[gmx-users] Force-switch algorithm

[Mark Abraham]

Hi,

On Tue, Jan 5, 2016 at 7:26 AM Mark Abraham <mark.j.abraham at gmail.com>
wrote:

> Hi,
>
> On Tue, Jan 5, 2016 at 4:46 AM Justin Lemkul <jalemkul at vt.edu> wrote:
>
>>
>>
>> On 1/4/16 10:41 PM, Pallavi Banerjee wrote:
>> > Hi Mark,
>> >
>> > Thank you for your reply.
>> >
>> > I want to implement this wacky idea because a part of the potential of
>> the
>> > system needs to be input in a tabulated form (referring to Iterative
>> > Boltzmann Inversion) and the other part (solvent-solvent) interaction is
>> > defined by sigma and epsilon, which I again input in the form of a table
>> > since GROMACS does not  support both tabulated and non-tabulated at the
>> > same time.
>> >
>> > So, then, am I right to understand that eqn 4.28 of manual-5.1 is the
>> > functional form of potential of the force-switch method, and that there
>> is
>> > no documentation provided for potential-switch?
>> >
>>
>> It may not be in the GROMACS manual, but is it not the same as what is in
>> Steinbach and Brooks JCC (1994) 15: 667-683, like the force-switch is?
>> This is
>> the CHARMM potential switching, which should be the same, IIRC.
>>
>
> Thanks, but
> https://github.com/gromacs/gromacs/blob/master/src/gromacs/gmxlib/nonbonded/nb_generic.cpp#L137 suggests
> they are not the same.
>

I did some cranial archaeology upon Erik, which led me to
commit 4e61f8f4b26da, which led me to
https://github.com/gromacs/gromacs/blob/master/src/gromacs/tables/forcetable.cpp#L961.
The potential-switch function is indeed in the spirit of Steinbach &
Brooks, but smoother because it has higher-order continuity. Several others
have also implemented switching functions in this way. (But I don't think
anybody now recommends you use it for Coulomb...)

I've documented the potential-switch modifier for the upcoming 5.1.2
release - see https://gerrit.gromacs.org/#/c/5526/

Mark