[gmx-users] equilibration: run time or number of steps?

[Justin A. Lemkul]

Juliette N. wrote:
> Hello,
> 
> For the equilibration one usually looks at the total energy or the 
> observable of interest to be independent of time. I wanted to figure out 
> when we are referring to equilibration which of the run time or n_steps 
> parameters are important. One could run 1,000,000 steps with dt of 0.001 
> ps or 0.002 ps ending up with 1 and 2 ns simulation times. Can we infer 
> that 2 ns trial is closer to equilibration or one needs to look at the 
> n_steps regardless of total run time. I hope my question in clear to 
> receive your feedback.
>  

I would say time is more important than the number of steps (if dt is altered 
accordingly).  With dt = 0.000000001 I can run a billion steps and only achieve 
1 ps of simulation time, which is not nearly enough for any relevant quantity to 
converge.  Stability, of course, is an issue and hence we cannot simply run a 
million steps with dt = 0.01 ps and expect an atomistic simulation to be stable.

Proper equilibration is done such that the observables of interest have 
converged.  What those observables are depend entirely upon the system in 
question, but you certainly have to have stable energy, temperature, pressure, 
etc (depending on the ensemble).

-Justin


-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================