[lammps-users] Issue with fix adapt (UNCLASSIFIED)

LAMMPS LAMMPS Mailing List Mirror
1

CLASSIFICATION: UNCLASSIFIED

Hi,

I think that fix adapt may have an issue with correctly scaling the tail correction term of a potential when using the hybrid option for pair styles. When using fix adapt/fep the tail correction scaling is correct. To demonstrate this issue in a barebones way, I grabbed the LJ benchmark input file from the LAMMPS distribution and made some edits as shown here:

2

fix adapt can do things that fix adapt/fep cannot do and there are some more (sometimes subtle) differences. fix adapt/fep was introduced, because it included some change that was not considered suitable for fix adapt. from then on the two fixes have diverged further.

from a cursory look at the code it is not clear why there should be a discrepancy for your example. perhaps steve has an idea for an explanation.

axel.

3

CLASSIFICATION: UNCLASSIFIED

Hi Axel,

Thanks for the quick response. I now see that I was mistaken in thinking that fix adapt was simply an older version of fix adapt/fep.

If replacing this portion of fix_adapt.cpp:

if (anypair) {

for (int m = 0; m < nadapt; m++) {

Adapt *ad = &adapt[m];

if (ad->which == PAIR) {

ad->pair->reinit();

}

}

}

With this line from fix_adapt_fep.cpp:

if (anypair) force->pair->reinit();

Then the discrepancy disappears for fix adapt and the tail correction is properly scaled. I hope this helps with tracking the issue.

Best,

Garrett

4

Now please try an unmodified version of fix adapt and re-issue the fix adapt command after you changed the pair style.
If that will also “cure” the problem, then we have identified the source of the problem.

5

CLASSIFICATION: UNCLASSIFIED

Hi Axel,

As instructed, I made these edits to the end of the input file to unfix the initial fix adapt and then reapply fix adapt after the changed pair style:

6

Thanks,

not sure, if we can look into this right away. Since fix adapt/fep seems to be sufficient for your use case, you should use that for now.
It would help if you could submit a summary of this discussion as a bug report issue on github at https://github.com/lammps/lammps/issues so that it won’t get lost as otherwise happen often with mailing list bug reports.

Axel.