arXiv subjects
Publications and source records attributed to G. Bhanot.
In June 2004 QCD was simulated for the first time at sustained speed exceeding 1 TeraFlops in the BlueGene/L supercomputer at the IBM T.J. Watson Research Lab. The implementation and performance of QCD in the BlueGene/L is presented.
We compute high temperature expansions of the 3-d Ising model using a recursive transfer-matrix algorithm and extend the expansion of the free energy to 24th order. Using ID-Pade and ratio methods, we extract the critical exponent of the specific heat to be alpha=0.104(4).
We consider 1+1-dimensional QCD coupled to Majorana fermions in the adjoint representation of the gauge group $SU(N)$. Pair creation of partons (fermion quanta) is not suppressed in the large-$N$ limit, where the glueball-like bound states become free. In this limit the spectrum is given by a linear \lc\ Schr\" odinger equation, which we study numerically using the discretized \lcq. We find a discrete spectrum of bound states, with the logarithm of the level density growing approximately linearly with the mass. The wave function of a typical excited state is a complicated mixture of components with different parton numbers. A few low-lying states, however, are surprisingly close to being eigenstates of the parton number, and their masses can be accurately calculated by truncated diagonalizations.
On simple cubic lattices, we compute low temperature series expansions for the energy, magnetization and susceptibility of the three-state Potts model in D=2 and D=3 to 45 and 39 excited bonds respectively, and the eight-state Potts model in D=2 to 25 excited bonds. We use a recursive procedure which enumerates states explicitly. We analyze the series using Dlog Pade analysis and inhomogeneous differential approximants.
We describe a method to study the rate at which modes decorrelate in numerical simulations. We study the XY model updated with the Metropolis and Wolff dynamics respectively and compute the rate at which each eigenvector of the dynamics decorrelates. Our method allows us to identify the decorrelation time for each mode separately. We find that the autocorrelation function of the various modes is markedly different for the `local' Metropolis compared to the `non-local' Wolff dynamics. Equipped with this new insight, it may be possible to devise highly efficient algorithms.