Paper detail

Monte Carlo Algorithms For Reduced Lattices, Mixed Actions, And Double-Trace Deformations

We construct efficient Monte Carlo updating algorithms for two classes of pure SU(N) lattice gauge actions with non-linear dependence on the link variables. Our construction generalises the method of auxiliary variables used by Fabricius and Haan in the framework of Eguchi-Kawai models. We first review the original Fabricius-Haan method of constructing a pseudo-heatbath algorithm for fully reduced models, and discuss its extension to lattices with any number of reduced directions. We then use a similar method to construct updating algorithms for generic SU(N) mixed Wilson actions. We construct explicit examples of algorithms for Wilson actions whose plaquettes are in an irreducible representation of SU(N) with N-ality up to 3. We also construct updating algorithms for the lattice version of centre-stabilised SU(N) Yang-Mills theories defined on R^{d-1} x S^1, including the case of a fully reduced compact direction. We simulate the new algorithms and show that they are, in general, significantly more efficient than their Metropolis counterparts.