I am developing a new fix, say fix custom_force, that adds forces on atoms. For that I need to store some atom attributes which I do by using fix property/atom within my fix custom_force. Now I also have to add and delete atoms and need to write a new fix, say fix add_delete, that does that based on certain criteria. My question is once I delete and add atoms (i.e. change the total atom count), how does the fix property/atom, initialised and defined within fix custom_force, behave in such a case. Is there a way to retain the information for old atoms and add new information for the new atoms?
When you delete atoms, their data will be inaccessible and eventually remove or reused. When you add atoms, new storage will be allocated and initialized to 0.
You would have to program that yourself, since only you know which old atom should provide the information for a new atom. Keep in mind that per-atom data is distributed across the parallel domains, so it may be tricky to do this unless you have extra communication and replicate the deleted atom data across all MPI processes.