Questions and problems?

What dielectric constant with water?

I would very much appreciate receiving AMBER users' opinions on the best way to set the dielectric multiplicative constant (DIELC) when running simulations explicitly in water and using eedmeth=5 (IDIEL=1 for earlier versions and sander_classic) and periodic boundary conditions. Assuming that the dielectric constant of water is 80, how does one set up the simulation for this?

I find the dielectric in molecular simulations a rather interesting object as well. The general concensus is that if you've including water explicitly then you should be using a dielectric of 1 since the explicit water will produce the required dielectric of 80 on its own accord. If you are not including explicit water then using a straight dielectric of 80 ( ie, not distance dependent ) seems to be invalid for three reasons.

  1. It's a time averaged quantity. That is the dielectric of water is frequency dependent and hence should be different for different frequency motions.
  2. It's a space averaged quantity, ie macroscopic. So it shouldn't be applied between things only nm apart, which is obvious in the limiting case of two atoms only a few angstrongs apart and hence with a vacuum between them.
  3. Bound surface water around a protein does not have the same dielectric as bulk water, so shouldn't be modelled as such.
In the case of excluding explicit water the distance dielectric model seems to be the most fashionable, but as a physicist I've always been worried about a model that breaks Gauss' Law and produces an unphysical model. But then if the model agrees with enough experimental data, it's a good model !

On the distance dielectric model, for a particular protein BPTI, Guenot and Kollman showed in this paper,

J. Comp. Chem, 14(3),295--311 (1993)

that the distance dielectric model with a finite cutoff provided a closer average conformation to the explicit water model with an infinite cut-off than did any of the explicit water models with finite cut-offs. So the distance dielectric model gets a vote of confidence.