Re: simulated annealing

From: Brian Radak (
Date: Thu Apr 07 2016 - 10:00:03 CDT

On 04/07/2016 08:14 AM, faride badalkhani wrote:
> Dear all,
> I need to pick your brain about annealing procedure because it is my
> first time I want to run an annealing simulation. I need to use
> heating steps of 50K, from 300K to 700K and then same cooling
> procedure from 700K down to 300K with 50K steps. Now I have some
> questions:
> 1) I do not know how many steps should I consider for each temperature.
There is no objective way to determine this - it all depends on what
kind of resources you have. I personally like >= 1 ps/K, but that is
totally arbitrary and just happens to sound nice to me.
> 2) Is it necessary to perform a NPT MD simulation after each energy
> minimization at each temperature? If yes how should I apply it in the
> configuration file?
Probably not. I don't see much point in minimizing at each temperature
change either. Remember, water boils (though TIPnP models usually don't
boil like real water), so running NpT will likely not accomplish what
you want.
> 3) which one is more reasonable? performing annealing procedure just
> for my structure in vacuum or for molecule+water box system?
I would say that annealing in vacuum is pretty much the same as any
simulation in vacuum. Keep in mind, if you have multiple associated
molecules, they will likely dissociate at high temperatures and this is,
again, probably not what you want.
> 4) could you give me a template of simulated annealing configuration file?
Here's a quick script that I wrote the other day to ramp up the
temperature along a ladder at 550ps intervals (I wanted to scan the
approximate average energy at each temperature):

set nsteps 275000 ;# at 2 fs timestep, this is 550ps
set Told 298.0
set Ts {303.2 308.4 313.8 319.3 324.8 330.5 336.2 342.1 348.0}

langevinTemp $Told
run $nsteps
foreach T $Ts {
   rescalevels [expr {sqrt($T/$Told)}] ;# this should nominally cut down
on re-equilibration
   langevinTemp $T
   run $nsteps
   set Told $T


> Regards,
> Farideh

Brian Radak
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory
9700 South Cass Avenue, Bldg. 240
Argonne, IL 60439-4854
(630) 252-8643

This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:20:19 CST