1: \documentstyle[preprint,aps,epsf]{revtex}
2: \begin{document}
3: \tightenlines
4: \draft
5: \title{\bf Self-Organisation to Criticality in a
6: System without Conservation Law}
7:
8: \author{Stefano Lise}
9:
10: \address{Department of Mathematics, Huxley Building, Imperial College
11: of Science, Technology, and Medicine, London UK SW7 2BZ \\}
12: \date{\today}
13:
14: \maketitle %\parskip 2ex
15:
16: \begin{abstract}
17: We numerically investigate the approach to the stationary state in the
18: nonconservative Olami-Feder-Christensen (OFC) model for earthquakes.
19: Starting from initially random configurations, we monitor the average
20: earthquake size in different portions of the system as a function of time
21: (the time is defined as the input energy {\em per site} in the system).
22: We find that the process of
23: self-organisation develops from the boundaries of the system and it is
24: controlled by a dynamical critical exponent $z\simeq 1.3$ that appears
25: to be universal over a range of dissipation levels of the local dynamics.
26: We show moreover that the transient time of the system $t_{tr}$ scales with
27: system size $L$ as $t_{tr} \sim L^z$. We argue that the (non-trivial) scaling
28: of the transient time in the OFC model is associated to the establishment of
29: long-range spatial correlations in the steady state.
30: \end{abstract}
31:
32: \vspace{0.3cm}
33: {PACS numbers: 05.65.+b, 45.70.Ht, 89.75.Fb}
34:
35: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
36:
37: \section{Introduction}
38:
39: The idea of self-organised criticality (SOC) was introduced by Bak, Tang and
40: Wiesenfeld (BTW)~\cite{btw} as a possible paradigm for the widespread
41: occurrence in nature of scale free phenomena. It refers to the intrinsic
42: tendency of extended, non-equilibrium systems to spontaneously self-organise
43: into a dynamical critical state. In general, SOC systems are driven externally
44: at a very slow rate and relax with bursts of activity (avalanches) on a very
45: fast (almost instantaneous) time scale.
46: The standard signature of SOC is a power law distribution of avalanche sizes
47: and in this sense is the system said to be critical. Typical physical
48: realizations of this phenomena includes, among others, earthquakes,
49: forest fires and biological evolution (for a review, see e.g.,
50: Ref.~\cite{bak_book,jen_book}).
51:
52: A number of simple lattice models have been developed to test the
53: applicability of SOC to a variety of complex interacting dynamical
54: systems~\cite{bak_book,jen_book}.
55: In general these models reach a stationary critical state after a
56: sufficiently long transient time. Not much attention though has been paid
57: to the self-organisation process and studies have mainly concentrated on the
58: properties of different models at stationarity.
59: This is partly justified by the fact that some of the most studied models,
60: such as the BTW~\cite{btw} and the Zhang~\cite{zhang} models, display a
61: very simple transient time behaviour. Indeed in these models the relaxation
62: time does not scale with system size (time is defined as the input energy per
63: site in the system)~\cite{achille}. On the other hand, there are also models
64: with a more complex behaviour (see, e.g., Ref.~\cite{sibani}).
65:
66: In this paper we investigate the approach to the stationary state in
67: the so-called Olami-Feder-Christensen (OFC) model for earthquake
68: dynamics~\cite{ofc}.
69: This model has in recent years attracted a considerable deal of attentions
70: especially because it has been proposed as an example of a system displaying
71: SOC behaviour even with a non-conservative
72: dynamics~\cite{grassberger,socolar,middleton,ceva}.
73: One of the most important question in this field is indeed whether a conserving
74: local dynamics is a necessary condition for SOC~\cite{hwa,grinstein}. For
75: example, it is well known that the BTW model is subcritical if dissipation is
76: introduced~\cite{manna}. The presence of criticality in the non-conservative
77: OFC model is still debated~\cite{carvalho,kim}. Recent numerical
78: investigations, though, have shown that the model on a square lattice displays
79: scaling behaviour, up to lattice sizes presently accessible by computer
80: simulations~\cite{lisepac1,lisepac2}.
81:
82: The present investigation complements previous analysis of the OFC model which
83: were based on the study of the probability distribution for earthquake sizes.
84: It provides further numerical support in favour of criticality in the
85: non-conservative regime. Indeed, we will show that the model displays a non
86: trivial transient time behaviour: the relaxation time scales with system
87: size and it is controlled by a dynamical critical exponent $z$ that appears
88: to be universal over a range of dissipation levels of the local dynamics.
89: Moreover we will establish the presence of long range spatial correlations in
90: the system. In so doing, we will be able to gain some insight into the
91: mechanisms behind criticality in non-conserving systems, mechanisms that are
92: very different from those at work in systems with a conservation law.
93:
94:
95: The plan of the paper is as follows. In section II we describe the model
96: and briefly summarise previous foundings relevant to our investigation.
97: In section III we define the quantities of interest and present the results
98: of our numerical study. Finally, in section IV we discuss our main
99: conclusions.
100:
101:
102: \section{The model}
103:
104: The OFC model is a coupled map lattice model, where to each site $(i,j)$
105: of a square lattice of linear size $L$ is associated a real variable
106: $F_{ij}$. In the initial state, at time $t=0$, the values of the $F_{ij}$
107: are chosen randomly in the uniform interval $(0,F_{c})$.
108: Subsequently the variables evolve according to the following two-steps
109: dynamical rules: (i)
110: if all sites in the system are stable (i.e., $F_{ij} < F_{c}$), they
111: increase simultaneously and uniformly at a constant rate
112: \begin{equation}
113: \label{slow_dyn}
114: \frac{\partial F_{ij}(t)}{\partial t}=v ;
115: \end{equation}
116: (ii) as soon as one of them reaches the threshold value $F_{c}$, the
117: uniform driving is stopped and an ``earthquake'' starts:
118: \begin{equation}
119: \label{fast_dyn}
120: F_{ij} \ge F_{c} \Rightarrow \left\{ \begin{array}{l}
121: F_{ij} \rightarrow 0 \\
122: F_{nn} \rightarrow F_{nn} + \alpha F_{ij}
123: \end{array} \right.
124: \end{equation}
125: where ``$nn$'' denotes the set of nearest neighbour sites of $(i,j)$ and
126: $\alpha$ is a parameter that controls the level of conservation of the
127: dynamics ($\alpha=1/4$ corresponds to the conservative case).
128: The ``toppling'' rule (\ref{fast_dyn}) can possibly produce a chain reaction,
129: which ends when there are no more unstable sites in the system. At that point,
130: the uniform growth (\ref{slow_dyn}) starts again.
131: In the following we will assume, without loss of generality, a unit growth
132: rate, i.e. $v=1$.
133: A crucial point in the description of the model is the choice of boundary
134: conditions and, in accordance with previous investigations, we will consider
135: open boundary conditions.
136: These conditions imply that sites close to the boundaries topple according to
137: (\ref{fast_dyn}) but have a smaller coordination number.
138:
139: There is a clear separation of time scales in the system: earthquakes occur
140: instantaneously on the slow time scale of the driving.
141: The time in the system is therefore set by the slow time variable $t$.
142: By construction, moreover, the time coincides with the input energy {\em per
143: site} in the system. We will make use of this latter observation when we
144: will try to compare the behaviour of the OFC model with other models, such
145: as the BTW model or the Zhang model.
146:
147:
148: After a sufficiently long transient time, the system settles into a
149: stationary state, where the statistical properties of the model (e.g. the
150: probability distribution for earthquake sizes) do not depend on time.
151: In the BTW model and in the Zhang model, the transient time is relatively
152: brief and does not scale with system size. On average, an input energy
153: proportional to system size is needed to reach the steady
154: state~\cite{achille}.
155: On the contrary, transient times in the OFC model are known to be extremely
156: long, especially for large lattices.
157: For example, it was claimed in Ref.~\cite{grassberger} that $4 \times 10^8$
158: earthquakes are not enough to reach stationarity in a system of size $L=200$
159: for $\alpha=0.1$.
160: This conclusion was reached by observing the very slow convergence of the
161: mean earthquake size to an asymptotic value during the transient time.
162: In Ref.~\cite{grassberger} though, a systematic investigation of the
163: relaxation times for different $\alpha$ and $L$ was not attempted.
164: A more quantitative approach to the problem was proposed in
165: Ref.~\cite{middleton} by Middleton and Tang (MT).
166: According to these authors a ``self-organised'' region develops first close
167: to the boundaries and propagates thereafter into the bulk of the system.
168: The distance from the boundaries of the invasion front grows with time
169: as a power law, $d(t) \sim t^{\gamma (\alpha)}$, with
170: $\gamma = 0.23 \pm 0.08, 0.63 \pm 0.08$ for $\alpha =0.07, 0.15$,
171: respectively. The system reaches stationarity when the SOC region crosses
172: the whole sample ($d(t) \sim L$). Assuming that the power law behaviour
173: of $d(t)$ holds till saturation, than the transient time
174: of the system should scale as $t_{tr} \sim L^{1/\gamma (\alpha)}$.
175: More recently, it has been suggested in Ref.~\cite{ceva} that two distinct
176: relaxation times exist in the system, associated respectively with the power
177: law region and the ``tail'' (induced by finite size effects) of the
178: distribution of earthquake sizes. According to this study, the former should
179: stabilise much faster than the latter.
180:
181:
182:
183: \section{Results}
184:
185: Most of the studies on the OFC model at stationarity have concentrate
186: on the probability distribution of earthquake sizes, $P_L (s)$, where $L$ is
187: the size of the system and $s$ is the total number of sites that topple
188: during an earthquake. This probability distribution does not show
189: simple finite size scaling, at least in the range of lattice sizes accessible
190: to simulations at present~\cite{lisepac1}.
191: In a recent paper~\cite{lisepac2}, we have focused instead on the properties
192: of earthquakes confined within a fictitious subsystem of linear size
193: $\lambda$ (see fig.~\ref{scheme}). The model is driven according to its
194: usual dynamics but only those particular earthquakes that are entirely
195: contained within the subsystem are counted. We have shown that if
196: $\lambda$ is sufficiently smaller than $L$ the size distribution for this
197: subset of earthquakes, $P_{conf} (\lambda,s)$, obeys ordinary finite size
198: scaling, i.e. $P_{conf} (\lambda,s)\simeq \lambda^{-\beta}f(s/\lambda^D)$,
199: where the exponents $\beta=3.6$ and $D=2$ are universal over a range of
200: values of $\alpha$.
201:
202: In this work we want to address the issues briefly summarised in the
203: previous section concerning the approach to stationarity in the OFC model.
204: In order to be able to formulate scaling hypothesis and make use of
205: collapse plots, we will proceed in a way similar to that of
206: Ref.~\cite{lisepac2}. We will consider earthquakes localised within given
207: subsystems, in particular subsystems placed (a) at the boundaries and (b) at
208: the centre of the system.
209: As it is a prohibitive task to determine the time evolution of the entire
210: distribution $P_{conf} (\lambda,s)$, we will restrict ourselves to the mean
211: earthquake size, $<s>_{\lambda,L} (t)$, and, in general, to $q$-th moment
212: (up to $q=4$) of the distribution, $<s^q>_{\lambda,L} (t)$.
213: To determine numerically these quantities we have run several simulations
214: with different initial conditions, partitioning the time into bins of size
215: $\Delta t$. Let $n$ be the number of earthquakes occurring
216: between time $t-\Delta t/2$ and $t+\Delta t/2$ in a given realization of
217: the system and let $s_1, \ldots, s_n$ be the sizes of these earthquakes.
218: Then we define
219: \begin{equation}
220: <s^q>_{\lambda,L} (t) \simeq
221: \frac{<\left[ s_1^q + \ldots + s_n^q \right]_{t-\Delta t/2}^{t+\Delta t/2}>}
222: {<\left[ n \right]_{t-\Delta t/2}^{t+\Delta t/2}>}
223: \end{equation}
224: where $< \ldots>$ denotes an average over different realizations of the
225: system, i.e. over different initial conditions. For each system, the parameter
226: $\Delta t$ has been choose small with respect to the transient time
227: $t_{tr}$ but large enough to collect reasonably accurate statistics.
228:
229:
230: We consider first the case of subsystems placed adjacent to a boundary of
231: the system, in a symmetric position with respect to the corners.
232: In figure 2 we report $<s>_{\lambda,L}$ as a function of time for
233: $\alpha =0.18$ and some $\lambda$ and $L$. We observe that if the linear
234: dimension $\lambda$ of the box is sufficiently smaller than the linear
235: dimension of the system $L$ (approximately $L \ge 4 \lambda$) then the curve
236: $<s>_{\lambda,L} (t)$ becomes indistinguishable for different $L$.
237: This has been verified also for other values of $\alpha$ and $\lambda$.
238: We will therefore denote with $<s>_{\lambda} (t)$ the mean earthquake size
239: in this limit.
240: It is already visible from figure 2 that the relaxation time of
241: $<s>_{\lambda} (t)$ increases with $\lambda$. This is an indication in support
242: of the scenario proposed by MT. Regions close to the boundaries reach
243: stationarity sooner, signalling that an invasion front is moving toward the
244: bulk of the system. Some of MT conclusions nonetheless will have to be
245: modified as we will show later.
246:
247:
248: In order to describe quantitatively the invasion from the boundaries of
249: the self-organised region we make the following simple scaling hypothesis
250: \begin{equation}
251: \label{bound_scal}
252: <s>_{\lambda} (t)= \lambda^{\eta} F(t/\lambda ^z)
253: \end{equation}
254: where $\eta$ and $z$ are two suitable critical exponents. In particular
255: $z$ is a dynamical critical exponent that should satisfy $z=1/\gamma$, where
256: $\gamma$ is the ``invasion'' exponent as defined by MT.
257: In the limit of $t \rightarrow \infty$ the scaling function $F(x)$ saturates
258: to a constant, implying that the exponent $\eta$ is related to the finite
259: size exponents of the probability distribution $P_{conf}(\lambda,s)$ by the
260: relation $\eta =2 D -\beta \approx 0.4$. In figure \ref{fig_3} we report
261: collapse plots of the form (\ref{bound_scal}) for various values of $\alpha$.
262: We observe that a reasonably good collapse could be obtained for all the
263: $\alpha$ if we choose the universal exponent $z=1.3 \pm 0.1$. We are therefore
264: led to conclude that a universal exponent $z$ exist contrary to the claims
265: by MT.
266:
267: We consider next subsystems of different sizes placed at the centre of the
268: system. We observe that the relaxation time does not depend on the size of
269: the subsystems (see fig.~\ref{fig_4}a).
270: This confirm that the self-organisation mechanism develops from the
271: boundaries and that the system enters stationarity when the self-organised
272: region span the whole system. Only when the bulk of the system is reached by
273: the self-organised region is stationarity settled so that concentric
274: subsystem of different sizes will inevitably reach stationarity at the same
275: time.
276: We have also tested the relaxation times for higher moments of the avalanche
277: probability distribution to see whether different parts of the distribution
278: (e.g. power law part and the ``tail'') have different relaxation times as
279: suggested in Ref.~\cite{ceva}. In our investigation though we have not
280: observed any significant difference in the relaxation times associated with
281: different moments. We report as an example in fig.~\ref{fig_4}b the
282: behaviour of the first, second and fourth moment in a particular case.
283:
284:
285: The scaling equation (\ref{bound_scal}) suggests that the transient time in
286: the OFC model scales with system size as $t_{tr} \sim L^z$. One way to test
287: this is by comparing the time behaviour of the average earthquake size in a
288: central subsystem of size $\lambda$ for different system sizes $L$. Indeed the
289: asymptotic value $<s>_{\lambda,L} (t \rightarrow \infty)$ should not depend on
290: $L$. We report as example the case for $\alpha=0.18$ in fig.~\ref{fig_5},
291: where we have rescaled the time by a factor $L^z$. The curves show some noisy
292: behaviour, due to the difficulties in
293: collecting good statistics (relatively few earthquakes occur in the bulk
294: of the system as compared to the boundaries). Nonetheless the value deduced
295: for the exponent, $z \approx 1.3$, is consistent with the determination made
296: through the analysis of the earthquakes occurring at the boundaries. We have
297: obtained similar results also for other values of $\alpha$. In addition,
298: besides the average earthquake size, we have considered also the time
299: behaviour of other quantities such as the roughness of the energy landscape
300: (in analogy to surface growth problems) and the number of earthquakes per
301: unit time. All these different quantities on average reach stationary at the
302: same time.
303:
304:
305: The algebraic divergence of the relaxation time with system size reflects
306: the presence of long-range spatial correlations in the stationary state.
307: Indeed if correlations were only short range, than one would expect that the
308: transient time would not scale with system size. This is for example the case
309: for the BTW model in $d$ dimensions, where the height-height correlations are
310: algebraic but decays as fast as $r^{-2d}$ ($r$ being the distance between two
311: sites)~\cite{majumdar}. We have measured for various $L$ and $\alpha$ the
312: probability distribution of the spatially averaged force in the system
313: \begin{equation}
314: \label{M}
315: M=\frac{1}{L^2} \sum_{i,j=1}^{L} F_{i,j}
316: \end{equation}
317: In a system with sufficiently short range correlations, this probability
318: distribution would tend, in the limit $L \rightarrow \infty$, to a gaussian
319: distribution around the mean due to the central limit theorem (this is indeed
320: what results in the BTW model). Let $<M>$ and $\sigma _M$ be respectively
321: the average and the standard deviation of the distribution. In figure
322: \ref{fig_6} we have plotted $\log(\sigma _M P(M))$ versus
323: $(M-<M>)/\sigma _M$ for various $L$ and $\alpha$. Using these
324: coordinates a gaussian function would result in an inverse parabola.
325: For each $\alpha$ value the data of figure \ref{fig_6} collapse on a single
326: function, which is clearly not gaussian (deviations from gaussianity are more
327: pronounced for increasing $\alpha$ values). This indicates that the central
328: limit theorem does not hold in this case, not even for large $L$, suggesting
329: that long range algebraic correlations are present and therefore the
330: sum (\ref{M}) can not be decomposed into a sum of independent terms.
331: This observation is in agreement with the results reported in
332: Ref.~\cite{grinstein} where the presence of long-range spatial correlations
333: were deduced from the behaviour of a suitably defined susceptibility,
334: $\chi \equiv (L \sigma _M)^2$. It was claimed that $\chi$ diverges as $L^2$
335: and correspondingly that $\sigma _M$ is, to leading order, independent on $L$
336: (if $M$ was a sum of uncorrelated variables, $\sigma _M$ would decrease as
337: $1/L$).
338: In our investigation we have found that $\sigma _M$ slightly increases with
339: $L$ but asymptotically tends to a constant value, in accordance with
340: Ref.~\cite{grinstein} ($M$ is a bounded variable so $\sigma _M$ can not grow
341: indefinitely).
342:
343:
344:
345: \section{Conclusions}
346:
347: In conclusion, in this paper we have examined the process of self-organisation
348: in the OFC model. By considering earthquakes confined within a given subsystem
349: we have been able to clarify some of the issues related to this problem. In
350: accordance with Middleton and Tang~\cite{middleton} we have found that SOC
351: develops first close to the boundaries and subsequently invades the
352: interior of the system. The invasion process is controlled by a dynamical
353: critical exponent, $z \simeq 1.3$, which, contrary to previous claims, is
354: universal over a range of values of the dissipation level of the local
355: dynamics.
356: We have shown moreover that the transient time in the system scales with
357: system size as $t_{tr} \sim L^z$. This is a peculiarity of the OFC model as
358: other ``sandpile-like'' models (e.g. the BTW and the Zhang models) do not
359: display any scaling in the transient time. We have associated this feature
360: with the presence of long-range spatial correlations in the stationary state.
361:
362: Our findings are in general agreement with recent works on the OFC
363: model~\cite{lisepac1,lisepac2}. Indeed we have provided complementary evidences
364: (not based on the probability distribution for earthquake sizes) that the
365: model is critical even in a non conservative regime. Moreover it confirms
366: that there is universality in the system and that finite-size scaling can be
367: recovered by considering subsystems whose linear extent is sufficiently small.
368:
369: Finally, it is interesting to remark that the probability distribution for the
370: spatially averaged force in the system is somewhat reminiscent of a
371: probability distribution observed in a confined turbulent flow
372: experiment~\cite{bramwell} (BHP). As a term of comparison we have reported
373: in fig.~\ref{fig_6} the BHP functional form over-imposed to the curve for
374: $\alpha=0.21$. Attempts to link SOC systems to turbulent phenomena have long
375: been suggested, but only recently this has been put on a firmer
376: basis~\cite{mario}.
377:
378: \medskip
379:
380: We thank M.~Paczuski for helpful conversations and H.J.~Jensen for a critical
381: reading of the manuscript. This work was supported by the EPSRC (UK),
382: Grant No. GR/M10823/01 and Grant No. GR/R37357/01.
383:
384: \begin{thebibliography}{99}
385:
386:
387: \bibitem{btw} P. Bak, C. Tang, and K. Wiesenfeld, Phys. Rev. Lett. {\bf59},
388: 381 (1987); Phys. Rev. A. {\bf 38}, 364 (1988).
389:
390: \bibitem{bak_book}
391: P. Bak,
392: {\it How Nature Works: The Science of Self-Organized Criticality}
393: (Copernicus, New York, 1996).
394:
395: \bibitem{jen_book}
396: H. Jensen,
397: {\it Self-Organized Criticality}
398: (Cambridge University Press,New York, 1998).
399:
400: \bibitem{zhang}
401: Y.-C. Zhang, Phys. Rev. Lett. {\bf 63}, 470 (1989).
402:
403: \bibitem{achille}
404: A. Giacometti and A. Diaz-Guilera,
405: Phys. Rev. E {\bf 58}, 247 (1998)
406:
407: \bibitem{sibani}
408: P. Sibani and C.M. Andersen,
409: Phys. Rev. E {\bf 64}, 021103 (2001)
410:
411: \bibitem{ofc}
412: Z. Olami, H.J.S. Feder, and K. Christensen,
413: Phys. Rev. Lett. {\bf 68}, 1244 (1992);
414: K. Christensen and Z. Olami,
415: Phys. Rev. A {\bf 46}, 1829 (1992).
416:
417: \bibitem{grassberger} P. Grassberger, Phys. Rev. E {\bf 49}, 2436 (1994).
418:
419: \bibitem{socolar} J.E.S. Socolar, G. Grinstein, and C. Jayaprakash,
420: Phys. Rev E, {\bf 47}, 2366 (1993).
421:
422:
423: \bibitem{middleton}
424: A. A. Middleton and C. Tang,
425: Phys. Rev. Lett. {\bf 74}, 742 (1995).
426:
427: \bibitem{ceva} H. Ceva, Phys. Lett. A, {\bf 245}, 413 (1998).
428:
429: \bibitem{hwa}
430: T. Hwa and M. Kardar,
431: Phys. Rev. Lett. {\bf 62}, 1813 (1989).
432:
433: \bibitem{grinstein}
434: G. Grinstein, D.-H. Lee, and S. Sachdev,
435: Phys. Rev. Lett. {\bf 64}, 1927 (1990).
436:
437: \bibitem{manna}
438: S.S. Manna, L.B. Kiss, and J. Kertesz,
439: J. Stat. Phys. {\bf 61}, 923 (1990).
440:
441: \bibitem{carvalho}
442: J.X. Carvalho and C.P.C. Prado,
443: Phys. Rev. Lett. {\bf 84}, 4006 (2000)
444:
445: \bibitem{kim}
446: K. Christensen, D. Hamon, H.J. Jensen, and S. Lise,
447: Phys. Rev. Lett. {\bf 87}, 039801 (2001);
448: J.X. Carvalho and C.P.C. Prado,
449: Phys. Rev. Lett. {\bf 87}, 039802 (2001).
450:
451: \bibitem{lisepac1}
452: S. Lise and M.Paczuski, Phys. Rev. E, {\bf 63}, 036111 (2001).
453:
454: \bibitem{lisepac2}
455: S. Lise and M.Paczuski, Phys. Rev. E, {\bf 64}, 046111 (2001).
456:
457: \bibitem{majumdar}
458: S.N. Majumdar, and D. Dhar,
459: J. Phys. A, {\bf 24}, L357 (1991)
460:
461: \bibitem{bramwell}
462: S.T. Bramwell, P.C.W. Holdsworth, and J.-F. Pinton,
463: Nature {\bf 396}, 552 (1998).
464:
465: \bibitem{mario}
466: M. De Menech and A.L. Stella,
467: cond-mat/0103601
468:
469: \end{thebibliography}
470:
471:
472: \begin{figure}[hb]
473: \narrowtext
474: \epsfxsize=2.5in
475: \centerline{\epsffile{scheme.eps}}
476: \protect\vspace{0.2cm}
477: \caption[1]{\label{scheme}
478: Schematic representation of earthquakes entirely confined within a subsystem
479: of linear size $\lambda$ (dashed line). Toppling sites are denoted with a
480: cross.
481: }
482: \end{figure}
483:
484: \begin{figure}[hb]
485: \narrowtext
486: \epsfxsize=4.in
487: \centerline{\epsffile{fig_2.eps}}
488: \caption[1]{\label{fig_2}
489: Average earthquake size $<s>_{\lambda,L}$ in a subsystem placed at the
490: boundary as a function of time; $\alpha =0.18$ and, from bottom to top,
491: $\lambda=32$, $\lambda=64$ and $\lambda=128$.
492: }
493: \end{figure}
494:
495: \begin{figure}[hb]
496: \narrowtext
497: \epsfxsize=4.5in
498: \centerline{\epsffile{fig_3.eps}}
499: \caption[1]{\label{fig_3} Collapse plots of $<s>_{\lambda} (t)$ for a
500: subsystem placed at the boundary and for (a) $\alpha=0.15$, (b) $\alpha=0.18$
501: and (c) $\alpha=0.21$. The value of the dynamical critical exponent is
502: $z=1.3$.
503: }
504: \end{figure}
505:
506: \begin{figure}[hb]
507: \narrowtext
508: \epsfxsize=4.in
509: \centerline{\epsffile{fig_4.eps}}
510: \caption[1]{\label{fig_4} Time dependence of the average earthquake sizes in
511: subsystems placed at the centre of the system for $\alpha=0.18$ and $L=256$;
512: (a) average earthquake size $<s>_{\lambda,L}$ for, from bottom to top,
513: $\lambda=16,32,64,128,256$ and (b) $q-th$ moment of the distribution for
514: $\lambda=32$ and, from bottom to top, $q=1,2,4$.
515:
516:
517: }
518: \end{figure}
519:
520: \begin{figure}[hb]
521: \narrowtext
522: \epsfxsize=4.in
523: \centerline{\epsffile{fig_5.eps}}
524: \caption[1]{\label{fig_5}
525: Collapse plots of $<s>_{\lambda, L} (t)$ for a subsystem of size $\lambda=32$
526: placed at the centre of a system of size $L$; the conservation parameter is
527: $\alpha=0.18$. The value of the dynamical critical exponent is $z=1.3$.
528: }
529: \end{figure}
530:
531: \begin{figure}[hb]
532: \narrowtext
533: \epsfxsize=4.in
534: \centerline{\epsffile{fig_6.eps}}
535: \caption[1]{\label{fig_6}
536: Rescaled probability distribution of the spatially averaged force in the system
537: for, from bottom to top, $\alpha=0.21, 018, 0.15$. $<M>$ and $\sigma _M$ are
538: respectively the average and the standard deviation of the distribution. The
539: top and bottom curves have been shifted by a factor of $10$ respectively up
540: and down for visual clarity. Squares represent the BHP (rescaled) probability
541: distribution observed in an experiment on turbolence~\protect\cite{bramwell}.}
542: \end{figure}
543:
544:
545:
546: \end{document}
547: