AMBER Archive (2005)

Subject: Re: AMBER: Comparing EELEC from amber and that calculated in the trivial fashion.

From: David A. Case (case_at_scripps.edu)
Date: Mon Nov 28 2005 - 00:12:55 CST


On Sun, Nov 27, 2005, Pavan G wrote:
>
> I am trying to reproduce the value of EELEC for the first time step
> recorded in a md run:
> " NSTEP = 100 TIME(PS) = 1626.100 TEMP(K) = 298.76 PRESS = -18.6
> Etot = -76468.3208 EKtot = 35432.1212 EPtot = -111900.4421
> BOND = 22540.2259 ANGLE = 2334.1316 DIHED = 2373.1196
> 1-4 NB = 749.3074 1-4 EEL = 24187.2168 VDWAALS = 20399.8387
> EELEC = -184484.2821 EHBOND = 0.0000 RESTRAINT = 0.0000
> EKCMT = 10710.9469 VIRIAL = 10864.5298 VOLUME = 383436.0683
> Density = 1.0553
> Ewald error estimate: 0.3583E-05"
>
> This is a run containing ~40,000 atoms. I got the charges of the atoms
> from prmtop file and distances using the coordinates of the mdcrd
> file. Now, the crudest way I could calculate the Electrostatic
> interaction is to pick 2 atoms from the 40,000 and calculate energy
> using
> E = charge.1 * charge.2 / distance
> REF: http://amber.scripps.edu/Questions/units.html
> But this obviously includes atoms which are covalently bonded and have
> hydrogen bonds.
> This gives me a number which is ~12 times the number by amber(EELEC =
> -184484.2821).
>
> Now, where is the mistake. What set of atoms should I consider/not consider ?

If you really want to do this, start wtih a (gas-phase) system of a dozen
atoms or so, not a periodic simulation with 40,000 atoms. Figure out what all
of the atom pairs are, and try to reproduce the math. Remember that 1-4
interactions are divided by 1.2.

To make a start with periodic systems, you need to learn first about Ewald
summations and periodic imaging. There are lots of good textbooks that cover
this material. The calculation of the electrostatic energy for the sort of
system you have is really quite complex. You might want to think about
studying the tinker code first: one of the goals it has is to provide a pretty
straightforward computer implementation of these sorts of ideas.

...good luck...dac

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu