skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Efficient implementation of the Lanczos method for magnetic systems

Abstract

Numerically exact investigations of interacting spin systems provide a major tool for an understanding of their magnetic properties. For medium size systems the approximate Lanczos diagonalization is the most common method. In this article we suggest two improvements: efficient basis coding in subspaces and simple restructuring for openMP parallelization.

Authors:
 [1]; ;  [2]
  1. Universitaet Bielefeld, Fakultaet fuer Physik, Postfach 100131, D-33501 Bielefeld (Germany), E-mail: jschnack@uni-bielefeld.de
  2. Universitaet Osnabrueck, Fachbereich Physik, D-49069 Osnabrueck (Germany)
Publication Date:
OSTI Identifier:
21159424
Resource Type:
Journal Article
Resource Relation:
Journal Name: Journal of Computational Physics; Journal Volume: 227; Journal Issue: 9; Other Information: DOI: 10.1016/j.jcp.2008.01.027; PII: S0021-9991(08)00033-8; Copyright (c) 2008 Elsevier Science B.V., Amsterdam, The Netherlands, All rights reserved; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; APPROXIMATIONS; IMPLEMENTATION; MAGNETIC PROPERTIES; MATHEMATICAL SPACE; SPIN

Citation Formats

Schnack, Juergen, Hage, Peter, and Schmidt, Heinz-Juergen. Efficient implementation of the Lanczos method for magnetic systems. United States: N. p., 2008. Web. doi:10.1016/j.jcp.2008.01.027.
Schnack, Juergen, Hage, Peter, & Schmidt, Heinz-Juergen. Efficient implementation of the Lanczos method for magnetic systems. United States. doi:10.1016/j.jcp.2008.01.027.
Schnack, Juergen, Hage, Peter, and Schmidt, Heinz-Juergen. 2008. "Efficient implementation of the Lanczos method for magnetic systems". United States. doi:10.1016/j.jcp.2008.01.027.
@article{osti_21159424,
title = {Efficient implementation of the Lanczos method for magnetic systems},
author = {Schnack, Juergen and Hage, Peter and Schmidt, Heinz-Juergen},
abstractNote = {Numerically exact investigations of interacting spin systems provide a major tool for an understanding of their magnetic properties. For medium size systems the approximate Lanczos diagonalization is the most common method. In this article we suggest two improvements: efficient basis coding in subspaces and simple restructuring for openMP parallelization.},
doi = {10.1016/j.jcp.2008.01.027},
journal = {Journal of Computational Physics},
number = 9,
volume = 227,
place = {United States},
year = 2008,
month = 4
}
  • A grid method for obtaining eigensolutions of bound systems is presented. In this, the block-Lanczos method is applied to a Chebyshev approximation of exp(- H/{Delta}), where {Delta} is the range of eigenvalues we are interested in. With this choice a preferential convergence of the eigenvectors; corresponding to low-lying eigenvalues of H is achieved. The method is used to solve a variety of one-, two-, and three-dimensional problems. To apply the kinetic energy operator we use the fast sine transform instead of the fast Fourier transform, thus fulfilling, a priori, the box boundary conditions. We further extend the Chebyshev approximation tomore » treat general functions of -matrices, thus allowing its application to cases for which no analytical expressions of the expansion coefficients are available. 23 refs., 2 figs., 6 tabs.« less
  • This paper analyzes a few methods based on the Lanczos algorithm for solving large sparse symmetric linear systems with several right-hand sides. The methods examined are suitable for the situation when the right sides are not too different from one another, as is often the case in time-dependent or parameter-dependent problems. We propose a theoretical error bound for the approximation obtained from a projection process onto a Krylov subspace generated from processing a previous right-hand side.
  • Let L be a real linear operator with a positive definite symmetric part M. In certain applications a number of problems of the form Mv = g can be solved with less human or computational effort than the original equation Lu = f. An iterative Lanczos method, which requires no information a priori on the spectrum of the operators, is derived for such problems. The convergence of the method is established with the assumption only that M/sup -1/L is bounded. If M/sup -1/L differs from the identity mapping by a compact operator, the convergence is shown to be superlinear. Themore » method is particularly well suited for large sparse systems arising from elliptic problems. Results from a series of numerical experiments are presented. They indicate that the method is numerically stable and that the number of iterations can be accurately predicted by the error estimate. 1 figure, 1 table.« less
  • Let L be a real linear operator with a positive definite symmetric part, M. In certain applications a number of problems of the form MV = g can be solved with less human or computational effort than the original equation Lu = f. An iterative Lanczos method, which requires no a priori information on the spectrum of the operators, is derived for such problems. The convergence of the method is established with the assumption only that M-/sup 1/L is bounded. If M-/sup 1/L differs from the identity mapping by a compact operator, the convergence is shown to be superlinear. Themore » method is particularly well suited for large sparse systems arising from elliptic problems. Results from a series of numerical experiments are presented. They indicate that the method is numerically stable and that the number of iterations can be accurately predicted by that error estimate. 1 figure, 1 table.« less
  • An implementation of the Lanczos algorithm for the exact diagonalization of the two dimensional Hubbard model on a 4[times]4 square lattice on the Connection Machine CM-2 system is described. The CM-2 is a massively parallel machine with distributed memory. The program is written in C/PARIS. This implementation minimizes memory usage by generating the matrix elements as needed instead of storing them. The Lanczos vectors are stored across the local memory of the processors. Using translational symmetry only, the dimension of the Hilbert space at half filling is more than 10 million. A speed of about 2.4 min per iteration ismore » achieved on a 64K CM-2. This implementation is scalable. Running it on a bigger machine with more processors speeds up the process. The performance analysis of this implementation is shown and discuss its advantages and disadvantages are discussed.« less