Effective potential at each charge site

[peterbygrave]

Does this library provide the periodic effective potential matrix (i.e. the 1/r_{ij} matrix for molecular system) at each atomic/charge site in the system?

In other libraries e.g. OpenMM, sampling the potential (i.e. get q_j/r_{ij} at site i) at a charge site just returns infinity and it would be awesome to have this feature.

Many thanks!

[andysim]

Hi @peterbygrave. That would be very easy to add in a janky way for sure; all of the components needed are there. The potential has three components: direct, reciprocal and self terms. The reciprocal and self terms are already implemented (the self term corrects for the fact that we can't simply exclude the "probe" atom from the potential the way we can in the pairwise loops for direct space terms).

The formula for the direct term is very simple, but implementing it properly is a little trickier. The problem is figuring out all the pairs in a smart way. The proper solution is to use some kind of a non bonded list generator to keep everything O(N) in code, and apply the minimum image convention to correctly wrap around periodic boundaries. I didn't add any code like this to avoid scope creep; from there it's a slippery slope to becoming OpenMM itself! Hacking in an O(N^2) solution that doesn't use a list generator should be relatively easy for me to do and, depending on your target application, it might be sufficient. Adding some simple list generating code wouldn't be too difficult on top of that, but I would rather try without it first, to avoid coding up unnecessary features.

What kind of system size are you talking about, and is this something that you would want to evaluate just a handful of times, or many?

[peterbygrave]

So we are doing QM/MM and periodic electrostatics can quickly get out of control, but also can have a variety of system sizes. We have a canonical Ewald scheme, but this won't suffice as system sizes grow. Currently O(N^2) would be absolutely great. Could you let me know which function would give the reciprocal potential? This will give me a good place to start learning the interface. Is there anything I can do to help?

…

On Sat, 12 Sep 2020, 14:11 Andy Simmonett, ***@***.***> wrote: Hi @peterbygrave <https://github.com/peterbygrave>. That would be very easy to add in a janky way for sure; all of the components needed are there. The potential has three components: direct, reciprocal and self terms. The reciprocal and self terms are already implemented (the self term corrects for the fact that we can't simply exclude the "probe" atom from the potential the way we can in the pairwise loops for direct space terms). The formula for the direct term is very simple, but implementing it properly is a little trickier. The problem is figuring out all the pairs in a smart way. The proper solution is to use some kind of a non bonded list generator to keep everything O(N) in code, and apply the minimum image convention to correctly wrap around periodic boundaries. I didn't add any code like this to avoid scope creep; from there it's a slippery slope to becoming OpenMM itself! Hacking in an O(N^2) solution that doesn't use a list generator should be relatively easy for me to do and, depending on your target application, it might be sufficient. Adding some simple list generating code wouldn't be too difficult on top of that, but I would rather try without it first, to avoid coding up unnecessary features. What kind of system size are you talking about, and is this something that you would want to evaluate just a handful of times, or many? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACHGZPR65ESTM5IIPCIZ6MTSFNXORANCNFSM4RHK3VOA> .