# [lammps-users] Equilibration problem

Hi Steve,
Is there any mistake in the gaussian distribution velocity creation in the code? I removed that statement and ran the code. The code equilibrates at ~60K satisfactorily.Any idea, how I can get it to do the same at 300K?

Regards,
Nandu Gopan

Hi Steve,
Is there any mistake in the gaussian distribution velocity
creation in the code? I removed that statement and ran the code. The code

no.

equilibrates at ~60K satisfactorily.Any idea, how I can get it to do the
same at 300K?

is this with nve or nvt?

you seem to be repeatedly ignoring two possible scenarios:

1) you initial system configuration has a high potential energy.
this will be converted to kinetic energy until an equilibrium
is reached. fix nvt is not a good method to get rid of that
kind of energy. better to find out why your system has that
much potential energy and adjust your initial config.

2) if your time step is too large, you will generate energy in
every step. that can add up to quite a bit.

you alway have to keep in mind the GIGO principle. if any of your
input data are bad, it is not the MD code's fault.

cheers,
axel.

Dear Axel,
Thanks for the inputs. The equilibration that I talked about is done using nve. And regarding the initial configuration; I did use the minimization code to minimize the initial energy.

Regards,
Nandu Gopan

I doubt there is a mistake with that. Right after you create velocities,
on step 0 of your run, LAMMPS prints out the temperature of the atoms.
Is that temp the target temp, regardless of whether you used gaussian or
uniform distribution? It should be (at least to several decimal
places accuracy).
If it is, then the velocity create did the right thing.

Steve

Dear Steve,
Yes, the temperature at step zero is the target temperature.Thanks for the input on checking the velocity command.

Regards,
Nandu Gopan

Dear Axel,
Thanks for the inputs. The equilibration that I talked about is done
using nve. And regarding the initial configuration; I did use the
minimization code to minimize the initial energy.

yes, but a minimization can only take you to the next _local_
energy minimum. it can only find global minima for small systems
(molecules) and even there it can fail to get the global minimum.
that is in particular the case, if you have a highly symmetric
initial geometry that not consistent with your potential. once
you "wiggle" this configuration a bit, it may easily drop into
a different minimum. thus my mentioning of having to do multiple
"quenches", i.e. removal of excess kinetic energy plus optionally
more minimization runs.

cheers,
axel.

Dear Axel,
Thanks for your suggestions. I do have a highly symmetric initial structure; I am trying the multiple ‘quenches’ now.This entire post been really informative. I have never had to encounter such issues before, for I have never tried modelling such a system.Thanks for your patient explanations.

Regards,
Nandu Gopan