langevin damping parameter

Hi Steve,

Thanks for your reply. I know that the NVT docs pages lists 100 timesteps
as a good damping time. I was hoping someone could suggest -why- 100
timesteps and what -good- means regarding a damping constant.

Is there a physical justification for damping constant that relates to
experiment or first principles result? I am not sure how to justify a
choice of damping constant beyond modeling considerations (i.e. "small"
fluctuations, short equilibration time).

Any paper suggestions are greatly appreciated!
Thanks again,
John

Like most thermostats, 100 timesteps is a good
damping time for temperature. See the fix nvt
doc page for additional discussion.

Steve

Hey John,
The following reference has a pretty good discussion of the things to consider when selecting the coupling strength.

Reference Type: Book Chapter
Author: Hünenberger, Philippe H.
Editor: Dr. Holm, Christian
Editor: Prof. Dr. Kremer, Kurt
Primary Title: Thermostat Algorithms for Molecular Dynamics Simulations
Book Title: Advanced Computer Simulation
Book Series Title: Advances in Polymer Science
Copyright: 2005
Publisher: Springer Berlin / Heidelberg
Isbn: 978-3-540-22058-9
Start Page: 130
End Page: 130
Volume: 173
Url: http://dx.doi.org/10.1007/b99427
Doi: 10.1007/b99427

Look at section 3.2.

Salomon

Hi Steve,

Thanks for your reply. I know that the NVT docs pages lists 100 timesteps
as a good damping time. I was hoping someone could suggest -why- 100
timesteps and what -good- means regarding a damping constant.

Is there a physical justification for damping constant that relates to
experiment or first principles result? I am not sure how to justify a
choice of damping constant beyond modeling considerations (i.e. "small"
fluctuations, short equilibration time).

ab initio has nothing to do with it. stat mech
is more where you have to look (as usual).

apart from proper statistical mechanical justifications,
you can make some handwaving considerations.

- there is a difference between equilibration and production,
  for the former you generally want to be on the shorter side
  of an "optimal" value for the later on the longer side.

- there are limitations to certain thermostats, especially
  nose-hoover chains are a bit sensitive by construction.
  other thermostats (langevin, dpd/tstat) have additional
  objectives/properties and have uses besides maintaining
  a constant temperature.

- thermostats can hide bad dynamics. one always have
  to make sure that an MD works well without without
  the thermostat, when only a "weak" coupling is desired.

- looking at the total temperature of a system can be
  misleading. careful people also monitor equipartitioning
  of the kinetic energy.

HTH,
    axel.

Thanks Salomon, that's very helpful!
Best,
John