LJ potential with group option is broken in master

brucefan1983 commented 2 years ago

Need to fix this before releasing GPUMD-3.4.

brucefan1983 commented 2 years ago

We use atom symbols in model.xyz now and the code will check if all the atoms in model.xyz are allowed according to a potential file. This will break the hybrid-potential scheme, or we have to increase the complexity of the code to recover the hybrid-potential scheme. In my opinion, the hybrid-potential scheme is rather empirical and is usally not well justified.
Perhaps we have to leave GPUMD-v3.3.1 as the last version that have the best support for the emperical potentails. We have to accept this change, if we want to make the code clean and tidy, in terms of both coding and usage.
Do you agree with this? @erhart1

brucefan1983 commented 2 years ago

Solved in PR #261

brucefan1983 / GPUMD