Speaker
Description
In recent years multigrid algorithms have dramatically reduced the cost of generating gauge field ensembles and quark propagators for lattice simulations including light quarks described by the Wilson and Wilson-clover fermion actions. As a result, we have observed in recent calculations of nuclear physics at the physical pion mass that assembling correlation functions from quark propagators is an increasingly costly aspect of these calculations. In this talk we will discuss a sparsening algorithm for building correlation functions describing multi-body systems of nucleons. This algorithm works by first block averaging lattice quark propagators, producing sparsened quark propagators defined on a coarsened lattice, and then computing correlation functions from these sparsened propagators at reduced computational cost. We have explored this approach by analyzing the low energy QCD spectrum, including systems as large as $ ^{4}_{2} \mathrm{He} $, on a single $ 32^{3} \times 48 $ Wilson-clover lattice ensemble with $ m_{\pi} \approx 800 $ MeV. We find that the ground state masses and binding energies we extract are consistent between correlation functions constructed from sparsened or full propagators. In addition, while we observe small, systematic biases in excited states for the sparsened correlation functions, we also find that these biases can be removed by computing an inexpensive correction term.
