[CPMD-list] convergence of metadynamics runs?

[Nisanth Nair]

Hi,

> 
> I am curious to know if anyone can shed some light on how to determine if 
> a metadynamics simulation has converged or not. 

It is hard to answer this question without knowing in detail what you are 
simulating, and what you are looking at, details of your simulation set up 
etc. So the answers here can not be complete. 

My advice here will be to 
learn and get experience for some simple case. A pleanty of 
practical information and tips are in 
Ensing et al. J. Phys. Chem. B. (2005) 109(14) ; 6676-6687

>To paraphrase that, when 
> does one stop a metadynamics simulations. 

If there is two minima A and B, and a transition A->B is observed, you can 
use this simulation for estimating the free energy barrier for A->B 
transition. If you need the relative free 
energies between A and B, you need to wait until B->A occurs.

>The literature I have read 
> indicates that it is critical to stop it at the right point otherwise the 
> system might get into exploring unphysical minima. 

what do you mean by unphysical minima?

you can continue as much as you can; but when computing the free energies, 
you can trim the colvar_mtd file and parvar_mtd file to the point you 
want.

> Is anyone aware of any 
> quantitative measure to determine when/if the free energy surface has 
> converged or not?

It is not possible to converge free energies below than a kT. Upto that 
you can. Use small gaussians and restart before the transition occurs and 
repeat simulations until you see convergence (if affordable). 

> 
> i) A recrossing back into an "old minimum" a good indication of 
> convergence (or close to convergence) ?

see above.

> 
> ii) Evaluating error bars in the free energy value from the hills along 
> the trajectory and along bins and checking to see that they remain 
> reasonably constant over blocks?

sorry, i couldn't understand!

best,
Nisanth


> 
> I'd be happy to hear other insights or opinions.
> 
> Ali
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://cpmd.org/mailman/listinfo/cpmd-list
> 

=========================================================================== 
 Nisanth N. Nair
 Lehrstuhl fuer Theoretische Chemie   Tel: ++49 (0)234 32 22121
 Ruhr-Universitaet Bochum                Fax: ++49 (0)234 32 14045
 Universitaetstrasse 150
 D-44801 Bochum, Germany
 Email: nisanth.nair at theochem.ruhr-uni-bochum.de
 Web  : http://www.theochem.ruhr-uni-bochum.de
===========================================================================