[gmx-developers] virial

[Erik Lindahl]

Hi Jason,

On Nov 21, 2006, at 9:41 PM, Jason de Joannis wrote:

> From the manual and the code, it looks like Gromacs uses the atomic
> virial. I think that is besides the point.
>
> Why does Gromacs use f[i] which (I think) contains all bonded and  
> nonbonded forces? After all the derivative resp. volume scaling  
> vanishes for some of the bonded forces.

Forces are forces are forces, which give rise to pressure. There is  
no fundamental physical difference between bonded and nonbonded  
interactions.

Gromos (and some other codes) use whole-molecule-virials to reduce  
the effect of some numerical errors, but there are other drawbacks  
with this - primarily performance.

>
> Also, the manual describes the SHAKE virial but what about LINCS?

We use "Shake virial" as a generic name for virial contributions from  
constraints, so it's identical for SHAKE and LINCS.

>
> Also, where can I find a little more info on the fr->fshift  
> contribution.
> On the mailing list archive someone explained that this comes from  
> forces on the 27 unit cells. Is that explained in the manual or  
> somewhere?
>

It's mentioned in the manual where we derive the expression for  
single-sum-virial. The performance advantage of the single sum is  
also the reason why we prefer the atomic virial, btw!

Cheers,

Erik