dimensional oscillator array problem is
474: related to the classical types of quenched random interactions
475: that have been introduced in the past in the context of the $XY$ model.
476: Once this point is clarified and once we will explain how the non-equilibrium
477: steady state properties of two dimensional arrays of dissipatively
478: coupled array of oscillators can be understood studying the
479: equilibrium properties of an effective $XY$ model in the
480: presence of quenched random interactions of the (DM) type, we will
481: critically reconsider the Migdal Kadanoff position space renormalization
482: group calculation of reference \cite{nineb} to
483: establish/check whether and how random (DM) interactions affects
484: algebraic order, proposing a new set of recursion relations to evaluate
485: the corresponding phase diagram. We hence suggest the
486: correct recursions that can answer the question of how quenched interactions
487: affects (BKT) order according to the (MK) approximation, a method that
488: has been shown to describe low dimensional systems with quenched
489: disorder in a simple and effective way.
490: We believe that the phase diagram of the $XY$ model, in the
491: presence of random (DM) interactions within the MK approximation is
492: consistent with the reentrant behavior originally
493: predicted within real space renormalization group theory \cite{nineteen}.
494: We suggest however that reentrance reduces
495: at very low temperatures, so that the phase boundary intercept the zero
496: temperature axis at finite values of the disorder strength
497: (with vertical slope).
498: In order to obtain the phase diagram, we write
499: generalized recursion relations for a discretized clock state model,
500: following the work of reference \cite{seven}, including bond
501: and (DM) randomness in a consistent way, at any finite $Q$ values
502: of the clock state parameter. Our recursion
503: relations reduce to the well known random bond Ising model for values of the
504: clock state parameter $Q=2$ and reduces to the (MK) recursion relations of
505: the pure $XY$ model at large values of the clock
506: state parameter $Q$, when no disorder is present.
507: Without loss of continuity we are able to interpolate between the
508: Random Bond Ising model (RBIM) \cite{twentythreeb},
509: the pure $XY$ model at large values of $Q$ and no disorder as well as
510: the general case of both exchange and (DM) random interactions being
511: present, ultimately proposing a novel framework to study the random gauge
512: glass model, where both exchange and (DM) interactions coexists
513: and are related by the proper initial condition in the renormalization
514: group flows.
515:
516: The relevance of these models and of the above questions to the
517: oscillator array problem can be explained as follows.
518: Let us consider the following effective hamiltonian
519:
520: \begin{equation}
521: - \beta { \cal H} = \sum_{ \langle i,j \rangle} J_{ij} \vec{S}_i
522: \vec{S}_j +
523: \sum_{ \langle i,j \rangle} D_{ij} \hat{z} \cdot \vec{S}_i \times \vec{S}_j
524: \label{DM}
525: \end{equation}
526:
527: In the classical formulation of the two dimensional $XY$ model one
528: expands around small values of subsequent phase shifts and then
529: imposes the periodicity of the phase variables, leading to the canonical
530: Coulomb gas formulation \cite{twentythreec}, so the Kosterlitz
531: Thouless renormalization group equations can be derived. Similarly, expanding the
532: second term related to the random (DM) interaction, an effective
533: Coulomb gas with random dipole moments can be obtained \cite{nineteen}.
534: Following these steps and averaging over the disorder either with the
535: replica method or alternatively \cite{nineteen,twentyone} recursion relations
536: of the Kosterlitz Thouless type that include the presence of disorder have
537: been obtained and a reentrant phase diagram in the space of temperature
538: and random strength has been predicted.
539:
540: The above hamiltonian, considering the case of no disorder being present in
541: the exchange interaction $J_{ij}=J_o$ in eq. (\ref{DM}), usually referred
542: as the two dimensional ferromagnet with Dzyaloshinskii-Moriya (XYDM)
543: interactions, is characterized by equation of motions that, to leading
544: order in the adjacent phase shift $\theta_i-\theta_j$
545: corresponds the the two dimensional Kuramoto model, as a simple
546: calculation of the equations of motion reveals (see e.g. eq. (9.7) in
547: reference \cite{twentyfourb}). We note that the intrinsic frequencies
548: $\omega_i$ are related to the quenched dipole distribution according to
549: $\omega_i =\sum_{j\in {\cal L}_j} q_{ij}$, where $q_{ij}$ are
550: the effective dipoles related to the original (DM) interaction.
551:
552: The system of oscillators with distributed, quenched frequencies has
553: been related to the case of identical oscillators with an effective
554: quenched dipole field, that induces a relative frequency mismatch.
555: According to the above hamiltonian (\ref{DM}), and assuming no
556: disorder present in the exchange interaction, the problem relates,
557: as one can check explicitly writing the Langevin equations
558: associated with (\ref{DM}), to the lattice oscillator problem
559: in two dimensions.
560:
561: It is then reasonable to ask how to write the proper recursion relations
562: within the the Migdal Kadanoff approximation, corresponding to the above
563: hamiltonian system (\ref{DM}).
564:
565: The equilibrium properties of model (\ref{DM}) relates to the steady
566: state solution of the two dimensional oscillator model we originally
567: intended to study. In the next section we present a study of the
568: two dimensional $XY$ model with quenched exchange and/or random (DM)
569: interactions within the MK approximation.
570: Not only our finding will be relevant to the
571: oscillator array problem but to many other physical problems
572: to which the two-dimensional $XY$ model in the presence
573: of disorder of the type dictated by equation (\ref{DM}) is known to relate.
574:
575: \section {The Migdal Kadanoff method}
576:
577: In the classical formulation of the pure $XY$ model with the Migdal Kadanoff
578: renormalization group scheme \cite{sixb}, renormalization
579: group recursion relations for the effective coupling of combined bonds, within
580: a Fourier mode formulation are considered.
581: At low temperatures the approximation recovers effective algebraic order,
582: the potential flows to a Villain form \cite{sixc} and the system is
583: characterized by an infinite correlation length, even though has vanishing
584: magnetization. The only drawback of the (MK) approximation is that after
585: a sufficiently large number of renormalization group steps, the recursion
586: relations flows to a high temperature disordered sink at any finite
587: temperature so that true fixed line behavior is not present in the
588: strict sense. The number of iterations required for this to occur is
589: however diverging below the transition point so that effective algebraic
590: order is successfully described by the approximation.
591: It is then reasonable to ask
592: how to reformulate the classical (MK) approach to the $XY$ model in the
593: presence of quenched randomness, due to the overall success of the (MK)
594: approximation in the context of disordered Ising type of models.
595:
596: \subsection{The discrete Clock State}
597:
598: Meanwhile a second, interesting formulation of the (MK) approximation for
599: the two dimensional pure $XY$ model has been suggested in the past \cite{seven}.
600: Rather than the usual Fourier version of the (MK) recursion relations,
601: one considers a discrete clock state model, at integer finite values of the
602: clock state parameter $Q$. The above formulation has been given for the
603: pure $XY$ model in two dimensions. Attempts in order to include the
604: presence of (DM) interactions have been made \cite{twentyfive,twentyseven,twentyeight}
605: but we are not aware of any study where correct recursion
606: relations have been written for this case.
607: Another delicate point is how one treats the recursion
608: relations in the case of disorder (both in the exchange and (DM)
609: interactions). We treat disorder in the same way it has been considered
610: in the past for the (RBIM), namely within a binning procedure, avoiding
611: to resort to random pool methods \cite{twentyseven,twentyeight}, our method
612: being simply a deterministic one (despite we are dealing with
613: disorder). Our recursion relations and algorithmic method reduce to the
614: calculation we did in the past for the (RBIM) for values of the clock state
615: parameter $Q=2$.
616: The computational effort grows linearly with $Q$.
617: As we will see for the simple case of the pure model, effective
618: algebraic order is observed already for values of
619: the clock state parameter as small as $Q=32$.
620: In order to treat random exchange and (DM) interactions of both sign
621: on equal footing we need to consider a renormalization group rescaling
622: length $b=3$. This last choice makes the algorithmic solution of the recursion
623: relations in the presence of disorder quite slow, since triple
624: convolutions have to be considered. A similar issue has been discussed
625: already for the simple case of the $Q=2$ (RBIM) \cite{twentythreeb}.
626:
627: \subsection{The Recursion Relations}
628:
629: As mentioned above we consider the case of discrete phase angles
630: $\theta_i=2 \pi q_i/Q, q_i=0,\cdots,Q-1$.
631:
632: The hamiltonian reads
633: \begin{equation}
634: - \beta { \cal H} = \sum_{ \langle i,j \rangle} J_{ij} \cos(\theta_i-\theta_j)
635: + \sum_{ \langle i,j \rangle} D_{ij} \sin(\theta_i-\theta_j).
636: \end{equation}
637: Under renormalization group transformation, we consider two generalized
638: couplings between neighbouring sites $J_{ij}$ and $D_{ij}$ independently.
639: The renormalization group recursion relations read
640:
641: \begin{eqnarray}
642: J'(q)&=& \frac{1}{2} \log(R(Q,q) R^{ \dagger}(Q,q))-G'(Q) \nonumber \\
643: D'(q)&=& \log(R(Q,q)/R^{ \dagger}(Q,q)) \nonumber \\
644: G'(Q)&=& \frac{1}{2Q} \sum_{q=0}^{Q-1}\log(R(Q,q)R^{ \dagger}(Q,q))
645: \label{RR}
646: \end{eqnarray}
647: where interaction terms are always considered modulo $Q$, and
648: where the captive renormalization group constant imposes the constrain
649: $\sum_q J'(Q,q)=0$ and a similar constrain for $D'(Q,q)$ equally applies.
650: The renormalization group polynomials, for the case of a
651: rescaling length factor $b=2$ read:
652:
653: \begin{eqnarray}
654: R(Q,q) = \sum_{l=0}^{Q-1} e^{\tilde{J}_{12}(l)+\tilde{J}_{23}(q+l)+\tilde{D}_{12}(l)-\tilde{D}_{23}(q+l)} \nonumber
655: \\
656: R^{ \dagger}(Q,q) = \sum_{l=0}^{Q-1} e^{\tilde{J}_{12}(l)+\tilde{J}_{23}(q+l)-\tilde{D}_{12}(l)+\tilde{D}_{23}(q+l)}
657: \label{RGpols}
658: \end{eqnarray}
659: where the ``bond-moved'' exchange interactions are
660: \begin{equation}
661: \tilde{J}_{ij}=\sum_{n=1}^bJ_{i_n j_n},
662: \end{equation}
663: together with a similar expression for the (DM) interactions $\tilde{D}_{ij}$.
664: These recursion relations might seem at first sight rather similar to the ones
665: discussed in the past \cite{seven}, however a important symmetry property is
666: included in the above recursion relations, that was not discussed in
667: previous formulations. A set of recursion relations of the
668: (MK) type, where the two interactions were treated separately as above,
669: was discussed, but only in the so called harmonic approximation
670: \cite{twentyfive,twentyfiveb,twentysix}.
671:
672: The above recursion relations will be firstly discussed for a
673: renormalization group rescaling length $b=3$, in the absence of disorder, so
674: we can check that the onset of effective algebraic order described
675: by the (MK) method is properly recovered within the discrete clock state
676: formulation and we can determine the effect of an uniform (DM)
677: interaction on the pure $XY$ model. We will then consider the
678: case of small clock state $Q$ values, and include the presence of
679: disorder, checking that the above recursions reproduce, as expected,
680: the phase diagram for the (RBIM) at $Q=2$ precisely, before returning to the
681: ultimate issue of large values of $Q$ and disorder being present,
682: corresponding to the $XYSG$, $XYDM$ and $RGG$ models discussed above.
683:
684: Attempts in this direction were already considered in \cite{seven,twentyseven}
685: except we are now able to incorporate the presence of (DM)
686: interactions in a consistent way.
687: The recursion relations (\ref{RR}), despite their
688: simplicity, are the only possible ones that properly reflects
689: the symmetries of the original hamiltonian, as a direct analysis
690: reveals. Note that, as also explicitly stated
691: in reference \cite{seven}, the symmetry properties of the generalized
692: potential there proposed were lost under renormalization group transformation
693: in the case of (DM) interactions being present, something that should
694: simply not occur. The fact that the recursion relations
695: written in the past were not complete is probably the reason behind the
696: erratic behaviour of the renormalization group flows, also reported in
697: \cite{twentyeight}, whenever interactions of the (DM) type were
698: considered. Any conclusion on the shape of phase diagram of the $XYDM$
699: problem based on these recursion relations \cite{nineb} should then
700: be taken with care.
701:
702: Differently, in our method the potential maintains its symmetry
703: properties under renormalization group transformation
704: (see e.g. Fig. \ref{XYDMpotential}), as it should
705: be under any (symmetry) group transformation.
706: Once the novelty of the above recursions is understood,
707: we need to discuss how to implement them algorithmically, in the
708: presence of quenched random interactions.
709: The way to treat disorder from the algorithmic point of view is a delicate
710: issue we should also discuss. As explained above we use a
711: $deterministic$ method based on a binning procedure, that is known to
712: respect the symmetries of the renormalization group
713: transformation, as already discussed in the context of the (RBIM), that
714: corresponds to $Q=2$ in the above recursions.
715:
716: \section{Identical oscillators}
717:
718: We first show the results we obtained for the pure model (and $b=3$),
719: at large finite values of $Q$, checking that the results of the
720: (MK) approximation in the $XY$ model are properly reproduced within the
721: discretization scheme here considered and discussing the role of (DM)
722: uniform interaction terms, before returning on the issue of disorder,
723: we will be able to include via the proper choice of the initial
724: conditions of the renormalization group flows.
725:
726: \begin{figure}
727: \includegraphics*[scale=0.7,angle=0]{fig4.eps}
728: \caption{The fixed Villain potential observed for the pure $XY$ model, at
729: temperature value $T=0.5$ for a renormalization group rescaling
730: length $b=3$, in the absence of (DM) interactions.
731: Crosses shows the fixed potential in the discrete scheme
732: with $Q=32$, while the continuous line shows the results obtained for
733: the value of the clock state parameter $Q=512$.}
734: \label{XYpotential}
735: \end{figure}
736:
737: We do not report explicitly the form of the renormalization group
738: polynomials for the case of $b=3$, but they clearly include a double sum,
739: rather than the single sum in eq.(\ref{RGpols}), as
740: usually discussed in the context of the (RBIM) \cite{twentythreeb}, as one
741: needs to decimate an even number of sites at each RG step in order to
742: treat interactions of both signs in the same manner, so to produce
743: a phase diagram that will have the proper symmetry properties.
744: Going back to the pure case, when no randomness is present, we expect
745: that the location of the Kosterlitz Thouless transition, within the (MK)
746: approximation, depends on the decimation parameter $b$, as can be seen
747: in Fig. \ref{iter}, showing the number of iterations needed before
748: the above recursions (\ref{RR}) flow to the high temperature disordered sink.
749: Note that in the case of the pure $XY$ model without (uniform) (DM) terms,
750: one recovers the usual Villain potential \cite{sixc} for any value
751: of the clock state parameter $Q \ge 16$. We show, for the pure
752: case, both the fixed Villain potential we obtain for $Q=32$ (blue crosses) as
753: well as $Q=512$ (continuous red line), in Fig. \ref{XYpotential}.
754:
755: \begin{figure}
756: \includegraphics*[scale=0.7,angle=0]{fig5.eps}
757: \caption{The two components of the complex fixed potential, observed
758: for the pure $XY$ model, at temperature value $T=0.5$ for a
759: renormalization group rescaling length $b=3$, in the presence of uniform
760: (DM) interactions, obtained for the clock state parameter value $Q=512$.
761: Clearly, under renormalization group transformation, the exchange
762: interaction is symmetric around $ \pi$, while the (DM) generalized
763: interaction is antisymmetric. Fix-line behavior as in the case
764: where (DM) interactions are absent is observed.}
765: \label{XYDMpotential}
766: \end{figure}
767:
768: These results shows that the clock state
769: discrete approximation is a very robust one, in that effective algebraic
770: order is found at relatively small values of the clock state parameter,
771: and there is no need to increase the values of $Q$ of the order of
772: $10^3$, as done in reference \cite{seven}.
773: This last remark will be crucial when dealing with
774: the disordered case, even though the computational effort of the algorithm we
775: will introduce in the next section ultimately scales in a linear way with
776: $Q$, while it scales as a cubic power of the number of bins we will use
777: to discretize the quenched probability distribution(s) of the exchange
778: and (DM) interactions at each renormalization group step.
779:
780: \begin{figure}
781: \includegraphics*[scale=0.6,angle=270]{fig6.eps}
782: \caption{The Number of Iterations for the pure $XY$ model
783: with $b=2$ and $b=3$. The results shows that effective algebraic order
784: is observed in that the increase in the iteration step number is rather
785: sharp around the effective critical transition point.The results seems
786: to suggest that the location of the transition point depend, as occurs
787: in Ising type of models within the (MK) approximation, on the
788: renormalization group rescaling length.}
789: \label{iter}
790: \end{figure}
791:
792: In the presence of (DM) interactions the complex potential
793: converges to a symmetric and an antisymmetric part ,as in the
794: original interaction eq. (\ref{XYDMpotential}).
795: Note that the symmetry (antisymmetry) of the two terms in the initial
796: hamiltonian are preserved by the above recursion relations (\ref{RR}),
797: and that is because we wrote $two$ distinct recursion relations for
798: each term, since the renormalization group polynomials have to be
799: regarded, in the presence of (DM) interactions,
800: effectively as complex quantities. As in the case of the pure $XY$
801: model, effective algebraic order is observed below the
802: (BKT) transition point, that do not changes in the presence
803: of uniform (DM) interactions, as expected.
804:
805: Before we conclude this section on the (MK) clock state approach to the
806: pure $XY$ model with uniform (DM) interactions, we want to add
807: that, as stressed above, the location of the (BKT) transition do not
808: changes importantly for values of the clock state parameter $Q>32$,
809: so we can safely consider the disordered case considering finite
810: values of $Q$ in this range of values.
811:
812: For the case of the clock state parameter
813: $Q=4$ we observe a ferromagnetic transition point at $T_c \simeq 2.078$
814: that decreases for increasing values of the relative interaction
815: strength $D/J$, as can be seen in Fig. \ref{DJ}.
816:
817: \begin{figure}
818: \includegraphics*[scale=0.7,angle=0]{fig7.eps}
819: \caption{The location of the critical transition point for the clock
820: state parameter value $Q=4$, when no disorder is present, and where
821: uniform (DM) interactions are considered. The $x$-axis shows the
822: relative strength of the two interactions, as chosen in the initial
823: conditions or the recursions (\ref{RR}).}
824: \label{DJ}
825: \end{figure}
826:
827: \section{The Role of Disorder}
828:
829: In order to test our recursion relations and algorithmic methods,
830: we begin reconsidering the simple (RBIM) case, that is included in the
831: above equations at values of the clock state parameter $Q=2,4$, for
832: vanishing values of the (DM) interaction strength $D$.
833: As expected we recover the reentrant phase diagram that has been
834: discussed at length in the past. Clearly, at $Q=2,4$, gauge invariance
835: is respected, so we plot the Nishimori line \cite{twentythree} and note
836: that, as already found in the past, the phase boundary reaches the highest
837: value of the antiferromagnetic bond concentration on this line.
838: Note that, at $D=0$ only, the $Q=2$ and $Q=4$ models coincide up to a
839: redefinition of temperature so we discuss the latter case only.
840: Interestingly, high precision measures at very low temperatures
841: shows that reentrance, that is clearly present as reported
842: already \cite{twentythreeb}, diminishes gradually at very low temperatures,
843: so that our results are consistent with the expectation that the phase
844: boundary is vertical, but at $low$ temperatures only.
845: Simply this do not occur all the way up to the Nishimori line, according
846: to our findings. This interesting reentrant behaviour that gradually
847: reduces to intercept the zero temperature axis with a vertical slope has also
848: been observed recently in the context of (gauge invariant) Gallager
849: codes, and we conjecture that the qualitative behaviour of the phase boundary
850: we are here discussing applies to all spin systems for higher values of
851: $Q$, whenever gauge invariance is respected, except the ferromagnetic
852: phase is replaced by a (BKT) phase for values of the clock state
853: parameter $Q \ge 16$.
854: It is important to mention at this point that any initial condition in
855: the (RG) flows for $Q=2,4$, close to the boundary and above the
856: Nishimori line, will flow to the finite temperature unstable fixed point,
857: while any initial condition below the Nishimori line will flow to the
858: strong coupling low temperature fixed point already discussed in
859: the past \cite{twentythreeb}, a phenomenon known as strong violation
860: of universality.
861:
862: The phase diagram above (Fig. \ref{PD1}) clearly do not relate to
863: the original $XY$ model we planning to consider, but we performed
864: this calculation to test the algorithm that implements the recursion
865: relations in the presence of disorder and we will be
866: able to consider large values of $Q$ of the (DM) type at a later stage,
867: changing without loss of continuity
868: the parameters in the initial condition of the RG flows. Moreover,
869: it is quite interesting to study the clock state model
870: at small finite values of the parameter $Q$ with disorder, since it is
871: a well defined problem that interpolates between the Ising and $XY$
872: type of models. As found in the pure case, when disorder is present
873: in the exchange interaction, we observe ferromagnetic order at low
874: temperatures for the values $Q=2,4,8$, while effective algebraic order
875: appears at values of the clock state parameter $Q \ge 16 $.
876: This means that the recursion relations (\ref{RR}) flows to a
877: ferromagnetic fixed point for values of $Q \le 8$, while at $ Q \ge 16$
878: we observe the usual quasi-fixed line behaviour discussed in the context
879: of the $XY$ model ( see Fig. \ref{fig10}), and we do not expect things
880: to change importantly for higher values
881: values of $Q$, as we have seen already for the pure case, meaning that
882: the discretization scheme converges rapidly to the original problem
883: of continuous degrees of freedom.
884: For the case of $Q=8$, and when DM interactions are not present
885: we computed, for a bimodal exchange interaction, the phase diagram
886: in Fig.\ref{PD2}.
887: In this case gauge invariance is not present
888: \cite{twentythree} and reentrance do not occur, as expected.
889: On the other side we still
890: observe ferromagnetic and paramagnetic order, divided by the phase
891: boundary (red thick line in Fig. \ref{PD1}), where a multicritical
892: point occur. As in the $Q=2,4$ case,
893: any initial condition close to the phase boundary and above the
894: multicritical point flows to an unstable finite temperature fix
895: point, while any initial condition close to the boundary and
896: below the multicritical point flows to a strong coupling zero
897: temperature distribution, even though gauge symmetry is not present.
898: This implies that at small, finite values of the clock state parameter
899: the strong violation of universality, discussed in the past for the
900: (RBIM) \cite{twentythreeb}, also occurs in the clock state model,
901: for values of the clock state parameter $Q \le 8$.
902:
903: \begin{figure}
904: \includegraphics*[scale=0.6,angle=0]{fig8.eps}
905: \caption{
906: The Phase Digram for the clock state parameter value $Q=4$, without
907: (DM) interactions and in the presence of bimodal exchange interactions,
908: with an antiferromagnetic bond concentration p. The inset shows the
909: reentrant behavior of the phase boundary around the multicritical point.
910: Gauge invariance is present in the model and so is reentrance.}
911: \label{PD1}
912: \end{figure}
913:
914: \begin{figure}
915: \includegraphics*[scale=0.8,angle=0]{fig9.eps}
916: \caption{The Phase Digram for the clock state parameter value $Q=8$, without
917: (DM) interactions and in the presence of bimodal exchange interactions,
918: with an antiferromagnetic bond concentration $p$.
919: Gauge invariance is not present in this problem, since (DM) interactions
920: are set to zero. Reentrance is not expected in this case and the
921: Nishimori symmetry line is not present. We observe however the presence
922: of a multicritical point. Any initial condition above the multicritical
923: point and close to the phase boundary, flows to a finite temperature
924: unstable fix point, as seen in the inset above, while any initial
925: condition below the multicritical point (indicated my the cross) and
926: close to the boundary flows to a strong coupling, low temperature fix
927: distribution, as in the second inset. The insets plot, as a
928: function of the RG iteration number the Average Potential, defined as
929: $ \sqrt{ \sum_q |J(Q,q)|^2} $. }
930:
931: \label{PD2}
932: \end{figure}
933:
934: In the case of larger values of the clock state parameter, e.g. $Q=16$,
935: and still at $D=0$, we do not observe the recursion relations (\ref{RR})
936: to flow to a ferromagnetic fix point anymore, but rather quasi-fixed line
937: behavior of the (BKT) type appears at finite values of the
938: antiferromagnetic bond concentration, as seen in Fig. \ref{fig10}.
939: \begin{figure}
940: \includegraphics*[scale=0.8,angle=0]{fig10.eps}
941: \caption{ Number of iterations required to flow to the paramagnetic
942: sink, as a function of temperature for the
943: clock state parameter value $Q=16$ for different values of the
944: antiferromagnetic bond concentration $p=0.0, p=0.01, \cdots,0.07$.
945: The figure indicates that effective algebraic order is observed
946: for finite, small values of the disorder strength in the exchange
947: interaction. From the location of these lines it is possible to
948: reconstruct the phase diagram of the $XYSG$ problem at $Q=16$.}
949:
950: \label{fig10}
951: \end{figure}
952:
953: We are considering in fact the discrete version of the $XY$ spin glass problem.
954: No gauge invariance nor reentrance is expected in this case, and
955: we observe effective algebraic order at low values of the bimodal
956: exchange interaction, while a paramagnetic phase occurs at higher
957: values of disorder and temperature. Clearly no spin-glass order is
958: observed in two dimension everywhere in the phase diagram.
959:
960: Global phase diagrams for the
961: $XYSG$ and $XYDM$ problems are being evaluated for values of the clock state
962: parameter $Q=16,32,64$ and will be presented elsewhere.
963:
964: We did not consider yet the final case where both randomness
965: in the exchange and (DM) interactions are present, even though we are
966: aware that this was the main question in order to discuss oscillator
967: arrays, and leave this question for future work.
968:
969: We believe we have set up a consistent machinery
970: to evaluate the phase diagram in the general case of large values of
971: $Q$ and disorder of both types and results in this direction will
972: be able to answer on the issue of reentrance for continuous
973: spin models, at least within the MK approximation.
974:
975: When both disorder in the exchange and (DM) interactions is
976: present, the renormalization group recursions involve a two dimensional
977: probability distribution, so that the binning technique required is slightly
978: more complicated than the one considered in the absence of (DM) interactions.
979: A similar type of situation has been considered
980: for the Random Field Random Bond (RFRB) model, where three dimensional
981: probability distributions were considered, and it is just
982: a technical issue (but rather tedious) to extend the techniques
983: developed in that case to the case of random exchange and random (DM)
984: interactions, at finite, large values of the clock state parameter $Q$.
985:
986: \section{CONCLUSIONS AND FUTURE PERSPECTIVES}
987:
988: We conclude with a few considerations and predictions on
989: what we expect to see in the general case of large values of the
990: clock state parameter $Q$ and disorder of the (DM) and exchange type,
991: according to the preliminary results we obtained so far.
992:
993: \begin{figure}
994: \includegraphics*[scale=0.6,angle=0]{fig11.eps}
995: \caption{ Qualitative behaviour of the phase diagram in the case
996: of the two dimensional ferromagnet with (DM) random interactions
997: of width $ \Delta$. (A) Real space renormalization \cite{nineteen},
998: suggesting reentrance that extends all the way down to vanishing
999: disorder strength. (B) Absence of reentrance and vertical boundary
1000: at finite values of the disorder strength as suggested, e.g. in
1001: reference \cite{twentythree}. (C) Reentrant behavior of the
1002: phase boundary and a (BKT) phase that extend at finite values
1003: of the disorder strength for low temperature values. }
1004: \label{PD4}
1005: \end{figure}
1006:
1007: If both randomness in the exchange and (DM) interaction are considered
1008: in a way such that gauge invariance is recovered, we expect to observe a (BKT)
1009: phase and reentrant behavior of the phase boundary of the type discussed
1010: above. Note that the exchange randomness is usually considered as
1011: irrelevant, when (DM) random interactions are present, so similar
1012: conclusions should apply to the (XYDM) problem.
1013:
1014: We expect the phase boundary to reenter below the Nishimori line,
1015: and finally to intercept the zero temperature axis with vertical slope,
1016: at finite, $non-vanishing$ values of the disorder strength.
1017: This would imply that (AO) is stable against the intrinsic frequency
1018: distribution and it is reasonable to ask whether one can observe
1019: order of the (BKT) in arrays of two dimensional, non-identical
1020: oscillators, where technically one should consider temperature to be
1021: small. Most likely, the reason why previous
1022: calculations within the (MK) approximation have not been able to show the
1023: reentrance predicted is simply because the recursion
1024: relations considered in \cite{seven,twentyeight} were not taking properly
1025: into account the (DM) interactions, as stressed above.
1026: This is likely the reason why the (MK) phase diagram
1027: discussed in \cite{nineb} is not reentrant, as we also observe for the
1028: case of $Q=8$ in Fig. (\ref{PD4}) where the (DM) interaction is
1029: explicitly set to zero and gauge invariance is not present.
1030: A careful analysis of the recursion relations (\ref{RR}) in the presence
1031: of (DM) interactions, and the corresponding phase diagram, both
1032: at small and large values of the clock state parameter $Q$ is in
1033: progress and will be presented shortly.
1034:
1035: The results we expect can be summarised as follows. The phase diagram
1036: is reentrant as predicted long ago, but reentrance (see curve B in
1037: Fig. \ref{PD4}) reduces and the phase boundary between the (BKT) phase
1038: and the paramagnetic phase intercept, with vertical slope, the zero
1039: temperature axis at finite values of the disorder strength. This
1040: intermediate scenario would be consistent both with the reentrant behavior
1041: predicted by classical renormalization group arguments \cite{nineteen}
1042: and with the results indicating that the phase boundary intercepts
1043: the zero temperature axis with vertical slope \cite{twentythree}.
1044:
1045: \begin{acknowledgments}
1046: The author wish to thank Prof. David Lowe and Prof. David Saad for useful
1047: discussions and comments. The author benefited also from brief
1048: but interesting discussions with colleagues during the conference
1049: in honour of Professor Sherrington, on the occasion of his 65th
1050: birthday.
1051:
1052: The author acknowledge support from EPSRC.
1053: \end{acknowledgments}
1054:
1055: \begin{references}
1056: \bibitem{zero} Y. Kuramoto, Lecture Notes in Physics,
1057: Vol. 39. International Symposium on Mathematical Problems in
1058: Theoretical Physics, ed. H. Araki (Springer, New York), 420 (1975).
1059: \bibitem{zerob}P. C Matthews, R. Mirollo and S. H. Strogatz, Physica D {
1060: \bf 52}, 293 (1991).
1061: \bibitem{zeroc}L.L. Bonilla, C.J. P\'erez Vicente and R. Spigler,
1062: Physica D { \bf 113}, 79 (1998).
1063: \bibitem{one} S. H. Strogatz and R. Mirollo, J. Phys. A { \bf 21} L699, (1988).
1064: \bibitem{two} H. Sakaguchi and Y. Kuramoto, Prog. Theor. Phys. { \bf 76},
1065: 576 (1986).
1066: \bibitem{twob}H. Sakaguchi, S. Shinomoto and Y. Kuramoto,
1067: Prog. Theor. Phys. { \bf 77}, 1005 (1987).
1068: \bibitem{three} H. Daido, Phys. Rev. Lett., { \bf 61}, 231 (1988).
1069: \bibitem{threeb}M. Antoni and S. Ruffo, Phys. Rev. E { \bf 52}, 2361 (1995).
1070: \bibitem{threec}N.D. Mermin H. Wagner, Phys. Rev. Lett. {\bf 22}, 1133
1071: (1966).
1072: \bibitem{threed} H.E. Stanley and T.A. Kaplan, Phys. Rev. Lett. { \bf 17}, 913 (1966).
1073: \bibitem{threee} V.L. Berezinskii Sov. Phys-JETP { \bf 32}, 493 (1970);
1074: J.M. Kosterlitz and D.J. Thouless, J. Phys. C { \bf 6}, 1181 (1973);
1075: J.M. Kosterlitz, J. Phys. C { \bf 7}, 1046 (1974).
1076: \bibitem{four}T. Risler, J. Prost and F. Julicher, Phys. Rev. Lett. {
1077: \bf 93}, 175702-1 (2004).
1078: \bibitem{five} P. Coullet, J. Lega, B.Houchmanzadeh and D. Lajzerowicz,
1079: Phys. Rev. Lett. { \bf 65}, 1352 (1990).
1080: \bibitem{fiveb}P.C. Hohenberg and B.I. Halperin, Rev. Mod. Phys. { \bf
1081: 49}, 435 (1977).
1082: \bibitem{six} M.Y. Choi and S. Kim, Phys. Rev. B {\bf 44}, 10411 (1991).
1083: \bibitem{sixb} J.V. Jos\'e, L.P. Kadanoff, S. Kirkpatrick and D. Nelson,
1084: Phys. Rev. B { \bf 16}, 1217 (1977).
1085: \bibitem{sixc}J. Villain, J. Phys. (Paris) { \bf 36}, 581 (1975).
1086: \bibitem{seven}M. Cieplak, J.R. Banavar, M.S. Li and A. Khurana,
1087: Phys. Rev. B { \bf 45}, 786 (1992).
1088: \bibitem{eight}D. Andelman and A.N. Berker, Phys. Rev. B { \bf 29}, 2630
1089: (1984); S.R. McKay and A.N. Berker, J. Appl. Phys. { \bf 64}, 5785 (1988).
1090: \bibitem{nine}A.N. Berker and L.P. Kadanoff, J.Phys. A { \bf 13}. L259 (1980).
1091: \bibitem{nineteen}M. Rubinstein, B. Shraiman and D. R. Nelson,
1092: Phys. Rev. B { \bf 27}, 1800 (1983).
1093: \bibitem{twentytwo} E. Granato and J.M. Kosterlitz, Phys. Rev. Lett. {
1094: \bf 62}, 823 (1989).
1095: \bibitem{nineb}T. Nattermann, S. Sheidl, S.E. Korshunov and M.S. Li,
1096: J. Phys. I France {\bf 5}, 565 (1995); S.E. Korshunov and
1097: T. Nattermann, Phys. Rev. B { \bf 53}, 2746 (1996).
1098: \bibitem{twentyfour}N. Akino and J.M. Kosterlitz, Phys. Rev. B { \bf
1099: 66}, 054536 (2002).
1100: \bibitem{ten} A. Pikovsky, M. Rosenblum and J. Kurths, $Synchronization,
1101: A~universal~ concept~ in~ non-linear ~sciences$, Cambridge University
1102: Press, Cambridge UK (2001).
1103: \bibitem{eleven}I.S. Aranson and L. Kramer, Rev. Mod. Phys. { \bf 74},
1104: 99 (2002).
1105: \bibitem{fourteen} M.C. Cross, A. Zumdieck, R. Lifshitz and J.L. Rogers,
1106: Phys. Rev. Lett. { \bf 93}, 224101-1 (2004).
1107: \bibitem{twelve}V. Amnegaokar and S. Teitel, Phys. Rev. B { \bf 19},
1108: 1667 (1979).
1109: \bibitem{thirteen}B. Yurke, A.N. Pargellis, T. Kovacs and D.A. Huse,
1110: Phys. Rev. E { \bf 47}, 1525 (1993).
1111: Cross, Phys. Rev. E { \bf 64}, 056140-1 (2001).
1112: \bibitem{fifteen}B.J. Kim, P. Minnhagen and P. Olsson, Phys. Rev. B {
1113: \bf 59}, 11506 (1999).
1114: \bibitem{sixteen}R. Gupta, J. DeLapp, G.G. Batrouni, G.C. Fox,
1115: C.F. Baillie and J. Apostolakis, Phys. Rev. Lett. { \bf 61}, 1996 (1988).
1116: \bibitem{seventeen}T. Ohta and D. Jasnow, Phys. Rev. B { \bf 20}, 139 (1979).
1117: \bibitem{eighteen}M. Hasenbush, J. Phys. A { \bf 38}, 5869 (2002).
1118: \bibitem{twenty}A. Houghton, R.D. Kenway and S.C. Ying, Phys. Rev. B {
1119: \bf 23}, 298 (1981); Y.Y. Goldschmidt and A. Houghton, Nucl. Phys. B { \bf
1120: 210}, 155 (1982).
1121: \bibitem{twentyone}J. L. Cardy and S. Ostlund, Phys. Rev. B { \bf 25},
1122: 6899 (1982).
1123: \bibitem{twentythree} Y. Ozeki and H. Nishimori, J. Phys. A { \bf 26},
1124: 3399 (1993).
1125: \bibitem{twentyfourb}Y. Y. Goldschmidt and B. Schaub, Nucl. Phys. B{
1126: \bf 251}, 77 (1985).
1127: \bibitem{twentyeight} M.J.P. Gingras and E.S. Sorensen, Phys. Rev. B {
1128: \bf 57}, 10264 (1998).
1129: \bibitem{twentythreeb}G. Migliorini and A.N. Berker, Phys. Rev. B { \bf
1130: 57}, 426 (1998).
1131: \bibitem{twentythreec}P. Minnaghen, Rev. Mod. Phys. { \bf 59}, 1001 (1987).
1132: \bibitem{twentyfive} M.J.P. Gingras, Phys. Rev. B { \bf 43}, 13747
1133: (1991).
1134: \bibitem{twentyfiveb} B.W. Morris, S.G. Colborne, M.A. Moore, A.J. Bray and
1135: J.Canisius, J.Phys. C {\bf 19}, 1157 (1986).
1136: \bibitem{twentysix} J.R. Banavar and A.J. Bray, Phys. Rev. B { \bf 38},
1137: 2564 (1988).
1138: \bibitem{twentyseven} F. Krzakala, Europhys.Lett. {\bf 66}, 847 (2004).
1139:
1140: \end{references}
1141:
1142: \end{document}
1143: