Inconsistency of fix spring and fix smd cfor

LAMMPS LAMMPS Development
,
1

Dear all,

I am interested in running diffusion simulations of molecules through porous frameworks by using smd/umbrella sampling. I have borrowed some impiration from the tutorial of Simon Gravelle on Free Energy calculations (Free energy calculation -) and also tried to use the fix smd in cfor mode with the keyword tether enabled. From my understanding the two fixes should do the same in the way I specified them but I get significantly different values when computing the PMF with the wham algorithm.
In the zipfile I have included the data files and the input and in each subfolder the results of the smd/spring forced diffusion. For simplicity a have applied the fix nvt only to the diffusing molecule.
smd_spring_issue_LAMMPS.zip (130.0 KB)

Judging from the results it seems that the fix smd gives the more realistic results since diffusion is experimentally observed and the values of the spring-simulation seem to be to large.

I am unsure if I am overlooking some differences how each fix works or if this is a bug. I am using the 29Aug2024 version for consistency reasons.

Maybe someone knows what is going on here?
Kind regards
Jakob

2

They don’t do the same. With fix spring the target distance between the two tethered points for computing the force remains constant with fix smd not.

3

In the input you attached, the fix spring and fix smd fixes exert completely different forces on your system:

The fix spring would exert a group-total restoring z-force of 20/angstrom times the z-distance between the group’s center of mass and the tether’s z. (Force units = 1 kcal/mol/angstrom ~ 0.7 pN.)

The fix smd would exert a group-total constant z-force of 20 pointing towards the tether. (The name “constant force” was not chosen to be creative or interesting.)

So not only are they different force modes – you would expect the force magnitudes to be different, and if you have debiased both distance distributions in the same way to recover the PMF, one of those PMFs is mathematically incorrect, since the biases imposed are not the same.