[gmx-developers] Re: The development of a new implementation

[Michael Shirts]

Hi, Ramon-

> Until now, I am implementing for myself an iterative scheme. I don't think
> that it is worth of being contributed (correct me otherwise). I am thinking
> about implementing a non iterative scheme. But it is quite non trivial. I
> am not sure about how can one have a reversible Parrinello barostat without
> iterating. I guess that there must be a method published. I will look for
> bibliography.

You're right, it's not 100% reversible -- you make a slight
approximation for the new velocity of the box size in order to predict
the resized velocities of the particles. This is very close to the
correct answer, and NPH simlations run with essentially constant
enthalpy for many nanosecons. Perhaps we should add an optional
iteration step just in case this isn't good enough for some
applications.

Best,
Michael