Open Harrykjg-physics opened 2 years ago
Hi and thanks for your interest!
It's not in our immediate plans. The way I've been thinking about that is by extending the basis set to spin degrees of freedom; ie G_vectors would be (0,0,1,1,2,2) etc instead of (0,1,2), and there'd be a corresponding list of spins (-1,1,-1,1) etc. The Hamiltonian should just need a change in the kinetic energy, and we should change the way we represent densities to be 2x2 matrices. So I'd say medium goal: relatively straightforward in principle but there's a bunch of places in the code that need to be extended. We'd welcome any help there of course!
Is it possible to predict Non-colinear magnetic structures in an end-to-end manner(through AD maybe) ?
What do you mean by predicting structures in an end-to-end manner?
Update here: thanks to @epolack in https://github.com/JuliaMolSim/DFTK.jl/pull/851 we will get multicomponent support soon, then it should be a relatively simple matter to add the requisite terms in the Hamiltonian
See discussion https://github.com/JuliaMolSim/DFTK.jl/issues/134#issuecomment-1853704449, apparently there's something nontrivial to do with pseudos
I just want to know that adding spin-orbit coupling and non-linear magnetism are your plans or not in the near future. Is it a relatively easy or hard goal? Is it possible to predict Non-colinear magnetic structures in an end-to-end manner(through AD maybe) ?