Dresden 2011 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

O: Fachverband Oberflächenphysik

O 48: Focussed session: Theory and computation of electronic structure: new frontiers IV (jointly with HL, DS)

O 48.1: Topical Talk

Wednesday, March 16, 2011, 15:00–15:30, TRE Phy

Progress in diffusion quantum Monte Carlo calculations — •Richard Needs — Cavendish Laboratory, J J Thomson Avenue, Cambridge CB3 0HE, UK

My group has developed the CASINO code [1] for performing variational and diffusion quantum Monte Carlo calculations. Fixed-node diffusion quantum Monte Carlo is the most accurate method known for calculating the energies of large many-particle quantum systems. The key ingredient is an accurate trial many-body wave function which controls the statistical efficiency and accuracy of the calculations. Accurate wave functions can be obtained by building correlation effects on top of mean field descriptions such as density functional theory or Hartree-Fock theory. About 80% of the correlation energy can typically be included by multiplying the mean-field determinant by a Jastrow factor which is small when electrons are close together and tends to unity at large separations. Such wave functions provide an excellent description of electron correlation in closed shell molecules but are often not much better than density functionals for small open shell systems. The wave functions of open shell systems can, however, be greatly improved by introducing more determinants, pairing functions, and backflow transformations, and extremely good results can be obtained. The calculations are expensive but the polynomial scaling with system size allows calculations for 1000 or more particles. The discussion of the methodology will be illustrated by recent applications to atoms, molecules and extended systems.

[1] http://www.tcm.phy.cam.ac.uk/~mdt26/casino2.html