LAMMPS WWW Site - LAMMPS Documentation - LAMMPS Mailing List Archives
[lammps-users] Moltemplate OPLS-AA improper parameters
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[lammps-users] Moltemplate OPLS-AA improper parameters

From: Andrew Jewett <jewett@...1937...>
Date: Sat, 2 Sep 2017 22:16:54 -0700

On Fri, Sep 1, 2017 at 6:45 AM, David Huang <david.huang@...3703...> wrote:
> Firstly, I would like you to thank you for making the Moltemplate tool
> available. It has been invaluable to me. The reason I'm writing is because
> of what appears to me to be a couple of errors in the specification of
> improper torsions for the OPLS-AA force field in, but perhaps I am
> misunderstanding something.
> Firstly, the "harmonic" improper_style is used, when the OPLS-AA force field
> uses the same type of function for impropers as the AMBER force field, which
> is equivalent to the "cvff" improper_style in LAMMPS.

Jason Lambert wrote the conversion script.  ("")  I
don't know why Jason chose "harmonic" instead of "cvff" but I strongly
suspect it is because improper_style "harmonic" can be used with GPU
acceleration (harmonic/kk), and "cvff" can not (yet). Since, for small
angles the "harmonic" function is a pretty good approximation of the
cosine function used in "improper_style cvff", this is a pretty
important reason to keep things as they are.
    If the entire purpose of the improper interaction is to keep the 4
atoms coplanar (which it is), then hopefully it does not matter too
much if you use one or the other (as long as I did not screw up the
magnitude of the spring constants.  I think they are correct, but it's
possible they might be off by a factor of 2.  Let me know if you spot
something.)  I should also admit that we also ignore the
"Urey-Bradley" terms in the OPLS force field.  (Although I think there
is only one of them, and apparently most people don't bother with
these terms unless they care about the high frequency vibrational

   Nevertheless, if you have reason to believe that it's important to
use a cosine function instead of a quadratic one, I can make an
alternative version of the file which uses "cvff" instead.
Nobody has asked for this so far, but perhaps they are just not as
diligent as you.  I'm happy to do this.

> Secondly, the
> parameters for the improper torsions do not seem to match those in the
> papers in which the OPLS-AA force field is specified - in particular, the
> improper torsion parameters are apparently from the original AMBER paper,
> and the latest revision of OPLS (OPLS3) (published in
> appears to use similar
> parameters. I realise you have taken these parameters from the TINKER
> parameter file for OPLS, but they do not seem to be correct.

The OPLSAA parameters in the "" file were taken from this file:
Here's an excerpt from the most recent version of that file:
"The parameters supplied with TINKER are from "OPLS All-Atom Parameters
for Organic Molecules, Ions, Peptides & Nucleic Acids, July 2008" as
provided by W. L. Jorgensen, Yale University during June 2009. These
parameters are taken from those distributed with BOSS Version 4.8."

It's really confusing.  There are multiple sets of OPLSAA parameters,
as they have changed over the years.  I am under the impression that
they update the parameters frequently and they don't always publish a
new paper when they do this (but perhaps I'm wrong).  The parameters
have definitely changed a lot from the original OPLS papers.

I don't know if this helps, but here is a link to the parameters used
in BOSS 4.9.  Hopefully they are similar to BOSS 4.8.  The list of
parameters is located on pages 66-74 of this file:

Unfortunately, I could not tell if there are any _improper_ torsion
parameters in this file (or just ordinary torsion parameters).

If you can find a recent source for OPLSAA parameters which disagrees
with the moltemplate parameters in "" file, let me know and I
will definitely fix it.  There are only 5 improper interactions in
this version of OPLSAA.  If that's all that's wrong, it is easy to fix
them manually.

Thanks for the kindly words about moltemplate.