ct7b00352_si_001.pdf (132.19 kB)
Treating Subvalence Correlation Effects in Domain Based Pair Natural Orbital Coupled Cluster Calculations: An Out-of-the-Box Approach
journal contribution
posted on 2017-06-12, 00:00 authored by Giovanni Bistoni, Christoph Riplinger, Yury Minenkov, Luigi Cavallo, Alexander A. Auer, Frank NeeseThe validity of the
main approximations used in canonical and domain
based pair natural orbital coupled cluster methods (CCSD(T) and DLPNO-CCSD(T),
respectively) in standard chemical applications is discussed. In particular,
we investigate the dependence of the results on the number of electrons
included in the correlation treatment in frozen-core (FC) calculations
and on the main threshold governing the accuracy of DLPNO all-electron
(AE) calculations. Initially, scalar relativistic orbital energies
for the ground state of the atoms from Li to Rn in the periodic table
are calculated. An energy criterion is used for determining the orbitals
that can be excluded from the correlation treatment in FC coupled
cluster calculations without significant loss of accuracy. The heterolytic
dissociation energy (HDE) of a series of metal compounds (LiF, NaF,
AlF3, CaF2, CuF, GaF3, YF3, AgF, InF3, HfF4, and AuF) is calculated at
the canonical CCSD(T) level, and the dependence of the results on
the number of correlated electrons is investigated. Although for many
of the studied reactions subvalence correlation effects contribute
significantly to the HDE, the use of an energy criterion permits a
conservative definition of the size of the core, allowing FC calculations
to be performed in a black-box fashion while retaining chemical accuracy.
A comparison of the CCSD and the DLPNO-CCSD methods in describing
the core–core, core–valence, and valence–valence
components of the correlation energy is given. It is found that more
conservative thresholds must be used for electron pairs containing
at least one core electron in order to achieve high accuracy in AE
DLPNO-CCSD calculations relative to FC calculations. With the new
settings, the DLPNO-CCSD method reproduces canonical CCSD results
in both AE and FC calculations with the same accuracy.