Methods of Moller-Plesset Perturbation Theory

The simplest method of MP perturbation theory, MP2, is of less computational demand compared to CCSD(T) and QCISD(T). Although it is often assumed to recover most of the electron correlation energy [9], unfortunately, it may give poor results for graphene-containing systems. 

The overestimation of E of benzene dimers by MP2 is known (see, e.g., [10] and [11]) further, MP2 overestimates (, when one or morex-system takes part in intermolecular interaction [6]. MP3, on the other hand, underbinds x-containing systems [6]. None of this is observed when only o-bonds exist in the system, and MP2-predicted is sfightly higher (see (, definition adopted in this chapter in Section 11.3.1) than the value predicted by CCSD(T) (as notable examples, compare MP2-predicted j for intermolecular complexes described in [6] and [12]). 

QCISD(T) are costly (they scale as Af ) and are difficult to use for modeling systems of reasonable size. A rare example of QCISD(T) usage to model benchmarking supramolecular systems of more than 50 non-hydrogen atoms is available in an L7 set (see Section 11.2.3.2 for discussion). 

Canonical MP-methods probably have some unavoidable difficulties with graphene systems, caused by their intrinsic approach. MP2C/CBS and MP2.5/CBS perform well however, they are to some extent costly compared to DFT methods that usually scale as or at least N.  

---