# The difference between molpro and mrcc

2 years 1 month ago #879 by jamcas
I found a question that puzzled me, that is, for the same closded shell system, the triples (T) contribution for mrcc and molpro is totally different:.

molpro:
E(CCSD);-2980.779852732739
CCSD(T) correction:-0.324599006747
E(CCSD(T):-2981.10445174

mrcc:
E(CCSD);-2980.779847028194
CCSD(T) correction:-0.369297289025
E(CCSD(T):-2981.149144317219

My input file:
basis=aug-cc-pvdz
calc=ccsd(t)
scftype=rhf
dft=user
2
1.0 b88
1.0 pw91c
mem=60GB
dfbasiscor=none
dfbasis_scf=none
ccmaxit=50000
scfmaxit=20000
mult=1
charge=-1
scfdamp=0.7
scflshift=5
verbosity = 3

unit=angs
geom=xyz
7

C                -0.35087300    1.64374600    0.00250500
C                 1.01404000    1.70160700    0.00213000
C                 1.01401300   -1.70160600   -0.00211500
C                -0.35091000   -1.64372500   -0.00255200
V                 0.54132900    0.00190000   -1.06181900
V                 0.54127000   -0.00191700    1.06185300
V                -1.42858300    0.00001200   -0.00002500

I also tested N2 and I got the same result:

molpro:
E(CCSD);-109.278009380004
CCSD(T) correction:-0.015147862165
E(CCSD(T):-109.29315724

mrcc:
E(CCSD);-109.278009034846
CCSD(T) correction:-0.015291141772
E(CCSD(T):-109.293300176618

2 years 1 month ago #880 by Nike
Dear jamcas,
1) One difference is in the integral transformation for frozen core systems, but in your closed-shell system I would hope the difference wouldn't be this big.
2) I believe MRCC does UCCSD whereas in MOLPRO you can specify RCCSD or UCCSD, so you might want to show us your input file (it won't be very big anyway).
3) Are your SCF (Hartree-Fock) energies the same?
With best wishes,
Nike

2 years 1 month ago #881 by jamcas
2. This is my molpro input:
memory,300,m
file,2,r-m1-34n-avdz-ccsdt.wfu,old

gprint,basis,orbitals,civector;
gthresh,printci=0.05;

geometry={angstrom;
C                -0.35087300    1.64374600    0.00250500
C                 1.01404000    1.70160700    0.00213000
C                 1.01401300   -1.70160600   -0.00211500
C                -0.35091000   -1.64372500   -0.00255200
V                 0.54132900    0.00190000   -1.06181900
V                 0.54127000   -0.00191700    1.06185300
V                -1.42858300    0.00001200   -0.00002500
}

basis={
c=avdz
v=avdz
}

{rks,b,pw91c;
closed,47;
occ,47;
wf,94,1,0;
ACCU,14;
}

{rccsd(t);
shift,1,1;
core,31;
closed,47;
occ,47;
wf,94,1,0;
}

3.Yes, both the reference energy and the ccsd correlation energy are same.
Reference energy
molpro:-2979.279074961804
mrcc:-2979.279083044167

2 years 1 month ago #882 by kallay
Probably the reason is that the contribution of the occupied-virtual block of the Fock-operator to the (T) correction is also calculated in molpro. This contribution is zero for canonical HF. It is not zero for a KS reference but it is usually small.

Best regards,
Mihaly Kallay

2 years 1 month ago #883 by jamcas
Thanks for your reply! How to get a open-shell singlet solution in mrcc? I know gaussian can performe stable check, but I don't seem to see similar keywords in mrcc.

2 years 1 month ago #884 by kallay
You can get an open-shell singlet state with MCSCF.

Best regards,
Mihaly Kallay