PAO
References: Schuett2018, Berghold2002
Polarized Atomic Orbital Method [Edit on GitHub]
Subsections
Keywords
Keyword descriptions
- CG_INIT_STEPS
Type: integer
Default: 2Description: Number of steepest descent steps before starting the conjugate gradients optimization.
- CG_RESET_LIMIT
Type: real
Default: 1.00000000E-001Description: The CG is reseted if the cosine of the angle between the last search direction and the new gradient is larger that the limit.
- CHECK_GRADIENT_TOL
Type: real
Default: -1.00000000E+000Description: Tolerance for check of full analytic gradient against the numeric one. Negative values mean don’t check at all.
- CHECK_UNITARY_TOL
Type: real
Default: -1.00000000E+000Description: Check if rotation matrix is unitary. Negative values mean don’t check at all.
- EPS_PAO
Type: real
Default: 1.00000000E-005Description: Convergence criteria for PAO optimization.
Mentions: ⭐PAO-ML
- EPS_PGF
Type: real
Description: Sets precision for potential and descriptor matrix elements. Overrides DFT/QS/EPS_DEFAULT value.
- IGNORE_CONVERGENCE_FAILURE
Type: logical
Default: F
Lone keyword: T
Usage: IGNORE_CONVERGENCE_FAILURE logical_valueDescription: If true, only a warning is issued if a PAO optimization has not converged. By default, a run is aborted if the required convergence criteria have not been achieved.
- LINPOT_INITGUESS_DELTA
Type: real
Default: 0.00000000E+000Description: Eigenvalue threshold used for calculating initial guess.
- LINPOT_PRECONDITION_DELTA
Type: real
Default: 0.00000000E+000Description: Eigenvalue threshold used for preconditioning.
- LINPOT_REGULARIZATION_DELTA
Type: real
Default: 0.00000000E+000Description: Eigenvalue threshold used for regularization.
Mentions: ⭐PAO-ML
- LINPOT_REGULARIZATION_STRENGTH
Type: real
Default: 0.00000000E+000Description: Strength of regularization on linpot layer.
- MAX_CYCLES
Type: integer
Default: 75Description: Maximum number of PAO line search cycles for a given hamiltonian.
Mentions: ⭐PAO-ML
- MAX_PAO
Type: integer
Default: 1000Description: Maximum number of PAO basis optimization steps.
Mentions: ⭐PAO-ML
- MIXING
Type: real
Default: 5.00000000E-001Description: Mixing fraction of new and old optimizied solutions.
Mentions: ⭐PAO-ML
- NUM_GRADIENT_EPS
Type: real
Default: 1.00000000E-008Description: Step length used for the numeric derivative when checking the gradient.
- NUM_GRADIENT_ORDER
Type: integer
Default: 2Description: Order of the numeric derivative when checking the gradient. Possible values are 2, 4, and 6.
- OPTIMIZER
Type: enum
Default: CGDescription: Optimizer used to find PAO basis.
Valid values:
CGConjugate gradient algorithmBFGSBroyden-Fletcher-Goldfarb-Shanno algorithm
- PARAMETERIZATION
Type: enum
Default: ROTINVDescription: Parametrization of the mapping between the primary and the PAO basis.
Valid values:
ROTINVRotational invariant parametrization (machine learnable)FOCKFock matrix parametrizationGTHParametrization based on GTH pseudo potentialsEXPOriginal matrix exponential parametrizationEQUIVARIANTEquivariant parametrization
- PENALTY_DISTANCE
Type: real
Default: 1.00000000E-001Description: Distance at which approaching eigenvalues are penalized to prevent degeneration.
- PENALTY_STRENGTH
Type: real
Default: 5.00000000E-003Description: Strength of the penalty term which prevents degenerate eigenvalues.
- PRECONDITION
Type: logical
Default: F
Lone keyword: TDescription: Apply a preconditioner to the parametrization.
- PREOPT_DM_FILE
Type: string
Description: Read pre-optimized density matrix from given file.
- REGULARIZATION
Type: real
Default: 0.00000000E+000Description: Strength of regularization term which ensures parameters remain small.
- RESTART_FILE
Type: string
Description: Reads given files as restart for PAO basis