Next: 2. Methods
Up: Performance of the Effective-characteristic-polynomial
Previous: Performance of the Effective-characteristic-polynomial
1. Introduction
Recently, methods have been
discussed for the computation of correlation energy estimators based on
Møller-Plesset (MP) perturbation theory that
may also be regarded as accelerating the
convergence of the MP series
[1],[2],[3],[4],[5],[6],[7],[8],[9],[10].
In [8], some methods
were discussed that are based on MP calculations of fourth order (MP4)1:
- The Feenberg energy of fourth order F4
[11],[12],[10],[8],
- The Padé approximant [2/2]
2
[13],[14]
- The 2 approximation that is
computed as the zero of an effective characteristic polynomial [15],[16],[17],[18],[19],[20],[21],[22]
of degree 2.
All these approximations are calculated from the terms of the MP
series with negligible extra effort.
Explicit formulas for these methods are given in Section 2. All
the methods are size-extensive [8],[9].
Also, test calculations were reported in [8]
for a rather large number of small molecules (BH, HF, CH2, H2O,
NH2,
NH3, CO, C2H2, O3, CN) for which Full Configuration
Interaction (FCI) or Coupled-Cluster (CC) including Single (S), Double (D) and
Triple (T) excitations, i.e., CCSDT results are available, mainly for basis
sets of double zeta (DZ) or DZ plus polarization (DZP) quality.
It was shown that (for the treated cases) the 2 method yields very good
approximations for the energy if the values of F4, [2/2] and 2
are sufficiently close
together. If the latter criterion is satisfied, all three methods
improve the MP4 values considerably. The above
criterion to accept the result of the perturbation calculation is especially
important since it is well-known that the quality of the MP results
deteriorates for greater distances from the equilibrium geometry. Thus, the
criterion allows to judge the quality of the MP series.
The criterion will be further discussed in Section 2.
In the present contribution, we report further studies of the performance of
the 2 and also
the F4 and [2/2] methods. In particular, the dependence on the choice of
the basis sets is important for the application of the methods. We limit
attention to diatomic systems. Also, we report
some results concerning the quality of potential energy surfaces and the
calculation of spectroscopic constants.
Next: 2. Methods
Up: Performance of the Effective-characteristic-polynomial
Previous: Performance of the Effective-characteristic-polynomial
Herbert H. H. Homeier (herbert.homeier@na-net.ornl.gov)