The problem of the critical slowing-down affecting local processes is of fundamental importance in the study of second-order phase transitions, as increasing the size of the system in order to reduce finite-size effects has the disadvantage of requiring a far larger number of moves to reach thermal equilibrium. Indeed the correlation time usually increases as with or greater; since, to be accurate, the simulation time must be, this is a major limitation in the size of the systems that can be studied through local algorithms. SW algorithm was the first to produce unusually small values for the dynamical critical exponents: for the 2D Ising model ; for the 3D Ising model, as opposed to for standard simulations.
Description
The algorithm is non-local in the sense that in a single sweep of moves a collective update of the spin variables of the system is done. The key idea is to take an additional number of 'bond' variables, as suggested by Fortuin and Kasteleyn, who mapped the Potts model onto a percolation model via the random cluster model. Consider a typical ferromagnetic Ising model with only nearest-neighbor interaction.
Starting from a given configuration of spins, we associate to each pair of nearest neighbours on sites a random variable which is interpreted in the following way: if there is no link between the sites and ; if, and are connected. These values are assigned according to the following probability distribution:
; ; ; ; where is the ferromagnetic interaction intensity. This probability distribution has been derived in the following way: the Hamiltonian of the Ising model is , and the partition function is . Consider the interaction between a pair of selected sites and and eliminate it from the total Hamiltonian, defining Define also the restricted sums: Introduce the quantity ; the partition function can be rewritten as Since the first term contains a restriction on the spin values whereas there is no restriction in the second term, the weighting factors can be interpreted as probabilities of forming/not forming a link between the sites: The process can be easily adapted to antiferromagnetic spin systems, as it is sufficient to eliminate in favor of .
After assigning the bond variables, we identify the same-spin clusters formed by connected sites and make an inversion of all the variables in the cluster with probability 1/2. At the following time step we have a new starting Ising configuration, which will produce a new clustering and a new collective spin-flip.
Correctness
It can be shown that this algorithm leads to equilibrium configurations. The first way to prove it is using the theory ofMarkov chains, either noting that the equilibrium maps into itself, or showing that in a single sweep of the lattice there is a non-zero probability of going from any state of the Markov chain to any other; thus the corresponding irreducible ergodic Markov chain has an asymptotic probability distribution satisfying detailed balance. Alternatively, we can show explicitly that detailed balance is satisfied. Every transition between two Ising configurations must pass through some bond configuration in the percolation representation. Let's fix a particular bond configuration: what matters in comparing the probabilities related to it is the number of factors for each missing bond between neighboring spins with the same value; the probability of going to a certain Ising configuration compatible with a given bond configuration is uniform. So the ratio of the transition probabilities of going from one state to another is since. This is valid for every bond configuration the system can pass through during its evolution, so detailed balance is satisfied for the total transition probability. This proves that the algorithm works.
Efficiency
Although not analytically clear from the original paper, the reason why all the values of z obtained with the SW algorithm are much lower than the exact lower bound for single-spin-flip algorithms is that the correlation length divergence is strictly related to the formation of percolation clusters, which are flipped together. In this way the relaxation time is significantly reduced. The algorithm is not efficient in simulating frustrated systems.