Concentrating on zero temperature Quantum Monte Carlo calculations of electronic systems, we give a general description of the theory of finite size extrapolations of energies to the thermodynamic limit based on one and two-body correlation functions. We introduce new effective procedures, such as using the potential and wavefunction split-up into long and short range functions to simplify the method and we discuss how to treat backflow wavefunctions. Then we explicitly test the accuracy of our method to correct finite size errors on example hydrogen and helium many-body systems and show that the finite size bias can be drastically reduced for even small systems.
PACS numbers:Quantum Monte Carlo (QMC) methods allow us to calculate the energy per particle E N of a finite system containing N particles, with N 10 3 for almost all simulations of electronic systems [1,2]. However, for extended systems, we are often interested in scaling to the thermodynamic limit, E ∞ ; this scaling is one of the major source of bias in Quantum Monte Carlo calculations of electronic structure. In practice, extrapolation is often performed numerically by assuming simple functional forms for E N as a function of 1/N , often inspired by results of approximate theories, such as KohnSham DFT [3][4][5] or from the behavior of approximate many-body calculations, e.g. from RPA calculations [6]. These heuristic extrapolations can be dangerous and introduce a possible systematic bias, as the exact ground state energy, as well as other properties, are in general not a simple analytic function of 1/N . In fact, the scaling function will depend on the electronic state, for example, it will be different in a metal and an insulator, and can depend on the form of the trial wave function underlying the QMC calculation. In addition, within variational approaches, the amount that the variational energy is above the exact energy may depend on the system size because of the values of the variational parameters. This introduces a further source of error in a purely numerical extrapolation. Projection methods can reduce this bias, since they are closer to the true ground state energy, but in practice it can be a difficult problem to ensure a uniform convergence concerning projection time or population size with respect to the system size [7].In this paper we present a general theory for understanding the finite size bias of QMC calculations. Although we concentrate on electronic systems where finite size effects represent one of the major limitations, our approach applies equally well to other quantum systems with different interactions and dimensionality, including bosonic ones. As we will show, the leading order size effects can be understood by looking at the analytical structure of the trial wavefunction [8,9] which is -at least partially -determined by singularities of the Hamiltonian and/or the boundary conditions [10][11][12]. Different types of wavefunction will, in general, have different types of size effects. In particular, we show that backflow...