1: \documentstyle[aps,multicol,psfig]{revtex}
2:
3: \draft
4:
5: \begin{document}
6:
7: \title{Non conservative Abelian sandpile model with BTW toppling rule}
8:
9: \author{Alexei V\'azquez$^{1,2}$}
10:
11: \address{$^1$ Abdus Salam International Center for Theoretical Physics\\
12: Strada Costiera 11, P.O. Box 586, 34100 Trieste, Italy}
13:
14: \address{$^2$ Department of Theoretical Physics, Havana University\\
15: San L\'azaro y L, Havana 10400, Cuba}
16:
17: \maketitle
18:
19: \begin{abstract}
20:
21: A non conservative Abelian sandpile model with BTW toppling rule introduced in
22: [Tsuchiya and Katori, Phys. Rev. E {\bf 61}, 1183 (2000)] is studied. Using a
23: scaling analysis of the different energy scales involved in the model and numerical
24: simulations it is shown that this model belong to a universality class different
25: from that of previous models considered in the literature.
26:
27: \end{abstract}
28:
29: \pacs{64.60.Lx, 05.70.Ln}
30:
31: \begin{multicols}{2}\narrowtext
32:
33: \section{Introduction}
34:
35: Recently Tsuchiya an Katori \cite{katori00} have introduced a non conservative
36: Abelian sandpile model with a toppling rule similar to that of the well known Bak,
37: Tang and Wiesenfeld (BTW) sandpile model \cite{bak87}. The model is defined in a
38: square lattice of size $L$ and an integer energy profile $z_{ij}$ is considered.
39: Sites with energy below the threshold $z_c=4\alpha\zeta$ are stable. If the energy
40: at any given site $(i,j)$ exceeds this threshold the site transfer energy to its four
41: nearest neighbors following the toppling rule: $z_{ij}\rightarrow z_{ij}-z_c$,
42: $z_{i\pm1j}\rightarrow z_{i\pm1j}+\zeta$ and $z_{ij\pm1}\rightarrow
43: z_{ij\pm1}+\zeta$. $\zeta$ is an integer number and $\alpha$ is such that $z_c$ is
44: also integer. The boundaries are assumed to be open and the system is
45: perturbed by adding a unit of energy at a site selected at random and letting it
46: evolve until an equilibrium configuration is reached.
47:
48: On each toppling event an amount of energy $\epsilon=4\zeta(\alpha-1)>0$ is
49: dissipated. For $\alpha=1$ the model is conservative, it is just the BTW model but
50: with a different scale of toppling and energy addition. In the BTW model
51: \cite{bak87} the energy added to perturb the system is 1 and on toppling an active
52: site transfers an energy 1 to each neighbor, they are of the same order. In the
53: model defined above the energy added is still 1 but the energy transferred on
54: toppling is $\zeta$. In the limit $\alpha=1$ and $\zeta=1$ the BTW model
55: \cite{bak87} is recovered while for $\alpha=1$ and $\zeta\gg1$ it is similar to the
56: BTW model but with a uniform driving.
57:
58: In the BTW model ($\alpha=1$ and $\zeta=1$) the avalanches can be decomposed in a
59: sequence of sub avalanches called waves \cite{ivashkevich94} with well defined
60: finite size scaling properties. On the contrary, the distribution of the overall
61: avalanche size $s$ is better described using a multi-fractal analysis
62: \cite{tebaldi99}. The break down of the finite size scaling has been recently shown
63: to be a consequence of the existence of correlations in the sequence of waves
64: \cite{demenech00}.
65:
66: For $\alpha>1$ the model is non conservative but still Abelian \cite{katori00}. In
67: the thermodynamic limit $L\rightarrow\infty$, exact calculations by Tsuchiya and
68: Katori yield the mean avalanche size (including avalanches with size zero) $\langle
69: T\rangle=\epsilon^{-1}$ \cite{katori00}. Since $\epsilon=4\zeta(\alpha-1)$ they
70: concluded that in the limit $\alpha\rightarrow1$ $\langle T\rangle$ diverges.
71: However, as it is shown in sec. \ref{sec:scaling} this conclusion is wrong because
72: $\alpha$ cannot goes to zero in an arbitrary way, in order to satisfy the constraint
73: that $z_c=4\alpha\zeta$ remains integer. Here it is demonstrated that
74: $\epsilon\geq1$ and, therefore, $\langle T\rangle\leq1$ for all possible values of
75: $\zeta$ and $\alpha$.
76:
77: The main goal of this work is to investigate the scaling properties of this non
78: conservative BTW like model in the limit $\alpha\rightarrow1$. The main questions
79: are related with the existence or not of criticality in the conservative limit
80: $\alpha\rightarrow1$ and if in this limit one recovers the conventional BTW model
81: $\alpha=1$. From the analysis of the energy scales involved in the model ( sec.
82: \ref{sec:scaling}) and numerical simulations (sec. \ref{sec:numerical}) it is
83: concluded that the model is critical when $\alpha\rightarrow1$ but it does not
84: belong to the universality class of the BTW model. Its relation with other non
85: conservative models with BTW like toppling rule introduced in the literature
86: \cite{manna90,ghaffari97,chessa98} is also discussed (sec. \ref{sec:numerical}).
87:
88: \section{Scaling laws}
89: \label{sec:scaling}
90:
91: In this section some scaling laws are derived based on the energy scales involved in
92: the model. The main idea of this approach is that the balance between input and
93: dissipation of energy determine the scaling of some magnitudes with the dissipation
94: per toppling event, following the general guidelines introduced by Vespiganani {\em
95: et al} \cite{vespignani97}. For this purpose the avalanches are assumed to be
96: instantaneous and the analysis is focused in the time scale of the driving field. On
97: each step one adds 1 unit of energy and measure the toppling activity and the energy
98: dissipated. On each toppling event an amount of energy $\epsilon=4\zeta(\alpha-1)$
99: is locally dissipated while an amount $4\zeta$ is transferred to nearest neighbors.
100: For boundary sites part of the energy is also dissipated through the boundary.
101:
102: Let $G(r)$ be the Green function \cite{dhar89}, the average number of toppling
103: events at a distance $r$ from the site where the energy was added. Close to $r=0$
104: the effect of local dissipation gives an small contribution and the main energy
105: scale is given by the transport of the energy from active sites to their nearest
106: neighbors. On the contrary, far from $r=0$ the effects of the local dissipation
107: becomes more important. How far will depend on certain correlation length $\xi$,
108: such that for $r\ll\xi$ transport is more important than local dissipation while for
109: $r\gg\xi$ the opposite occurs.
110:
111: Thus, there are two characteristic lengths in this model: the system size $L$ and
112: the correlation length $\xi$. The analysis developed above is valid in the
113: thermodynamic limit $L\gg\xi$. In this case the dissipation through the boundary of
114: the system is negligible in comparison with the energy dissipated on each toppling
115: event. In such a situation the only way to reach an stationary state is to balance
116: the input of energy from the driving field with the energy locally dissipated.
117: Moreover, since $\xi$ is the only characteristic length $G(r)$ is expected to
118: satisfy the scaling law
119: \begin{equation}
120: G(r)=r^{\eta-d}{\cal F}(r/\xi),
121: \label{eq:1}
122: \end{equation}
123: where $\eta$ is an scaling exponent and $d$ is the spatial dimension.
124:
125: The amount of energy $\delta E_d(r)$ locally dissipated inside an hyper-circle of
126: radius $r$ is
127: \begin{equation}
128: \delta E_d(r)\propto\epsilon\int_0^rd\rho\rho^{d-1}G(\rho)
129: \propto \epsilon\xi^\eta f(r/\xi),
130: \label{eq:2}
131: \end{equation}
132: where $f(x)=\int_0^x dy y^{\eta-1}{\cal F}(y)$ and the second proportionality is
133: obtained using eq. (\ref{eq:1}). On the other hand, the average energy transported
134: through its boundary $\delta E_t(r)$ is given by
135: \begin{equation}
136: \delta E_t(r)\propto\zeta r^{d-1}\frac{dG}{dr}(r)
137: \propto\zeta\xi^{\eta-2}g(r/\xi),
138: \label{eq:3}
139: \end{equation}
140: where $g(x)=d[x^{\eta-2}{\cal F}(x)]/dx$ and the second proportionality is obtained
141: using eq. (\ref{eq:1}). The correlation length $\xi$ can be defined as the radius
142: $r$ at which this two contribution becomes of the same order. With this definition
143: and equating eqs. (\ref{eq:2}) and (\ref{eq:3}) with $r=\xi$ it results that
144: \begin{equation}
145: \xi\sim\left(\frac{\zeta}{\epsilon}\right)^{\nu},\ \ \ \ \nu=\frac{1}{2}.
146: \label{eq:4}
147: \end{equation}
148:
149: On the other hand, on each step 1 unit of energy is added and in average the amount
150: $\epsilon\langle T\rangle$ is dissipated, where $\langle T
151: \rangle\propto\int_0^\infty dr r^{d-1}G(r)$ is the mean avalanche size counting
152: even the case when it has size 0. Equating this two contributions it results that
153: \begin{equation}
154: \langle T\rangle=\frac{1}{\epsilon}=\frac{1}{4\zeta(\alpha-1)}.
155: \label{eq:5}
156: \end{equation}
157: Moreover, using eq. one obtains
158: (\ref{eq:1})
159: \begin{equation}
160: \eta=0.
161: \label{eq:6}
162: \end{equation}
163:
164: Eq. (\ref{eq:5}) reproduces the exact result by Tsuchiya and Katori. The present
165: approach is however based on more general arguments an can be easily adapted to any
166: sandpile model with local dissipation. The same argument (energy balance) has been
167: previously used by Vespiganani {\em et al} \cite{vespignani97} to understand the
168: scaling properties of other sandpile models with local dissipation. Here, a slightly
169: different approach has been considered where the new parameter $\zeta$, the ratio
170: between the energy received from nearest active neighbors and from the driving
171: field, has been taken into account.
172:
173: From eq. (\ref{eq:5}) Tsuchiya and Katori concluded that when $\alpha\rightarrow1$
174: $\langle T\rangle$ diverges. However, this conclusion is not valid if
175: $z_c=4\zeta\alpha$ is restricted to be an integer number. To show this let us write
176: $\alpha=1+\epsilon/4\zeta$ which follows from eq. (\ref{eq:5}). But
177: $4\zeta\alpha=4\zeta+\epsilon$ is restricted to take integer values. With $\zeta$
178: being an integer number the only way to satisfy this requirement is that $\epsilon$
179: is also integer, i.e. $\epsilon=1,2,3,\ldots$. Then, since the smaller non negative
180: integer is 1 it is concluded that $\epsilon\geq1$ and, therefore, from eq.
181: (\ref{eq:5}) $\langle T\rangle\leq1$, i.e. it is bounded.
182:
183: Nevertheless, the correlation length $\xi$ in eq. (\ref{eq:4}) does not only depend
184: on $\epsilon$ but also on $\zeta$. For fixed $\epsilon$ it diverges in the limit
185: $\zeta\rightarrow\infty$ and the model is critical. The real control parameter is
186: then $\epsilon_{eff}=\epsilon/\zeta$, i.e. the energy dissipated per toppling event
187: relative to the characteristic energy scale of transport $\zeta$. Although this
188: result is in complete agreement with the field theory approach of Vespignani {\em et
189: al} \cite{vespignani97} the fact that $\langle T\rangle$ does not diverges when
190: $\epsilon_{eff}\rightarrow0$ ($\zeta\rightarrow\infty$) excludes this model from
191: their analysis.
192:
193: Moreover, in previous sandpile models conservation implies the scaling law $\langle
194: s\rangle\sim \xi^2$, where $\langle s\rangle$ is the mean avalanche size excluding
195: those with size 0 \cite{vespignani97}. To investigate the validity of such
196: scaling relation for the present model let us take into account that $\langle
197: s\rangle$ is related to $\langle T\rangle$ through the expression
198: \begin{equation}
199: \langle s\rangle=\langle T\rangle/P_a,
200: \label{eq:7}
201: \end{equation}
202: where $P_a$ is the probability to obtain an avalanche with non zero size. In the
203: models considered by Vespignani {\em et al} \cite{vespignani97} $\zeta=1$ and,
204: therefore, from eqs. (\ref{eq:4}), (\ref{eq:5}) and (\ref{eq:7}) it results that
205: $\langle s\rangle\sim \xi^2/P_a$. Moreover, in this models $P_a$ has a finite value
206: and, therefore, one obtains the mentioned scaling law $\langle s\rangle\sim\xi^2$.
207:
208: On the contrary, in the model considered here $\langle s\rangle$ can not be related
209: to $\xi$ using these arguments. For fixed $\epsilon$ from eqs. (\ref{eq:5}) and
210: (\ref{eq:7}) one obtains that $\langle s\rangle \sim 1/P_a$. Thus, from the energy
211: balance invoked above we cannot say anything about the scaling of $\langle s\rangle$
212: with $\xi$ (an exponent 2 will be an accidental coincidence) and, therefore, this
213: model belongs to a new universality class.
214:
215: \section{Numerical simulations and discussion}
216: \label{sec:numerical}
217:
218: In this sections results obtained from numerical simulations of the model studied
219: above are presented. The simulations were performed using $\epsilon=1$, $L=4096$ and
220: $\zeta$ ranging from $2^0$ to $2^{10}$ ($\epsilon_{eff}=1/\zeta$ ranging from 1 to
221: $2^{-10}$). For these values the condition $L\ll\xi$ was observed to be satisfied.
222: Statistics was taken over $10^8$ avalanches after the system reached the stationary
223: state.
224:
225: Before entering in the analysis of the statistics of the avalanches let us check the
226: validity of eq. (\ref{eq:5}). The log-log plot of $\langle s\rangle$ vs. $\zeta$ is
227: shown in fig. \ref{fig:1}. A clear linear behavior is observed for
228: $\log_{10}\zeta\geq5$ suggesting that above this value simple scaling applies. On
229: top of this points the numerically computed values of $1/P_a$ are plotted obtaining
230: an overlap in agreement with eq. (\ref{eq:5}). If the scaling relation $\langle
231: s\rangle \sim\xi^2$ where valid, using eq. (\ref{eq:4}), $\langle
232: s\rangle\sim\zeta$. However, a linear fit to this log-log plot gives a slope
233: $\sim0.9$.
234:
235: The fact that this scaling relation thus not hold is clearly shown in fig.
236: \ref{fig:2}, where the stationary energy distribution is shown. As it can be seen
237: $P_a=P_{z_c-1}$ does not decreases as $1/\zeta$ but slower, which explains why
238: $\langle s \rangle$ increases with $\zeta$ with a slope smaller than 1. The rest of
239: the distribution scales like $1/\zeta$ which is just a consequence of the
240: increase of the density of possible values of $z$.
241:
242: The avalanche statistics will be characterized by: the number of toppling events $s$
243: and steps $t$ required to reach an stable configuration, the number of sites $a$
244: "touched" by the avalanche, and the characteristic radius of the cluster formed by
245: these sites $r$. The main goal of the simulations is to determine the probability
246: densities $p_x(x,\zeta)$ ($x=s,t,a,r$) in the stationary state.
247:
248: One can easily see that $s=a$, in other words sites topples only once within an
249: avalanche. In this model, as a difference with the original BTW model, only one wave
250: of topplings takes place. The first wave if generated from an initial site with
251: height $z=z_c=4\zeta+\epsilon$. When this site topples it transfers an amount equal
252: to $z_c$ to its nearest neighbors and, therefore, ends with energy $z=0$. The best
253: we can have to obtain a second toppling at this site is that its four nearest
254: neighbors also become active. In such a case the initial side will receive
255: $4\zeta<z_c$ units of energy, which is not enough to make it active again. Hence, no
256: second wave will be obtained yielding $s=a$.
257:
258: Since the waves are known to satisfy well defined finite-size scaling properties and
259: in the present model an avalanche is made by one wave it is expected that the
260: distributions $p_x(x,\zeta)$ also satisfy a finite-size scaling. However, the
261: scaling exponents will not necessarily be those obtained for the scaling of waves
262: because, in the present model, conservation does not introduce any scaling relation
263: among the different scaling exponents.
264:
265: If finite size scaling applies then these densities will satisfy
266: \begin{equation}
267: p_x(x,\zeta)=x^{-\tau_x}{\cal G}[x/x_c(\zeta)],
268: \label{eq:8}
269: \end{equation}
270: where $\tau_x$ is the power law exponent characterizing the self-similar regime and
271: $x_c$ is a cut off above which the distribution deviates from a power law and has a
272: fast decay given by ${\cal G}$. The validity of this scaling form is supported by
273: the numerical results. The cut off $x_c$ is determined by the existence of the
274: characteristic length $\xi\sim\zeta^\nu$ and is expected to scale as
275: $x_c\sim\xi^{D_x}\sim\zeta^{d_x}$, where $d_x=D_x\nu$ is an effective fractal
276: dimension.
277:
278: To compute the exponents $\tau_x$ and $d_x$ the moment analysis technique
279: introduced by De Menech {\em et al} \cite{demenech98} is used. The moments of the
280: probability density in eq. (\ref{eq:8}) are given by
281: \begin{equation}
282: \langle x^q\rangle=\int_0^\infty dx p(x) x^q\sim \zeta^{\sigma_x(q)},
283: \label{eq:9}
284: \end{equation}
285: where
286: \begin{equation}
287: \sigma_x(q)=(1-\tau_x)d_x+d_xq.
288: \label{eq:10}
289: \end{equation}
290: The last equivalence in eq. (\ref{eq:9}) is valid for values of $q$ not to small,
291: for which the precise form of $p_x(x,\zeta)$ at small $x$ is not important.
292:
293: $\sigma_x(q)$ can be determined from a linear fit to the log-log plot of $\langle
294: x^q\rangle$ vs. $\zeta$. The resulting values using $\zeta=2^5,\ 2^6,\ldots,\
295: 2^{10}$ are shown in fig. \ref{fig:3}. In all cases ($x=s,t,r$) for $q$ larger than
296: 1 a well defined linear dependence is observed. From the linear fit (see eq.
297: \ref{eq:10}) to these plots the exponents $\tau_x$ and $d_x$ are computed. The
298: results are shown in table \ref{tab:1}.
299:
300: The exponent $\nu$ is very close to $1/2$ in very good agreement with the scaling
301: arguments of previous section. On the other hand, $d_s$ is quite close to 1 which
302: implies that the avalanche size (or area) scale as $s\sim r^2$, i.e. avalanches are
303: compact $D_s=2$. With this value, the scaling relation $(2-\tau_s)D_s=2$ yields the
304: power law exponent $\tau_s=1$ which is clearly in disagreement with the value
305: computed numerically. The reason for this result is that conservation does not
306: introduce any scaling relation as it generally occurs in sandpile models
307: \cite{vazquez00}.
308:
309: The exponents computed using the moment analysis technique can be checked using
310: rescaled plots of the integrated distribution $P_x(x,\zeta)=\int_x^\infty
311: dxp_x(x,\zeta)$. The resulting plots are shown in figs. \ref{fig:4}, \ref{fig:5},
312: and \ref{fig:6}. The scaling works quite good supporting the validity of the
313: reported exponents.
314:
315: In the literature we can find other sandpile models with local dissipation and BTW
316: like toppling rule \cite{manna90,ghaffari97,chessa98}. In the models considered in
317: \cite{manna90} and \cite{ghaffari97} the energy profile is continuous and the
318: dissipation rate per toppling event $\epsilon$ is a control parameter which can take
319: any real value and, therefore, can be tuned to zero. Another feature of these models
320: is that only one wave of toppling can take place and, therefore, for any finite
321: $\epsilon$ the model is in a different universality class from that of the BTW
322: model.
323:
324: On the other hand in \cite{chessa98} the energy profile is discrete as in the
325: original BTW model at the prize of introducing stochasticity in the model. In this
326: case with a probability $p$ energy is fully dissipated yielding an average
327: dissipation per toppling event $\epsilon=2dp$. Clearly $p$ may take any real
328: variable between 0 and 1 and, therefore, also in this case the dissipation per
329: toppling event can be fine tuned to zero. As a difference with the models described
330: in the previous paragraph, in this case multiple toppling of a site within an
331: avalanche is possible, which make it closer to the original BTW model. Moreover, the
332: use of finite size scaling techniques can be also questioned and a multi-fractal
333: analysis may be more appropriate \cite{vespignani00}, which is another
334: characteristic feature of the BTW model \cite{tebaldi99}. All this elements together
335: with the numerical results reported in \cite{chessa98} suggest that in the limit
336: $p\rightarrow1$ ($\epsilon\rightarrow0$) this model belong to the same universality
337: class of the BTW model.
338:
339: A common feature of all this models \cite{manna90,ghaffari97,chessa98} is that
340: $\langle s\rangle\sim\epsilon^{-1}$ as predicted by the field theory approach of
341: Vespignani {\em et al} \cite{vespignani97}, leading to the scaling relation
342: $(2-\tau_s)D_s=2$. On the contrary, in the present model the scaling of $\langle
343: s\rangle$ with $\epsilon_{eff}$ is not known and conservation does not introduce the
344: above scaling relation. Hence, the model introduced by Tsuchiya and Katori belongs
345: to different class among sandpile models.
346:
347: \section{Summary and conclusions}
348:
349: A non conservative Abelian sandpile model with a BTW like toppling rule has been
350: studied. The model can be though as the only possible generalization of the BTW
351: model to include local dissipation without introducing stochasticity in the toppling
352: rule and keeping a discrete energy profile. However, the scaling approach and
353: numerical the simulations reported here show that it does not belong to the
354: universality class of the BTW model, not even to the universality class of any
355: sandpile model previously considered in the literature.
356:
357: \section*{Acknowledgements}
358:
359: I thanks R. Pastor Satorras and A. Vespignani for useful comments and discussion.
360: The numerical simulations where performed using the computing facilities at the
361: ICTP.
362:
363:
364: \begin{thebibliography}{99}
365:
366: \bibitem{katori00} T. Tsuchiya and M. Katori, Phys. Rev. E {\bf 61}, 1183 (2000).
367:
368: \bibitem{bak87} P. Bak, C. Tang, and K. Wiesenfeld, Phys. Rev. Lett. {\bf 59},
369: (1987); Phys. Rev. A {\bf 38}, 364 (1988).
370:
371: \bibitem{ivashkevich94} E. V. Ivashkevich, D. Ktitarev, and V. B. Priezzhev,
372: Physica A {\bf 209}, 347 (1994); D. V. Ktitarev, S. L\"ubeck, P. Grassberger, and
373: V. B. Priezzhev, Phys. Rev. E {\bf 61}, 81 (2000).
374:
375: \bibitem{tebaldi99} C. Tebaldi, M. De Menech, and A. L. Stella, Phys. Rev. Lett.
376: {\bf 83}, 3952 (1999).
377:
378: \bibitem{demenech00} M. De Menech and A. L. Stella, arXiv:cond-mat/0002310.
379:
380: \bibitem{manna90} S. S. Manna, L. B. Kiss, and J. Ker\'esz, J. Stat. Phys. {\bf 61},
381: 923 (1990).
382:
383: \bibitem{ghaffari97} P. Ghaffari, S. Lise, and H. J. Jensen, Phys. Rev. E {\bf 56},
384: 6702 (1997).
385:
386: \bibitem{chessa98} A. Chessa, E. Marinari, A. Vespignani, and S. Zapperi, Phys.
387: Rev. E {\bf 57}, R6241 (1998).
388:
389: \bibitem{vespignani97} A. Vespignani and S. Zapperi, Phys. Rev. Lett. {\bf 78}, 4793
390: (1997); Phys. Rev. E {\bf 57}, 6345 (1998); A. Vespignani, R. Dickman, M. A.
391: Mu\~noz, and S. Zapperi, Phys. Rev. Lett. {\bf 81}, 5676 (1998).
392:
393: \bibitem{dhar89} D. Dhar and R. Rammaswamy, Phys. Rev. Lett. {\bf 63}, 1659 (1989);
394: for a review see cond-mat/9909009.
395:
396: \bibitem{demenech98} M. De Menech, A. L. Stella, and C. Tebaldi, Phys. Rev. E {\bf
397: 58}, R2677 (1998).
398:
399: \bibitem{vazquez00} A. V\'azquez and O. Sotolongo-Costa, J. Phys. A {\bf 32}, 2633
400: (1999).
401:
402: \bibitem{vespignani00} A. Vespignani, personal communication.
403:
404: \end{thebibliography}
405:
406: \begin{figure}
407: \centerline{\psfig{file=fig1.ps,width=3in,angle=-90}}
408: \caption{Log-log plot of the mean avalanche size (excluding avalanches with size
409: $s=0$) as a function $\zeta$. It can be clearly seen that it scales as $P_a^{-1}$,
410: the probability per unit step to obtain an avalanche with $s>0$. The line is a
411: linear fit to the high $\zeta$ interval.}
412: \label{fig:1}
413: \end{figure}
414:
415: \begin{figure}
416: \centerline{\psfig{file=fig2.ps,width=3in,angle=-90}}
417: \caption{Probability $P_z$ that a site has energy $z$ in the stationary state, for
418: different values of $\zeta=2^n$. $z$ is expressed in units of the threshold
419: $z_c=4\zeta+1$ while $P_z$ has being rescaled by an amount $\zeta$ because with
420: increasing $\zeta$ the density of $z/z_c$ values increases as $\zeta$.}
421: \label{fig:2}
422: \end{figure}
423:
424: \begin{figure}
425: \centerline{\psfig{file=fig3.ps,width=3in,angle=-90}}
426: \caption{Moment exponent $\sigma_x(q)$ for different values of $q$ and $x=s,t,r$.
427: The lines are linear fits [$\sigma_x(q)=(1-\tau_x)d_x+d_xq$] to the interval $1\geq
428: q\leq3$. The resulting exponents $\tau_x$ and $d_x$ are shown in tab. \ref{tab:1}.}
429: \label{fig:3}
430: \end{figure}
431:
432: \begin{figure}
433: \centerline{\psfig{file=fig4.ps,width=3in,angle=-90}}
434: \caption{Scaled plot of the integrated distribution of avalanche sizes (or area
435: since $s=a$ in this model) using the exponents displayed in tab \ref{tab:1}.}
436: \label{fig:4}
437: \end{figure}
438:
439: \begin{figure}
440: \centerline{\psfig{file=fig5.ps,width=3in,angle=-90}}
441: \caption{Scaled plot of the integrated distribution of avalanche durations using the
442: exponents displayed in tab \ref{tab:1}.}
443: \label{fig:5}
444: \end{figure}
445:
446: \begin{figure}
447: \centerline{\psfig{file=fig6.ps,width=3in,angle=-90}}
448: \caption{Scaled plot of the integrated distribution of avalanche radius using the
449: exponents displayed in tab \ref{tab:1}.}
450: \label{fig:6}
451: \end{figure}
452:
453: \begin{table}
454: \begin{tabular}{llllll}
455: $d_s$ & $d_t$ & $d_r=\nu$ & $\tau_s$ & $\tau_t$ & $\tau_r$\\
456: \hline
457: 0.994(5) & 0.630(5) & 0.495(5) & 1.11(1) & 1.16(1) & 1.14(1)\\
458: \hline
459: $D_s=d_s/\nu$ & $z=D_t=d_t/\nu$ & & & &\\
460: \hline
461: 2.01(1) & 1.27(1)
462: \end{tabular}
463: \caption{Scaling exponents obtained from linear fits
464: [$\sigma_x(q)=(1-\tau_x)d_x+d_xq$] to the data shown in fig. \ref{fig:3}.}
465: \label{tab:1}
466: \end{table}
467:
468: \end{multicols}
469:
470: \end{document}
471:
472: