[gmx-developers] a problem about the implement of shear flow in gromacs

[J Hu]

*Hello,*


*I **have recently tried to simulate a shear flow in a molecular dynamics
model by implementing the **Lees Edwards boundary (periodic shear flow)
conditions in GROMACS. However the *

*results I got are strange ...  *

*I added the effect of the shear, delta_U(y)=gamma*dy, where gamma = dU/dy
by modifying the coordinate equation dx/dt = ... + gamma*dy  in the
x-direction so that coordinate change along the x direction becomes
x(t+dt) = x(t) + dt*[v(t) +gamma*dy]. The *
*calculation of the molecular dynamics velocity, v(t)  from the pair
potential was kept the same as usual from Newton's second law without any
change.*


*To compare with the analytical shear flow solution U(y), I then wrote a
routine that averages all atoms velocities in time as well as in the
homogeneuous x&z directions and so obtained the velocity profile along the
y-direction (the  velocity **information was directly obtained from the gro
files). However, strangely, the resulting profile of the mean atom
velocity, which I expected to be flat around zero, turned out to have a
negative slope, -gamma, as if the model tried to compensate for the
positive shear flow I was trying to impose. This effect is very consistent
for a range of different shears -- from moderate to very small gamma. So I
thought this is due to the thermostat or some other intrinsic GROMACS
feature which checks the atom velocities/deviations and *

*holds the velocities to correspond to the uniform state, hence, creates an
"anti-shear" effect.Did someone encounter a similar problem in GROMACS and
know how to disable these stabilising  features of *
*GROMACS?  Or, does someone know a simpler method how to generated a shear
flow in GROMACS ?*


*Thanks a lot.*


*Best wishes*


*Jin*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20161004/808af16a/attachment.html>