Discretization errors in the spectrum of the Hermitian Wilson-Dirac operator

I study the leading effects of discretization errors on the low-energy part of the spectrum of the Hermitian Wilson-Dirac operator in infinite volume. The method generalizes that used to study the spectrum of the Dirac operator in the continuum, and uses partially quenched chiral perturbation theory for Wilson fermions. The leading-order corrections are proportional to a(2) (a being the lattice spacing). At this order, I find that the method works only for one choice of sign of one of the three low-energy constants describing discretization errors. If these constants have the relative magnitudes expected from large N-c arguments, then the method works if the theory has an Aoki phase for m similar to a(2), but fails if there is a first-order transition. In the former case, the dependence of the gap and the spectral density on m and a(2) are determined. In particular, the gap is found to vanish more quickly as m(pi)(2)-> 0 than in the continuum. This reduces the region where simulations are safe from fluctuations in the gap.