Help needed displaying correct atom locations

Hello all,

I am attaching The modified Monte Carlo code for energy minimization (from lammps example) used on a known stable Cristobalite Structure for Quartz. The code gives the following result corresponding to the attached input file and parameters:

image.png

Note that the Minimum energy of perfect lattice is corresponding to the stable energy of the already optimized structure. On moving the original atom locations in the input file by 0.01 on multiple atom locations , the code gives the same minimum perfect lattice energy. Unfortunately, this isnt true for displacement of atoms >0.01 A. I have been attempting to display the atom locations corresponding to the minimum energy of the perfect lattice using " “write_dump all xyz system.xyz modify sort id” &". The above line keeps displaying the final location of the lattice corresponding to the final energy -7354 as shown above, instead of -5421. This claim can further be verified by the xyz values and the video generated. Please provide guidance on where to put the xyz location generating command to output the right atom locations.

In order to find the minimum energy upon annealing is increasing kT value sufficient or something else needed?

PS- potential used in this code is corresponding to the attached paper.

cristobalite (beta) Coordinates.txt (2.93 KB)

in.quartz_GP_Si (3.84 KB)

data.G8SiO2 (982 Bytes)

Guillot_Paper.pdf (606 KB)

system.xyz (667 Bytes)

two comments on this:
a) if you want the coordinates corresponding to the energy stored in the variable emin, you need to output them right where it is computed (after the first run 0).
b) you have a high dimensional potential hypersurface which has many minima. the minimization algorithms will only find a local minimum, not the global minimum. thus the more you perturb the initial structure, the more likely it becomes that minimization will “find” a different local minimum. depending on the composition and geometry of your initial geometry (which according to your input has not been minimized or relaxed, so whether that corresponds to the “perfect lattice” depends on how you have constructed the initial geometry). in the example that you modified, the initial geometry is a 2d lattice of lennard-jones particles, for which the initial structure can be easily seen to be the perfect lattice, since there is effectively only one free parameter.

your other questions are about understanding of the method you are using and are thus irrelevant to LAMMPS and thus off-topic for this mailing list. i suggest you discuss with your adviser/supervisor, as that is the person you should be discussing your understanding of the science of your research with.

axel.

image.png