search

cortner / SKTB.jl

Other
9 stars 4 forks source link

spd Model #6

Open cortner opened 7 years ago

cortner commented 7 years ago

I seem to have broken the spd model :(

cortner commented 7 years ago

actually, the spd model matches the one in Atoms.jl, so if the current one is broken, then so is the one in Atoms.jl. What is going on???

cortner commented 6 years ago

@jarvist as I mention here I don't fully trust my spd implementation. If you notice anything in the band diagrams you are looking at that seems wrong, would you be willing to report it here?

jarvist commented 6 years ago

Will do. (Haven't checked anything yet, but doing so is certainly high on the agenda.)

cortner commented 6 years ago

thank you

cortner commented 6 years ago

@jarvist Did you ever convince yourself whether (or not) the spd implementation is correct?

jarvist commented 6 years ago

Have not convinced myself yet, but it will be done.

cortner commented 6 years ago

it isn't urgent yet for me, but it will become urgent in about 1-2 months' time. How would you actually go about testing the SK implementation?

jarvist commented 6 years ago

I'm planning to cross-validated against TBE with an identical model. you can punch out the Hamiltonian matrix elements from TBE to check, as well as compare the band structure. Sticking points are getting the full-blown band structure code added to TightBinding.jl rather than my lightweight hack for HCP metals, and implementing an identical model in both codes.

cortner commented 6 years ago

but with the NRLTB data that should in principle be ok. Just one warning: I changed the cut-off a little bit (I think it is maybe an option) since the standard NRLTB cut-off is not C1, i.e., it doesn't actually go to zero smoothly at the cut-off radius