[gmx-users] energy minimization

Erik Lindahl lindahl at csb.stanford.edu
Tue Feb 3 20:37:01 CET 2004


On Feb 3, 2004, at 7:58 PM, Madhuri Agashe wrote:

> While doing energy minimization, what is the effect of specifying 
> nsteps?
> Does nsteps decide the number of steps for which minimization will run?
>

The maximum number of steps the minimizer will run, unless it converges 
to machine precision first.

"Steepest descent" should really only be used as a quick-and-dirty 
minimizer to get to a state where you can start the dynamics. If you 
are actually trying to find a local minimum of the structure it is much 
better to use "cg" (conjugate gradients) or "l-bfgs" (Low-memory 
Broyden-Fletcher-Goldfarb-Shanno)

Cheers,

Erik




More information about the gromacs.org_gmx-users mailing list