You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for reporting this! It should be fixed now in 867b741.
This turned out to be a subtle but very serious issue specific to large multi-orbital models. In order to reduce RAM usage when building large systems, pybinding slices the work up into smaller pieces and buffers some intermediate hopping data. All the system slices have the same size except for the last one. This wasn't a problem for single-orbital models, but multi-orbital ones needed a buffer layout change in addition to the size change for that last slice. This is why the bug appeared in your example only after a certain size and for one side of the system (the last slice).
As seen in this gist a symmetry present in the local dos suddenly breaks down at a given system size.
The text was updated successfully, but these errors were encountered: