Deviation too large for solvent
Clicked A Few Times
3:44:05 PM PST - Wed, Nov 20th 2013
Hi,

I'm running QMMM calculations using nwchem and having a problem with the error when trying to run QMMM dynamics:

Deviation too large for solvent 598 *:: 598

I'm wondering what the number above implies. It doesn't seem to be a residue no. of water that is having a problem to me. I examined the structure and found nothing weird to this residue number. Indeed, no too close contacts between water molecules themselves or water molecule and protein residues were spotted. I have searched and followed a couple of suggestions to cope with this problem from other threads but it doesn't seem to completely rid of the problem. Running constant energy as suggested by Marat in http://nwchemgit.github.io/Special_AWCforum/st/id300/%22Deviation_too_large_for_so... doesn't help in my case. Increasing the dimension box as recommended by TBS in http://nwchemgit.github.io/Special_AWCforum/st/id772/%220%3A_Deviation_too_large_f... does help sometimes. However, increasing the box dimension might makes the box not rectangular any more when using periodic boundary condition. I'm wondering and concerned whether this effect is going to cause any differences in result.

I can send you some output if it would be helpful.

Best,
Tee

Clicked A Few Times
2:01:02 AM PST - Tue, Nov 26th 2013
same problem
Hi,

I encounter the same problem and searching the solution. I am appreciated if you can share me with your experience.

Did you increase the box dimension directly in the rst file? I want to know how long the box dimension you increased to make it go through the computation. In my simulation, I change the box dimension from 8 angstrom to 8.3, but still error. So the question is how long we can increase?

Thanks for your time.

Best,

Jingbo

Clicked A Few Times
1:04:58 PM PST - Tue, Nov 26th 2013
Hi Jingbo,

From my experience, This error comes from 2 reasons. The first one is the dimension box is to tight as suggested by TBS. Yes, you can increase dimension box from the restart file directly or recreate the new restart file. There is no rule about how much we should increase the box. In my case, I tried 0.2 nm to each dimension and that did fix my problem. Please keep in mind that increasing too much will cause the water box not rectangular any more. In case you use PBC, I'm not sure this will affect the result or not so you need to be aware of that. If this cannot solve the problem, there is likely to be a chance that there is something bad happen to the solvent. Have you ever extracted the pdb file from the restart file to see what might go wrong for water molecules? From my experience, I always spot some waters clumping together (getting too close). From time to time, this can be fixed if you run constant energy MD or optimize the structure. This may drive those clumping waters apart but unfortunately sometimes it does not help either.


My another question is how long do you equilibrate you system at once. If you equilibrate the system for a long time at once (e.g. > 10 ps), there is likely to be a chance that waters will be clumping together. This sounds like a bug in MD module as I talked to Marat. My advice is to equilibrate the system with shorter number of steps (5 ps or 5000 steps) at the time. For example, you want to equilibrate the system for 20 ps, you better do it for 5 ps 4 times rather than doing it 20 ps at once. So far, this seems to fix the problem for clumping waters, one source of 'Deviation too large for solvent'.

Hope this might help fix your problem. I also had a hard time figuring out how to completely get rid of this problem too.





Quote:Wjb0920 Nov 26th 9:01 am
Hi,

I encounter the same problem and searching the solution. I am appreciated if you can share me with your experience.

Did you increase the box dimension directly in the rst file? I want to know how long the box dimension you increased to make it go through the computation. In my simulation, I change the box dimension from 8 angstrom to 8.3, but still error. So the question is how long we can increase?

Thanks for your time.

Best,

Jingbo

Clicked A Few Times
7:41:47 PM PST - Tue, Nov 26th 2013
Thanks
Dear Tpirojsi,


Here I provide another way to solve the problem. The increase of cutoff from 0.9nm to 1.2nm in MD directives group, and also change bqzone from 9 to 12 in qmmm directive group may go through the simulation without error of deviation too large.

Best,

Jingbo

Clicked A Few Times
12:34:13 PM PST - Wed, Dec 4th 2013
Hi Jingbo,

Glad for having someone to discuss the issue with. For some reason, your approach doesn't help in my case but it's good to know and I'll keep that in mind too. This error seems to be a common error that I think coming from many reasons. Other than what I explained above, when doing FEP calculation I also still have seen this error arising even though the waters were in the perfect shape and place, i.e. no clumping waters or bad contacts. My solution was rewriting restart files from NEB beads with the 'center' keyword in prepare module and using the new restart files to run FEP calculations. Also make sure that we use these new restart files in gerenating xyzi and esp files too.

UPDATE Although the approach sometimes fixes my problem in case this error arises in FEP calculation, it still doesn't seem to completely get rid of the problem.

Best,
Tee

Quote:Wjb0920 Nov 27th 2:41 am
Dear Tpirojsi,


Here I provide another way to solve the problem. The increase of cutoff from 0.9nm to 1.2nm in MD directives group, and also change bqzone from 9 to 12 in qmmm directive group may go through the simulation without error of deviation too large.

Best,

Jingbo

Clicked A Few Times
3:13:35 PM PST - Wed, Dec 4th 2013
Please ignore this message. It seems this approach (using 'center' in rewriting restart files) might cause a huge error in free energy calculation because the QM region of each bead will not be the same (their positions are shifted). Sorry for any confusion.


Quote:Tpirojsi Dec 4th 7:34 pm
Hi Jingbo,

Glad for having someone to discuss the issue with. For some reason, your approach doesn't help in my case but it's good to know and I'll keep that in mind too. This error seems to be a common error that I think coming from many reasons. Other than what I explained above, when doing FEP calculation I also still have seen this error arising even though the waters were in the perfect shape and place, i.e. no clumping waters or bad contacts. My solution was rewriting restart files from NEB beads with the 'center' keyword in prepare module and using the new restart files to run FEP calculations. Also make sure that we use these new restart files in gerenating xyzi and esp files too.

UPDATE Although the approach sometimes fixes my problem in case this error arises in FEP calculation, it still doesn't seem to completely get rid of the problem.

Best,
Tee

Quote:Wjb0920 Nov 27th 2:41 am
Dear Tpirojsi,


Here I provide another way to solve the problem. The increase of cutoff from 0.9nm to 1.2nm in MD directives group, and also change bqzone from 9 to 12 in qmmm directive group may go through the simulation without error of deviation too large.

Best,

Jingbo


Forum >> NWChem's corner >> QMMM