[gmx-developers] virial symmetry

[Alexander Herz]

Hi guys,

I'm having some problems with the walls feature (new in gromacs 4.0).

If I understand correctly then I do get a symmetric virial tensor even for
systems which are inhomogenous as long only pair potentials are used

(Theta=-0.5*sum(r_ij*F_ij) where r_ij is r_j-r_i (so the distance
between to particles acting on each other)
and F_ij is the force acting on them;with F_ij=-F_ji the contributions
cancel out and the viral is symmetric )

Now if I use 2 walls with different density I get an asymmetric virial
and looking at the wall's source
it seams that for walls not r_ij is used but absolute wall coordinate
(box_zz for wall2 and 0 for wall 1
giving a zero contribution to the viral for wall1). Also for the
individual atoms it seams the absolute position
is used rather than r_ij (looking at calc_vir() in calcvir.c).

So why do the walls use the absolute z coordinate instead of
particle_z-wall_z and why
do I get an asymmetric virial for walls with different density but not
for the same
system without walls which is still asymmetric in z (some crystal
occupies the lower z half off the system, the upper
half is filled with water).

Thx,
Alex