Polish dielectric example and improve chi0 computation #286
Conversation
| @@ -180,7 +186,7 @@ is false, only compute excitations inside the provided orbitals. | |||
| # ∑_{n,m != n} (fn-fm)/(εn-εm) ρnm <ρmn|δV> | |||
| ρnm = conj(ψnk_real) .* ψmk_real | |||
| weight = dVol * dot(ρnm, δV) | |||
| δρk .+= real(ratio .* weight .* ρnm) | |||
| δρk .+= (n == m ? 1 : 2) * real(ratio .* weight .* ρnm) | |||
There was a problem hiding this comment.
Can you add a comment either here or in the loop above? for m = n:nbands is very very easy to confuse with for m=1:nbands
There was a problem hiding this comment.
Also fuse the multiplication
| @@ -19,7 +19,7 @@ script: | # Install dependencies and test with coverage | |||
|
|
|||
| # Some optional python packages via conda | |||
| using Conda | |||
| for pkg in ["ase", "matplotlib =3.2.2"] | |||
| for pkg in ["nomkl", "ase", "matplotlib =3.2.2"] | |||
There was a problem hiding this comment.
Add a comment for why this is necessary?
| dρ = apply_χ0(scfres.ham, scfres.ψ, scfres.εF, scfres.eigenvalues, from_real(basis, dv)) | ||
| Kdρ = apply_kernel(basis, dρ; ρ=scfres.ρ) | ||
| vec(dv - Kdρ.real) | ||
| # Apply ε = 1 - χ0 (vc + fxc) |
There was a problem hiding this comment.
Eps is the dielectric operator, and applies to potentials conventionally. Call that epsT or something?
| end | ||
| end | ||
| end | ||
| howmany = 5 | ||
| arnoldi(eps_fun, vec(randn(size(scfres.ρ.real))), howmany) | ||
| arnoldi(eps_fun, vec(randn(size(scfres.ρ.real))); howmany=5) |
There was a problem hiding this comment.
We should replace that by krylovkit's native implementation anyway now that they have the "greedy" option. @gkemlin is doing that, we'll just merge later
| @@ -130,7 +133,7 @@ Solve the Kohn-Sham equations with a SCF algorithm, starting at ρ. | |||
| # Callback is run one last time with final state to allow callback to clean up | |||
| info = (ham=ham, basis=basis, energies=energies, converged=fpres.converged, | |||
| ρ=ρout, eigenvalues=eigenvalues, occupation=occupation, εF=εF, | |||
| n_iter=n_iter, ψ=ψ, stage=:finalize) | |||
| n_iter=n_iter, ψ=ψ, diagonalization=info.diagonalization, stage=:finalize) | |||
There was a problem hiding this comment.
Diagonalization_info? We probably need a more consistent naming for nested named tuples. Eg args for input and info for output, so this could be diag_info and the input parameters diag_args
No description provided.