• Aucun résultat trouvé

The perimeter in site directed percolation. Mean perimeter expansions

N/A
N/A
Protected

Academic year: 2021

Partager "The perimeter in site directed percolation. Mean perimeter expansions"

Copied!
5
0
0

Texte intégral

(1)

HAL Id: jpa-00210290

https://hal.archives-ouvertes.fr/jpa-00210290

Submitted on 1 Jan 1986

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

The perimeter in site directed percolation. Mean perimeter expansions

J.A.M.S. Duarte, H.J. Ruskin

To cite this version:

J.A.M.S. Duarte, H.J. Ruskin. The perimeter in site directed percolation. Mean perimeter expansions.

Journal de Physique, 1986, 47 (6), pp.943-946. �10.1051/jphys:01986004706094300�. �jpa-00210290�

(2)

The perimeter in site directed percolation. Mean perimeter expansions

J. A. M. S. Duarte (*)

Solid State Physics, Imperial College, London SW7 2AZ, U.K.

and H. J. Ruskin (+)

School of Quantitative Methods, N.I.H.E. Glasnevin, Dublin 9, Republic of Ireland

(Reçu le 18 octobre 1985, accepte le 29 janvier 1986)

Résumé.

2014

Nous analysons des séries pour le périmètre moyen et la taille moyenne en percolation dirigée, séries

dont nous avons obtenu des termes nouveaux. Nous obtenons des estimations plus précises pour le seuil de perco- lation critique pc dans le cas des réseaux triangulaires et cubiques. Dans le cas du problème de percolation de site

avec seconds proches voisins sur le réseau carré, nous trouvons pc

=

0,4965 ± 0,002.

Abstract.

2014

Exact expansions for susceptibility

2014

like mean perimeter series for directed percolation and extended

mean size series are analysed, on two and three dimensional lattices. The critical threshold pc estimates are refined for the triangular and simple cubic lattices. On the square next-nearest-neighbour site problem pc is estimated as

pc

=

0.4965 ± 0.002.

Classification

Physics Abstracts

05.50

Introduction.

Studying the critical behaviour of the moments of the cluster size distribution is one of the standard ways of

establishing a set of critical exponents for the percola-

tion transition in any dimension. Consider directed site percolation, where clusters spread from a given origin along preferred orientations of the axes, between sites that are occupied with probability p. In this

special case it is clear that the various moments M ‘( p)

can be expanded in low density series (small values ofp)

from a knowledge of the total numbers of configura-

tions of finite clusters starting from the origin (g sf)

and we have

where s is the size of the cluster, and t (the other essen-

tial characteristic of the cluster) is the perimeter count,

(*) On sabbatical leave from : Laborat6rio de Fisica, Faculdade de Ciencias, Universidade do Porto, 4000 Porto, Portugal.

(+) Work carried out in part at Trinity College Dublin 2, Eire.

i.e. the number of bounding lattice sites along the

allowed orientation of the axes (see Fig. 1. A of

Ref. [10]). The gst are crucial to the determination of such exact series expansions. In this paper we shall

concentrate on the analyses of first moment expan- sions (i

=

1 in Eq. (1)). This gives the average number of sites connected to the origin, or its « mean cluster

size » S(p), at a given p. This quantity diverges near

the percolation threshold Pc with the leading critical exponent y

However, knowledge of the gst also enables alterna- tive calculations of an equivalent « susceptibility >>

-

like quantity, the mean perimeter T(p) by taking

Consider the critical behaviour of the correspond- ing low-density series expansions. From the sum rule

for the occupation of the origin,

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01986004706094300

(3)

944

Fig. 1.

-

Pad6 plot of y versus Pc for the square site problem

from the mean size (2022) and mean perimeter series (+) [i/j ] - j degree of the denominator, i degree of the numera-

tor (the mean size series are not identified for easier reading).

one obtains by taking derivatives w.r.t. p,

and, clearly, near Pc :

Expansions for S(p) have been published and ana- lysed for y in dimensions 2 to 5 (the upper critical dimension for directed percolation) [ l, 2]. They follow

as a by-product from a knowledge of the gst histo- grams given in [3] and in the reference [10]. From equation (6) T(p) must have the same leading ’ diver-

gence associated with the exponent y. Therefore t, the perimeter in the percolation sense, is somehow a

measure of the cluster size and the sums in equation (1) (for i

=

1) and equation (3) are dominated, for large

s, by the same value of the perimeter as shown in the reference [10]. In fact, at fixed size s the average peri-

meter value is proportional to the size with coeffi- cient (1 - Pc)/Pc :

and these contributions dominate in equation (6)

the summations on s near PC.

We reencounter in this context a situation known from undirected percolation and by now well-docu-

mented [4, 5]. The use of alternative measures of cluster size was explored in reference [6] (and also in [7]) to

obtain alternative susceptibility series.

Series expansions for T ( p) are of a comparable length to those for S(p), when these are obtained from the gst

-

they can alternatively be obtained by a

combination of transfer-matrix and graph-theoretical

methods (see [1]). Note that expanding the gst (up to

order s) through order s + 2, then (conf. Eq. (4)),

and

where

so that, using equations (9) (or Eq. (5)), 2 more terms

can be added to S(p) and one more term to T( p), if

gs+2 (the total number of clusters with s + 2 sites) is

known. This is particularly useful for the directed percolation site problems where exact expressions

are known for square, triangular, simple cubic (to

order 17) [8] and body-centred cubic (to order 42)

animals.

We have used this additional data to obtain expres- sions for T( p) and for S(p). These are given in tables I

and II and we describe the results of this numerical

study in the following section.

1.1 PADÉSlUDIESOF Pc AND y.

-

Strong non-physical singularities make ratio methods of the usual type,

or even refined modifications [9] almost useless for the present series. We have therefore formed dlog Pad6 approximants to them and figures 1 and 2 present the situation for the square and simple cubic lattices for

S(p) and bi! T(p). The approximants for the square

mean size are, of course, found in [ 1 ] and the mean perimeter pole locations fall on a smooth line almost

super-imposed on them. Both the y value in 2 dimen-

sions, y

=

2.269 ± 0.018 and the Pc estimates from

series, 0.7055 ± 0.0005 are in agreement with the

present alternative evidence from T(p).

(4)

Table I.

-

Mean size series for the various lattices ((*)

-

added coeffcients). On the b.c.c. all

=

173.737 and a12

=

514.988 (earlier terms in [2]).

Fig. 2.

-

Pad6 plot of y versus p, for the simple cubic site problem from the mean size (2022) and mean perimeter series (+) [ilj] - j degree of the denominator, i degree of the

numerator.

Table II.

-

Mean perimeter series for the various lattices.

(5)

946

For the square next-nearest-neighbour lattice the

much shorter series we have obtained are in excellent agreement with the value for y. The poles for both S(p) and bit T( p) fall once again nearly on the same

line and we estimate

for this problem.

On the triangular lattice the pole patterns tend to

concentrate around values of both y and Pc higher

than the existing estimates. Also, the lines for the two

alternative susceptibilities remain somewhat apart.

We refine the current estimate for Pc to

on the basis of the new longer series.

Of the 2 three-dimensional lattices, it is the simple

cubic which exhibits the most difficult features. i) More

than 13 terms are required for the singularities to

exceed the central estimates for Pc or y (see Fig. 2) but higher order approximants then become concentrated

on this higher region. ii) A more serious gap is appa- rent between both lines of singularities (for S(p) and bit T(p)) than was found to be the case for the trian- gular lattice, although the common artificial exten- sion obtained by forming dlog Pad6s to 1 + T(p) brings the perimeter and mean size singularities closer.

The mean size singularities concentrate around /?c = 0.4355 whereas perimeter singularities favour (with

the present extension) 0.4345. We estimate

from the present evidence. On the b.c.c. lattice the series are only marginally longer than those in [2]

and the singularities for T(p) are less regularly distri-

buted than those for the s.c. case. The Pc value of 0.344 ± 0.004 in [2] is still adequate.

2. Summary.

The singularity analyses for mean perimeter expan-

sions in two and three-dimensional directed percola-

tion were found to be in agreement with a susceptibi- lity - like divergence at the critical threshold. Based

on the extended perimeter polynomials of [3] and a preceding paper [10], series for the mean size have also been extended (by a significant number of terms

for the triangular and simple cubic site problems).

The leading divergence in both types is, of course, the same.

Acknowledgments.

We are indebted to D. Dhar for information (on the

b.c.c. numbers) J. Adler for discussions and to the B.

Council and the Gulbenkian Foundation for funding.

One of us (J. A. M. S. D.) thanks N. Rivier and the Solid State Group at I.C. for their kind hospitality.

References

[1] DE BELL, K., ESSAM, J. W., J. Phys. A 16 (1983) 385.

[2] DE BELL, K., ESSAM, J. W., J. Phys. A 16 (1983) 3553.

[3] DUARTE, J. A. M. S., Portugal. Phys. 15 (1984) 119.

[4] STAUFFER, D., Phys. Rep. 54 (1979) 1.

[5] STAUFFER, D., Introduction to Percolation Theory (London, ed. Taylor and Francis) 1985.

[6] DUARTE, J. A. M. S., RUSKIN, H. J., Zeit. Phys. B 46 (1982) 225.

[7] PRIVMAN, V., VAGNER, I. D., Zeit. Phys. B 50 (1983) 353.

[8] DHAR, D., Phys. Rev. Lett. 51 (1983) 853.

[9] PEARCE, C. J., Adv. Phys. 27 (1978) 89.

[10] DUARTE, J. A. M. S., J. Physique (Paris) submitted

for publication.

Références

Documents relatifs

square site problem is given in table II.. - Final section of perimeter/size ratio esti- mates for square site animals.. the first correction to the

2014 Exact expansions for susceptibility 2014 like mean perimeter series for directed percolation and extended.. mean size series are analysed, on two and three

One identity relates the values associated with diagrams for the spin-spin correlation function to the values associated with diagrams for the free energy.. The

We notice that the fact that equality in (10) implies that u is one- dimensional is a specific feature of the Gaussian setting, and the analogous statement does not hold for the

These features lead to the design of a topology optimization algorithm suitable for perimeter dependent objective functionals.. Some numerical results on academic problems

Abstract For the symmetric two-state vertex model on the honeycomb lattice we construct a series expansion of the free energy which, at finite order, depends on free gauge

Abstract: Some structural break techniques defined in the time and frequency domains are presented to explore, at the same time, the empirical evidence of the mean and covariance

In this paper we consider a Lagrange Multiplier-type test (LM) to detect change in the mean of time series with heteroskedasticity of unknown form.. We derive the limiting