[gmx-users] Re: Advantage of NPT over NVT?

[Mark Abraham]

Vitaly V. Chaban wrote:
>> Vitaly V. Chaban wrote:
>>> If your model gives a correct fluid density with NPT, use NPT. If it
>>> does not, then use NVT.
>>> This is a quick practical guide, not not read much. :)
>> Sounds dangerous :-) If the model produces a wrong density, you'd want
>> to be reasonably confident your observable is not linked to the
>> underlying cause, before just proceeding at NVT... Sounds like a really
>> easy criticism for a reviewer to knock back a paper on.
> 
> Mark,
> 
> It sounds dangerous if the model is not thoroughly tested. If you know
> that the model is OK using NVT, but it gives a bit wrong density with
> NPT - this is an argument to select the first ensemble. Such behaviour
> is peculiar to a big number of "rigid particle" models.

Sure. Having demonstrated it's OK is the important thing. The original 
context was advice to a newcomer, who might well not appreciate that 
point. :-)

Mark