[gmx-users] short and long range coulomb with pme

[Erik Lindahl]

nicolis at guideo.fr wrote:

>Hello,
>I simulate a CA2+ in a 3 nm SPC water box with pme (actually in different box sizes to see the effect on free energy) without any counter ion.
>When I check the interactions with g_energy I get constant 0 for the LR coulomb interactions for CA-CA, CA-SOL and SOL-SOL, but I get a value for the total LR_Coulomb.
>If I check the SR Coulomb or the LJ, I get the totals equal to the sum of the CA-CA, CA-SOL and SOL-SOL contributions.
>Why this is not the case for LR coulomb? Am I missing soemthing about the parameters rlist and rcoulomb? Should I give a value to rcoulomb-switch?
>
>Thank you,
>Yiannis
>
No, the problem is that it isn't possible to separate group 
contributions in the PME algorithm. I'm pretty sure it says that 
somewhere in the manual too :-)

PME works by first spreading all charges onto a grid and then doing a 
3D-FFT of the grid and convoluting it. If we wanted group contributions 
we would have to spread
each group onto a separate grid, perform separate FFT's and convolute 
all possible combinations of groups. That would be prohibitively 
expensive, although it will probably be possible in the next version of 
Gromacs just for fun :-)

Cheers,

Erik