[gmx-developers] likely an error in gromacs

[Dongsheng Zhang]

Dear GMX developers,

I used a user-defined potential for non-bonded interactions. By chance I
found there was likely an error in potential energy calculation. To make
the story short, I will just tell you what different tests I have done.

In the test, I calculated the potential energy between one pair by using
user-defined potential (tabulated potential). The coordinates of these
two particles are:
42.04700       37.95200       42.47800
42.99200       37.88300       42.77700

The distance between them is 0.99357250826 (got from my own code), it is
between 0.992 and 0.994

I varied the tabulated potential to see how gromacs and my own code to
correspond:

Table one (only relevant part): (To make it simple, I set the second
derivative term zero)

0.992	0	0	0		0		0	0
0.994	0	0	10		0		0	0

The result from my code is 7.86254128, The result from gromacs (by
defining energy groups to get the interaction of one pair) is 7.802608

Table two:

0.992	0	0	0		0		0	0
0.994	0	0	100		0		0	0

The result from my code is 7.86254128e+01. The result from gromacs is
78.562798, which is not 10 times of 7.802608.



Could anyone give me a hint what happens in gromacs? Thank you a lot!


All the best!


Dongsheng