| 12:25:01 PM PDT - Fri, Sep 6th 2013 |
|
Hi Huub,
Thanks for the suggestion. I tried this and the program has not crashed but I am dubious that any projection has occurred. The reason I am suspicious is that the molecule I am working with is an open shell doublet and the SCF can converge erroneously to a higher energy solution. With a small basis set it has no problem in finding the correct solution and I can tell it is the correct solution by looking at the S^2 value. For the smaller basis set (6-311G**) the S^2 value is 0.7553 and the total energy is -398.225929335025 Hartree. If I perform the calculation with the larger basis set (6-311+G(2d,2p)) without any reading in of the vectors then the S^2 value is 1.1236 and the total energy is -398.113823075940 ---->>>> yes that is HIGHER in energy than it is with the smaller basis set <<------ This is because the SCF is stuck in a meta-stable solution. This has nothing to do with loose convergence criteria either, for example this is the final iteration:
41 -398.1138230759 8.43D-08 9.58D-09 605.6
So this is where the idea for projection comes in. If I try your suggestion above I get the same answer as if I didn't read anything in. To me this suggests that no projection has happened. Is there any other things that I could do here?
The documentation you sent me has the following directive for projection:
vectors input project "small basis" small.movecs
but if I do this the NWchem crashes
Thanks Geoff.
|