Begging the first question, the neutralization bug itself, would you
consider fixing this problem in one of the following updates or
versions, or is it something implemented by the method the system
charges are calculated and cannot be fixed?
Thanks,
Alican
On Mon, 15 Sep 2014 12:23:17 -0400, Jason Swails wrote:
> On Mon, 2014-09-15 at 11:45 -0400, Marcus Arieno wrote:
>> Out of curiosity, what type of simulation artifacts would one expect
>> if the periodic system carries a net charge? I suspect such an issue
>> might give slight density problems if running NVT simulations for a
>> long time, but NPT should be more or less fine.
>
> I'll assume you're referring to "with Particle-Mesh Ewald", because
> the
> answer to this question depends heavily on the model you're using for
> the long-range electrostatics. I'll also assume you meant a problem
> with NPT rather than NVT, since density fluctuations can only occur
> with
> the former. And indeed you're right -- the most common artifacts are
> pressure-related as I recall (although NVT is susceptible to some
> artifacts as well).
>
> The standard PME implementation results in what is called a "net
> neutralizing plasma" for charged systems (I believe this is a
> consequence of omitting the k=0 term in the reciprocal sum which is 0
> for a neutral system or infinite for a charged system). This is
> formally equivalent to introducing a "background" charge that
> neutralizes the entire system and smearing it evenly across all of
> the
> atoms. For small unit cells, you see a bigger effect than for larger
> ones. For example, see DOI 10.1063/1.476320 from ~15 years ago.
>
> It also seems that for highly non-homogeneous systems (like lipid
> bilayers, micelles, etc.), this correction can actually increase the
> preference for the ion to embed itself inside a low-dielectric region
> (i.e., inside the bilayer)! See, for example, DOI 10.1021/ct400626b.
>
> In practice, this correction is often used for calculations in which
> the
> total charge is not fixed (a prime example being pKa calculations,
> such
> as the model I presented in the article DOI 10.1021/ct401042b) and it
> yields satisfactory behavior.
>
> I think the discussion in the second article I linked is particularly
> instructive, and if you simulate a small system you should be more
> careful.
>
> Hope this helps,
> Jason
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Sep 16 2014 - 09:00:02 PDT
