Systematic improvement of wavefunctions in the variational Monte Carlo method for the t–J model Original Research Article

We present an algorithm to systematically improve wavefunctions in the variational Monte Carlo method for the t–J model. The ground-state wavefunction is approximated by a linear combination of single-particle states with Gutzwiller projection. The single-particle states are successively generated according to the numerical path-integral renormalization group (PIRG) procedure. In the initial step, the present method reduces to the usual variational Monte Carlo method. As the number of single-particle states increases, the wavefunction converges to the ground-state wavefunction. Applying this method to the two-dimensional t–J model, we have confirmed that the wavefunction, which is composed of a small number of single-particle states, gives a good estimate of the ground-state energy. This algorithm does not suffer from the negative sign problem and can be applied to the models with strong correlations.