Simulationen der QCD mit dynamischen Overlap-Fermionen

点击翻译按钮获取中文摘要

Some of the main systematic uncertainties in lattice hadron phenomenology are caused by the lack of chiral symmetry of typical fermion actions used. Thus, it is of central importance for our Forschergruppe to proceed towards simulations with e.g. dynamical Overlap fermions. So far only quenched simulations were feasible, primarily due to the sheer costs to evaluate the matrix sign function as well as due to algorithmic difficulties with the stability of the Hybrid Monte Carlo (HMC) scheme, handling zero-crossing of eigenmodes of the Wilson operator. Our QCD-group at Wuppertal/ZAM has carried out simulations with dynamical fermions over more than 10 years (SESAM project, T(L project). As far as the Overlap fermion discretization is concerned, we have introduced the optimal Zolotarev scheme for the sign function which turns out to be well suited for the HMC algorithm. Recently, we have achieved substantial improvements to the nested inversions with a factor 5 for HMC and 10 for propagator computations. We have already implemented a 2-flavor HMC-code with explicit treatment of zero crossing for one eigenmode. In this sub-project we aim at the generation and evaluation of dynamical Overlap configurations with two and three light sea quarks. We will develop a new one-flavor chirally projected HMC code to accelerate topological fluctuations in the very light quark regime. We will extend the HMC to improved gauge actions and the two-mode crossing problem. Furthermore, we will carry out a pilot study aiming for sea quarks as light as possible on lattices of size 163x32, in coordination with the other sub-projects, paricular sub-project. 1. We will extend the truncated eigenmode approximation to Overlap fermions, computing disconnected diagrams for ¿¿ and all-to-all propagators for string breaking. The generated configurations will also be used by the other members of the Forschergruppe. The precise set of oberservables evaluated on them will, however, depend on the efficiency of the algorithm we will achieve.