[gmx-users] Brownian and Langevin Dynamics
Marc Baaden
baaden at smplinux.de
Fri Mar 8 02:16:24 CET 2002
Hi,
I am playing a bit with Brownian and Stochastic dynamics and have a couple
of questions. Firstly concerning Brownian dynamics:
- is there a general consensus, when one can assume diffusive motion to be
predominant and thus use the BD equations ? (I am fine with ions, but my
question relates eg to mobile extracellular loops of a membrane protein,
does it make sense to sample their motion with BD or is SD required ?)
- concerning the bd_fric parameter and the time step to use for BD, I was
wondering wether there is any significance in the absolute values ? If
I understand the BD equation used in Gromacs correctly, the relation of
timestep compared to the friction constant determines the largest 'leap'
the moving particle can take, depending on the forces it experiences.
Eg, if I keep the same ratio dt/bd_fric, and increase the time-step let's
say from 0.005 ps to 0.05, 0.5 .. I should get the same trajectory. Is
that right ? (assuming the random number generator would produce the same
series of random numbers going into the calculation of the random force)
If it is right, has anybody looked into what sensible values are ? (I
remember Erik's mail saying something like 3000 .. but that must then
depend on dt .. and what justification can one give ?)
Another implication would also be that the actual simulation 'time' has no
physical sense .. (?)
- is there a rule of thumb when one constant bd_fric or one mass weighted
coefficient should be used ?
Concerning SD:
- referring to the question about bd_fric/dt for BD, am I right that for
SD the actual timestep is again insignificant and only the ratio counts ?
(in both dissipation and collision term bd_fric and dt should cancel out)
- with respect to bd_fric, I have seen people suggesting rather to
use a weighting factor depending on the number of neighbors (cf 'exposed
area'), are there any plans of implementing such a thing or is it obsolete ?
Thanks a lot.
Marc
More information about the gromacs.org_gmx-users
mailing list