MP2 Correlation energy too big
Gets Around
10:22:34 PM PDT - Fri, May 3rd 2013
I run RI-MP2 calculation of cubane molecule (C8H8) in a different bases.
cc-pCVTZ, cc-pCVQZ looks good, but with cc-pCV5Z basis I have got unrelailable MP2 energy.

my input file is (molecular geometry is almost equilibrium):
title "cubane cc-pV5Z MP2-RI energy"

geometry
 symmetry Oh
 H                     1.41063149     1.41063149     1.41063149
 C                     0.78371973     0.78371973     0.78371973
end

basis
 H library cc-pV5Z
 C library cc-pCV5Z
end

basis "ri-mp2 basis"
 H library cc-pV5Z-ri
 C library cc-pV5Z-ri
end

task rimp2


output energy is:
          -------------------------------------------
          SCF energy                -307.507255162945
          Correlation energy       -4369.270156917300
          Total MP2 energy         -4676.777412080244
          -------------------------------------------

Gets Around
6:54:42 AM PDT - Mon, May 6th 2013
Use of spherical functions can do the trick!

title "cubane cc-pV5Z MP2-RI energy"

geometry
symmetry Oh
H 1.41063149 1.41063149 1.41063149
C 0.78371973 0.78371973 0.78371973
end

basis spherical
H library cc-pV5Z
C library cc-pCV5Z
end

basis "ri-mp2 basis"
H library cc-pV5Z-ri
C library cc-pV5Z-ri
end

scf
direct
end

task rimp2


output energy is:
-------------------------------------------
SCF energy                -307.507287526028
Correlation energy          -1.815754093360
Total MP2 energy          -309.323041619388
-------------------------------------------



Gets Around
12:15:58 AM PDT - Fri, Jun 7th 2013
Dunning basis sets are designed to be used with spherical functions so the cartesian functions cause very bad linear dependency problems in some cases. Spherical functions aren't a trick; they are the right way to run this calculation.


Forum >> NWChem's corner >> Running NWChem