Analysis of the errors in explicitly correlated electronic structure theory

[Astronuc]

PCCP Hot paper: Theoretical breakthrough - analysis of the errors in explicitly correlated electronic structure theory

1. Could you explain the significance of your article to the non-specialist?

Methods to study the electronic structure of molecules are of paramount importance to chemists of all kinds. Explicitly correlated methods hold the promise of chemical accuracy at low computational cost. But the most applicable (so-called R12) methods have not lived up to expectations. Here for the first time a comprehensive analysis of the errors in R12 methods reveals the most important source of error --- and points towards a new generation of high-accuracy methods that are applicable to large chemical systems.

2. What has motivated you to conduct this work?

Over the past few years much attention has been focused on successive possible sources of error in R12 methods, looking first at the approximation of 3-electron integrals [JCP 116, 6397 (2002); JCP 119, 5358 (2003); CPL 395, 190 (2004)].

This work confirmed that the accuracy of the 3-electron integrals was not the problem, so attention turned to the Brillouin condition approximations [Valeev, to be submitted, 2005]. These approximations also proved to have little effect.

At around the same time it was suggested that the form of the correlation factor, which until recently had been dogmatically fixed as r_12 was of great importance [FRM, ACS meeting, August 2004] and at the same meeting Ten-no provided convincing evidence of the merits of a Slater-type correlation factor [CPL 398, 56 (2004)]. We decided that the time had come for an accurate determination of the relative impact of all of the approximations made in this class of theories.

3. Where do you see this work developing in the future?

This work points towards a new class of electronic structure methods that combine the accuracy of optimized correlation factors with the efficiency of the R12 class of methods. Investigations towards this goal are already underway in a number of research groups.

4. Are there any particular challenges facing future research in this area?

The work presented, and much of the existing research in this field, is restricted to the MP2 level of theory. A central challenge is to extend the these successes to coupled cluster methods.

Analysis of the errors in explicitly correlated electronic structure theory
Andrew J. May, Edward Valeev, Robert Polly and Frederick R. Manby, Phys. Chem. Chem. Phys., 2005