量子电子学报, 2014, 31 (1): 80, 网络出版: 2014-02-26
关于海森堡模型中一种并行算法实现的讨论
Discussion on implementation of a parallel algorithm about Heisenberg model
摘要
使用并行算法(简称Z分法)Fortran编程计算获取海森堡模型位型[N,k] (N为海森堡链总格点数, k为格点中自旋向上的电子数)最小本征值的最短时间。使用置换群方法产生模型的能量矩阵, 将能量矩阵对角化所得到的本征值构成数据群,采用Z(Z=1,2…)分法Fortran编程计算获得群中 最小数据的最短(或最长)时间。结果表明:同一位型[N,k], 使用2分法获取模型位型[N,k]最小 本征值的时间最长,而不等分或满等分(此时Z=1或位型[N,k]的矩阵维数)时的时间最短且二者相等;对于不同位型[N,k], 当N(k)同, k(N)增大且Z相同时,获取模型最小本征值的最短时间增加。
Abstract
The shortest time of the minimum eigenvalue of [N,k] of Heisenberg model (N is the total number of sites of Heisenberg chain , k is the number of electrons at site spin up) were obtained using parallel algorithm (Z equisection method, Z is from 1 to a, a is the number of the eigenvalue of [N,k]) in Fortran program. The energy matrix of [N,k] was produced by permutation group. The eigenvalues were obtained by diagonalling the energy matrix. The shortest (or longest) time of the minimum eigenvalue of [N,k] was obtained from the data group being made up of the eigenvalues using Z equisection method. The results show that the time is the shortest and same when Z is 1 or the number of the eigenvalue of [N,k]. When N(k) are same, k(N) increases and Z is same, the time acquisiting the minimum eigenvalue of [N,k] increases.
黄敏, 韩文娟, 刘海. 关于海森堡模型中一种并行算法实现的讨论[J]. 量子电子学报, 2014, 31(1): 80. HUANG Min, HAN Wen-juan, LIU Hai. Discussion on implementation of a parallel algorithm about Heisenberg model[J]. Chinese Journal of Quantum Electronics, 2014, 31(1): 80.