Abstract

First-principles quantum Monte Carlo (QMC) techniques, such as variational quantum Monte Carlo (VMC) and diffusion quantum Monte Carlo (DMC), are among the state-of-the-art numerical methods used to obtain highly accurate many-body wave functions. These methods are especially useful when tackling challenging cases such as low-dimensional materials¹ because QMC is no longer dependent on any semi-empirical exchange-correlation functions. We have been intensively improving a QMC code “TurboRVB,” which has been mainly developed by Prof. Sandro Sorella (SISSA)[2]. I am going to talk about two recent improvements in the QMC algorithm.

The first topic is about all-electron calculations. Although it is convenient to replace core electrons in QMC calculations as in DFT , such replacement sometimes induces nontrivial biases. All-electron calculations in QMC are not as widely used as in DFT because the computational cost scales with Z^5.5−6.5, where Z is the atomic number. We have recently developed new algorithms to drastically decrease computational costs of all-electron DFT (valid only for QMC )[3], and all-electron lattice regularized diffusion monte Carlo (LRDMC)[4,5]. I will present basic ideas of the new algorithms and show several applications such as a binding energy calculation of the sodium dimer³.

The second topic is about a workflow system for QMC optimizations. We are currently developing a python wrapper for TurboRVB, which is called Genius-TurboRVB (g-turbo), in order to “automatize” the complicated optimization procedure of a many-body wave function. The wrapper also makes it much easier to prepare input files, to analyze output files, and to perform advanced calculations. I will present fundamental features and several applications of the wrapper, for example, a phonon dispersion calculation of a solid⁶.

[1] S. Sorella, et al. Phys. Rev. Lett. 121, 066402 (2018).

[2] S. Sorella, https://people.sissa.it/~sorella/web, accessed 4 August (2019).

[3] K. Nakano, et al. J. Chem. Theory Comput. 15, 4044-4055 (2019).

[4] M. Casula, et al. Phys. Rev. Lett. 95, 100201 (2005).

[5] K. Nakano, et al. to be submitted to Phys. Rev. Lett.

[6] K. Nakano, et al. in preparation.