Should I use two T-coupling groups for ice+graphene?

User discussions
1

GROMACS version: 2023

Hi everyone.
I was running the growth of ice on a substrate surface (graphene). I found it so different between using two T-coupling groups and using one T-couping group.

When I set the substrate and ice as two independent T-coupling groups, the bottom layer of ice (the layer in contact with substrate) is hard to maintain its original crystal, or to say the bottom layer of ice melts quickly.

Tcoupl = V-rescale
tau_t = 0.2 0.2
tc_grps = SOL non-water
ref_t = 180 180

When I put the whole system into one T-couping group, the ice crystal was well maintained.

Tcoupl = V-rescale
tau_t = 0.2
tc_grps = system
ref_t = 180

The substrate is set to be frozen in these two cases

freezegrps = non-water
freezedim = Y Y Y

I want the bottom layer of ice to maintain its crystal structure, but I am not sure which one is correct? And what makes it so diffenet when seperate it into two T-coupling groups?

Thank you so much! I will really appreciate it if you can do me this favor!

2

I would think that there should be no difference between the two considering that all atoms in graphene are frozen.

Can you report here what mdrun reports for nrdf for the two groups and the single group in the md.log files?

Note that a tau_t of 0.2 is very short. Likely this doesn’t lead to any issues, but I would suggest to use tau_t=1

3

Thank you so much for your reply!

I don’t quite understand what “nrdf” stands for, but I try to search it in both of the log files.
In the two T-coupling groups, it shows:

grpopts:
nrdf: 37821 0
ref-t: 180 180
tau-t: 0.2 0.2

In the one T-coupling group, it shows:

grpopts:
nrdf: 19206.8
ref-t: 180
tau-t: 0.2

I will show the snapshots of bottom layer ice of these two situations after 50 ns run:

图片
图片931×438 230 KB

Do we have any creterian for the tau_t? It is said that the smaller tau_t, the better temperature control. Isn’t 0.2 better than 1.0?

Thank you once again for your reply!

4

nrdf stands for number of degrees of freedom. I would expect the number to be approximately the same, but they clearly are not. I suppose the number of degrees of freedom is too high when using two groups. Are the T-coupling groups really the only thing you changed between the two setups?

5

I reproduced the issue. This is a bug in GROMACS. You can use one t-coupling group for the whole system to avoid this bug.

6

My last answer was incorrect. The seem to be two issues with counting the number of degrees of freedom. One with frozen atoms and constraints and another one with frozen atoms being in the same group as non-frozen atoms.

The only way I can get correct counts for my test system is when using two separate T-coupling groups and removing all constraints between frozen atoms.

7

I am sure T-coupling groups really the only difference.

Thank you so much! And by the way, are you one of the GROMACS developers? I really appreciate you!

8

I have removed the frozon group and I think I should use two T-coupling groups. I post my result here:

f4435e09111b46916fa426206e54f5a
f4435e09111b46916fa426206e54f5a1233×339 96.8 KB

No freeze substrate, it seems to be normal. The nrdf 111134 and 37821+73314 are almost the same. When freeze substrate, 19206.8 =! 37821+0. But two T groups one should be right because it is 37821, as the same when no freeze.

I think there is something wrong when there are freezon atoms in one T group.

9

Yes. I found the issue and have a fix. The issue appears when fully frozen atoms are in the same T-coupling group as the non-frozen atoms. Using two temperature coupling groups avoids this issue.

It seems that constraints do not affect this at all. My previous comment on that was incorrect.

10

The issue and link to the fix are here: