[gmx-users] Using thermostats to create temperature gradient in system

Suzen, Mehmet msuzen at gmail.com
Sun Apr 26 18:09:08 CEST 2015

Hi Agnivo,

Temperature gradient means non-equilibrium MD (NEMD)

See notes from Prof. Martini:

What observable would you like to measure? Lets say you want to
measure observable A.

One "procedure" I can think of:

1.  You can run all system at T=300K with NVT. After equilibrium,
continue running all system at T=400K with freezing solvent group.
See, to how to freeze a group,

To get a good statistic, you need to repeat this procedure K times as
"solvent thermal wall" configuration  would be
different at each simulation.

100ps? Where did you get that? You need to see how observable A
behaves to see if 100ps is enough, <A> time average. And see how <A>
behaves in K different cases.

2. Similar to previous procedure but this time after freezing solvent
group, run system in NVE. Again, you need to repeat this
many times to get a good statistic. Similarly monitor observable A to
see if 500ps is enough.

However,  I suggest you to talk to someone having NEMD experience as
this procedure is just a heuristic approach.


More information about the gromacs.org_gmx-users mailing list