Nucleon Matrix Elements at Physical Pion Mass and Cost Comparison

TL;DR

This paper presents a lattice QCD calculation of nucleon matrix elements at physical pion mass, compares computational costs across different fermion discretizations, and discusses efficiency improvements in overlap fermion calculations.

Contribution

It provides the first physical pion mass nucleon matrix elements with cost analysis comparing domain-wall, overlap, twisted mass, and clover fermions, highlighting efficiency improvements.

Findings

Calculated nucleon axial charge $g_A^3$ as 1.18(4).

Determined quark momentum fraction $raket{x}_{u-d}$ as 0.170(14).

Overlap fermions can be as efficient as twisted mass and clover with improvements.

Abstract

We report a lattice calculation of nucleon forward matrix elements on a $48^3 \times 96$ lattice at the physical pion mass and a spatial size of 5.5 fm. The $2+1$ flavor dynamical fermion configurations are generated with domain-wall fermions (DWF) and the overlap fermions are adopted for the valence quarks. The isovector $g_A^3$ and $g_S^3$, and the connected insertion part of $g_S^0$ are reported for three source-sink separations. With local current, we obtain $g_A^3 = 1.18(4)$ from a two-state fit. For the quark momentum fraction $\langle x \rangle_{u-d}$, we have included smaller lattices (i.e. $24^3 \times 64$ and $32^3 \times 64$ lattice with pion mass at 330 and 290 MeV respectively) for a fit which includes partially quenched cases as well as finite volume and continuum corrections. A global fit with perturbative renormalization gives $\langle x \rangle_{u-d} (\overline{MS},\, \mu = 2\, {\rm GeV}) = 0.170(14)$. We made a cost comparison of calculating the nucleon matrix elements with those from the twisted mass fermion on similar sized lattice at the physical pion point and the domain-wall fermion calculation on the same DWF lattice. We also compare cost with the clover fermion calculation on similar sized lattice at about the same quark mass. The comparison shows that with several improvements, such as many-to-all correlator with grid source and low-mode substitution in the connected insertion and low-mode average in the quark loop can make the overlap as efficient as the twisted-mass and clover fermions in calculating the three-point functions. It is more efficient than the DWF. When the multi-mass feature is invoked, the overlap can be more efficient in reaching the same precision than the single mass comparison made so far.