abinit/doc/topics/_SCFAlgorithms.md

49 lines
1.8 KiB
Markdown

---
description: How to select the SCF algorithm
authors: XG
---
<!--- This is the source file for this topics. Can be edited. -->
This page gives hints on how to select the SCF algorithm with the ABINIT package.
## Introduction
Self-Consistent Field calculations allow to determine the solution of the
Kohn-Sham equations, ending with converged "self-consistent" wavefunctions,
density, and Kohn-Sham potentials.
Different algorithms can be chosen to
converge to the solution of this set of equations.
The input variables [[iscf]] governs
the choice of density/potential
self-consistency algorithms, while [[wfoptalg]] focuses on the determination
of the wavefunction through the solution of the Schrodinger equation with fixed
Kohn-Sham potential.
In the first class of algorithms, selected by [[iscf]]), Pulay
mixing is one of the most efficient. Also, an efficient preconditioner will
speed up the convergence. Among different choices, a generalized Kerker
preconditioner is implemented, see [[diemac]], [[diemix]] and [[dielng]].
In order to perform a non-self-consistent calculations of wavefunctions and
corresponding eigenvalues in a fixed potential, as for representing a full
band structure, the loop over density/potentials self-consistency must be
disabled, for which [[iscf]]=-2 must be chosen.
Among the algorithms to find the wavefunctions, selected by [[wfoptalg]], the
conjugate-gradient and the LOBPCG ones are the favourite.
The RMM-DIIS algorithm is faster, but might be unstable. Use it for molecular dynamics run
or long geometry optimizations. Use the Chebyshev filtering for massive parallel runs.
Inner electronic eigenvalues can be computed thanks to the minimisation of the
residual with respect to a target energy value, see [[eshift]].
## Related Input Variables
{{ related_variables }}
## Selected Input Files
{{ selected_input_files }}