Re: Meaning of the warning: Low global exclusion count!

From: David Hardy (dhardy_at_ks.uiuc.edu)
Date: Thu Apr 22 2021 - 14:15:40 CDT

Hi Pawel,

I’m glad you found a workaround.

However, this does sound like suspicious behavior in NAMD. If you wouldn’t mind sharing your system with me, I’d be interested to see why the code isn't handling the Z=0 case.

Best regards,
Dave

--
David J. Hardy, Ph.D.
Beckman Institute
University of Illinois at Urbana-Champaign
405 N. Mathews Ave., Urbana, IL 61801
dhardy_at_ks.uiuc.edu, http://www.ks.uiuc.edu/~dhardy/
> On Apr 22, 2021, at 8:21 AM, Pawel Kedzierski <pawel.kedzierski_at_pwr.edu.pl> wrote:
> 
> Dear David,
> 
> Thank you for detailed explanation.
> Increasing the margin to e.g. 3.0, which is more than ring diameter, didn't help.
> After several more tests I think, I am triggering a bug in this global exclusion checksum calculation.
> 
> My grafene was generated idealized with all Z coordinates exactly 0.0, and that was the source of the problem.
> 
> When I started calculations using coordinates from final MD frame of previous job, so they were slightly distorted, NOT EVEN A SINGLE WARNING appeared. On the other hand, when I start from ideally flat sheet aligned to XY plane, I am getting this warning at virtually every step. The checksum have not to be calculated from scratch, only updated, and some initial mistake is never corrected later; apparently such update occurs once per cycle (20 steps in my case).
> The warnings are issued both during minimization and dynamics but namd continues anyway, I run tests for 100ps with no problem.
> 
> I also checked whether the problem is triggered due to zero Z thickness of the entire system; but adding a sodium ion like 5A above the graphene did not help.
> 
> So my workaround now is to generate the graphene coordinates with minimal random distortion (-0.05 to +0.05A range) out of the XY plane and voila, no more warnings :-D.
> 
> Thank you,
> Paweł
> 
> W dniu 21.04.2021 o 22:31, David Hardy pisze:
>> Hi Pawel,
>> 
>> NAMD calculates certain checksums to make sure that, due to its parallel decomposition, no terms are accidentally dropped during force evaluation. The “global exclusion” checksum counts the total number of non-bonded exclusions evaluated, where the exclusion policy is set in the config file (usually “exclude scaled1-4” for CHARMM FF). Before the simulation begins, NAMD counts the total number of non-bonded exclusions that it should be evaluating when computing the force. It will then report when an undercount occurs. The numbers reported are the checksum vs. the number of exclusions that should have been evaluated. Since this does not end your simulation, this must be occurring for you during minimization. If this were to occur during dynamics, NAMD would terminate with an error.
>> 
>> So why would this be occurring? Well, in order for an exclusion to be calculated, the pair of atoms being excluded must be in an adjacent pair of patches. It sounds like your graphene sheet has carbons spaced far enough away, that you have exclusions due to a covalently bonded chain of 3 or perhaps 4 carbon atoms that extend beyond an entire patch. To avoid these messages, one thing you might try doing is artificially increasing the patch size by setting “margin” to a nonzero value large enough to pad the patches so that these 3-chains or 4-chains of carbon atoms are kept in adjacent patches.
>> 
>> As to the “hydrogen group” designation, NAMD, for performance reasons, keeps each hydrogen atom together with its heavy parent atoms, and refers to the resulting clusters as “hydrogen groups.” In your case, every carbon atom would be its own hydrogen group (and migration group). Since this has important meaning for biomolecules, we report aspects of these clusters, such as the “114 hydrogen groups with all atoms fixed” message.
>> 
>> Best regards,
>> Dave
>> 
>> --
>> David J. Hardy, Ph.D.
>> Beckman Institute
>> University of Illinois at Urbana-Champaign
>> 405 N. Mathews Ave., Urbana, IL 61801
>> dhardy_at_ks.uiuc.edu <mailto:dhardy_at_ks.uiuc.edu>, http://www.ks.uiuc.edu/~dhardy/ <http://www.ks.uiuc.edu/~dhardy/>
>> 
>>> On Apr 21, 2021, at 6:01 AM, Pawel Kedzierski <pawel.kedzierski_at_pwr.edu.pl <mailto:pawel.kedzierski_at_pwr.edu.pl>> wrote:
>>> 
>>> Dear NAMD wizards,
>>> 
>>> The warning from the subject is a topic of many questions appearing on this list, but I haven't found a single answer which would explain its exact meaning and the conditions which trigger it. The usual answers are that it signs some steric or topology problems and such reasons lead to a quick crash due to divergence of movement integrations. But it does not apply to my case.
>>> 
>>> In a case I am playing with - a sheet of graphene - NAMD produces immense numbers of these warnings, which seem to be false. The dynamics of the system is stable, no crash, setting cutoff and pairlist well above the system size does not help. These messages are just annoying filling 90% of the log size.
>>> 
>>> I want to use this graphene example as exercise for my students, and this flood of warnings not only makes it look unprofessional, but what troubles me more is that I am unable to explain it. Specifically:
>>> 
>>> what exactly is "excluded globally" and what are the numbers reported?
>>> are there any other conditions apart from cutoff, pairlistdist and steric problems, which I can change to silence this warning?
>>> The first number reported in the warnings changes regularly but the difference between them is rather huge, from 2340 vs 4593 when they first appear during minimization to for example 3759 vs 4593 at the end of dynamics.
>>> 
>>> From my current experience, the combination of graphene and CHARMM forcefield is not optimal. I already learned that due to L-J interactions between bound carbons (in-plane) I was getting huge excess of VDW energy; this was mostly corrected with setting the 1-4 L-J parameters to 0, but possibly all in-plane L-J interactions should be discarded. Basically I use a copy of the parameters for standard CA type (which is also duplicated for CGenFF multi-ring ligands) and it provides all necessary parameters, I only changed the 1-4 part. Note that this type uses regular dihedrals instead of impropers to stiffen the ring plane, so I have generated the PSF with all planar dihedrals but without impropers.
>>> 
>>> I will be grateful for any insight.
>>> 
>>> With regards,
>>> 
>>> Paweł Kędzierski
>>> 
>>> P.S. Summary of the system for which I reported the numbers above. Note there are no hydrogens so "114 hydrogen groups" is confusing. Here I fixed atoms on the edge of the sheet but without fixing the flood of warnings is still there, only the numbers change.
>>> 
>>> Info: STRUCTURE SUMMARY:
>>> Info: 603 ATOMS
>>> Info: 867 BONDS
>>> Info: 1659 ANGLES
>>> Info: 3186 DIHEDRALS
>>> Info: 0 IMPROPERS
>>> Info: 0 CROSSTERMS
>>> Info: 0 EXCLUSIONS
>>> Info: 114 FIXED ATOMS
>>> Info: 1467 DEGREES OF FREEDOM
>>> Info: 603 HYDROGEN GROUPS
>>> Info: 1 ATOMS IN LARGEST HYDROGEN GROUP
>>> Info: 603 MIGRATION GROUPS
>>> Info: 1 ATOMS IN LARGEST MIGRATION GROUP
>>> Info: 114 HYDROGEN GROUPS WITH ALL ATOMS FIXED
>>> Info: TOTAL MASS = 7242.63 amu
>>> Info: TOTAL CHARGE = 0 e
>>> 
> <pawel_kedzierski.vcf>

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST