Re: [AMBER] which solvent model to use with solvateCap

From: Jason Swails <>
Date: Mon, 27 Jul 2015 10:19:10 -0400

On Mon, 2015-07-27 at 11:12 +0000,
> Dear Amber users,
> I would like to use the TIP4P solvent model with the amber force
> field. After some reading I found that using the TIP4P-Ew was
> recommended when running PME simulations.
> However I also run non periodic simulations (using solvateCap in tleap
> to create for instance a solvent sphere that I constrain later in the
> simulation). Can I also use the same water parameters (TIP4P-Ew) for
> such simulations or shall I switch to normal TIP4P in this case?

Solvent caps are not frequently used for much anymore. (The newest
paper I saw from Google Scholar matching "solvent cap molecular
dynamics" was from 2001.) PME is often cheaper for large systems even
if you need more water than the solvent cap by virtue of using a O(N
logN) algorithm instead of a O(N^2) one. Add in the fact that neither
pmemd.cuda nor pmemd/pmemd.MPI are capable of running with a solvent
cap, and it's even more likely from a practical perspective for PME to
actually be *faster* than a solvent cap simulation (pmemd.cuda is ~1-2
orders of magnitude faster than sander and pmemd is ~2x faster with
better parallel scaling).

Long story short, I don't think this question has been thoroughly
investigated (if investigated at all). If I had to guess which one
would be better, I'd probably still pick TIP4Pew (it gives better
properties under a more theoretically rigorous treatment of long-range
electrostatics). But it's really just my intuition applied to a model
that I never use, so take that for what it's worth.


P.S. My expectation is, actually, that it would make no discernible
difference, and the effect of TIP4P vs. TIP4Pew would be tiny compared
to the effect of the solvent cap.

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Mon Jul 27 2015 - 07:30:02 PDT
Custom Search