| 12:41:55 AM PDT - Mon, May 13th 2019 |
|
Dear NWChem-Team
For our research we require gradients at CCSD(T) level of theory and NWChem TCE's is one of the few options available.
The specific question is about the functionality of TCE to read SCF vectors from a previous calculation, e.g. first a SCF calculation is carried out selecting a smaller number of processors and the resulting MO vectors are then read by a second CCSD(T) job assigned to a much larger number of nodes (due to the effort associated to the correlated treatment).
This feature to read MO vectors is listed under the SCF "VECTORS" keyword, but it is not mentioned for the TCE module.
Is there way I can (i) make the TCE read precomputed MO vectors and (ii) skip the SCF-step and directly launch a correlated
calculation like MP2, CC, etc.?
Possible the second point (skipping the SCF part) could be realised by setting the SCF iteration to 1 and catching the error message via the python interface (similar as the example in the manual):
>
> python
> try:
> task_energy('scf')
> except NWChemError, message:
> print 'Error from NWChem ... ', message
> end
>
> task python
>
Thanks for your comments and suggestions.
Best wishes,
Thomas
|