[lammps-users] Quench method

LAMMPS LAMMPS Mailing List Mirror
1

Hi all,
I have some questions. How can i use the quench method with conjugate gradient for static molecular in lammps to do minimisation?

Tanck

2

Hi all,
I have some questions. How can i use the quench method with conjugate gradient for static molecular in lammps to do minimisation?

you have to explain in more detail what you are after. all features in LAMMPS are documented so the conjugate gradient minimization works as it is described and has the options and features the documentation mentions. if you want to do something else you either have to figure out how you can do this through other functionality or modify the source code and adapt it for your needs.

axel.

3

Thanck Axel,

I want to do minimization with fix box/relax and conjugate gradient. The problem is i have some problems like “linesearch alpha is zero”. So i try to do the same think with fix npt (the command that i use is "fix 1 all npt temp 1e-6 1e-6 16 tri 0.0 0.0 1000 ")and remove fix box/relax. My question is that, this commande fix at 10⁻6 K it is the same like fix box/relax? I have this question because i do molecular static and i dont know if fix npt at 10⁻6K it is a molecular static?

sorry if my question is so i am new to lammps. Thank you
4

Thanck Axel,

I want to do minimization with fix box/relax and conjugate gradient. The problem is i have some problems like “linesearch alpha is zero”. So i try to do the same think with fix npt (the command that i use is "fix 1 all npt temp 1e-6 1e-6 16 tri 0.0 0.0 1000 ")and remove fix box/relax. My question is that, this commande fix at 10⁻6 K it is the same like fix box/relax?

no. fix npt is a variable cell time integrator and only works with “run” not with “minimize”. the documentation for it clearly says so.

why do you think that “linesearch alpha is zero” is a problem? if it happens for a non-minimal structure then it is usually an indication that there is some other problem with your input, e.g. an overflow or a divergence in the force computation.

if your minimization is “stuck” and you are are certain to not have reached the minimum there are multiple things you can do:

  • use displace_atoms with a suitably small (but not too small) random displacement and re-minimize. that is also a good test for the stability of a minimum
  • switch between minimization algorithms, e.g. do some steps with steepest descend before trying again with conjugate gradient
  • change the length of the timestep. this can have an impact on minimizer algorithms (aka minimizer styles). you need to study their documentation carefully to see which

I have this question because i do molecular static and i dont know if fix npt at 10⁻6K it is a molecular static?

the term “a molecular static” makes no sense to me.

axel.

5

Thank Axel.