Replica exchange and multiple time-stepping

From: Niklaus Johner (nij2003_at_med.cornell.edu)
Date: Mon Mar 05 2012 - 10:19:10 CST

Hi all,

This is just a warning concerning replica exchange in namd:

I think steps_per_run has to be set to something that is divisible by 10!

Otherwise the potential energy that is used to calculate the exchanges might be wrong. I think this issue is linked to multiple time-stepping and to the fact that the ouputEnergies is set in the replica_exchange.tcl script to [expr $steps_per_run / 10] which truncated to an integer, so that you will have the output of the energy not coinciding with the the time-step at which the energy is evaluated.

I obtained the correct energies both by using a steps_per_run that is divisible by 10 or by changing line 87 of replica_exchange.tcl to {replica_eval null "outputEnergies [expr $steps_per_run ]"}.

I hope that helps and any input on the reasons for this "error" are welcome. Also, a warning about this in the manual or in the example tcl script for replica exchange would be useful.

Best,

Niklaus

Niklaus Johner
Weill Cornell Medical College
Harel Weinstein Lab
Department of Physiology and Biophysics
1300 York Avenue, Room D-501
New York, NY 10065

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:17 CST