Fast and accurate correlation energies with RPA + SOSEX

The Random Phase Approximation (RPA) is a state-of-the art method for the accurate description of electron correlation effects, which does not suffer from the divergences of MP2 for large molecules and small HOMO-LUMO gaps. However, the success of the RPA relies on the cancellation of errors. These errors stem from the presence of unphysical terms in the RPA energy expression, leading to strongly overestimated total correlation energies. By adding second-order screened exchange (SOSEX) to the RPA (RPA+SOSEX), this issue is substantially reduced. RPA+SOSEX often leads to major improvements over RPA alone.

RPA+SOSEX comes with the same canonical scaling than MP2, but its high prefactor often makes the method too expensive for practical applications. In recent work, the statically screened G3W2 contribution to the electronic self-energy has been suggested as a way to obtain another cost-efficient alternative to RPA+SOSEX[1].

Benchmark results for a wide range of bonding characteristics have now demonstrated that this new correlation energy expression provides the same accuracy as RPA+SOSEX at MP2 cost[2]. Especially for charged excitations and non-covalent interactions, this new energy expression leads to major improvements over RPA and is on par in accuracy with modern double hybrid functionals, while not having to use any empirical parameters [2].

Learn more? Check out Arno’s webinar in October!



  • Check out more news items and research highlights.
    Why not explore AMS yourself with a free 30-day trial?
    Subscribe to our newsletter (~6 times a year) to keep up to date about news, job openings, functionality and events such as webinars and workshops!

[1] A, Förster, L. Visscher, Exploring the statically screened G3W2 correction to the GW self-energy: Charged excitations and total energies of finite systems, PRB, (2022) 105, 125121
[2] A. Förster, Assessment of the Second-Order Statically Screened Exchange Correction to the Random Phase Approximation for Correlation Energies, JCTC, (2022

Key concepts