[gmx-users] Ewald potential correction term?

[Erik Lindahl]

Rui Qiao wrote:

> Dear all:
>         I look into the code on Ewald method, and it seems that when there
> is non-zero system dipole, one should apply a correction term on the
> potential energy. Gormacs does implement this in function
>
>         ewald_LRcorrection()    in file: ewald_util.c
>
>         but two issues are:
>         1. In the manual, it seems that such option is not documented as a
> mdrun option;
>         2. In the implementation, the correction term is
>            -1/(6*epison*V)*|M|^2
>            but in all other papers, it seems that it should be
>            +1/(6*epison*V)*|M|^2
>            could this be a bug?

Hm - I recall David and me talking around this about a year ago. Of course
it *could* be a bug, but it is kind of tricky since this is sort of a
compensation
for an effect... I wont swear by it,  but if I recall correctly I had it the
other way
around first, until David noted that his simulations became more stable when
the
sign was reversed and went on to show that it was indeed right too...

David, do you remember the discussion?

Cheers,

Erik