MP2FIT auxillary basis set for DFMP2


#1

I would like to run DFMP2 calculations using MP2FIT auxiliary basis sets. Is that possible to do that using psi4? I can do that using Molpro via the following input file, and basically want to do the same using psi4.

===Molpro input ============================================
memory,500,m

geomtyp=xyz
geometry={
10

C -1.888896 -0.179692 0.000000
O -1.493280 1.073689 0.000000
O -1.170435 -1.166590 0.000000
H -2.979488 -0.258829 0.000000
H -0.498833 1.107195 0.000000
C 1.888896 0.179692 0.000000
O 1.493280 -1.073689 0.000000
O 1.170435 1.166590 0.000000
H 2.979488 0.258829 0.000000
H 0.498833 -1.107195 0.000000
}

basis,aV5Z

DF-HF,DF_BASIS=aV5Z/JKFIT
DF-MP2,DF_BASIS=aV5Z/MP2FIT

======================================================


#2

The MP2FIT aux.basis sets are named with -RI suffix in PSI4

They are usually automatically assigned if you request a DF-MP2 calculations, but can be explicitly requested as well:
set df_basis_mp2 aug-cc-pV5Z-RI


#3

@hokru is correct. SCF and MP2 are density-fitted by default in psi4, so energy('mp2/aug-cc-pV5Z') will apply fitting basis sets according to http://psicode.org/psi4manual/master/basissets_byfamily.html.


#4

I am trying to reproduce Lori’s Molpro 10 calculations for formic acid dimer in this paper: J. Chem. Theory Comput. 2014, 10, 49−57.

here is what I get from Molpro 18 and Psi4 for the DF-MP2/a5Z energies

paper | -379.14598836
molpro18 | -379.14598701
psi4 | -379.36355182

I am wondering if the RI basis sets in psi4 are the same as MP2FIT ones in Molpro, why Psi4 numbers do not agree with the other two. What is the underlying reason if it is not the basis set?

here is how my Psi4 input looks like:

=======Psi4 input=======================
molecule {
0 1
C -1.888896 -0.179692 0.000000
O -1.493280 1.073689 0.000000
O -1.170435 -1.166590 0.000000
H -2.979488 -0.258829 0.000000
H -0.498833 1.107195 0.000000
C 1.888896 0.179692 0.000000
O 1.493280 -1.073689 0.000000
O 1.170435 1.166590 0.000000
H 2.979488 0.258829 0.000000
H 0.498833 -1.107195 0.000000
}

memory 60 GB

set df_basis_mp2 aug-cc-pV5Z-RI

energy(‘MP2/aug-cc-pV5Z’)

=======================================


#5

Different auxiliary basis sets shouldn’t cause an error on the order of 0.2 kcal, let alone hartrees, for a system this size. You’d need @loriab to know for sure, but my hunch is that the difference is with the frozen core approximation. A quick skim of the methods section doesn’t mention the approximation either way, but it’s safe to assume the frozen core approximation is used if you see basis sets that look like pVXZ instead of pCVXZ. Psi does not invoke the frozen core approximation by default, but Molpro 10 does, and I imagine Molpro 18 does as well.


#6

I did a “set freeze_core true” option in the psi4 input and now all the numbers agree.

units are hartree:
original paper ====> -379.14599
molpro18 ====> -379.14599
psi4 ====> -379.36355
psi-4 frozen core ====> -379.14599