SCF convergence for post SCF calculations

Just Got Here

9:18:10 AM PDT - Wed, Nov 2nd 2011
I'm wondering if I can change the default SCF convergence threshold when doing post Hartree-Fock calculations. The default tolerance is 1e-6. If I change this to 1e-5 in my input and do an SCF only calculation (task scf), then the value of 1e-5 is used. However if I set the SCF tolerance to 1e-5 in my input and request an MP2 or CCSD run, the default value of 1e-6 is used during the SCF regardless of my setting in the input file. Is there something that I am doing wrong?

Forum Vet

3:07:05 PM PDT - Wed, Nov 2nd 2011
You are not doing anything wrong. To run MP2 or CCSD your SCF orbitals need to be reasonably converged, and we have set the convergence threshold (tolerance) for that to be 1e-6. I.e., at the beginning of an MP2 or CCSD task, the code will check if the convergence is accurate enough, and if not it will tighten the convergence and rerun the SCF. Unfortunately, right now this is hardcoded, and cannot simply be overwritten by an environment variable. Bert Quote:Detaylor Nov 2nd 4:18 pm I'm wondering if I can change the default SCF convergence threshold when doing post Hartree-Fock calculations. The default tolerance is 1e-6. If I change this to 1e-5 in my input and do an SCF only calculation (task scf), then the value of 1e-5 is used. However if I set the SCF tolerance to 1e-5 in my input and request an MP2 or CCSD run, the default value of 1e-6 is used during the SCF regardless of my setting in the input file. Is there something that I am doing wrong?