Playing with the kinetic term in the HMC

TL;DR

This paper explores modifying the Hamiltonian Monte Carlo algorithm for QCD simulations by replacing Gaussian momentum distributions with Lorentz distributions to potentially improve efficiency.

Contribution

It introduces a novel approach of using Lorentz distributions for momentum generation in HMC, aiming to enhance simulation performance.

Findings

Lorentz distribution can impose a dynamic cutoff on coordinate changes

Preliminary results in SU(2) gauge theory show promise

Potential for improved HMC efficiency in QCD simulations

Abstract

The HMC algorithm, combining the advantages of molecular dynamics and Monte-Carlo methods, is the most efficient algorithm to simulate QCD including the effects of sea quarks. In the standard approach momentum fields are generated with a Gaussian probability density. In this work I will explore another possibility. By using a Lorentz distribution one can dynamically impose a cutoff in the rate of change of the coordinates that potentially could have a better behavior. I will present some results in pure SU(2) gauge theory.