Physical Review Letters

ScalingAnsatzfor Swendsen-Wang Dynamics

W. Klein, Tane S. Ray, Pablo Tamayo

Published Jan 9, 1989• 37 citations

Abstract

Using cluster and percolation concepts, we obtain a relation between relaxation times at critical points obtained from a new dynamical model introduced by Swendsen and Wang and those obtained from Glauber dynamics. This relation, which involves only static critical exponents, provides new insight into the physics underlying cluster acceleration methods such as the one introduced by Swendsen and Wang.

DOI: 10.1103/physrevlett.62.163

Related papers

Fixed-cluster acceleration algorithm for spin systems

Tanniemola B. Liverpool, Sharon C. Glotzer

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topicsMay 1, 199615 citations

Dynamic exponents for potts model cluster algorithms

Paul David Coddington, CLIVE F. BAILLIE

We have studied the Swendsen-Wang and Wolff cluster update algorithms for the Ising model in 2, 3 and 4 dimensions. The data indicate simple relations between the specific heat and the Wolff autocorrelations, and between the magnetization and the Swendsen-Wang autocorrelations. This implies that the dynamic critical exponents are related to the static exponents of the Ising model. We also investigate the possibility of similar relationships for the Q-state Potts model.

Nuclear Physics B - Proceedings SupplementsJan 1, 19921 citation

Empirical relations between static and dynamic exponents for Ising model cluster algorithms

Paul David Coddington, CLIVE F. BAILLIE

Physical Review LettersFeb 17, 199245 citations

Mean-field study of the Swendsen-Wang dynamics

Tane S. Ray, Pablo Tamayo, W. Klein

Physical Review AJun 1, 198944 citations

Dynamic critical exponents for Swendsen–Wang and Wolff algorithms obtained by a nonequilibrium relaxation method

Jianqing Du, Bo Zheng, Jian‐Sheng Wang

arXiv:cond-mat/0603038Journal of Statistical Mechanics Theory and ExperimentMay 11, 200614 citations

Theory for the dynamics of clusters near the critical point. I. Relaxation of the Glauber kinetic Ising model

Kurt Binder, Dietrich Stauffer, Heiner Muller-Krumbhaar

Physical review. B, Solid stateDec 1, 197598 citations

Physical Review Letters

ScalingAnsatzfor Swendsen-Wang Dynamics

W. Klein, Tane S. Ray, Pablo Tamayo

Published Jan 9, 1989• 37 citations

Publisher

Abstract

DOI: 10.1103/physrevlett.62.163

Related papers

Fixed-cluster acceleration algorithm for spin systems

Tanniemola B. Liverpool, Sharon C. Glotzer

A recently introduced cluster acceleration algorithm for spin models [J. Machta et al., Phys. Rev. Lett. 75, 2792 (1995)] based on invasion percolation is generalized to temperatures away from the critical point using concepts of random bond percolation. The generalized algorithm is equivalent to an "optimized" version of the Swendsen-Wang cluster algorithm [R. H. Swendsen and J. S. Wang, Phys. Rev. Lett. 58, 86 (1987)], and we demonstrate its success and speed at all temperatures for the $d=2$ Ising model. We argue that the previously discussed connection to self-organized criticality may be viewed as one limit of the fixed-cluster algorithm.A recently introduced cluster acceleration algorithm for spin models [J. Machta et al., Phys. Rev. Lett. 75, 2792 (1995)] based on invasion percolation is generalized to temperatures away from the critical point using concepts of random bond percolation. The generalized algorithm is equivalent to an "optimized" version of the Swendsen-Wang cluster algorithm [R. H. Swendsen and J. S. Wang, Phys. Rev. Lett. 58, 86 (1987)], and we demonstrate its success and speed at all temperatures for the $d=2$ Ising model. We argue that the previously discussed connection to self-organized criticality may be viewed as one limit of the fixed-cluster algorithm.

Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topicsMay 1, 199615 citations

Dynamic exponents for potts model cluster algorithms

Paul David Coddington, CLIVE F. BAILLIE

Nuclear Physics B - Proceedings SupplementsJan 1, 19921 citation

Empirical relations between static and dynamic exponents for Ising model cluster algorithms

Paul David Coddington, CLIVE F. BAILLIE

We have measured the autocorrelations for the Swendsen-Wang and the Wolff cluster update algorithms for the Ising model in two, three, and four dimensions. The data for the Wolff algorithm suggest that the autocorrelations are linearly related to the specific heat, in which case the dynamic critical exponent is ${\mathit{z}}_{\mathrm{i}\mathrm{n}\mathrm{t},}$ ${\mathit{E}}^{\mathrm{W}}$ =\ensuremath{\alpha}/\ensuremath{\nu}. For the Swendsen-Wang algorithm, scaling the autocorrelations by the average maximum cluster size gives either a constant or a logarithm, which implies that ${\mathit{z}}_{\mathrm{i}\mathrm{n}\mathrm{t},}$ ${\mathit{E}}^{\mathrm{SW}}$ =\ensuremath{\beta}/\ensuremath{\nu} for the Ising model.We have measured the autocorrelations for the Swendsen-Wang and the Wolff cluster update algorithms for the Ising model in two, three, and four dimensions. The data for the Wolff algorithm suggest that the autocorrelations are linearly related to the specific heat, in which case the dynamic critical exponent is ${\mathit{z}}_{\mathrm{i}\mathrm{n}\mathrm{t},}$${\mathit{E}}^{\mathrm{W}}$=\ensuremath{\alpha}/\ensuremath{\nu}. For the Swendsen-Wang algorithm, scaling the autocorrelations by the average maximum cluster size gives either a constant or a logarithm, which implies that ${\mathit{z}}_{\mathrm{i}\mathrm{n}\mathrm{t},}$${\mathit{E}}^{\mathrm{SW}}$=\ensuremath{\beta}/\ensuremath{\nu} for the Ising model.

Physical Review LettersFeb 17, 199245 citations

Mean-field study of the Swendsen-Wang dynamics

Tane S. Ray, Pablo Tamayo, W. Klein

Nonuniversal dynamics introduced by Swendsen and Wang [Phys. Rev. Lett. 58, 86 (1987)] is studied in a mean-field Ising model. Numerical simulation data indicate that z, the dynamical critical exponent, is 0.98\ifmmode\pm\else\textpm\fi{}0.08. A modified version of this dynamics in which only large clusters are flipped apparently gives the same exponent, although the amplitude is reduced by a factor of 1/2. In addition, a theoretical argument is presented which gives z=1 and provides insight into the dynamics.Nonuniversal dynamics introduced by Swendsen and Wang [Phys. Rev. Lett. 58, 86 (1987)] is studied in a mean-field Ising model. Numerical simulation data indicate that z, the dynamical critical exponent, is 0.98\ifmmode\pm\else\textpm\fi{}0.08. A modified version of this dynamics in which only large clusters are flipped apparently gives the same exponent, although the amplitude is reduced by a factor of 1/2. In addition, a theoretical argument is presented which gives z=1 and provides insight into the dynamics.

Physical Review AJun 1, 198944 citations

Dynamic critical exponents for Swendsen–Wang and Wolff algorithms obtained by a nonequilibrium relaxation method

Jianqing Du, Bo Zheng, Jian‐Sheng Wang

Using a nonequilibrium relaxation method, we calculate the dynamic critical exponent z of the two-dimensional Ising model for the Swendsen–Wang and Wolff algorithms. We examine dynamic relaxation processes following a quench from a disordered or an ordered initial state to the critical temperature T c , and measure the exponential relaxation time of the system energy. For the Swendsen–Wang algorithm with an ordered or a disordered initial state, and for the Wolff algorithm with an ordered initial state, the exponential relaxation time fits well to a logarithmic size dependence up to a lattice size L = 8192. For the Wolff algorithm with a disordered initial state, we obtain an effective dynamic exponent z exp = 1.19(2) up to L = 2048. For comparison, we also compute the effective dynamic exponents through the integrated correlation times. In addition, an exact result of the Swendsen–Wang dynamic spectrum of a one-dimensional Ising chain is derived.Using a nonequilibrium relaxation method, we calculate the dynamic critical exponent z of the two-dimensional Ising model for the Swendsen–Wang and Wolff algorithms. We examine dynamic relaxation processes following a quench from a disordered or an ordered initial state to the critical temperature T c , and measure the exponential relaxation time of the system energy. For the Swendsen–Wang algorithm with an ordered or a disordered initial state, and for the Wolff algorithm with an ordered initial state, the exponential relaxation time fits well to a logarithmic size dependence up to a lattice size L = 8192. For the Wolff algorithm with a disordered initial state, we obtain an effective dynamic exponent z exp = 1.19(2) up to L = 2048. For comparison, we also compute the effective dynamic exponents through the integrated correlation times. In addition, an exact result of the Swendsen–Wang dynamic spectrum of a one-dimensional Ising chain is derived.

arXiv:cond-mat/0603038Journal of Statistical Mechanics Theory and ExperimentMay 11, 200614 citations

Theory for the dynamics of clusters near the critical point. I. Relaxation of the Glauber kinetic Ising model

Kurt Binder, Dietrich Stauffer, Heiner Muller-Krumbhaar

A semiphenomenological cluster theory is developed for dynamic critical properties, which is not limited to small deviations from equilibrium. Explicit numerical expressions are derived for linear and nonlinear response functions of the kinetic Ising model, which are compatible with dynamic scaling and recent calculations of dynamic critical exponents. An uniaxial magnet (or Ising system) and its critical fluctuations are approximated by a "droplet" or "cluster" model, the critical part of the free energy being the sum of the contributions of clusters each containing $l$ spins. These clusters are assumed to grow and shrink with a phenomenological rate $\ensuremath{\propto}{l}^{r}$, $r\ensuremath{\approx}1$. These rates are compatible with direct Monte Carlo simulations in two and three dimensions. The resulting cluster reaction and diffusion equation is an approximation to the master equation of the Glauber kinetic Ising model (in which the magnetization is not conserved), and has the same structure as used in nucleation theories. In linear response to space- and time-dependent fields the relaxation times of energy and magnetization are expressed as triple integrals. All relaxation times diverge as ${|1\ensuremath{-}\frac{T}{{T}_{c}}|}^{\ensuremath{-}(2\ensuremath{-}r)\ensuremath{\beta}\ensuremath{\delta}}$. This treatment is consistent with dynamic scaling and universality, and is more general than the conventional (Van Hove) theory of critical slowing down. Also the (smaller) exponents found for the response to localized variations ("autocorrelation functions") agree with dynamic scaling. The wave-vector dependence of the relaxation times arises from the static correlation function only, and not from the cluster diffusion term. If the equilibrium cluster distribution is assumed to be that of the Fisher droplet model, and for special values of $r$ , the (discrete) eigenvalue spectrum and eigenfunctions (generalized Laguerre polynomials) of the cluster reaction equation are found explicitly. Relaxation functions and frequency-dependent susceptibilities are expressed by hypergeometric series, even in the case of nonlinear response. At the critical point, the spectrum becomes continuous and the exponential decay for large times is replaced by power-law behavior. Similarly, it is found that dynamic scaling applies also to the nucleation rate for this model, and the scaling exponent of the nucleation rate is $2\ensuremath{-}\ensuremath{\alpha}+(2\ensuremath{-}r)\ensuremath{\beta}\ensuremath{\delta}$.A semiphenomenological cluster theory is developed for dynamic critical properties, which is not limited to small deviations from equilibrium. Explicit numerical expressions are derived for linear and nonlinear response functions of the kinetic Ising model, which are compatible with dynamic scaling and recent calculations of dynamic critical exponents. An uniaxial magnet (or Ising system) and its critical fluctuations are approximated by a "droplet" or "cluster" model, the critical part of the free energy being the sum of the contributions of clusters each containing $l$ spins. These clusters are assumed to grow and shrink with a phenomenological rate $\ensuremath{\propto}{l}^{r}$, $r\ensuremath{\approx}1$. These rates are compatible with direct Monte Carlo simulations in two and three dimensions. The resulting cluster reaction and diffusion equation is an approximation to the master equation of the Glauber kinetic Ising model (in which the magnetization is not conserved), and has the same structure as used in nucleation theories. In linear response to space- and time-dependent fields the relaxation times of energy and magnetization are expressed as triple integrals. All relaxation times diverge as ${|1\ensuremath{-}\frac{T}{{T}_{c}}|}^{\ensuremath{-}(2\ensuremath{-}r)\ensuremath{\beta}\ensuremath{\delta}}$. This treatment is consistent with dynamic scaling and universality, and is more general than the conventional (Van Hove) theory of critical slowing down. Also the (smaller) exponents found for the response to localized variations ("autocorrelation functions") agree with dynamic scaling. The wave-vector dependence of the relaxation times arises from the static correlation function only, and not from the cluster diffusion term. If the equilibrium cluster distribution is assumed to be that of the Fisher droplet model, and for special values of $r$, the (discrete) eigenvalue spectrum and eigenfunctions (generalized Laguerre polynomials) of the cluster reaction equation are found explicitly. Relaxation functions and frequency-dependent susceptibilities are expressed by hypergeometric series, even in the case of nonlinear response. At the critical point, the spectrum becomes continuous and the exponential decay for large times is replaced by power-law behavior. Similarly, it is found that dynamic scaling applies also to the nucleation rate for this model, and the scaling exponent of the nucleation rate is $2\ensuremath{-}\ensuremath{\alpha}+(2\ensuremath{-}r)\ensuremath{\beta}\ensuremath{\delta}$.

Physical review. B, Solid stateDec 1, 197598 citations

d=2

Ising model. We argue that the previously discussed connection to self-organized criticality may be viewed as one limit of the fixed-cluster algorithm.

{\mathit{z}}_{\mathrm{i}\mathrm{n}\mathrm{t},}

{\mathit{E}}^{\mathrm{W}}

=\ensuremath{\alpha}/\ensuremath{\nu}. For the Swendsen-Wang algorithm, scaling the autocorrelations by the average maximum cluster size gives either a constant or a logarithm, which implies that

{\mathit{z}}_{\mathrm{i}\mathrm{n}\mathrm{t},}

{\mathit{E}}^{\mathrm{SW}}

=\ensuremath{\beta}/\ensuremath{\nu} for the Ising model.

l

spins. These clusters are assumed to grow and shrink with a phenomenological rate $\ensuremath{\propto}{l}^{r}$, $r\ensuremath{\approx}1$. These rates are compatible with direct Monte Carlo simulations in two and three dimensions. The resulting cluster reaction and diffusion equation is an approximation to the master equation of the Glauber kinetic Ising model (in which the magnetization is not conserved), and has the same structure as used in nucleation theories. In linear response to space- and time-dependent fields the relaxation times of energy and magnetization are expressed as triple integrals. All relaxation times diverge as ${|1\ensuremath{-}\frac{T}{{T}_{c}}|}^{\ensuremath{-}(2\ensuremath{-}r)\ensuremath{\beta}\ensuremath{\delta}}$. This treatment is consistent with dynamic scaling and universality, and is more general than the conventional (Van Hove) theory of critical slowing down. Also the (smaller) exponents found for the response to localized variations ("autocorrelation functions") agree with dynamic scaling. The wave-vector dependence of the relaxation times arises from the static correlation function only, and not from the cluster diffusion term. If the equilibrium cluster distribution is assumed to be that of the Fisher droplet model, and for special values of

r

, the (discrete) eigenvalue spectrum and eigenfunctions (generalized Laguerre polynomials) of the cluster reaction equation are found explicitly. Relaxation functions and frequency-dependent susceptibilities are expressed by hypergeometric series, even in the case of nonlinear response. At the critical point, the spectrum becomes continuous and the exponential decay for large times is replaced by power-law behavior. Similarly, it is found that dynamic scaling applies also to the nucleation rate for this model, and the scaling exponent of the nucleation rate is $2\ensuremath{-}\ensuremath{\alpha}+(2\ensuremath{-}r)\ensuremath{\beta}\ensuremath{\delta}$.