-
|
Hi! I am currently working on amorphous materials, and I would love to get recommendations on which toolkit I should use for generating amorphous structure. The details are below. Generate “amorphous” metal oxides and perform their DFT or DeepMD relaxation. Here, I want to make thousands of random amorphous structures (up to 150 atoms) with high structural diversity but with moderately realistic starting structures so that the relaxation process isn’t too slow. The relaxed structures will be fingerprinted to assess their structural diversity. One scheme would be a high-temperature MD run followed by relaxation of various MD timesteps. Another scheme would be an explicit structure creator. I appreciate any input on this! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Sorry, I don't know of any toolkits that generate amorphous structures easily... All methods that I know of involve a simulated annealing step, which is normally done with DFT. This guide gives a good overview of the process. Since you want to generate thousands of candidates, the best route is probably to do some quick simulated annealing using an empirical potential before doing a more robust DFT relaxation. Maybe use a basic lennard jones or EMT potential in ASE (see their options here) combined with ASE's MD module (here) to quickly and roughly anneal the structures. You could also randomize the pre-annealed input structures depending on how diverse you want the amorphous ones to be. |
Beta Was this translation helpful? Give feedback.
Sorry, I don't know of any toolkits that generate amorphous structures easily... All methods that I know of involve a simulated annealing step, which is normally done with DFT. This guide gives a good overview of the process.
Since you want to generate thousands of candidates, the best route is probably to do some quick simulated annealing using an empirical potential before doing a more robust DFT relaxation. Maybe use a basic lennard jones or EMT potential in ASE (see their options here) combined with ASE's MD module (here) to quickly and roughly anneal the structures. You could also randomize the pre-annealed input structures depending on how diverse you want the amorphous ones to be.