[gmx-users] Long Range dispersion correction with Lennard-Jones PME

[Joel Jose Montalvo Acosta]

Dear David,

Thanks for your answer.

Following the papers you suggested me (also a recent paper titled
"Direct-Space Corrections Enable Fast and Accurate Lorentz− Berthelot
Combination Rule Lennard-Jones Lattice Summation". JCTC 2015 *11* (12),
5737-5746.) plus your explanations I could understand a bit more the
implementation of LJ-PME in Gromacs, at least in the case where geometric
combination rule is used for both LJPME and the force field (as e.g.,
OPLS).  However, I still have some questions about the use of arithmetic
combination rule with LJPME.

In a first case, when both the LJPME and the force field use the arithmetic
combination rule (as Charmm FF), the dispersion correction is zero, because
LJPME is exact in this condition and there is not need to apply any
dispersion correction (this is the same result when geometric combination
rule is used for both LJPME and force field as you indicated before),
however this is impractical because the arithmetic combination rule for
LJPME is very slow. Then, to have a good efficiency, LJPME should be used
with the geometric combination rule even with force fields which use the
arithmetic combination rule (as Charmm). In this second case, a very small
dispersion correction is computed (as you indicated), but I still don't
know how this contribution is obtained (I could not reproduce this value by
hand using a model system), could you provide me more detail about the
computation of <C6> and the dispersion correction for this case?. Also, Is
it necessary to add this (small) dispersion correction for this second case?

Thank you in advances,

Joel



2016-07-20 19:21 GMT+02:00 David van der Spoel <spoel at xray.bmc.uu.se>:

> On 20/07/16 17:01, Joel Jose Montalvo Acosta wrote:
>
>> Dear Gromacs users and developers,
>>
>> I want to understand how the long-range dispersion correction is
>> implemented in gromacs when van der waals (vdw) interactions are computed
>> with cut-off or PME, so I started reading the section 4.9.1 in the gromacs
>> (version 5.1.2) manual to check the involved formulas. Then, I did some
>> tests using a model system composed by 2 argon atoms and computing the
>> Lennard-Jones (LJ) contribution applying cut-off (without shift the vdw
>> potential, ie., using vdwtype=cut-off and vdw-modifier=none in the mdp
>> file) with and without dispersion correction (DispCorr=Ener and
>> DispCorr=no). After, I computed by hand the dispersion contribution to the
>> potential energy for this system. Finally, the values for the dispersion
>> contribution obtained from gromacs and by hand were equal.
>>
>> Next, I tried a second test with the same model and same conditions but
>> using PME instead cut-off to treat the vdw interactions with and without
>> the dispersion correction. In this second test, the dispersion
>> contribution
>>  computed by gromacs was 0. I expected this result because this correction
>> is suitable when vdw interactions are computed with cut-off and the radial
>> distribution function outside the cutoff is assumed equal to 1. Thus, I
>> though it looks incompatible to use dispersion correction (DispCorr=Ener)
>> and PME for computing vdw interactions.
>>
>> However, using a real system as a protein or ligand in water and applying
>> PME and the dispersion correction for vdw interactions, gromacs is
>> computing a dispersion correction contribution, which is unexpected
>> according to the previous tests done before. For this system, gromacs
>> prints in the log file the average dispersion constant (<C6>) which I
>> could
>> not reproduce manually following equation 4.169 in the gromacs (v. 5.1.2)
>> manual. I don't know how gromacs is computing this <C6> value for this
>> system with PME.
>>
>> Finally, My questions are:
>> 1. How does gromacs compute the dispersion correction when vdw
>> interactions
>> are computed with PME?
>>
> The result for the dispersion correction with LJPME are zero when you use
> a geometric combination rule, since the LJPME is exact in principle.
> If you use the arithmetic combination rule (e.g. Charmm FF) and LJPME is
> told to use the geometric combiation rule (for efficiency) the dispersion
> correction estimates the difference in dispersion between the two. Usually
> this number is very small.
>
> I suggest you read these two papers:
> - Lennard-Jones Lattice Summation in Bilayer Simulations Has Critical
> Effects on Surface Tension and Lipid Properties
> Christian L. Wennberg, Teemu Murtola, Berk Hess, and Erik Lindahl  J.
> Chem. Theory Comput. 2013, 9, 3527−3537
>
> - Nina M. Fischer, Paul J. van Maaren, Jonas C. Ditz, Ahmet Yildirim and
> David van der Spoel: Properties of Organic Liquids when Simulated with
> Long-Range Lennard-Jones interactions J. Chem. Theory Comput. 11 pp.
> 2938-2944 (2015)
>
> 2. Is it right to apply this correction when vdw interactions is computed
>> with PME? if the answer is not, it would be nice if gromacs prints a
>> warning message indicating this incompatibility when both options are
>> used.
>>
>> More info would be good indeed.
> Maybe you can file a documentation request on http://redmine.gromacs.org
>
>
>> Thank you for your help
>>
>> Joel Montalvo Acosta
>> PhD student at University of Strasbourg
>>
>>
>
> --
> David van der Spoel, Ph.D., Professor of Biology
> Dept. of Cell & Molec. Biol., Uppsala University.
> Box 596, 75124 Uppsala, Sweden. Phone:  +46184714205.
> spoel at xray.bmc.uu.se    http://folding.bmc.uu.se
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.