Storing structures so only diff in energy need to be calced - feature request?

icet
1

Is there an option to keep a history of analysed structures or just note the change in structure to the CE to make the CE to only analyse the change in the structure and not the whole structure?
I believe it would speed up my setup drastically.

I got rather big structures, and I am doing a KMC of LiNiO2 I would even like to go bigger with the structures, but ce.predict() needs quite some time. I would like to speed that up but didn’t find a cache option to store the global crystal and only analyse the change (or the space around the change - inside the biggest cutoff of the CE)

2

Hello,

I dont think there is any support for this using ce.predict(), however the ClusterExpansionCalculator does this internally when performing swaps. It has a `use_local_energy_calculator` kwarg which defaults to True, and then just calculates the local change in the energy when a MC-trial-step is performed via `_calculate_partial_change` function.

So probably you could adapt and reuse that functionality for your use-case

3

Ah thanks; I’ll look into the code of that and try that