[gmx-users] Setting correct non-bonded and pair (LJ) parameters in ffnonbonded.itp

Justin Lemkul jalemkul at vt.edu
Fri Jul 20 14:16:45 CEST 2018

On 7/20/18 2:42 AM, Apramita Chand wrote:
> Dear All,
> In my previous posts, I had written about errors arising upon changing the
> combination rule and using sigma and epsilon values. Dr Lemkul was kind
> enough to point out that incorrect formatting of the ffnonbonded.itp might
> cause errors and that the [nonbond params] as well as [pairtypes] section
> need to be supplied with the parameters.
> I have some doubts regarding the calculations of the [nonbond params] and
> particularly the [pairstype] section. I have gone through the archives and
> it has certainly helped me to understand the logic better but still some
> doubts persist as below
> 1.  As per manual, The combination rule 1 states that geometric mean of C6
> and C12 of (respective atomtypes section) of two atoms should give C6ij and
> C12ij which is included in the [nonbond params] section. In ffnonbonded.itp
> of gromos53a6ff, I see that C12(OM)=7.4149321e-07 and C12(O)=1e-06 give
> C12(OM O)= 8.611e-07. But when I calculate C12 (OA O), using
> C12(OA)=1.505529e-06 and C12(O)=1e-06, I get the value C12=1.227e-06 while
> the C12 mentioned for OA O interactions in the itp file =1.38651e-06.
> The interaction parameters C6ij and C1ij in most other cases also don't
> match with the geometric mean rule.
> What could be the reason for such discrepancy? I have already run
> simulations trusting this ffnonbonded.itp file to be correct (using
> combination 1). Are those wrong?

Again, as I've said a few times: what you are seeing in [nonbond_params] 
is explicitly *not* what comes from combination rules. You will not be 
able to compute those values from the parameters in [atomtypes], and 
this is by design. GROMOS uses a complex matrix of different combination 
rules for different parameters, so what you find in [nonbond_params], by 
definition, overrides whatever you might compute from combination rules 
(whichever rule you may wish to employ).

> 2.  What is the formula for calculation of [pairtypes] section in the
> ffnonbonded.itp? I haven't found any concrete answers yet to this doubt
> yet.

Again, these are special 1-4 interactions that have empirically derived 
values. Any 1-4 interaction not listed here uses normal combination 
rules, but the values you see will not come from any predictable equation.


> I'm hoping once I am clear about how the calculations for C6ij and C12ij
> for [nonbond params] and [pairtypes] sections have been done, I will be
> able to calculate the same using sigma and epsilon values for combinations
> 2 and 3 as Justin has suggested.
> Looking forward to your help,
> Thanking you,
> Yours sincerely,
> Apramita Chand


Justin A. Lemkul, Ph.D.
Assistant Professor
Virginia Tech Department of Biochemistry

303 Engel Hall
340 West Campus Dr.
Blacksburg, VA 24061

jalemkul at vt.edu | (540) 231-3129


More information about the gromacs.org_gmx-users mailing list