[gmx-users] How to add another electrostatic summation method in Gromacs

[Mark Abraham]

Shuangxing Dai wrote:
> Thank you for your help. I attached the page of the form of potential. 
> The equation (5.13) gives all the potential energy of neighbours of one 
> ion in a truncation sphere.  Rc is cut-off. Since this potential depends 
> on all the neighbours in a cut-off distance, I cannot give the pair 
> interaction because in this method, only the total potential energy of 
> one ion maks sense. In Ewald summation, both the real space and 
> reciprocal space summation can be written as the summation of pair 
> potential. But in wolf, I cannot do that.

It looks separable into a double sum over atom pairs from within a 
cutoff, since the value of that limit is constant.

> Since I cannot use Ewald without peoriodical boundary conditions while 
> this Wolf method can, so I cannot simply use Gromacs to do my 
> simulation. The question is which files needed to be modified if I want 
> to add a new one? Is there any direct way? Or I should read all the 
> whole code?

You can start by reading the kernel code generated by mknb during the 
make process and found in src/gmxlib/nonbonded/nb_kernel. If you play 
with the flags to mknb you can get comments embedded in the C files. 
Find the one that does Ewald without table lookups and consider how 
you'd need to change it to suit your needs.

Mark