[CP2K:5730] Re: Nickel cell optimization

[Noam Bernstein]

>  
> "This thing" is my version of CP2K. Yes, the "LANGEVIN" ensemble version is there, there's just no documentation of what that does to barostatics. I need both thermostat and isotropic barostat. If I select LANGEVIN as the ensemble, is that effectively well-damped NVT?

As I recall nvt vs npt is selected at the ensemble level, and is that's indeed the case the langevin ensemble might only be constant V.

I implemented AD_LANGEVIN, but I don't actually know much about how the rest of the dynamics in CP2K is implemented. I certainly never tested it together with a barostat. When I implemented barostats and thermostats in my own code, I was careful to come up with a consistent time propagation algorithm that included both. Assuming that the rest of the constant P dynamics (in CP2K) was implemented in a way that's reasonably thermostat agnostic, and since I based AD_LANGEVIN on the existing Nose and Langevin codes, it has a decent chance of working.  

There are also other stochastic thermostats like CSVR, which have been in the code longer and are presumably well tested, and might be decent for a small system.

What's keeping you from updating to a newer version, anyway?

   Noam