The version of lammps I am using is lammps-stable_12Dec2018
Hello, I have currently been using lammps to calculate elastic moduli of saddles and minima. One step in this elastic moduli is to allow the structure to minimize while allowing the box shape to change by using the box/relax function. We perform this on three different structures so we need to be able to make this calculation for each structure.
In order to have a structure in atom->x to use, my predecessor used a minimization command with number of steps equaling zero.
What I am not understanding is that this minimization command seems to change the final results.
For example, if I run the minimization command with zero steps twice I will get a completely different result then if I run it only once.
I looked at min.cpp and it seems that if 0 steps are used no minimization should take place. What could be causing this inconsistent result?
Thank you and please let me know if any clarification is need,
Alec
The version of lammps I am using is lammps-stable_12Dec2018
please consider upgrading. there have been improvements and bugfixes (in general and specifically to minimizations) since.
Hello, I have currently been using lammps to calculate elastic moduli of saddles and minima. One step in this elastic moduli is to allow the structure to minimize while allowing the box shape to change by using the box/relax function. We perform this on three different structures so we need to be able to make this calculation for each structure.
In order to have a structure in atom->x to use, my predecessor used a minimization command with number of steps equaling zero.
i don’t understand what you mean by that. it would be best, if you could provide an easy to read and fast to run minimal, small example demonstrating the workflow (without the need to be accurate). my suspicion is that you may confuse the meaning of the parameters of the minimize command.
What I am not understanding is that this minimization command seems to change the final results.
For example, if I run the minimization command with zero steps twice I will get a completely different result then if I run it only once.
you have to explain what you mean by “completely different”. in the past we have see calling people something completely different that was essentially the same.
how well a minimization can find a consistent minimum depends on a) how many degrees of freedom you have (i.e. size of the system and symmetry of interactions) b) and what minimization algorithm you use (some try to be more efficient by basing the next step on the history of previous steps and do that in different ways), and c) how noisy (or rugged) your potential hypersurface is.
you may get easily trapped in a local minimum. to confirm a minimum one can do “stability tests”, i.e. displace atoms by small random amounts and re-minimize. for a consistent minimum, the final coordinates should be the same within the available accuracy.