[gmx-users] "Correcting invalid box"

[Berk Hess]

> > I'm running gromacs-3.2.0beta1.
> > > The following message is appearing in md.log sporadically:
> > > Correcting invalid box:
> > old box (3x3):
> >    old box[    0]={ 7.43601e+00,  0.00000e+00,  0.00000e+00}
> >    old box[    1]={ 0.00000e+00,  7.43601e+00,  0.00000e+00}
> >    old box[    2]={ 3.71886e+00,  3.71886e+00,  5.25805e+00}
> > new box (3x3):
> >    new box[    0]={ 7.43601e+00,  0.00000e+00,  0.00000e+00}
> >    new box[    1]={ 0.00000e+00,  7.43601e+00,  0.00000e+00}
> >    new box[    2]={ 3.71886e+00, -3.71715e+00,  5.25805e+00}
> > > What does this mean? Should I be alarmed?
>No, at least I don't think so, maybe Berk can comment? Due to pressure
>coupling the box vectors have been modified, however, the length of the
>box[ZZ] vector after coupling is slightly longer than the others, due to
>numerical inaccuracy. This is corrected by this routine. What do your
>original box vectors look like?
>
> > > The simulation is being done with isotropic Parrinello-Rahman pressure
> > coupling, in a dodecahedron box.
> > > cheers,
> > > Malcolm

The box vectors do not satisfy the criteria on page 12 of the manual.
Therefore box 2 is modified by subtracting box 1, this does not alter
force or so in any way, it only makes the computation easier.

The problem is that the box vectors should always satisfy the criteria
when isotropic scaling is used. From the box you can see that
box[2][XX] and box[2][YY] are not equal to half box[0][XX] and box[1][YY].
I looked at the Parrinello-Rahman code and it looks incorrect as the
off-diagonal box elements change proportionally to the off-diagonal
compressibility, which is 0. My guess would be that all box elements
should use the same single compressibility value.

Erik, could you fix this?

Berk.

_________________________________________________________________
MSN Search, for accurate results! http://search.msn.nl