Berendsen or Langevin thermostat

Hi ,

i was wondering if i use a NVE integrator which is the most appropriate thermostat to use in order to compute the diffusion of some molecules in a CNT? I mean in Lammps is there any limitations like for example the Berendsen+canonical ensemble which cannot produce consistent trajectories or the nve+langevin which performs Brownian dynamics?

2011/10/13 Jim Jim <[email protected]>:

Hi ,
i was wondering if i use a NVE integrator which is the most appropriate
thermostat to use in order to compute the diffusion of some molecules in a
CNT? I mean in Lammps is there any limitations like for example the
Berendsen+canonical ensemble which cannot produce consistent trajectories or
the nve+langevin which performs Brownian dynamics?

it depends on the system. there is not enough information here.

are your talking about diffusion/adsorption of individual molecules?
or diffusion inside a CNTs that as immersed in a bulk liquid?

in the first case you don't want any thermostat at all after
seeding your system with a suitable amount of initial kinetic energy.

in the second case, you want to use the NVT integrator
with a longish coupling constant, after the system is well
equilibrated.

axel.

I am talking about diffusion/adsorption inside an infinite CNT (PBCs). I was about to agree with you that i do not need a thermostat but only giving the molecules an initial kinetic energy (although i am not 100% sure that i know the theory about why 100% ) but what confused me is that i read in a paper (very good journal and the author as well) performing a NVE + a thermostat for the same case.

2011/10/13 Jim Jim <[email protected]>:

I am talking about diffusion/adsorption inside an infinite CNT (PBCs). I was
about to agree with you that i do not need a thermostat but only giving the
molecules an initial kinetic energy (although i am not 100% sure that i know
the theory about why 100% ) but what confused me is that i read in a paper
(very good journal and the author as well) performing a NVE + a thermostat
for the same case.

a thermostat may have been used to counter insufficient energy conservation.

check out the time constant it is probably fairly long. if you want
to do long trajectories for a system with potentials that don't conserve
energy extremely well (or choose an aggressive time step in order
to trade better statistics for an error in dynamics). in that case,
i would prefer a berendsen thermostat over langevin since it
doesn't change the direction of velocities. i would make a series
of test runs to determine, how much thermalization is needed
and try to keep it a small as possible.

axel.

If at some point later on, you become interested in thermal
equilibrium averages, people use Langeven over the Berendsen
thermostat. (The Berendsen thermostat has caused us all kinds of
problems with weighted-histogram analysis/WHAM, and replica
exchange/REMD, and other related sampling/analysis methods.)
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2699284/
I know this comment is irrelevant to your current question about
dynamics and diffusion.