# [gmx-users] Virial from non-bonded forces

Omololu Akin-Ojo prayerz4users at yahoo.com
Tue Feb 10 16:01:42 CET 2009

Hi,

I have two main issues:

1. In Appendix B of the GROMACS user manual, there is a discussion of the calculation of the virial. The top of the second page of the appendix (just before B.1.2) states, "In a triclinic system there are 27 possible images of i, ...".
My question is: In this case, there should also be 27 possible forces on particle  j due to i. Granted, only the force corresponding to the shortest i-j distance will be non-zero but when the forces are multiplied by \delta_i, (in Eq. B.11) the correct force (out of all the 27 possible ones) should be multiplied by the corresponding \delta_i.
In other words, if we label each of the 27 possible \delta_i by n ((\delta_i)^n), we should do the same with the forces F_i  ((F_i)^n) and the sum in B.11 should also contain a sum over n.
For ver. 3.3.1, I see that in src/mdlib/calcvir.c each of the different (\delta_i)^n is used but for fixed i, the same force F_i is used irrespective of the value of n.
(See also, J. Chem. Phys. vol. 117, p2449 (2002), the 3rd line from the bottom of the 2nd column on the first page.)

2. How does GROMACS calculate the virial for charged virtual sites using the Ewald summation (or SPME)? One needs to redistribute the forces before calculating the virial but if one uses this approach one will need not only 27 possible \delta_i but, technically, an infinite number.