Hi Steve, Aidan,
Tilt is not flipped in the fix box/relax for triclinic box whose skew becomes too large during minimization at specified stress.
LAMMPS does not give an error when using fix box/relax even if the skew becomes larger than half length. However, if I were to dump the file after minimization and use that file as the input to another simulation, I get the message that the skew is too large.
This error is fixed if I run a fix deform (with delta 0.0, for example) after the minimization using box/relax and before dumping.
Thank you,
Kedar.
Fix npt (dynamics) throws an error when this occurs. Fix box/relax
does not, but it could. At some point we've talked about allowing
the system to flip when it happens (like fix deform does). But often
this happens b/c the user has setup an invalid problem. E.g. trying
to impose a shear stress on a liquid, which will never come to
equilibrium with fix box/relax. For now, I added a NOTE to
the fix box/relax doc page, similar to what fix npt has.
What is your system, where it is valid for it to shear so far during
minimization?
Steve
Hi Steve,
Thank you for your response.
You are right that in a single shear such a thing should not be expected. I do consecutive minimizations at incremental stresses on a multilayer bimetallic system.
In such a set up, may be the shear may get too large eventually ?
Kedar.
I think that fix/box relax only makes sense for systems with a well-defined
zero stress ground state. Then box/relax can be used to examine how applied
stress elastically perturbs the ground state cell parameters. Once you
introduce enough strain to create plastic deformation (long before the box
flips, I expect), mininimization and box relax are probably not appropriate,
because there are a large number of possible minima. Instead, you should be
using dynamic sampling of the energy landscape using fix npt, or fix deform.
Aidan
Thank you Aidan for your response.
In this study, I am not interested in the stress ranges that lead to cases such as plastic deformation. I stumbled upon such a response but I was not expecting LAMMPS to throw an error while reading a file that was just dumped by LAMMPS.
Thank you,
Kedar.
I stumbled upon such a response but I was not
expecting LAMMPS to throw an error while reading a file that was just dumped
by LAMMPS.
I agree that is counter-intuitive. What I was trying to say is
that the way we handle this with dynamics (fix npt) vs minimization
is to throw an error and stop the code. In your case we are not
doing that but it is leading to a geometry that the data file
doesn't support. So would you rather have an error, or jump
thru the reformatting hoop you are currently doing?
Steve
Hi Steve, Aidan,
For me, the way it is currently - error when loading next - is fine. As Aidan mentioned in his previous email, in configurations where the tilt is too large, the stress can not be expected to be the desired values. I reject all configurations that have stresses different from what I wanted them to have. I also have several other checks to see if the configurations are meaningful/useful.
My original setup involves post processing of the file dumped by LAMMPS and then calling LAMMPS again after I generate another input file. It took me a while to figure out the error since this was the last place I looked. I was using the fix deform for debugging purpose only.
Now that I know what the error is, I can continue with what I was doing without any issues. I wanted to bring this "issue" to your attention in case you were not already aware.
I thank you both very much for your quick responses and offer to make modifications. I really appreciate it.
Sincerely,
Kedar.