Cubic-scaling algorithm and self-consistent field for the random-phase approximation with second-order screened exchange

TL;DR

This paper introduces a cubic-scaling algorithm for RPA+SOSEX that reduces computational cost and memory requirements, enabling more efficient and accurate electron correlation calculations in quantum chemistry.

Contribution

The authors develop a new algorithm that reduces RPA+SOSEX scaling from fifth to cubic order and introduces a self-consistent field approximating BCCD theory, improving efficiency and accuracy.

Findings

Achieves $ ext{O}(n^3)$ scaling for RPA+SOSEX

Demonstrates comparable accuracy to traditional methods in H$_2$ dissociation

Verifies improved scaling with H$_n$ ring systems

Abstract

The random-phase approximation with second-order screened exchange (RPA+SOSEX) is a model of electron correlation energy with two caveats: its accuracy depends on an arbitrary choice of mean field, and it scales as $\mathcal{O}(n^5)$ operations and $\mathcal{O}(n^3)$ memory for $n$ electrons. We derive a new algorithm that reduces its scaling to $\mathcal{O}(n^3)$ operations and $\mathcal{O}(n^2)$ memory using controlled approximations and a new self-consistent field that approximates Brueckner coupled-cluster doubles (BCCD) theory with RPA+SOSEX, referred to as Brueckner RPA (BRPA) theory. The algorithm comparably reduces the scaling of second-order M$\mathrm{{\o}}$ller-Plesset (MP2) perturbation theory with smaller cost prefactors than RPA+SOSEX. Within a semiempirical model, we study H$_2$ dissociation to test accuracy and H$_n$ rings to verify scaling.