I can help with only some of this...
----- Original Message -----
From: Eric Shamay <[email protected]>
Date: Wednesday, October 6, 2010 18:35
Subject: [gmx-users] New shell water model, tabulated bonding interactions, and
the documentation.
To: [email protected]
> Over the past couple days I've attempted to implement the water model
> developed by Ishiyama & Morita (J. Phys. Chem. C 2007, 111, 721-737) using a
> combination of the shell polarization technique and tabulated potentials for
> the bonded terms, among other things. Before elaborating below, the
> documentation I found was pretty weak for the tabulated bond potentials, and
> forum posts were limited in scope, so I'd like to gather some information
> from those in the know such that the wiki and documentation may be updated on
> this front. Furthermore, as I slowly understand and make progress, I'll most
> likely post follow-ups with more questions to round out the discussion.
> From the paper above, the water model is composed of 5 sites - Oxygen
(O), Hydrogen (2x H), center of mass (G), and an auxiliary site (M) along the
bisector a distance 0.021 nm from the O. All the nonbonding terms are given,
and the intramolecular potential is as follows:
> U_intra_i = Sum(n=2,6: K_n*(r1^n + r2^n)) + 1/2*K_theta*r3^2 +
K_r_theta*r3*(r1+r2) + K_r_r'*r1*r2
> Where r1 and r2 are the O-H bond displacements from the equilibrium
value, and r3 is the H-H displacement. The first term is a sum with force
constants K_2, K_3,...,K_6 multiplied by the corresponding bond displacement
exponentials, and is where I planned to use a tabulated bond potential a la the
'-tableb' option for mdrun. The higher-order terms have been found necessary
for reproducing the OH vibrational red-shifting in calculated IR, Raman, and
SFG spectra. The 2nd term is the standard harmonic term but for the H-H bond.
The 3rd and 4th terms are bond-angle and bond-bond cross terms, and are readily
available in the gromacs suite already.
> Turning to the tabulated potentials, this is where my questions
begin. I've calculated the 3 columns needed in a table_b0.xvg type of file: x,
f(x), f'(x) -- corresponding to the bondlength, potential, and first derivative
of the potential (a.k.a. the force), respectively. In doing so I've assumed
units of nm, kJ/mol/nm, and kJ/mol/nm^2, respectively.
Yes, per manual 4.2.13
> The function, f(x), is as follows:
> f(x) = K_2*x^2 + K_3*x^3 + K_4*x^4 + K_5*x^5 + K_6*x^6
> It's rather straightforward, and gives a clean analytical solution
for the first derivative:
> f'(x) = 2*K_2*x + 3*K_3*x^2 + 4*K_4*x^3 + 5*K_5*x^4 + 6*K_6*x^5
> Question: Is the analytical version what I sohuld put in the 3rd
column, or should I use a finite difference method? Using the analytical
version produced a warning that my potential derivative values are an average
of 164% off from the internally calculated ones - does that message mean
anything? Does it mean that I'm doing things right? Furthermore, why bother
with this 3rd column if it is something already calculated (and compared to) by
gromacs?
I'd expect you using analytic is best, and that GROMACS is doing an internal
finite difference calculation as a cross-check. I'm not sure why they might
disagree...
> Using a tabspace (resolution) of 1000 datapoints per nm, I
constructed a 3-column table, table_b0.xvg, and added the following to the .itp
file's [ bond ] section:
> [ bonds ]
> ; pair-potential harmonic bonds O-H terms
> 1 2 8 0 1.
> 1 3 8 0 1.
> (where 1 == Oxygen; 2,3 == hydrogens)
> This signifies that a lookup table will be used (with exclusions) and
that the potential will be multiplied by a force constant of 1.0 kJ/nm/mol --
effectively not changing the tabulated value.
> Question: Does this all sound sane so far?
Yes
> Is that value of 1.0 for the force constant actually keeping the potential
> term unaltered?
Unaltered from the table, yes.
>Furthermore, what changes if I use the function type 8(w/exclusions) vs. 9
>(w/o exclusions)?? What is being excluded if I use 8, and what does it mean if
>something is excluded from bonding terms? Where are the intra-molecular
>bonding exclusions defined?
Various nonbonded terms might get excluded, but not in your case. See manual
4.6.1 and 5.4
> Next step is to run the simulation. All is similar to previous
successful simulations in terms of my mdp input file, but I add the option of
'mdrun -tableb table.xvg' in order to use the tabulated bond potentials for my
O-H bonds.
> Question: is -tableb the only parameter needed to pass information
about the tabulated data for mdrun and simulations? Are any options necessary
in the .mdp run input file?
I would expect so. This will be straightforward to test. Run one with the old
non-table .itp, and another with the table .itp. If you're not convinced
something different is happening, adjust k for the tabulated interactions to
100 or something :-)
> Once I get some confirmation about the above I'll feel confident
enough to boldly stride towards finessing the water model .itp file for
sharing, and more coherent documentation. I'm sure others have come across this
issue before, and it would be great to have some type of example fully
illustrating use of the tabulated potential data in a simulation.
That would be good. I've created a stub section for you here
http://www.gromacs.org/Documentation/How-tos/Tabulated_Potentials, though you
may need to email Rossen to get access to edit.
Mark
--
gmx-users mailing list [email protected]
http://lists.gromacs.org/mailman/listinfo/gmx-users
Please search the archive at
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to [email protected].
Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
