Source-linked AI summary
A deterministic alternative to the full configuration interaction quantum Monte Carlo method
Norm M. Tubman, Joonho Lee, Tyler Y. Takeshita, Martin Head-Gordon, K. Birgitta Whaley
TL;DR
FCIQMC can suffer from biased sampling and expensive simulations, motivating a deterministic alternative for efficiently identifying important determinants. The paper develops ASCI, applies it to ground and excited states, and reports chemical-accuracy results for difficult Cr2 and C2 benchmarks.
Problem
FCIQMC has limitations including biased sampling and comparatively expensive simulations, motivating an alternative to stochastic determinant sampling.
Method
ASCI replaces stochastic sampling with a deterministic search that ranks important determinants, using core and target determinant subspaces and iterative diagonalization.
Results
ASCI achieved chemical accuracy for Cr2 with perturbation correction and calculated ground and excited-state energies for C2, while Cr2 determinant distributions remained important at high excitation levels.
Takeaways & Limitations
The results show that FCIQMC-like determinant-space exploration can be performed deterministically for ground and excited chemical states.
Takeaways & Limitations
Maintaining size consistency requires tdets and cdets to grow with system size, although cdets could be extrapolated for the systems studied.
Abstract
from arXiv · showhide
Development of exponentially scaling methods has seen great progress in tackling larger systems than previously thought possible. One such technique, full configuration interaction quantum Monte Carlo, is a useful algorithm that allows exact diagonalization through stochastically sampling determinants. The method derives its utility from the information in the matrix elements of the Hamiltonian, along with a stochastic projected wave function, to find the important parts of Hilbert space. However, the stochastic representation of the wave function is not required to search Hilbert space efficiently, and here we describe a highly efficient deterministic method to achieve chemical accuracy for a wide range of systems, including the difficult Cr$_{2}$ dimer. In addition our method also allows efficient calculation of excited state energies, for which we illustrate with benchmark results for the excited states of C$_{2}$.