On Thu, Jan 14, 2010, Alan wrote:
>
> However, what exactly means when I see something like this:
>
> [snip]
> sqm energy: 160 -238.2213 0.000754
>
> QMMM: WARNING!
> QMMM: Unable to achieve self consistency to the tolerances specified
> QMMM: No convergence in SCF after 1000 steps.
> QMMM: Job will continue with unconverged SCF. Warning energies
> QMMM: and forces for this step will not be accurate.
> QMMM: E = -0.2644E+07 DeltaE = -0.1397E-08 DeltaP = 0.3227E-08
> QMMM: Smallest DeltaE = -0.4657E-09 DeltaP = 0.3227E-08 Step = 3
>
> sqm energy: 170 -238.2214 0.001097
> sqm energy: 180 -238.2214 0.000298
> Final SCF energy is -238.22140565750306
This is OK. We don't have a universal good solution for 9000 molecules.
I would recommend applying the bugfix, hence stopping the calculation quickly
when you encounter a problem. You can then go back and run the ones that fail
with different criteria.
[Note that the message above means you *almost* converged, DeltaP was quite
small. Sometimes, however, convergence will fail quite badly, with DeltaP
of 0.1 or so. Then, it is useless to try to continue.]
> What exactly these lines means:
>
> QMMM: Job will continue with unconverged SCF. Warning energies
> QMMM: and forces for this step will not be accurate.
>
> Is it only "this step" that will be compromised? Or my whole calculation is
> doomed?
Just "this step".
>
> For the example above, if I use -ek "qm_theory='AM1', grms_tol=0.05,
> tight_p_conv=0, scfconv=1.d-8," I got NO Warnings, so would be a better
> result (Final SCF energy -235.67194987277617) than the original one above?
Since your energy here is higher than with a good convergence, the results
should be slightly worse. However, there may only be a negligible difference
in the am1-bcc charges -- it is the charges you want, not the energy.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jan 14 2010 - 09:00:03 PST
