[gmx-users] Periodic Molecule's Free Energy Calculation Error

[Justin Lemkul]

On 7/12/17 9:12 PM, Alex wrote:
> 
>>
>> I wonder if the "couple-intramol = yes" is a must. Does it have any influence 
>> on the output results if we turn off the intra-molecular non-bonded 
>> interactions of a large infinite molecule?
>>
> The answer to your question has nothing to do with Gromacs, but with 
> understanding the difference between crystals and biomolecules (for which 
> Gromacs was designed).

There is a functional difference between coupling intramolecular interactions 
and not, which will affect the computed free energy.  With couple-intramol = no, 
you get the correct vacuum state of the solute molecule; with couple-intramol = 
yes, you get perturbed intramolecular interactions.  This has important 
consequences for what one is trying to compute.

-Justin

> Also (unrelated), it is a common misconception to believe that PBC makes 
> something infinite -- the effective size of your system is entirely determined 
> by the supercell size (proof: consider the ripples in hBN and determine the 
> lowest wavelength of the ripple that can propagate -- it is commensurate with 
> the box size). In an infinite system, you can have an immensely long wave 
> (though not infinite, as shown by Landau a while back). PBC does not make 
> anything infinite, it is a mathematical way of avoiding surfaces.
>>
>> There is no universal force field for HBN, so I am using a modified 
>> gromos54a7_atb force field, i.e., manually adding the parameters for boron and 
>> nitrogen to the bonded & nonbonded .itp files.
> Oh, I know that there is no force fields for these structures. ;) My question 
> was about which Gromacs ff you were using to insert your parameters, and, most 
> importantly, where those parameters came from.
> 
>> The parameters are obtained from literature.
>>
> What literature? All bio-style ff adaptations of solid-state potentials (e.g. 
> Tersoff-Brenner for hBN) I am aware of make it very clear that "intramolecular" 
> interactions between atoms sharing up to a fairly distant covalently bound 
> neighbor are limited to bonds and angles. This comes from the math involved in 
> developing potentials for crystals. There was a recent question regarding this 
> very problem here, which was solved by setting a larger nrexcl value. In your 
> case, you solved it with turning off intramolecular coupling. In fact, if you 
> set your nrexcl to something like 4 or 5, you may not even need to turn off the 
> coupling. But then again, I don't know where the parameters came from.
> 
> Alex

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================