1: %\documentstyle [preprint,aps]{revtex}
2: \documentstyle[prb,aps]{revtex}
3:
4: \begin {document}
5:
6: %\draft
7:
8: \title {Growth model with restricted surface relaxation}
9:
10: \author {T. J. da Silva and J. G. Moreira}
11: \address {Departamento de F\'{\i}sica, Instituto de Ci\^encias Exatas,\\
12: Universidade Federal de Minas Gerais, Caixa Postal 702,\\
13: 30161-970, Belo Horizonte, MG - Brazil}
14:
15: \date {\today}
16: \maketitle
17:
18: \begin {abstract}
19: We simulate a growth model with restricted surface relaxation process
20: in $d=1$ and $d=2$, where $d$ is
21: the dimensionality of a flat substrate. In this model, each particle
22: can relax on the surface to a local minimum, as the Edwards-Wilkinson
23: linear model, but only within a distance $s$.
24: If the local minimum is out from this distance, the particle
25: evaporates through a refuse mechanism similar to the Kim-Kosterlitz
26: nonlinear model. In $d=1$,
27: the growth exponent $\beta$, measured from the temporal behavior
28: of roughness, indicates that in the coarse-grained limit, the linear term
29: of the Kardar-Parisi-Zhang equation dominates in short times
30: (low-roughness) and, in asymptotic times, the nonlinear term prevails.
31: The crossover between linear and nonlinear behaviors occurs in a
32: characteristic time $t_c$ which only depends on the magnitude of the
33: parameter $s$, related to the nonlinear term.
34: In $d=2$, we find indications of a
35: similar crossover, that is, logarithmic temporal behavior of roughness in short
36: times and power law behavior in asymptotic times.
37:
38: \end {abstract}
39: %\vskip 2truecm
40: \pacs {68.35.Fx, 05.70.Ln, 61.50.C}
41: %\vfil\noindent {PACS: 68.35.Fx, 05.70.Ln, 61.50.C}
42: \narrowtext
43: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname @twocolumnfalse\endcsname
44: \vskip2pc]
45: \newpage
46: %
47: %
48: %
49: % --------------------- begining --------------------------------------
50: %
51: %
52: %
53: \section {INTRODUCTION}
54: %
55: %
56: %
57: Self-affine interfaces generated by nonequilibrium surface growth
58: have been intensively studied in recent years [1-3]. Kinetic roughening models
59: as
60: ballistic deposition \cite{db}, Eden model \cite{eden}, solid-on-solid model
61: (SOS) with surface
62: relaxation \cite {ew,family86}, SOS with refuse \cite{kk} and SOS with
63: diffusion
64: \cite {wolfvillain} are some
65: examples of growth models that belongs to distinct universality classes.
66: Theoretically, these
67: growth processes are studied in three different schemes: through a continuum
68: description using Langevin-like equation and renormalization group analisys
69: for solving it; through numerical solutions of these equations; and
70: through computer simulations of discrete models.
71: The main goal is to obtain the universality class of a specific model and to
72: get
73: information about the presence of nonlinearities and broken symmetries.
74:
75: In computer simulation of lattice growth models, interfaces are described by
76: a discrete set $\{h_i(t)\}$ which represents the height of a site
77: $i$ at the time $t$. Such an interface has $L^d$ sites, where
78: $L$ is the linear size and $d$ is the dimension of the substrate.
79: The
80: roughness of the interface $\omega$ is defined as the root mean square of the
81: $\{h_i-{\bar h}\}$ distribution,
82: \begin{equation}
83: \omega^2(L,t)=\left<{1\over L^d}\sum_{i=1}^{L^d}(h_i-\bar h)^2\right>,
84: \end{equation}
85: where $\bar h$ is the mean height at time $t$ and the angular
86: brackets means the average over independent samples.
87:
88: The universality class of a discrete growth model is obtained through
89: the temporal and spatial behaviors of
90: roughness. In the most of kinetic roughening processes, which
91: starts at $t=0$ from a flat substrate, the temporal
92: behavior of roughness is described by the power law behavior,
93: $\omega(L,t)\sim t^{\beta}$, when $1\ll t\ll t_{\times}$, and its spatial behavior,
94: in the steady state, is described by
95: $\omega_{sat}(L)\sim L^{\alpha}$, for
96: $t\gg t_{\times}$. The exponents $\beta$ and $\alpha$ are the growth and roughness
97: exponents, respectivelly, and $t_{\times}$ is the saturation time.
98: These two behaviors are joined at the Family and Vicsek dynamical scaling
99: relation\cite{fv}
100: \begin{equation}
101: \omega(L,t)\sim L^{\alpha}f\left({t\over L^z}\right),
102: \end{equation}
103: where the function $f(x)$ must be $L$-independent. This function
104: scales as $f(x)\sim x^{\beta}$ for short times and tends to a constant in
105: the steady state. The dynamical exponent $z$ is related with $\alpha$ and
106: $\beta$ through the relationship $z=\alpha/\beta$ and it shows how the
107: saturation time depends on the system size $L$: $ t_{\times}\sim L^z$. Two of
108: these exponents predict which universality class a model belongs to.
109:
110: These universality classes are related to the dominant term of the stochastic
111: differential equation in the continuum limit. In the next section, we describe two
112: stochastic differential equations which represent two distinct universality classes.
113: The first is a linear equation proposed, in 1982, by Edwards and Wilkinson
114: (EW equation) \cite {ew}. The second is a nonlinear equation, introduced by
115: Kardar, Parisi and Zhang in 1986 (KPZ equation) \cite {kpz}. When the
116: nonlinear term of this equation is null, the EW linear equation is recuperated.
117:
118: In this article, we report on simulations of a growth model with
119: restricted relaxation process (called RR model), which was proposed
120: in order to study the crossover between the linear and nonlinear regimes
121: of the KPZ equation. The RR combines
122: features of the Edwards-Wilkinson (EW) model \cite {ew,family86}, related to
123: the EW linear equation, and the
124: Kim-Kosterlitz (KK) model \cite{kk}, related to the KPZ nonlinear equation.
125: The crossover between linear and nonlinear regimes of the
126: KPZ equation was studied numerically in $d=1$ and $d=2$
127: through numerical solutions of this equation where variations of the amount
128: of nonlinearity was allowed. Moser and Kert\'esz \cite{moser}
129: found, in $d=1$, the growth exponent $\beta=1/3$ for all values of the
130: nonlinearity. In $d=2$, the authors have found $\beta=0.240$, close to
131: the Kim and
132: Kosterlitz value \cite {kk}. So, a crossover between linear
133: and nonlinear regimes was not verified. Grossmann, Guo and Grant \cite {ggg}
134: also have obtained numerical solutions of KPZ equation where the surface
135: tension was fixed and the nonlinear parameter was continuously changed.
136: For small systems, the authors have found growth exponents in the interval
137: $1/4< \beta <1/3$. So, the crossover was characterized by
138: a continuum change in the growth exponent, which in $d=1$ means that the
139: dynamical exponent $z$ depends on the amount of the nonlinearity.
140: This continuous crossover in $d=1$ was also
141: verified by simulations on growth models \cite {yks,pelle}. Theoretically,
142: in $d=1$,
143: Nattermann and Tang \cite {nattang} have studied the KPZ equation in the
144: low nonlinear limit through renormalization group analisys and they have
145: obtained a different result: two behaviors separated by a characteristic
146: time $t_c$.
147: For $t\ll t_c$, the linear behavior of roughness was found, while for $t\gg t_c$,
148: the nonlinearity dominates.
149:
150: This article is organized in sections. The next section presents the
151: general approach for kinetic roughening and describes the EW and
152: KK models. In section III, we introduce the RR model and we show our
153: results which confirm the theoretical previsions of
154: Natterman and Tang \cite {nattang} and, in section IV, we
155: finally show our conclusions.
156: %
157: %
158: %
159: \section {THEORY AND DISCRETE MODELS}
160: %
161: %
162: %
163: In the continuum (coarse-grained) description, the interface motion is
164: described through Langevin-like equations [1-3],
165: \begin{equation}
166: {\partial h({\bf x},t)\over\partial t}=v_0+\eta({\bf x},t)+
167: \Phi\left[{h({\bf x},t)}\right]~~.
168: \end{equation}
169: In this equation, $v_0$ and $\eta({\bf x},t)$ are the deposition rate
170: and its noise, respectively.
171: This white noise has zero mean and variance given by
172: \begin{equation}
173: \left<\eta({\bf x},t)\eta({\bf x}^{'},t^{'})\right>=
174: D\delta^{d}({\bf x} -{\bf x}^{'})\delta(t-t^{'}).
175: \end{equation}
176: $\Phi\left[{h({\bf x},t)}\right]$ is the term related to the
177: correlations between neighbors which can have linear and nonlinear functions
178: of $h({\bf x},t)$.
179:
180: Our interest is focused on the SOS model with
181: surface relaxation \cite {ew,family86} and at the SOS model with
182: restriction \cite {kk}.
183: The SOS model with surface relaxation, introduced by Edwards and
184: Wilkinson \cite{ew} in 1982, is a random deposition of particles where
185: the difference of height constraint between the neighbors $\{j\}$ of a site
186: $i$ is given by
187: \begin{equation}
188: h_i-h_{\{j\}}< M,
189: \end{equation}
190: where $M$ is the parameter that controls the roughness. In this work, we
191: always use $M=1$.
192: If the height of the deposited particle on the site $i$
193: does not satisfy the height constraint, this particle must be
194: moved to a local miminum. Family \cite{family86} have obtained for this model,
195: in $d=1$, the exponents $\beta=0.25(1)$ and $\alpha=0.48(2)$. This result indicates
196: that this model, in a coarse-grained limit, belongs to the universality
197: class defined by the EW equation
198: \begin{equation}
199: {\partial h({\bf x},t)\over\partial t}=v_0+\eta({\bf x},t)+
200: \nu\nabla^{2}h({\bf x},t),
201: \end{equation}
202: where
203: the laplacian term is related to the surface relaxation process.
204: The exponents of this linear equation, obtained through Fourier
205: analysis \cite{ew,nattang},
206: are $\beta=1/4$, $\alpha=1/2$ and $z=2$ for $d=1$.
207: For $d=2$, which is the critical dimension of Eq.6,
208: these exponents are $\beta=\alpha=0$ which means that the roughness has
209: logarithimic behavior in space and time ($\omega^2\sim\log t$, for $t\ll L^z$,
210: and $\omega^2_{sat}\sim\log L$, for $t\gg L^z$). Both, the EW model and the
211: EW equation, generate gaussian height distributions as well.
212:
213: In the SOS model with restriction, particles are also randomly deposited
214: onto a substrate and the difference of height constraint is the same of
215: the EW model, Eq.5, but any kind of relaxation is allowed.
216: If the height of a deposited particle does not satisfy Eq.5, this choice
217: must be refused, that is, the particle evaporates.
218: This model was proposed in 1989 by Kim and Kosterlitz \cite{kk}
219: in order to study nonlinear kinetic roughening in high dimensions.
220: They numerically showed, in $d=1$, that this model, named KK model,
221: belongs to the universality class of
222: the well known KPZ equation, proposed by Kardar, Parisi and Zhang \cite {kpz}
223: \begin{equation}
224: {\partial h({\bf x},t)\over\partial t}=v_0+\eta({\bf x},t)+
225: \nu\nabla^{2}h({\bf x},t)+{
226: \lambda\over 2}\left[\nabla h({\bf x},t)\right]^{2}.
227: \end{equation}
228: The appearance of the nonlinear term
229: $\left[\nabla h({\bf x},t)\right]^{2}$ is due to the lateral
230: growth, that is, the dependence of the growth velocity on a local
231: normal of the growing interface,
232: or to the appearance of a perpendicular driven force that leads to a growth
233: velocity
234: greater or smaller than the deposition rate $v_0$. In the case of the
235: KK model, for example, the refuse mechanism makes the growth velocity
236: smaller than the deposition rate.
237: In $d=1$, the exponents of this equation \cite{kpz} are
238: $\beta=1/3$, $\alpha=1/2$ and $z=3/2$. In $d=2$, the analytical solution is
239: not known.
240: In $d=1$, numerical simulations of the KK model \cite {kk} indicate
241: $\beta=0.332(5)$ and the $\alpha$-exponent close to the expected value
242: $(\alpha=1/2)$ and, in $d=2$,
243: $\beta=0.250(5)$ and $\alpha=0.40(1)$.
244:
245: Eq.7 is not invariant under the $h\rightarrow -h$ transformation, which
246: means that the up-down symmetry is broken in surfaces generated by a
247: KPZ process. This fact leads to deviations in the gaussian caracter of
248: the height distributions which can be measured using other moments of
249: the distribution.
250: The Eq.1 can be generalized for any moments of height distribution as
251: \begin{equation}
252: W_q(L,t)=
253: \left<{1\over L^d}\sum_{i=1}^{L^d}(h_i-\bar h)^q\right>~~,
254: \end{equation}
255: where $q$ is the order of the moment. Note that the roughness
256: $\omega(L,t)$ is related to the second moment: $\omega^2(L,t)=W_2(L,t)$.
257: A growing profile has up-down symmetry when positive and negative local
258: curvatures are equals, and, in this case, $W_3(L,t)$ vanishes. On the other
259: hand, when asymmetries are present, $W_{3}(L,t)\neq 0$. The skewness,
260: defined by
261: \begin{equation}
262: S(L,t)={W_3(L,t)\over W_2^{3/2}(L,t)},
263: \end{equation}
264: is the ideal measurement of deviations from gaussian behavior. In the
265: case of the EW model, $S(L,t)=0$ always.
266: For systems in KPZ class in $d=1$, Krug {\it et al.} \cite {krugskew} have indicated
267: $|S(L,t)|\approx 0.28$ as an universal value in the transient state. In the
268: steady state, the profile shows a random-walk character, that is, the height
269: distribution is gaussian and $S_{sat}(L,t\gg t_{\times})=0$.
270: In $d=2$,
271: numerical simulations of the KK model indicate $S\approx -0.40$, in the
272: transient state, and $S_{sat}\approx 0.28$, in the steady state \cite {landau2,skew}.
273: %
274: %
275: %
276: \section {MODEL DESCRIPTION AND RESULTS}
277: %
278: %
279: %
280: In this article, we report on simulations of the growth model with
281: restricted surface relaxation model (RR model). In the EW model, each
282: incoming particle must
283: search the local minima when the height constraint (Eq.5)
284: is not satisfied. We introduce a parameter
285: $s$ that is the number of lattice units allowed for the relaxation process.
286: If the deposited particle does not find the minimum after $s$ relaxations,
287: then this choice must be refused, as in KK model. For $s=0$, the KK model is
288: recuperated and $s\rightarrow\infty$ yields the EW model.
289: %
290: %
291: \subsection {$d=1$ Results}
292: %
293: Figure 1 shows the log-log plot of roughness $\omega(L,t)$ vs time $t$
294: for a system with $L=10^5$ sites and $s=2$. The time unity means $L^d$
295: attempts of deposition.
296: The two straight lines in the figure are showing
297: the power-law fits with $\beta=0.249(1)$ $(1<t<10^3)$ and $\beta=0.332(1)$
298: $(10^3<t<10^5)$. The intersection of these two lines define the crossover time
299: $t_c$. This crossover is easy to understand considering the increase of
300: roughness: in short times, when the roughness is still small, particles do
301: not need to relax very much for seaching local minima, so the linear
302: behavior might dominate.
303: In large times, on the other hand, we note the appearance of relaxation
304: lenghts bigger than those observed in short times. As relaxation processes are
305: linked to refuse processes in this model, the system undergoes a crossover
306: to the nonlinear behavior. The crossover time $t_c$ is independent of the
307: system size $L$ and it is only a function of the parameter $s$.
308:
309: For a better understanding of this crossover we study the
310: statistics of relaxations in the EW model, where relaxations of all
311: sizes can occur.
312: At the time $t$, be
313: $\left<N_{k}(t)\right>$ the mean number of particles which diffused $k$
314: sites searching local minima. So, $k$ relaxations occur with
315: probability
316: \begin {equation}
317: P_k(t)={\left<N_{k}(t)\right>\over L^d}~~.
318: \end {equation}
319: These probabilities have an initial temporal dependence and a
320: steady state $P_k(\infty)$, whose values are shown in Table 1.
321: We note a strong decrease of the probability with the number
322: of sites diffused, which shows the rare ocurrences of
323: the relaxation with large relaxation lenghts.
324:
325: It is interesting to analyse the approach of
326: this probability to the steady state because each $P_k(t)$ has different
327: convergence times. For doing this, we define a normalised
328: probability of $k$ relaxations as
329: \begin {equation}
330: p_k(t)={P_k(t)\over P_k(\infty)}~~~.
331: \end {equation}
332: Figure 2 shows the plots of $p_k(t)$ vs $t$ for $k=0;1;2;3;4$
333: for the EW model with $L=10^5$. Note the differences
334: among the convergences to each steady state: the temporal behaviors of
335: $p_0$ and $p_1$ quickly go to its
336: steady values, while $p_2$ and $p_3$ tend to unity only at $t\approx 10^3$ and
337: $t\approx 10^4$, respectively. This fact suggests that large relaxation
338: lenghts might occur at large deposition times with small probabilities. So, drawing attention
339: to the curve $p_2(t)$ vs $t$, we observe that the saturation occurs
340: at $t\approx 10^3$. In the RR model with $s=2$, we estimate $t_c\approx 10^3$,
341: which indicates that the crossover from linear to nonlinear behaviors of the
342: RR model occurs, for a value of $s$, when $p_{k=s}(t)$ is time independent.
343:
344: This behavior of $p_k$ is responsible for the dependence of the crossover
345: time $t_c$ with the parameter $s$ in the RR model. Figure 3 shows clearly
346: this dependence with the plots of $\omega(L,t)/t^{\beta}$ vs $t$
347: for: (a) $\beta=1/4$ and (b) $\beta=1/3$, with $s=0,2,4$ and $L=10^5$. In (a),
348: the curve with $s=0$ always grows, while the curve for $s=4$ remains constant,
349: indicating that $\beta=1/4$, for this value of $L$, is the correct value
350: for $s=4$ (EW behavior)
351: and a noncorrect for $s=0$ (KK nonlinear behavior). In (b) similar conclusions
352: are obtained with $\beta=1/3$. For $s=2$, the initial linear and the
353: asymptotic KPZ behaviors are well observed in (a) and (b), respectively. Note
354: that the crossover for $s=2$ occurs when the $p_2(t)$ is time independent, that is,
355: when $p_2(t)\approx 1$, in Figure 2. For $s=4$, we observe only the linear behavior
356: because the total deposition time is smaller than the crossover time $t_c$.
357:
358: In order to do a more complete characterization of this crossover, we
359: also analyse the temporal behavior of the skewness $S(L,t)$.
360: Figure 4 shows plots of $S(L,t)$ vs $t$ for $L=10^5$ and several values of
361: the parameter $s$. For $s=0$, the skewness $S(t)$ goes
362: quickly to the KPZ transient value $S=-0.28$. As we have explained, $S\neq 0$
363: means that the interfaces have not up-down symmetry.
364: We find for $s=2$, a slower approach to the KPZ value, than observed for
365: $s=0$, indicating that the up-down symmetry is gradually lost when $s>0$.
366: In particular, for $s=3$, we clearly observe an initial behavior where
367: $S\approx 0$ and an approach faster to the KPZ value at
368: $t\approx 10^4$. The temporal dependence of $S(L,t)$, which is an independent
369: measurement of the universality class, also indicates an initial
370: linear behavior and the KPZ behavior in asymptotic times.
371: %
372: %
373: \subsection {$d=2$ Results}
374: %
375: It is also interesting to study this crossover in $d=2$, because changes in the
376: morphology are expected in this dimension. For the KK model, we have
377: power-law divergencies of roughness ($\omega^2\sim t^{0.50}$ and
378: $\omega^2\sim L^{0.78}$), while the EW model shows logarithmic divergences
379: ($\omega^2\sim\ln t$ and $\omega^2\sim\ln L$.
380:
381: In order to avoid saturation effects, we work with $L=2000$ (4$\times 10^6$ sites),
382: and we do simulations until $t=10^4$. Due to the computational cost,
383: we perform only two samples for each value of the parameter $s$ and,
384: consequently, the data quality in this section is poorer than in the previous
385: subsection. The crossover is analysed through a similar process done
386: in $d=1$.
387: We expect logarithmic behavior when the linear term dominates, so we define
388: \begin{equation}
389: Y_s(t)={\omega^2(t)-B_s\over A_s\ln(t)}~~,
390: \end{equation}
391: where $A_s$ and $B_s$ are the
392: $s$-dependent coefficients obtained through a logarithmic regression in the
393: interval $2<t<100$.
394: In Figure 5, we plot the temporal behavior of the function $Y_s(t)$,
395: in a semi-logarithmic scale.
396: If the temporal behavior of the roughness is
397: logarithmic, $Y_s(t)$ must be a constant, equal to one.
398: We note that this occurs for $s=\infty$, but, for $s=1$, this
399: constant behavior remains until $t\approx 10^2$. After this time, there is
400: a crossover to the power law behavior. However, until the studied time, it is not
401: possible to determine the exponent.
402:
403: In order to verify the existence of a power law behavior with the KPZ
404: exponent $\beta$, we need to analyse values
405: of the parameter $s$ smaller than $s=1$, that is,
406: we need to do continuum variations in $s$. So, we do this by assinging a
407: probability $s$ for the particle relax one lattice unit and a probability
408: $(1-s)$ for the particle be evaporated.
409: In figure 6, we show the graph of $\omega/t^\beta$ vs $t$, for $s=0$ (KK model) and
410: $s=0.1,0.3,0.5$. In this graph, it was used $\beta=1/4$ which is the
411: exponent for the KK model in $d=2$ \cite {kk}.
412: If the temporal behavior of the roughness has a power law behavior with this
413: exponent, the curves must be horizontal.
414: For the RR model ($s\neq 0$), the curves show asymptotic approaches to this nonlinear
415: behavior. For greater values of the parameter $s$, the crossover time to this
416: behavior occurs for times greater than the time studied ($t=10^4$).
417: %
418: %
419: %
420: \section {Conclusions}
421: %
422: %
423: %
424: We have studied the model with restricted surface
425: relaxation which combines the main features of the model with
426: surface relaxation (EW model) and the model with refuse (KK model).
427: A power-law temporal behavior of roughness with two growth exponents
428: was observed in $d=1$: The linear growth exponent, $\beta=1/4$, occurs in short
429: times and the nonlinear one, $\beta=1/3$, appears in the asymptotic limit.
430: This result suggests the following description: the linear
431: term of the KPZ equation (Eq.7) dominates in short times and the nonlinear
432: term dominates in asymptotic times. We also noted that the crossover time
433: $t_c$ is independent of the system
434: size $L$ and it is only function of the parameter $s$.
435: This result corroborates the
436: renormalization group solution made by Nattermann and Tang \cite {nattang},
437: where the KPZ equation with small nonlinear term was considered.
438: In $d=2$, we have found indications of the same kind of crossover:
439: A logarithmic temporal behavior of roughness in short times,
440: which is related to the linear EW equation, and
441: a power-law behavior with $\beta= 0.25$ in asymptotic times, related to
442: the nonlinear KPZ equation.
443: %
444: %
445: %
446: \bigskip
447:
448: \centerline {\bf Acknowledgments}
449: %
450: %
451: %
452: The authors would like to thank Rog\'erio Magalh\~aes Paniago for a critical
453: reading of the manuscript.
454: The simulations were made at an ensemble of Digital Alpha 500 Au of the
455: Departamento de F\'{\i}sica -UFMG and at a Sun HPC 10000 of the CENAPAD
456: MG-CO.
457: This work was supported by CNPq, Fapemig and Finep/Pronex, Brazilian agencies.
458:
459: %
460: %
461: % ----------------------- end --------------------------------
462: %
463: %
464: %
465:
466: \begin {thebibliography}{99}
467: %
468: \bibitem {meakin}
469: P. Meakin,
470: \newblock {\it Fractals, Scaling and Growth Far from Equilibrium},
471: \newblock Cambridge Univ. Press, Cambridge (1998).
472:
473: \bibitem {barab}
474: A.-L. Barab\'asi and H. E. Stanley,
475: \newblock {\it Fractal Concepts in Surface Growth},
476: \newblock Cambridge Univ. Press, Cambridge (1995)
477:
478: \bibitem {krug1}
479: J. Krug,
480: \newblock Adv. Phys. {\bf 46}, 139 (1997).
481:
482: \bibitem {db}
483: M. J. Vold,
484: \newblock J. Phys. Chem. {\bf 64}, 1616 (1960).
485: P. Meakin, P. Ramanlal, L. M. Sander and R. C. Ball,
486: \newblock Phys. Rev. A {\bf 34}, 5091 (1986).
487:
488: \bibitem {eden}
489: M. Eden,
490: \newblock Proc. Fourth Berkeley Symp. Mathematical Statistics and
491: Probability, Volume IV. Edited by L. Le Cam and J. Neyman:
492: Biology and Problems of Health (University of
493: California Press, Berkeley, 1961).
494:
495: \bibitem {ew}
496: S. F. Edwards and D. R. Wilkinson,
497: \newblock Proc. R. Soc. A {\bf 381}, 17 (1982).
498:
499: \bibitem {family86}
500: F. Family,
501: \newblock J. Phys A {\bf 19}, L441 (1986).
502:
503: \bibitem {kk}
504: J. M. Kim and J. M. Kosterlitz,
505: \newblock Phys. Rev. Lett. {\bf 62}, 2289 (1989).
506:
507: \bibitem {wolfvillain}
508: D. E. Wolf and J. Villain,
509: \newblock Europhys. Lett {\bf 13}, 389 (1990).
510:
511: \bibitem {fv}
512: F. Family and T. Vicsek,
513: \newblock J. Phys. A {\bf 18}, L75 (1985).
514:
515: \bibitem {kpz}
516: M. Kardar, G. Parisi and Y.-C. Zhang,
517: \newblock Phys. Rev. Lett. {\bf 56}, 889 (1886).
518:
519: \bibitem {moser}
520: K. Moser, J. Kert\'esz and D. E. Wolf,
521: \newblock Physica A {\bf 178}, 215 (1991).
522:
523: \bibitem {ggg}
524: B. Grossmann, H. Guo and M. Grant,
525: \newblock Phys. Rev. A {\bf 43}, 1727 (1991).
526:
527: \bibitem {yks}
528: H. Yan, D. Kessler and L. M. Sander,
529: \newblock Phys. Rev. Lett. {\bf 64}, 926 (1990).
530:
531: \bibitem {pelle}
532: Y. P. Pellegrini and R. Jullien,
533: \newblock Phys. Rev. Lett {\bf 64}, 1745 (1990);
534: \newblock Phys. Rev. A {\bf 43} 920 (1991).
535:
536: \bibitem {nattang}
537: T. Nattermann and L.-H. Tang,
538: \newblock Phys. Rev. A {\bf 45}, 7156 (1992).
539:
540: \bibitem {krugskew}
541: J. Krug, P. Meakin and T. Halpin-Healy,
542: \newblock Phys. Rev. A {\bf 45}, 638 (1992).
543:
544: \bibitem {landau2}
545: Y. Shim and D. P. Landau,
546: \newblock Phys. Rev. E {\bf 64} No 036110 (2001).
547:
548: \bibitem {skew}
549: C.-S. Chin and M. den Nijs,
550: \newblock Phys. Rev. E {\bf 59}, 2633 (1999);
551: M. Pr\"ahofer and H. Spohn,
552: \newblock Phys. Rev. Lett. {\bf 84}, 4882 (2000);
553:
554: \bibitem {krugmeakin}
555: J. Krug and P. Meakin,
556: \newblock J. Phys. A {\bf 23}, L987 (1990).
557:
558: \bibitem{ew2d}
559: S. Pal and D. P. Landau,
560: \newblock Physica A {\bf 267}, 406 (1999).
561:
562: \end {thebibliography}
563:
564: \newpage
565:
566: {\large \bf Figure captions}
567:
568: \bigskip
569:
570: {\bf Figure 1}
571:
572: The temporal behavior of the roughness $\omega(L,t)$ for the RR model
573: simulated at a substrate with $L=10^5$ sites and $s=2$, in a log-log plot.
574: The two straight lines are showing the
575: power-law fit results with $\beta=1/4$ (short times) and $\beta=1/3$
576: (asymptotic times). The crossover time $t_c$ is defined as the intersection
577: of these two lines, as indicated in the figure.
578:
579: {\bf Figure 2}
580:
581: The temporal behavior of normalised probability of $k$ relaxations
582: $p_k(t)$ for the EW model with $L=10^5$ sites for $k=0;1;2;3;4$, from top to below.
583:
584: {\bf Figure 3}
585:
586: The log-log plots of temporal behaviors of: (a) $\omega/t^{1/4}$
587: and (b)$\omega/t^{1/3}$ for $s=0;2;4$ and $L=10^5$.
588: We might observe, for $s=2$, when the
589: system drives away from the linear behavior in (a) and,
590: the arrival at the nonlinear KPZ behavior in (b). For $s=4$, the
591: system must approach to the KPZ behavior at deposition time
592: greater than $t=10^5$.
593:
594: {\bf Figure 4}
595:
596: The temporal behavior of the skewness $S(L,t)$ for $L=10^5$ for
597: several values of the parameter $s$. The two horizontal
598: lines show the value $S=0$ (EW value) and
599: $S=-0.28$ (KPZ value).
600:
601: {\bf Figure 5}
602:
603: The temporal behavior, in a semi-logarithmic scale, of the function
604: $Y_s(t)$ for $s=1$ and $s=\infty$ (EW) with $L=2000$. The dashed
605: horizontal line indicates the constant value $Y_s(t)=1$.
606:
607: {\bf Figure 6}
608:
609: The semi-logarithmic graph of $\omega(L,t)/t^{1/4}$ in function of
610: the time $t$ for $L=2000$ and some values of the parameter $s$.
611:
612:
613: \newpage
614: \centerline {TABLE 1}
615: \begin{center}
616: \begin{tabular}{lllllll} \hline \hline
617: $k$ &~~~~~~~~$P_k(\infty)$ \\ \hline
618: $0$ &~~~~~~~0.5813(3)\\
619: $1$ &~~~~~~~0.3481(3)\\
620: $2$ &~~~~~~~0.0616(1)\\
621: $3$ &~~~~~~~0.00812(2)\\
622: $4$ &~~~~~~~0.00085(1)\\
623: $5$ &~~~~~~~7(5)$\times 10^{-5}$\\
624: \hline \hline
625: \end {tabular}
626: \end {center}
627: The steady state values of probability of diffusion $k$ sites determined as
628: the mean value in the interval $10^4\leq t\leq 10^5$ for a EW system with
629: $L=10^5$.
630:
631:
632:
633: \end {document}
634:
635:
636: