Hi all,
The other day a colleague and I were working on a QM/MM calculation using Amber and PUPIL, and we got a large (~40 kcal/mol) discrepancy between, on the one hand, the SCF energy (reported as PUPESCF) in Amber, and on the other, the energy reported by Gaussian and converted from Hartrees into kcal/mol using current NIST conversion factors (see www.nist.gov).
It appears that most of the discrepancy comes from the value of EV_TO_KCAL in Amber, which is currently set to 23.061 kcal mol^-1 eV^-1, a value commented as "Dynamo's conversion". Using the current NIST values, though, the following conversion factors are obtained:
Thermochemical calorie (4.184 J/cal by definition): 23.0605497
20-degree calorie (c. 4.182 J/cal): 23.0715781
Other definitions of the calorie I've been able to find (e.g., the 15-degree and international steam table calories) contain more J/cal, and therefore will reduce the kcal mol^-1 eV^-1 if used. Given this observation, I guess that the thermochemical calorie has been used in Amber (also cf. the value of J_PER_CAL).
When the numbers involved are large (in the millions of kcal/mol), the difference of ~0.0005 J/cal becomes important.
EV_TO_KCAL seems to be used only when an energy value is coming down from QM/MM. Perhaps the Amber QM/MM algorithm has been optimised with that value in mind. Is it better, then, to revise the value for everything, or only for those calculations that use PUPIL?
Thanks,
Ben
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 02 2010 - 11:00:04 PDT
