Cluster algorithms: Difference between revisions

From SklogWiki
Jump to navigation Jump to search
m (→‎Geometric cluster algorithms: Added a recent publication)
 
(45 intermediate revisions by 5 users not shown)
Line 1: Line 1:
'''Cluster algorithms''' are mainly used in the simulation of [[Ising Models|Ising-like models]]. The essential feature is the use of collective motions
'''Cluster algorithms''' are mainly used in the simulation of [[Ising Models|Ising-like models]] using [[Monte Carlo|Monte Carlo]] methods. The essential feature is the use of collective motions of particles (spins) in a single [[Monte Carlo]] step.
of ''particles (spins)'' in a single Monte Carlo step.
An interesting property of some of these applications is the fact that the [[percolation analysis]] of the clusters can
An interesting property of some of these application is the fact that the [[percolation analysis]] of the clusters can
be used to study [[phase transitions]].
be used to study phase transitions.
== Swendsen-Wang algorithm ==
== Swendsen-Wang algorithm ==
As an introductory example we shall discuss the Swendsen-Wang technique (Ref 1) in the simulation of [[Ising Models]].
As an introductory example to the Swendsen-Wang algorithm we shall discuss the technique
=== Recipe ===
<ref>[http://dx.doi.org/10.1103/PhysRevLett.58.86  Robert H. Swendsen and Jian-Sheng Wang, "Nonuniversal critical dynamics in Monte Carlo simulations", Physical Review Letters '''58''' pp. 86-88 (1987)] </ref>
In one [[Monte Carlo]] step of the algorithm the following recipe is used:
in the simulation of the
[[Ising Models |Ising model]]. In one [[Monte Carlo]] step of the algorithm the following recipe is used:


* Consider every pair interacting sites (spins)
# Consider every pair of interacting sites (spins). In the current configuration the [[Intermolecular pair potential |pair interaction]] can be either negative: <math> \Phi_{ij}/k_B T= -K  </math> or positive <math> \Phi_{ij}/k_B T = + K </math>, depending on the product: <math> S_{i} S_{j} </math> (See the [[Ising Models |Ising model]] entry for notation).
 
#For pairs of interacting sites (i.e. nearest neighbours) with <math> \Phi_{ij}/k_B T < 0 </math>, [[random numbers |randomly]] create a bond between the two spins with a given probability <math> p </math>, where <math> p </math> will be chosen to be a function of <math> K </math>.
In the current configuration the [[Intermolecular pair potential |pair interaction]] can be either negative: <math> \Phi_{ij}/k_B T= -K  </math> or positive <math> \Phi_{ij}/k_B T = + K </math>,  
#The bonds generated in the previous step are used to build up clusters of sites (spins).
depending on the product: <math> S_{i} S_{j} </math> (See [[Ising Models]] for details on the notation)
#Build up the partition of the system in the corresponding clusters of spins. In each cluster all the spins will have the same state, either <math> S = 1 </math> or <math> S = -1 </math>.
 
#For each cluster, independently, choose at random with equal probabilities whether or not to flip (invert the value of <math> S </math>) the whole set of spins belonging to the cluster. The bonding probability <math> p </math> is given by: <math> p = 1 - \exp [ -2 K ] </math>.
* For pairs of interacting sites (nearest neighbors) with <math> \Phi_{ij}/k_B T < 0 </math> create a bond between the two spins with a given probability <math> p </math> (using [[random numbers]])
 
: <math> p </math> will be chosen to be a function of <math> K </math>  
* The bonds generated in the previous step are used to build up clusters of sites (spins).
 
* Build up the partition of the system in the corresponding clusters of spins.
 
In each cluster all the spins will have the same state (either <math> S = 1 </math> or <math> S = -1 </math>)
 
* For each cluster, independently, choose at random with equal probabilities whether to flip (invert the value of <math> S </math>) or not to flip the whole set of spins belonging to the cluster.
 
 
The bonding probability <math> p </math> is given by:
 
: <math> p = 1 - \exp [ -2 K ] </math>


== Wolff algorithm ==
== Wolff algorithm ==
See Ref 2 for details.
The procedure to create a given bond is the same as in the Swendsen-Wang algorithm. However in Wolff's method
The procedure to create a given bond is the same as in the Swendsen-Wang algorithm. However in Wolff's method
the whole set of interacting pairs is not tested to generate (possible) bonds. In stead, a single cluster
the whole set of interacting pairs is not tested to generate (possible) bonds. Instead, a single cluster
is built.  
is built. See
 
<ref>[http://dx.doi.org/10.1103/PhysRevLett.62.361 Ulli Wolff, "Collective Monte Carlo Updating for Spin Systems" , Physical Review Letters '''62''' pp. 361-364 (1989)]</ref>
* The initial cluster contains one site (selected at random)
for details.  
 
* Possible bonds between the initial site and  other sites of the system are tested:
 
The bonded sites are included in the cluster
 
* Then recursively, one checks the existence of bonds between the new members of the cluster and sites of the system to add, if bonds are generated,  new sites to the ''growing'' cluster, until no more bonds are generated.


* At this point, the whole cluster is flipped (see above)
#The initial cluster contains one site, which is selected at random.
#Possible bonds between the initial site and  other sites of the system are tested. Bonded sites are included in the cluster.
#Recursively, one checks the existence of bonds between the new members of the cluster and sites of the system to add, if bonds are generated,  new sites to the ''growing'' cluster, until no more bonds are generated.
#At this point, the whole cluster is flipped (see above).


== Invaded Cluster Algorithm ==
== Invaded Cluster Algorithm ==
(See Ref 3)
The purpose of this algorithm is to locate [[critical points]] (i.e. the critical temperature). So, in this case
 
the probability of bonding neighbouring sites with equal spins is not set ''a priori'' (See
The purpose of this algorithm is to locate [[critical points]] (critical temperature). So, in this case
<ref>[http://dx.doi.org/10.1103/PhysRevLett.75.2792    J. Machta, Y. S. Choi, A. Lucke,  T. Schweizer, and L. V. Chayes, "Invaded Cluster Algorithm for Equilibrium Critical Points", Physical Review Letters '''75''' pp. 2792-2795 (1995)]</ref>).
the probability of bonding neighboring sites with equal spins is not set ''a priori''.
 
The algorithm for an Ising system with [[periodic boundary conditions]] can be implemented as follows:
The algorithm for an Ising system with [[periodic boundary conditions]] can be implemented as follows:


Given a certain configuration of the system:
Given a certain configuration of the system:


* We consider the possible bonds on the system (pairs of nearest neighbours with favourable interaction)
#One considers the possible bonds in the system (pairs of nearest neighbours with favourable interaction).
#One assigns a [[random numbers |random]]  order to these possible bonds.
#The possible bonds are  ''activated''  in the order fixed in the previous step (the cluster structure is watched during this process).
#The bond activation stops when one cluster percolates through the entire system (i.e. considering the periodic boundary conditions the cluster becomes of infinite size).
#Every cluster is then is flipped with  probability 1/2, as in the Swendsen-Wang algorithm.
#An effective bond probability for the percolation threshold, <math> p_{per} </math> can be computed as <math> p_{per} = M_{act}/M </math> with <math> M_{act}</math> being the number of activated bonds when the first cluster percolates, and <math> M </math> is the number of possible bonds.
#The value of <math> p_{per} </math> (in one realisation, or the averaged value over the simulation, see  references for a practical application) can be related with the critical coupling constant, <math> k_c </math> as <math> p_{per} \approx  1 - \exp \left[ - 2 k_c \right]  </math>.


* Using [[random numbers]] we asign a random  order to these possible bonds
== Probability-Changing Cluster Algorithm ==
This method was proposed by Tomita and Okabe
<ref>[http://dx.doi.org/10.1103/PhysRevLett.86.572 Yusuke Tomita and Yutaka Okabe,  "Probability-Changing Cluster Algorithm for Potts Models", Physical Review Letters '''86''' pp. 572-575 (2001)]</ref>. This procedure is orientated towards computing [[critical points]].
It applies when the symmetry of the interactions imply that the critical
temperature is that in which the clusters, built using a Swendsen-Wang type algorithm, reach
the percolation threshold.
The simulation proceeds by a fine tuning of the temperature (or the coupling constant)
Given a configuration of the system and a current coupling constant <math> K_0 </math>:


* The possible bonds are being ''activated'' in the order fixed in the previous step (the cluster structure is watched during this process)
#One builds a bond realisation following the Swendsen-Wang strategy
#One establishes whether at least one of the cluster percolates through the whole system
#If percolation occurs one decreases the coupling constant (increase the temperature) by a small amount <math> K^{new} = K_0 - \delta K </math>
#If no percolation appears, the new value of the coupling constant is taken to be <math> K^{new} = K_0 + \delta K </math>, with <math>\delta K > 0 </math>.


* The bond activation stops when one cluster percolates through the entire system (i.e. considering the periodic boundary conditions
For small values of <math> \delta K </math> the value of <math> K </math>
the cluster becomes of inifinite size)
(after reaching the vicinity of the critical point) will show minor oscillations and the
 
results can be trusted to be those of an equilibrium simulation run. (note that [[Detailed balance |detailed balance]] is not
* Then,  every cluster (as in the Swendsen-Wang algorithm) is flipped with proability 1/2.
strictly fulfilled in this algorithm).
 
* An effective bond probability for the percolation threshold, <math> p_{per} </math> can be computed as:
 
: <math> p_{per} = M_{act}/M </math>
 
with <math> M_{act}</math> being the number of activated bonds when the first cluster ''percolates'', and <math> M </math> is the number
of possible bonds.
 
The value of <math> p_{per} </math> (in one realisation, or the averaged value over the simulation, See the references for the
practical application)
can be related with the critical coupling constant, <math> k_c </math> as:
 
: <math> p_{per} \approx  1 - \exp \left[ - 2 k_c \right] </math>
 
== Probability-Changing Cluster Algorithm ==
 
This method was proposed by Tomita and Okabe (See Ref 4) and a few details will be given here soon ....


== Beyond the Ising and Potts models ==
== Beyond the Ising and Potts models ==


The methods described so far, can be used with minor changes in the simulation of [[Potts model|Potts models]].
The methods described so far can be used, with minor changes, in the simulation of [[Potts model|Potts models]].
In addition, extensions have been proposed in the literature (References will be added one of these days)
In addition, extensions have been proposed in the literature  
to build up very efficient cluster algorithm to simulate more complex lattice systems ([[XY model]],  
to build up very efficient cluster algorithms to simulate more complex lattice systems (for example the [[XY model]],  
[[Heisenberg model]], etc).
[[Heisenberg model]], [[Lebwohl-Lasher model]]
<ref>[http://dx.doi.org/10.1103/PhysRevE.63.062702 N. V. Priezjev and Robert A. Pelcovits "Cluster Monte Carlo simulations of the nematic-isotropic transition" Physical Review E '''63''' 062702 (2001)]</ref>, etc.)


== Application to continuous (atomistic) models ==
== Application to continuous (atomistic) models ==
It is sometimes possible (and very convenient) to include cluster algorithms in the  simulation of
It is sometimes possible (and very convenient) to include cluster algorithms in the  simulation of
models with continuous trasnlational degrees of freedom. In most cases the cluster algorithm has
models with continuous translational degrees of freedom. In most cases the cluster algorithm has
to be complemented with other sampling moves to ensure ''ergodicity''. Examples:
to be complemented with other sampling moves to ensure [[Ergodic hypothesis |ergodicity]]. Examples:
 
* Spin fluids
* Spin fluids (more details soon)
* Binary [[mixtures]] having interaction symmetry
* Binary mixtures (with symmetry in the interactions) (examples soon)
* Continuous versions of the [[XY model]], [[Heisenberg model]], [[Lebwohl-Lasher model]], etc.
* Continuous version of XY, Heisenberg, and similar models
(Some references will be included soon)
 
In these cases, the usual approach is to combine one-particle moves (e.g. particle translations),  
In these cases, the usual approach is to combine one-particle moves (e.g. particle translations),  
with cluster procedures. In the cluster steps, multiparticle modification of -composition, orientations, etc.-
with cluster procedures. In the cluster steps, multiparticle modification of -composition, orientations, etc.-
is carried out.
is carried out.


== Other (not so smart) applications of cluster algorithms ==
== Geometric cluster algorithms ==
Geometric methods have been proposed for the efficient simulation of continuum fluids <ref>[http://dx.doi.org/10.1103/PhysRevLett.92.035504 Jiwen Liu and Erik Luijten, "Rejection-Free Geometric Cluster Algorithm for Complex Fluids", Physical Review Letters '''92''' 035504 (2004)]</ref> <ref> [http://dx.doi.org/10.1103/PhysRevE.71.066701 Jiwen Liu and Erik Luijten, "Generalized geometric cluster algorithm for fluid simulation",  Physical Review  E '''71''' 066701 (2005)]</ref>,
and have also been applied to simulations of [[mixtures]], <ref> [http://dx.doi.org/10.1063/1.1831274  Arnaud Buhot, "Cluster algorithm for nonadditive hard-core mixtures", Journal of Chemical Physics '''122''' 024105 (2005)] </ref>
such as [[colloids]] <ref>[http://dx.doi.org/10.1063/1.3495996 Douglas J. Ashton, Jiwen Liu, Erik Luijten, and Nigel B. Wilding "Monte Carlo cluster algorithm for fluid phase transitions in highly size-asymmetrical binary mixtures", Journal of Chemical Physics '''133''' 194102 (2010)]</ref>.
 
== Other applications of cluster algorithms ==
The cluster algorithms described so far are rejection-free methods, which means that every
new configuration generated throughout the sampling is accepted.
However, when the complexity of models increases, it becomes difficult to develop efficient
rejection-free algorithms. Nevertheless, in some cases it is still sometimes possible to build up quite efficient cluster algorithms.
 
Examples:
 
*Collective translations in the simulation of [[Micelles|micelles]]  <ref>[http://dx.doi.org/10.1021/j100189a030 David Wu, David Chandler and  Berend Smit, "Electrostatic analogy for surfactant assemblies", Journal of Physical Chemistry '''96'''  pp. 4077-4083 (1992)]</ref>
 
*Collective (cluster) translation/rotations in the simulation of the [[restricted primitive model|primitive model]] of electrolytes.<ref>
[http://dx.doi.org/10.1063/1.467770 Gerassimos Orkoulas and Athanassios Z. Panagiotopoulos, "Free energy and phase equilibria for the restricted primitive model of ionic fluids from Monte Carlo simulations", Journal of Chemical Physics '''101''' pp. 1452- (1994)]</ref>


* Monte Carlo simulation of atomistic systems with multiparticle moves:
*[[Monte Carlo|Monte Carlo]] simulation of atomistic systems with multiparticle moves.<ref>[http://dx.doi.org/10.1063/1.2759924  N. G. Almarza and E. Lomba "Cluster algorithm to perform parallel Monte Carlo simulation of atomistic systems", Journal of Chemical Physics '''127''' 084116 (2007)]</ref>.


[http://dx.doi.org/10.1063/1.2759924  N. G. Almarza and E. Lomba, ''Cluster algorithm to perform parallel Monte Carlo simulation of atomistic systems'', J, Chem. Phys (to appear) 2007 ]
*[[Monte Carlo|Monte Carlo]] simulation of [[Idealised models#'Hard' models | hard core models]] in the [[isothermal-isobaric ensemble|isothermal isobaric ensemble]].<ref>[http://dx.doi.org/10.1063/1.3133328 N. G. Almarza, "A cluster algorithm for Monte Carlo simulation at constant pressure", Journal of Chemical Physics '''130''', 184106 (2009) ]</ref>


== References ==
== References ==
#[http://dx.doi.org/10.1103/PhysRevLett.58.86  Robert H. Swendsen and Jian-Sheng Wang, "Nonuniversal critical dynamics in Monte Carlo simulations", Physical Review Letters '''58''' pp. 86 - 88 (1987) ]
<references/>
#[http://dx.doi.org/10.1103/PhysRevLett.62.361 Ulli Wolff, "Collective Monte Carlo Updating for Spin Systems" , Physical Review Letters '''62''' pp. 361 - 364 (1989) ]
[[category: computer simulation techniques]]
#[http://dx.doi.org/10.1103/PhysRevLett.75.2792    J. Machta, Y. S. Choi, A. Lucke,  T. Schweizer, and L. V. Chayes, "Invaded Cluster Algorithm for Equilibrium Critical Points" , Physical Review Letters '''75''' pp. 2792 - 2795 (1995)]
[[category: Monte Carlo]]
#[http://dx.doi.org/10.1103/PhysRevLett.86.572      Yusuke Tomita and Yutaka Okabe,  "Probability-Changing Cluster Algorithm for Potts Models", Physical Review Letters '''86''' pp. 572 - 575 (2001)]

Latest revision as of 11:43, 17 November 2010

Cluster algorithms are mainly used in the simulation of Ising-like models using Monte Carlo methods. The essential feature is the use of collective motions of particles (spins) in a single Monte Carlo step. An interesting property of some of these applications is the fact that the percolation analysis of the clusters can be used to study phase transitions.

Swendsen-Wang algorithm[edit]

As an introductory example to the Swendsen-Wang algorithm we shall discuss the technique [1] in the simulation of the Ising model. In one Monte Carlo step of the algorithm the following recipe is used:

  1. Consider every pair of interacting sites (spins). In the current configuration the pair interaction can be either negative: or positive , depending on the product: (See the Ising model entry for notation).
  2. For pairs of interacting sites (i.e. nearest neighbours) with , randomly create a bond between the two spins with a given probability , where will be chosen to be a function of .
  3. The bonds generated in the previous step are used to build up clusters of sites (spins).
  4. Build up the partition of the system in the corresponding clusters of spins. In each cluster all the spins will have the same state, either or .
  5. For each cluster, independently, choose at random with equal probabilities whether or not to flip (invert the value of ) the whole set of spins belonging to the cluster. The bonding probability is given by: .

Wolff algorithm[edit]

The procedure to create a given bond is the same as in the Swendsen-Wang algorithm. However in Wolff's method the whole set of interacting pairs is not tested to generate (possible) bonds. Instead, a single cluster is built. See [2] for details.

  1. The initial cluster contains one site, which is selected at random.
  2. Possible bonds between the initial site and other sites of the system are tested. Bonded sites are included in the cluster.
  3. Recursively, one checks the existence of bonds between the new members of the cluster and sites of the system to add, if bonds are generated, new sites to the growing cluster, until no more bonds are generated.
  4. At this point, the whole cluster is flipped (see above).

Invaded Cluster Algorithm[edit]

The purpose of this algorithm is to locate critical points (i.e. the critical temperature). So, in this case the probability of bonding neighbouring sites with equal spins is not set a priori (See [3]). The algorithm for an Ising system with periodic boundary conditions can be implemented as follows:

Given a certain configuration of the system:

  1. One considers the possible bonds in the system (pairs of nearest neighbours with favourable interaction).
  2. One assigns a random order to these possible bonds.
  3. The possible bonds are activated in the order fixed in the previous step (the cluster structure is watched during this process).
  4. The bond activation stops when one cluster percolates through the entire system (i.e. considering the periodic boundary conditions the cluster becomes of infinite size).
  5. Every cluster is then is flipped with probability 1/2, as in the Swendsen-Wang algorithm.
  6. An effective bond probability for the percolation threshold, can be computed as with being the number of activated bonds when the first cluster percolates, and is the number of possible bonds.
  7. The value of (in one realisation, or the averaged value over the simulation, see references for a practical application) can be related with the critical coupling constant, as .

Probability-Changing Cluster Algorithm[edit]

This method was proposed by Tomita and Okabe [4]. This procedure is orientated towards computing critical points. It applies when the symmetry of the interactions imply that the critical temperature is that in which the clusters, built using a Swendsen-Wang type algorithm, reach the percolation threshold. The simulation proceeds by a fine tuning of the temperature (or the coupling constant) Given a configuration of the system and a current coupling constant :

  1. One builds a bond realisation following the Swendsen-Wang strategy
  2. One establishes whether at least one of the cluster percolates through the whole system
  3. If percolation occurs one decreases the coupling constant (increase the temperature) by a small amount
  4. If no percolation appears, the new value of the coupling constant is taken to be , with .

For small values of the value of (after reaching the vicinity of the critical point) will show minor oscillations and the results can be trusted to be those of an equilibrium simulation run. (note that detailed balance is not strictly fulfilled in this algorithm).

Beyond the Ising and Potts models[edit]

The methods described so far can be used, with minor changes, in the simulation of Potts models. In addition, extensions have been proposed in the literature to build up very efficient cluster algorithms to simulate more complex lattice systems (for example the XY model, Heisenberg model, Lebwohl-Lasher model [5], etc.)

Application to continuous (atomistic) models[edit]

It is sometimes possible (and very convenient) to include cluster algorithms in the simulation of models with continuous translational degrees of freedom. In most cases the cluster algorithm has to be complemented with other sampling moves to ensure ergodicity. Examples:

In these cases, the usual approach is to combine one-particle moves (e.g. particle translations), with cluster procedures. In the cluster steps, multiparticle modification of -composition, orientations, etc.- is carried out.

Geometric cluster algorithms[edit]

Geometric methods have been proposed for the efficient simulation of continuum fluids [6] [7], and have also been applied to simulations of mixtures, [8] such as colloids [9].

Other applications of cluster algorithms[edit]

The cluster algorithms described so far are rejection-free methods, which means that every new configuration generated throughout the sampling is accepted. However, when the complexity of models increases, it becomes difficult to develop efficient rejection-free algorithms. Nevertheless, in some cases it is still sometimes possible to build up quite efficient cluster algorithms.

Examples:

  • Collective (cluster) translation/rotations in the simulation of the primitive model of electrolytes.[11]

References[edit]

  1. Robert H. Swendsen and Jian-Sheng Wang, "Nonuniversal critical dynamics in Monte Carlo simulations", Physical Review Letters 58 pp. 86-88 (1987)
  2. Ulli Wolff, "Collective Monte Carlo Updating for Spin Systems" , Physical Review Letters 62 pp. 361-364 (1989)
  3. J. Machta, Y. S. Choi, A. Lucke, T. Schweizer, and L. V. Chayes, "Invaded Cluster Algorithm for Equilibrium Critical Points", Physical Review Letters 75 pp. 2792-2795 (1995)
  4. Yusuke Tomita and Yutaka Okabe, "Probability-Changing Cluster Algorithm for Potts Models", Physical Review Letters 86 pp. 572-575 (2001)
  5. N. V. Priezjev and Robert A. Pelcovits "Cluster Monte Carlo simulations of the nematic-isotropic transition" Physical Review E 63 062702 (2001)
  6. Jiwen Liu and Erik Luijten, "Rejection-Free Geometric Cluster Algorithm for Complex Fluids", Physical Review Letters 92 035504 (2004)
  7. Jiwen Liu and Erik Luijten, "Generalized geometric cluster algorithm for fluid simulation", Physical Review E 71 066701 (2005)
  8. Arnaud Buhot, "Cluster algorithm for nonadditive hard-core mixtures", Journal of Chemical Physics 122 024105 (2005)
  9. Douglas J. Ashton, Jiwen Liu, Erik Luijten, and Nigel B. Wilding "Monte Carlo cluster algorithm for fluid phase transitions in highly size-asymmetrical binary mixtures", Journal of Chemical Physics 133 194102 (2010)
  10. David Wu, David Chandler and Berend Smit, "Electrostatic analogy for surfactant assemblies", Journal of Physical Chemistry 96 pp. 4077-4083 (1992)
  11. Gerassimos Orkoulas and Athanassios Z. Panagiotopoulos, "Free energy and phase equilibria for the restricted primitive model of ionic fluids from Monte Carlo simulations", Journal of Chemical Physics 101 pp. 1452- (1994)
  12. N. G. Almarza and E. Lomba "Cluster algorithm to perform parallel Monte Carlo simulation of atomistic systems", Journal of Chemical Physics 127 084116 (2007)
  13. N. G. Almarza, "A cluster algorithm for Monte Carlo simulation at constant pressure", Journal of Chemical Physics 130, 184106 (2009)