Generalised Spin Projection for Fermion Actions

TL;DR

This paper introduces a generalized spin projection method for fermion actions in lattice QCD, reducing computational costs associated with split-link actions like FLIC by optimizing spin projector usage.

Contribution

It presents a new approach to express split-link actions using spin projectors, minimizing additional computational costs and reviewing efficient evaluation of the clover term.

Findings

Reduced the additional cost of split-link actions to at most a factor of two.

Demonstrated an efficient method for evaluating the clover term.

Showed potential compute time savings with improved condition number actions.

Abstract

The majority of compute time doing lattice QCD is spent inverting the fermion matrix. The time that this takes increases with the condition number of the matrix. The FLIC(Fat Link Irrelevant Clover) action displays, among other properties, an improved condition number compared to standard actions and hence is of interest due to potential compute time savings. However, due to its two different link sets there is a factor of two cost in floating point multiplications compared to the Wilson action. An additional factor of two has been attributed due to the loss of the so-called spin projection trick. We show that any split-link action may be written in terms of spin projectors, reducing the additional cost to at most a factor of two. Also, we review an efficient means of evaluating the clover term, which is additional expense not present in the Wilson action.