Excited state dynamic-node diffusion Monte Carlo simulations

We describe a novel diffusion Monte Carlo algorithm for calculating excited state quantum wave functions and energies which requires little prior knowledge of nodal surfaces. An initial, guessed, nodal surface is refined during the simulation by balancing populations in different nodal regions. The method requires constraints on an excited system to ensure that it does not relax to a less energetic mode. So long as such constraints can be found, the method will allow complex nodal surfaces to evolve dynamically, and accurate energies of excited systems to be calculated. Results are given for a simple vibrational system, re-vibrational states of hydrogen fluoride, and excited vibrational states of the water molecule.