mkccsd geometry optimization


Just Got Here
Dear NWChem developers,

I've compiled 6.5 version but unfirtunatelly cannot run optimization at mkccsd level.
The same problem I've met with version 6.3 where after http://nwchemgit.github.io/images/Mrcctmp.patch.gz
patching situation even situation became worse - it could not reach convergence with mkccsd for hte startin geometry


input:

echo
start tce_mrcc_mkcc
memory stack 1000 mb heap 100 mb global 500 mb verify
geometry units au
H 0.00000000 -2.27289450 -1.58834700
O 0.00000000 0.00000000 -0.01350000
H 0.00000000 2.27289450 -1.58834700
end
basis spherical
O library cc-pvtz file /work/02018/sokovyty/Nwchem-6.5/src/basis/libraries/cc-pvtz
H library cc-pvtz file /work/02018/sokovyty/Nwchem-6.5/src/basis/libraries/cc-pvtz
end
charge 0
scf
rohf
singlet
thresh 1.0e-10
tol2e 1.0e-10
end
tce
mkccsd
thresh 1.0e-10
targetsym a1
maxiter 500
io ga
tilesize 18
end
mrccdata
se4t
root 1
cas 2 2 # Please make sure the references generated are correct.
end

task tce OPTIMIZE


error massage:


tce_hash: key not found                   0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
34: task tce OPTIMIZE
------------------------------------------------------------------------
------------------------------------------------------------------------
This error has not yet been assigned to a category
------------------------------------------------------------------------

Clicked A Few Times
MRCC formalisms have been designed for single point calculations only.

Best,
Karol

Gets Around
Quote:Karol Feb 13th 3:30 am
MRCC formalisms have been designed for single point calculations only.

Best,
Karol

It is not true. One can optimize geometry, saddle point and calculate frequencies with MRCC formalisms by using finite differences.
In this case I think MRCC iteration is unstable "Eigenvectors swaps", so you shell use another set of references orbitals (not cas 2 2).


Forum >> NWChem's corner >> NWChem functionality