[gmx-users] energy conservation: shift vs shifted user potential

[Anja Kuhnhold]

Dear gmx-users,

I have a problem concerning energy conservation when using user potentials (tables).
I'm using gromacs 4.5.4
I simulate a simple Lennard-Jones(6-12) +Fene polymer melt (1600 chains a 10 beads in a 26x26x26 periodic box).

I tried different vdwtypes (cutoff always 3.24):
The cut-off version does not conserve energy -- okay.
The shift and switch versions conserve energy -- fine.

Now I wanted to do the same with user tables:
Simple Lennard-Jones table gives really the same results as the cut-off version -- good.

But if I use a table with shifted Lennard-Jones potential it is not comparable to the shift version
and the energy is not conserved -- ?

I use a shift function as written in the manual (chapter 4.1.5) -- there must be a factor alpha added in the constants A and B --
(r1=0).

The parameters are the same for shift version and shifted user version.

Has someone an idea why the shifted user potential doesn't work in this way?

Here is the mdp:


integrator		= md-vv
dt			= 0.0035
nsteps			= 1000
nstxout			= 1
nstvout			= 1
nstfout			= 1
nstlog			= 1
ns_type			= grid
pbc			= xyz
rvdw			= 3.24
rlist			= 3.6
tcoupl			= no
tc-grps			= System
tau_t			= 2.0
ref_t			= 127.2717
vdwtype			= user;Shift
rcoulomb		= 3.6;2.24;1.12
coulombtype		= Cut-off
rvdw-switch		= 0.0

energygrps		= bead
energygrp_table		= bead bead


Thanks in advance
Anja