Petascale computing opens new vistas for quantum Monte Carlo

For many kinds of problem the accuracy of quantum Monte Carlo (QMC) is much better than that of density functional theory (DFT), and its scaling with number of atoms is much more favourable than that of high-level quantum chemistry. However, the widespread use of QMC has been hindered by the fact that it is considerably more expensive than DFT. We show here that QMC is very well placed to exploit the power of petascale supercomputers that are now becoming available, and we explain how this is opening up new scientific areas to investigation with QMC. We describe how we have been able to modify the Cambridge QMC code CASINO so that it runs with almost perfect parallel efficiency on 100000 cores and more on the JaguarPF machine at Oak Ridge. We also present illustrative results showing how QMC calculations run in this way are enabling us to go beyond the limitations of DFT in three important areas: the surface formation energies of materials, the adsorption energies of molecules on surfaces, and the energetics of water systems.