5.3.1. band_minimizer.auto

eminus.band_minimizer.auto(scf, W, Nit, cost=scf_step_occ, grad=get_grad_occ, condition=check_convergence, betat=3e-5, cgform=1)

Automatic precond. conjugate-gradient minimization algorithm for a fixed Hamiltonian.

This function chooses an sd step over the pccg step if the energy goes up.

Parameters:

• scf – SCF object.

• W – Expansion coefficients of unconstrained wave functions in reciprocal space.

• Nit – Maximum number of SCF steps.

Keyword Arguments:

• cost – Function that will run every SCF step.

• grad – Function that calculates the respective gradient.

• condition – Function to check and log the convergence condition.

• betat – Step size.

• cgform – Conjugate gradient form.

Returns:

Band energies per SCF cycle and optimized expansion coefficients.