nvt or npt ensemble

[Andrea Spitaleri]

Hi there,
in the last papers concerning metadynamics (jacs 2009 pietrucci et al, jctc 2009 pietrucci et al., jctc 2009 michel et al) the simulations have been carried out under NVT ensemble. is there any reason to choice NVT rather than NPT for metadynamics?
Any comment on this subject are very welcome.

Regards

Thanks in advance

Andrea

--
----------------------------------------------------------
http://sites.google.com/site/andreaspitaleri/
----------------------------------------------------------

[Fabio Pietrucci]

Dear Andrea,

as far as I know using NVT instead of NPT is not an issue specific of metadynamics, the choice is dictated by reasons common to normal MD: if you verified that volume and pressure do not play a particular role in your simulation, after equilibration you can also switch off the barostat.

regards
  
     Fabio

[Andrea Spitaleri]

Hi Fabio,
thanks for your reply. In fact I am trying to get deep into metadyamics simulation and I am testing different option too. In your paper, you did not mention whether you performed a Langevin MD or Brownian MD. Could you give some advices how these two approaches can influence the sampling in metadynamics? Also because the PLUMED gromacs tests are done with Langevin MD (integrator = sd).

thanks in advance

regards

andrea

2009/8/9 Fabio Pietrucci <fabio.p...@gmail.com>

[paolo raiteri]

Hi Andrea,

Conceptually there's no difference between a metadynamics simulation in the NVT or in the NPT ensemble.

However, in PLUMED no virial correction due the Gaussian bias has been implemented yet. 

This is often negligible if the CVs act on a small subset of atoms, which is often the case, but in principle by running a NPT metadynamics simulation without the virial correction will lead the system to equilibrate to a wrong pressure which changes as the bias changes.

To give you an example where this is crucial, think of inducing a first order phase transition in a crystal where all the atoms are involved into the definition of the collective variables and the pressure plays the key role.

Cheers

Paolo

[Giovanni Bussi]

On Mon, Aug 10, 2009 at 3:44 AM, paolo raiteri<paolo....@gmail.com> wrote:
> This is often negligible if the CVs act on a small subset of atoms, which is
> often the case, but in principle by running a NPT metadynamics simulation
> without the virial correction will lead the system to equilibrate to a wrong
> pressure which changes as the bias changes.

Moreover, the internal pressure of the system will depend on the CV.
Indeed, a good way to check this is to calculate the average internal
pressure selecting frames on the basis of the value of the CVs. The
more this pressure is uniform, the less is the systematic error due to
the neglection of the virial.

[Andrea Spitaleri]

Hi there,
thanks a lot to all!! I will try to metabolize what you said.

Regards

andrea

2009/8/10 Giovanni Bussi <giovann...@gmail.com>

[Fabio Pietrucci]

Hi Andrea,

in the papers we performed standard ("Newton's") MD, i.e. integrator=md in gromacs, plus of course the hills of metadynamics. Standard MD has the advantage of giving you trajectories that are similar to the real ones, even if the metadynamics bias perturbs them. The equilibrium kinetics can then be recovered a posteriori e.g. with the approach in PLoS Comput Biol 5(8): e1000452. About pros and cons of Langevin/Brownian MD someone else can perhaps give you advice.

regards

    fabio