# ITERATIVE_SOLVER

iterative_solver section [Edit on GitHub]

## Keyword descriptions

CONVERGE_BY_ENERGY: integer = 1

Defines the flavour of the iterative solver. [Edit on GitHub]

EARLY_RESTART: real = 5.00000000E-001

Restart early when the ratio unconverged vs lockable vectors drops below this threshold. [Edit on GitHub]

EMPTY_STATES_TOLERANCE: real = 0.00000000E+000

Additional tolerance for empty states. [Edit on GitHub]

ENERGY_TOLERANCE: real = 1.00000000E-002

Tolerance for the eigen-energy difference \f$|\epsilon_i^{old} - \epsilon_i^{new} | \f$ [Edit on GitHub]

EXTRA_ORTHO: logical = F

Lone keyword: T

Orthogonalize the new subspace basis functions one more time in order to improve the numerical stability. [Edit on GitHub]

INIT_EVAL_OLD: logical = T

Lone keyword: T

Initialize eigen-values with previous (old) values. [Edit on GitHub]

INIT_SUBSPACE: enum = LCAO

Valid values:

• LCAO

• RANDOM

Tell how to initialize the subspace. [Edit on GitHub]

LOCKING: logical = T

Lone keyword: T

Lock eigenvectors of the smallest eigenvalues when they have converged at restart. [Edit on GitHub]

MIN_NUM_RES: integer = 0

Minimum number of residuals to continue iterative diagonalization process. [Edit on GitHub]

NUM_SINGULAR: integer = -1

Number of singular components for the LAPW Davidson solver. [Edit on GitHub]

NUM_STEPS: integer = 20

Number of steps (iterations) of the solver. [Edit on GitHub]

RELATIVE_TOLERANCE: real = 0.00000000E+000

Relative tolerance for the residual L2 norm. (0 means this criterion is effectively not used. [Edit on GitHub]

RESIDUAL_TOLERANCE: real = 1.00000000E-006

Tolerance for the residual L2 norm. [Edit on GitHub]

SUBSPACE_SIZE: integer = 2

Size of the variational subspace is this number times the number of bands. [Edit on GitHub]

TYPE: enum = AUTO

Valid values:

• AUTO

• EXACT

• DAVIDSON

Type of the iterative solver. [Edit on GitHub]