Anthony D. Kennedy

Physical results from 2+1 flavor domain wall QCD and SU(2) chiral perturbation theory

We have simulated QCD using 2+1 flavors of domain wall quarks on a (2.74fm)3 volume with an inverse lattice scale of a?1=1.729(28) GeV. The up and down (light) quarks are degenerate in our calculations and we have used four values for the ratio of light quark masses to the strange (heavy) quark mass in our simulations: 0.217, 0.350, 0.617 and 0.884. We have measured pseudoscalar meson masses and decay constants, the kaon bag parameter BK and vector meson couplings. We have used SU(2) chiral perturbation theory, which assumes only the up and down quark masses are small, and SU(3) chiral perturbation theory to extrapolate to the physical values for the light quark masses. While next-to-leading order formulae from both approaches fit our data for light quarks, we find the higher order corrections for SU(3) very large, making such fits unreliable. We also find that SU(3) does not fit our data when the quark masses are near the physical strange quark mass. Thus, we rely on SU(2) chiral perturbation theory for accurate results. We use the masses of the ? baryon, and the ? and K mesons to set the lattice scale and determine the quark masses. We then find f?=124.1(3.6)stat(6.9)systMeV, fK=149.6(3.6)stat(6.3)systMeV and fK/f?=1.205(0.018)stat(0.062)syst. Using non-perturbative renormalization to relate lattice regularized quark masses to RI-MOM masses, and perturbation theory to relate these to MS¯ we find mMS¯ud(2GeV)=3.72(0.16)stat(0.33)ren(0.18)systMeV and mMS¯s(2GeV)=107.3(4.4)stat(9.7)ren(4.9)systMeV.

Cross Component Optimisation in a High Level Category-Based Language

High level programming languages offer many benefits in terms of ease of use, encapsulation etc. However, they historically suffer from poor performance. In this paper we investigate improving the performance of a numerical code written in a high–level language by using cross–component optimisation. We compare the results with traditional approaches such as the use of high performance libraries or Fortran. We demonstrate that our cross–component optimisation is highly effective, with a speed–up of up to 1.43 over a program augmented with calls to the ATLAS BLAS library, and 1.5 over a pure Fortran equivalent.

Accelerating Dynamical-Fermion Computations Using the Rational Hybrid Monte Carlo Algorithm with Multiple Pseudofermion Fields

There has been much recent progress in the understanding and reduction of the computational cost of the hybrid Monte Carlo algorithm for lattice QCD as the quark mass parameter is reduced. In this letter we present a new solution to this problem, where we represent the fermionic determinant using n pseudofermion fields, each with an nth root kernel. We implement this within the framework of the rational hybrid Monte Carlo algorithm. We compare this algorithm with other recent methods in this area and find it is competitive with them.

Exact 2+1 flavour RHMC simulations

We consider the Rational Hybrid Monte Carlo algorithm for performing exact 2+1 flavour fermion simulations. The specific cases of asqtad and domain wall fermions are considered. We find that in both cases the naive performance is similar to conventional hybrid algorithms.

The RHMC algorithm for 2 flavours of dynamical staggered fermions

Abstract We describe an implementation of the Rational Hybrid Monte Carlo (RHMC) algorithm for dynamical computations with two flavours of staggered quarks. We discuss several variants of the method, the performance and possible sources of error for each of them, and we compare the performance and results to the inexact R algorithm.

Tuning the strange quark mass in lattice simulations

Abstract QCD lattice simulations with 2 + 1 flavours typically start at rather large up-down and strange quark masses and extrapolate first the strange quark mass to its physical value and then the up-down quark mass. An alternative method of tuning the quark masses is discussed here in which the singlet quark mass is kept fixed, which ensures that the kaon always has mass less than the physical kaon mass. It can also take into account the different renormalisations (for singlet and non-singlet quark masses) occurring for non-chirally invariant lattice fermions and so allows a smooth extrapolation to the physical quark masses. This procedure enables a wide range of quark masses to be probed, including the case with a heavy up-down quark mass and light strange quark mass. Results show the correct order for the baryon octet and decuplet spectrum and an extrapolation to the physical pion mass gives mass values to within a few percent of their experimental values.

Cost of the generalised hybrid Monte Carlo algorithm for free field theory

Abstract We study analytically the computational cost of the generalised hybrid Monte Carlo (GHMC) algorithm for free field theory. We calculate the Metropolis acceptance probability for leapfrog and higher-order discretisations of the molecular dynamics (MD) equations of motion. We show how to calculate autocorrelation functions of arbitrary polynomial operators, and use these to optimise the GHMC momentum mixing angle, the trajectory length, and the integration stepsize for the special cases of linear and quadratic operators. We show that long trajectories are optimal for GHMC, and that standard HMC is more efficient than algorithms based on second order Langevin Monte Carlo (L2MC), sometimes known as Kramers equation. We show that contrary to naive expectations HMC and L2MC have the same volume dependence, but their dynamical critical exponents are z =1 and z =3/2, respectively.

Algorithms for dynamical fermions

This is the write-up of three lectures on algorithms for dynamical fermions that were given at the ILFTN workshop ‘Perspectives in Lattice QCD’ in Nara during November 2005. The first lecture is on the fundamentals of Markov Chain Monte Carlo methods and introduces the Hybrid Monte Carlo (HMC) algorithm and symplectic integrators; the second lecture covers topics in approximation theory and thereby introduces the Rational Hybrid Monte Carlo (RHMC) algorithm and ways of evading integrator instabilities by means of multiple pseudofermion fields; the third lecture introduces on-shell chiral (Ginsparg-Wilson) lattice fermions and discusses fivedimensional formulations for computing fermion propagators for such fermions.

Instability in the molecular dynamics step of a hybrid Monte Carlo algorithm in dynamical fermion lattice QCD simulations

We investigate instability and reversibility within hybrid Monte Carlo simulations using a nonperturbatively improved Wilson action. We demonstrate the onset of instability as tolerance parameters and molecular dynamics step sizes are varied. We compare these findings with theoretical expectations and present limits on simulation parameters within which a stable and reversible algorithm is obtained for physically relevant simulations. Results of optimization experiments with respect to tolerance parameters are also presented.

Accelerating staggered-fermion dynamics with the rational hybrid Monte Carlo algorithm

Improved staggered-fermion formulations are a popular choice for lattice QCD calculations. Historically, the algorithm used for such calculations has been the inexact R algorithm, which has systematic errors that only vanish as the square of the integration step size. We describe how the exact rational hybrid Monte Carlo (RHMC) algorithm may be used in this context, and show that for parameters corresponding to current state-of-the-art computations it leads to a factor of approximately seven decrease in cost as well as having no step-size errors.