[gmx-users] unexpected results with stochastic dynamics

[Michael Brunsteiner]

Hi everybody,

I just ran a number of simulations with two large and roughly spherical molecules
in a primitive model electrolyte (epsilon=80), they are constrained to the z-axes and
i calculate the PMF using the pull code restraining the distance between the
two molecules. integration is done with langevin dynamcs (gmx-4.5.3)

what i see is this: I start several runs all with the same initial orientations of the two
molecules only the distance varies, the dynamics seems to go well, but when i compare

the FINAL ORIENTATIONS of the two molecules after multiple runs at different
distances they are NEARLY THE SAME not matter which distance i choose.

below is the relevant part of my mdp file:


integrator               = sd
tau_t                    = 2.0
ref_t                    = 300
gen_vel                  = yes
gen_temp                 = 240.0
gen_seed                 = xxx ; this varies
tc_grps                  = System
nsteps                   = 200000
dt                       = 0.001

The forces depend on the distance of course (the molecules have
a heterogeneous charge distribution) and this is a highly chaotic system
therefore the probability that the conformations (basically the
rotational DOFs of the two molecules) are the same or very similar after
200 ps at 300K should basically be ZERO ... but this is what i see, 

its re-producible and happens at different distances AND with different
molecular topologies (charge distributions).


a possible explanation for this behaviour might be that of the terms 

in eqn. 3.108 only the velocity dependent and the random term are actually
considered but not force term AND gen_seed doesn't make any difference for 
SD ... then the small differences i see would be only from round-off errors ... ???

has anybody seen something similar ?

i'd appreciate any suggestions!

thanks
Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110429/53050234/attachment.html>