dxtb.config.ConfigSCF

class dxtb.config.ConfigSCF(*, strict=False, method=1, guess=0, maxiter=100, mixer=2, mix_guess=True, damp=0.5, damp_init=0.1, damp_dynamic=False, damp_dynamic_factor=0.99, damp_soft_start=True, damp_generations=5, damp_diagonal_offset=0.01, scf_mode=0, scp_mode=2, x_atol=0.0001, x_atol_max=1e-05, f_atol=0.0001, force_convergence=False, batch_mode=0, fermi_etemp=300.0, fermi_maxiter=200, fermi_thresh=None, fermi_partition=0, device=device(type='cpu'), dtype=torch.float32)

Bases: object

Configuration for the SCF.

All configuration options are represented as integers. String options are converted to integers in the constructor.

The settings for Fermi smearing are stored separately in the ConfigFermi class, which can be accessed via the fermi attribute.

Methods

info	Return a dictionary with the SCF configuration.

Attributes

strict	Strict mode for SCF configuration.
method	Integer code for tight-binding method.
guess	Initial guess for the SCF.
maxiter	Maximum number of SCF iterations.
mixer	Mixing scheme for SCF iterations.
mix_guess	Include the initial guess in the mixing scheme.
damp	Damping factor for the SCF iterations.
damp_init	Initial damping factor for the SCF iterations.
damp_dynamic	Whether to use dynamic damping in the SCF iterations.
damp_dynamic_factor	Damping factor for dynamic damping in the SCF iterations, i.e., when the norm of the error falls below a threshold.
damp_soft_start	If enabled, then simple mixing will be used for the first generations number of steps, otherwise only for the first (in Anderson mixing only).
damp_generations	Number of generations to use during mixing.
damp_diagonal_offset	Offset added to the equation system's diagonal's to prevent a linear dependence during the mixing process.
scf_mode	SCF convergence approach (denoted by backward strategy).
scp_mode	SCF convergence target (self-consistent property).
x_atol	Absolute tolerance for argument (x) in SCF solver.
x_atol_max	Absolute tolerance for max norm (L∞) of the error in the SCF.
f_atol	Absolute tolerance for function value (f(x)) the SCF solver.
force_convergence	Force convergence of the SCF iterations.
batch_mode	Batch mode for the SCF iterations.
fermi_etemp	Electronic temperature for Fermi smearing.
fermi_maxiter	Maximum number of iterations for Fermi smearing.
fermi_thresh	Threshold for Fermi iterations.
fermi_partition	Partitioning scheme for electronic free energy.
device	Device for calculations.
dtype	Data type for calculations.

Parameters:

• strict (bool)

• method (int)

• guess (int)

• maxiter (int)

• mixer (int)

• mix_guess (bool)

• damp (float)

• damp_init (float)

• damp_dynamic (bool)

• damp_dynamic_factor (float)

• damp_soft_start (bool)

• damp_generations (int)

• damp_diagonal_offset (float)

• scf_mode (int)

• scp_mode (int)

• x_atol (float)

• x_atol_max (float)

• f_atol (float)

• force_convergence (bool)

• batch_mode (int)

• fermi_etemp (float)

• fermi_maxiter (int)

• fermi_thresh (dict)

• fermi_partition (int)

• device (device)

• dtype (dtype)

info()

Return a dictionary with the SCF configuration.

Return type:

dict[str, Any]

Returns:

Dictionary with the SCF configuration.

Return type:

dict[str, Any]

batch_mode: int

Batch mode for the SCF iterations. - 0: Single system - 1: Multiple systems with padding - 2: Multiple systems with no padding (conformer ensemble)

damp: float

Damping factor for the SCF iterations.

damp_diagonal_offset: float

Offset added to the equation system’s diagonal’s to prevent a linear dependence during the mixing process. If set to None then rescaling will be disabled.

damp_dynamic: bool

Whether to use dynamic damping in the SCF iterations.

damp_dynamic_factor: float

Damping factor for dynamic damping in the SCF iterations, i.e., when the norm of the error falls below a threshold.

damp_generations: int

Number of generations to use during mixing. Defaults to 5 as suggested by Eyert.

damp_init: float

Initial damping factor for the SCF iterations.

damp_soft_start: bool

If enabled, then simple mixing will be used for the first generations number of steps, otherwise only for the first (in Anderson mixing only).

device: device

Device for calculations.

dtype: dtype

Data type for calculations.

f_atol: float

Absolute tolerance for function value (f(x)) the SCF solver.

fermi_etemp: float

Electronic temperature for Fermi smearing.

fermi_maxiter: int

Maximum number of iterations for Fermi smearing.

fermi_partition: int

Partitioning scheme for electronic free energy.

fermi_thresh: dict

Threshold for Fermi iterations.

force_convergence: bool

Force convergence of the SCF iterations.

guess: int

Initial guess for the SCF.

maxiter: int

Maximum number of SCF iterations.

method: int

Integer code for tight-binding method.

mix_guess: bool

Include the initial guess in the mixing scheme.

mixer: int

Mixing scheme for SCF iterations.

scf_mode: int

SCF convergence approach (denoted by backward strategy).

scp_mode: int

SCF convergence target (self-consistent property).

strict: bool = False

Strict mode for SCF configuration. Always throws errors if True.

x_atol: float

Absolute tolerance for argument (x) in SCF solver.

x_atol_max: float

Absolute tolerance for max norm (L∞) of the error in the SCF.