Hi,
I've just performed three calculations on the same transition metal containing molecule using EOM-CCSD (keyword CCSD), EOM-CCSDT (keyword CCSDT) and CREOMSD(T) (keyword CREOMSD(T)) and in all cases the same basis-set cc-pVDZ and the frozen core approximation. I'm a bit surprised by the results. The lowest excitation energies that I find are (when asking for 4 roots per irrep in CCSD and 3 roots per irrep in the other methods):
With EOM-CCSD:
2.34003 (b1)
3.05201 (a2)
3.19803 (b1)
3.36922 (a1)
with EOM-CCSDT
2.13482 (b1)
2.74294 (a2)
2.96335 (b1)
3.31140 (a1)
with CREOMSD(T)
2.62815 (b1)
3.35235 (a2)
3.44494 (b1)
3.46959 (a1)
Where for CREOMSD(T) I've given what is given behind "CR-EOMCCSD(T) excitation energy (eV)"
Now I would have naively assumed that the CREOMSD(T) values would lie in between those obtained by EOM-CCSD and EOM-CCSDT, and closer to the latter than the former, but that is clearly not the case. Am I missing something?
Also, in what I guess is the original paper for the CREOMSD(T) method (JCP, 2004, 120, 1715) there are different variants on the method discussed (Ia-III). This makes me wonder which is the variant included in NWChem.
Thanks in advance,
Martijn
P.S. this is my CREOMSD(T) input file:
START tio2_monomer
MEMORY 1800 MB
GEOMETRY units an autosym
Ti 0.0000000 0.0000000 0.6131572
O -1.3586178 0.0000000 -0.3066726
O 1.3586178 0.0000000 -0.3066726
END
BASIS
* library cc-pvdz
END
SCF
SINGLET
RHF
END
TCE
CREOMSD(T)
FREEZE CORE ATOMIC
NROOTS 3
thresh 1.0d-5
2eorb
2emet 13
attilesize 40a
tilesize 15
END
TASK TCE ENERGY
Where for EOM-CCSDT I replace the TCE block with:
TCE
CCSDT
FREEZE CORE ATOMIC
NROOTS 3
thresh 1.0d-5
tilesize 8
END
And for EOM-CCSD:
TCE
CCSD
FREEZE CORE ATOMIC
NROOTS 4
thresh 1.0d-5
2eorb
2emet 13
attilesize 40
END
|