[gmx-developers] possible problem in ewald_util.c?

[Mark Abraham]

On 15/11/2011 11:52 PM, David van der Spoel wrote:
> Hi,
>
> I just stumbled over this piece of code in the ewald correction routine:
>
>               rinv   = gmx_invsqrt(dr2);
>               rinv2  = rinv*rinv;
>               dr     = 1.0/rinv;
>               v      = gmx_erf(ewc*dr)*rinv;
>               vc     = qqL*v;
>
>
> Shouldn't that be an erfc rather than an erf?

No, that bit of code is correcting for atoms that should have been 
excluded from the reciprocal-space calculation for this particular atom 
because they were excluded from the real-space calculation (i.e. bonded 
neighbours). See eq 2.5 of the Essman SPME paper and the comment on the 
enclosing loop.

Mark