Re: [AMBER] Polarizability and induced dipoles

From: Ross Walker <>
Date: Tue, 15 Jun 2010 05:00:48 -0700

Hi Thomas,

> > Just so I understand this properly. The chamber patch will take the
> > polarizability data from the charmm force field file and include it
> in the
> > prmtop file. Are the implementations of the polarizable force fields
> in
> > charmm and amber identical?
> Polarizabilities are identical (Angstrom³)

What do you mean by this? You mean what ends up in the prmtop matches what
was in the psf?
> I would like to but,
> -) even though my CHARMM is reasonably up to date, I find myself unable
> to use
> the frcdump command (does one have to specify a certain keyword at
> compile
> time or something?)

I'm not sure on this, you may need the development version of Charmm. I'll
let Mark comment some more on this.

> -) the only polarization method available in both AMBER and CHARMM are
> inducible dipoles, however the CHARMM version cannot work with Ewald
> and as I
> understand it AMBER cannot work without it
> -) also if I am not mistaken pmemd cannot use polarisable force fields
> other
> than amoeba, or so I think I remember to have read somewhere (and my
> tests
> sure never worked, but that may be a different story)

My point is that if the method used for the inducible dipoles is different
does this still imply that exactly the same polarizabilities can be used?
E.g. from AMBER's FF02 (from parm99.dat) for various carbon's we have:

C 12.01 0.616 ! sp2 C carbonyl group
CA 12.01 0.360 sp2 C pure aromatic (benzene)
CT 12.01 0.878 sp3 aliphatic C

Where column 3 is the atomic polarizability in A^3. Note there is huge
variation based on atom type for the same element. Are the numbers in CHARMM
identical? I.e. does a CT in Charmm have a polarizability of 0.878. If it
does not then you need to look much more closely at the origin of these
numbers. I have, not myself looked at how these are obtained so can't
comment too much, but I would not be surprised if the values in CHARMM were
parameterized for the drude oscillator model while the ones in AMBER are for
the induced dipole model. If they are identical in the actual parameter
files (NOT the prmtop made from the psf) then you are probably okay,
although that assumes all the atom types match between the force fields. If
they are NOT the same then you will have to find a way of proving that it is
acceptable to use the CHARMM polarizabilities with the AMBER induced dipole
model. This will be pretty tough to do, probably it would be easier to just
implement the drude oscillator model in sander.

The key is you cannot just copy across the polarizabilities and expect them
to work if the underlying model is different. Thus before we add this to
chamber we will need to see the proof of why this is an acceptable thing to
do. If the parameters are the same in both force fields that would be
reasonable proof.

> However when polarising with induced dipoles in AMBER and with drude in
> at least mean squared displacement (of coordinates) matches reasonably
> well
> for a large system (27000 atoms in 2000 molecules).

"Reasonably well" - what does that mean? - You need to think up quantitative
ways to test this if you can't prove it analytically.
All the best

|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- |
| | |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

AMBER mailing list
Received on Tue Jun 15 2010 - 05:30:03 PDT
Custom Search