Nick Papior

@sofiasanz let me know if you have any questions, sisl has moved quite a bit... To a large extent it *should* be backwards compatible, but I would probably recommend to...

This article might be valuable: https://www.sciencedirect.com/science/article/pii/S0010465523000292?via%3Dihub

FYI did you see the btd in sisl/toolbox/btd? If your code enhances that one it could be of general use? ;-)

> @sofiasanz by the way, is the elec_idx arrays supposed to have the (1,n) shape, or is it me who have gotten the wrong impression at some point? :) The...

> @zerothi I just think it introduces a hidden transposition of the self-energies if it is indeed (1,n)? if it was (n,1) i think it would be fine :) its...

> I dont understand, is this related to the "batching" of the energy-points? :) Yes, but perhaps I misunderstood its purpose, I just saw you did `array_split`, but to me...

> Yes I added the `array_split` method to avoid building too large matrices and do it rather in "batches". This is only done for the NEQ integrals since is the...

> Yes zS - H - sum(Sigma) is build for all the energy-points given, but I guess it only stores the A_i, B_i and C_i blocks so this only scales...

I think an `SCFHamiltonian` is the way forward, I think SCF objects should *hold* the objects that they act on. The problem is that `Hamiltonian` should do 1 thing, and...

> Thanks, I had forgotten about [zerothi/sisl#97](https://github.com/zerothi/sisl/issues/97). Any particular inheritance in such a `SCFHamiltonian` class? Or is it rather a new one from scratch? (Not sure I understand what you...