5.13.7. localizer.get_wannier

eminus.localizer.get_wannier(atoms, psirs, Nit=10000, conv_tol=1e-7, mu=1, random_guess=False, seed=None)[source]

Steepest descent supercell Wannier localization.

This function is rather sensitive to the starting point, thus it is a good idea to start from already localized orbitals.

This optimizes the given orbitals under unitary constraint matrices, see IEEE Trans. Signal Process. 56, 1134.

Reference: Phys. Rev. B 59, 9703.

Parameters:

  • atoms – Atoms object.

  • psirs – Set of orbitals in real-space.

Keyword Arguments:

  • Nit – Number of iterations.

  • conv_tol – Convergence tolerance.

  • mu – Step size.

  • random_guess – Whether to use a random unitary starting guess or the identity.

  • seed – Seed to get a reproducible random guess.

Returns:

Localized orbitals.