[lammps-users] Annealing methods?

LAMMPS LAMMPS Mailing List Mirror
1

Hi Users,

I’m trying to “anneal” a simulation. Right now I’m doing this:

fix 1 all nvt temp 400.0 400.0 100.0
run 40000

unfix 1
fix 2 all nvt temp 350.0 350.0 100.0
run 30000

unfix 2
fix 3 all nvt temp 300.0 300.0 100.0
run 30000

Is there a better way to accomplish this? Also, as a secondary question, what does the “Tdamp” parameter do in the NVT fix?

Thank you for your time and help,

Casey

2

Hi Users,

I'm trying to "anneal" a simulation. Right now I'm doing this:

fix 1 all nvt temp 400.0 400.0 100.0
run 40000

unfix 1
fix 2 all nvt temp 350.0 350.0 100.0
run 30000

unfix 2
fix 3 all nvt temp 300.0 300.0 100.0
run 30000

Is there a better way to accomplish this? Also, as a secondary question,

ouch! yes, there is.
you use velocity create with suitably high initial value
and then fix nve in combination with fix viscous.
that is _the_ annealing procedure. if you want to make
it a bit more efficient, you can change the masses
of atoms to make them all the same.

what does the "Tdamp" parameter do in the NVT fix?

it is a time constant that determines how "aggressive"
the thermostat tries to bring the temperature to the
target value. for a well equilibrated system it is roughly
equivalent to the amount of time it takes to reach the
a target temperature.

axel.

3

Using NVT to anneal is fine. You can slowly ramp the temperature
over time to cool the system if you like. Tdamp is explained in great
detail on the fix nvt doc page.

Steve

4

Usually, I always use Tdamp = 10 * timestep for NVT. You can check the result by plotting MD steps vs temperature and see the fluctuation happen.

rizal

2011/3/18 Steve Plimpton <sjplimp@…33…24…>