Date: Wed, 24 Sep 2003 20:14:47 -0500
From: "Nathan A. Baker"
Subject: Re: AMBER: NaCl ion pairs in TIP3P water box

Thanks to all who replied; it appears this phenomenon may not be as
surprising as I thought. I've lumped the replies below as a summary
and included a few more comments. However, I do have a general
question: has anyone tried to calculate solubility products for salts
in TIP3P?

David A. Case (09-24-2003 14:09:06-0700):
>Nathan: I hope Tom Cheatham or Tom Darden will speak up here, but I suspect
>that this is a force-field "problem". Do the ion pairs break up if you
>soldier ahead and go to 300K, or do they remain? I don't have any explicit
>relevant experience, but I seem to have heard of similar problems through
>the grapevine. (If you haven't done so, ask Jay what he thinks...)

In fact, I started running the salt/water box becuase of a disaster
with a protein system at 800 mM NaCl. For the protein system, not
only did the ion pairs persist at 300 K but, after 500-1000 ps, they
nucleated larger "crystals" of salt. Some of these crystals contained
as many as 8 ions! Not surprisingly, this wreaked havoc with my free
energy calculation...

I haven't carried the salt/water box out much longer at 300 K; once I
started spotting 3-4 pairs again, I figured I was in trouble...
However, now that I know I'm not encountering anything too unexpected,
I'll pursue this trajectory a bit further.

I have definitely been talking with Jay about using AMOEBA on the same
system; hopefully polarizability could make the difference here.

Jiri Sponer (09-24-2003 23:41:27+0200):
>I think this could be due to atom-atom pair-additivity.
>There is no polarization and CT towards the (and through)
>the ion first coordination shell and actually I can imagine
>this type of a dysbalance as one of the possible outcomes.
>Point charge is another approximation that causes troubles
>with ions, especially anions.

That definitely seems likely; I'm planning to talk with Jay Ponder
about using his force field for this system.

>I would possibly try to reduce the charges below 1 in
>absolute value to see if any change due to a shift in compensation
>of errors.

Very interesting idea; I'll have to play around with the charges a
bit.

Bill Ross (09-24-2003 13:53:39-0700):
>> However, I invariably end up with ion pairs by the time I reach the
>> 150-200 K stage of this procedure.
>That's the norm. Same-charge ions form pairs with pairs of
>waters, e.g.
>
> H H
> O
> + +
> O
> H H
>
>Bill Ross

Interesting -- I do see some ion pairs like this but the majority are
NaCl pairs.

Ioana Cozmuta (09-24-2003 15:30:26-0700):
>I did some simulations with KCl at 1M and SPC/M water for about 1ns. I've
>started with two different configurations: the first in which ions are
>randomly placed in the unit cell. The second in which I study crystal
>dissolution. I end up with approximately same average behaviour thus I
>think that at high concentration pairing and even clustering of ions is
>not impossible and that's what some experimental chemist think as well.
>It's a dynamic equilibrium however, not all the time the same ions
>pair/cluster.

Yes; this does seem possible. However, I'm worried that the lifetime
of these pairs (and larger clusters) will be much longer than I can
simulate. Was 1 ns sufficient to observe the dissolution of the
crystal you were examining?

M. L. Dodson (09-24-2003 18:53:22-0500):
>
>This seems like a problem that should be solvable by ab initio
>methods, maybe?
>
>Bud Dodson

Or polarizable force fields! :)