Conventional and Explicitly Correlated ab Initio Benchmark Study on Water Clusters: Revision of the BEGDB and WATER27 Data Sets

Benchmark
ab
initio energies for BEGDB and WATER27 data sets have been re-examined
at the MP2 and CCSD­(T) levels with both conventional and explicitly
correlated (F12) approaches. The basis set convergence of both conventional
and explicitly correlated methods has been investigated in detail,
both with and without counterpoise corrections. For the MP2 and CCSD–MP2
contributions, rapid basis set convergence observed with explicitly
correlated methods is compared to conventional methods. However, conventional,
orbital-based calculations are preferred for the calculation of the
(T) term, since it does not benefit from F12. CCSD­(F12*) converges
somewhat faster with the basis set than CCSD-F12b for the CCSD-MP2
term. The performance of various DFT methods is also evaluated for
the BEGDB data set, and results show that Head-Gordon’s ωB97X-V
and ωB97M-V functionals outperform all other DFT functionals.
Counterpoise-corrected DSD-PBEP86 and raw DSD-PBEPBE-NL also perform
well and are close to MP2 results. In the WATER27 data set, the anionic
(deprotonated) water clusters exhibit unacceptably slow basis set
convergence with the regular cc-pVnZ-F12 basis sets,
which have only diffuse s and p functions. To overcome this, we have
constructed modified basis sets, denoted aug-cc-pVnZ-F12 or aVnZ-F12, which have been augmented with diffuse functions
on the higher angular momenta. The calculated final dissociation energies
of BEGDB and WATER27 data sets are available in the Supporting Information.
Our best calculated dissociation energies can be reproduced through
n-body expansion, provided one pushes to the basis set and electron
correlation limit for the two-body term; for the three-body term,
post-MP2 contributions (particularly CCSD-MP2) are important for capturing
the three-body dispersion effects. Terms beyond four-body can be adequately
captured at the MP2-F12 level.