"Forces" on BQs: Can they be stored in the RTDB
Clicked A Few Times
5:10:14 AM PDT - Thu, May 1st 2014
Hi all,

Is it possible that the "forces" on BQs could be saved to the RTDB in the future versions of NWChem? This does not seem to be the case in 6.3 or the current trunk version of qmmm_bq.F

And, as a passing observation, though they are labelled as "forces" on BQs, are they not the gradients?

[A two atom test (one QM, one BQ) gives me a "force" on the BQ which is the negative of the "gradient" on the QM atom... which would tell me that they are either both forces, or both gradients, but not one of each!]

Thanks,

Andy

Forum Regular
10:01:26 AM PDT - Thu, May 1st 2014
Hi Andy,

I think the reason that the forces/gradients on the BQ centers are not stored on the RTDB is that depending on the embedding model used QM/MM calculations can use very many BQ centers. Hence this data structure might be too big to conveniently dump it onto the RTDB. I would assume that this data is stored somewhere though and therefore there must be a way to get to it. Is there anything in particular you need this information for?

Best wishes, Huub

Clicked A Few Times
1:05:29 AM PDT - Fri, May 2nd 2014
Hi Huub,

I'm doing some housekeeping for the directly-linked interface of ChemShell/NWChem. I see the problem - yes the QM/MM models can have thousands of BQ centres.

Currently the forces are written to file, which is fine though there is a loss of accuracy with rounding, and it is a bit overly cumbersome as a write/read process.

It would be good if storing in the RTDB was perhaps optional, e.g. assuming some RTDB keyword is set, instead of writing to output? Or if the accuracy in the output could be set through an RTDB keyword?

What kind of limit on size do you assume for data stored in the RTDB? I guess this determines what can/cannot be stored there but I have no idea of the limits.

Thanks

Andy


Forum >> NWChem's corner >> NWChem functionality