THE METHOD OF MINIMAL REPRESENTATIONS IN 2D ISING-MODEL CALCULATIONS

被引：1

作者：

PARLETT, BN ^{[1
]}

HENG, WL ^{[1
]}

机构：

[1] UNIV CALIF BERKELEY,DEPT EECS,DIV COMP SCI,BERKELEY,CA 94720

来源：

JOURNAL OF COMPUTATIONAL PHYSICS | 1994年 / 114卷 / 02期

关键词：

D O I：

10.1006/jcph.1994.1164

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

We present a new method for approximating the partition function of 2D Ising models using a transfer matrix of order 2n. For n = 30 our current program took about 20 s on a Sparc station to obtain four correct decimals in the top two eigenvalues and 5 min for six correct decimals. Eigenvectors were computed at the same time. The temperature was within 3% of critical. The main idea is to force certain entries in vectors to have the same values and to find the crudest representation of this type that delivers the required accuracy. At no time does our program work with vectors with 2n entries. (C) 1994 Academic Press, Inc.

引用

页码：257 / 264

页数：8