"""Definition of the canonical annealing class."""
import numpy as np
from ase import Atoms
from ase.units import kB
from .. import DataContainer
from ..calculators import ClusterExpansionCalculator
from .thermodynamic_base_ensemble import ThermodynamicBaseEnsemble
[docs]
class CanonicalAnnealing(ThermodynamicBaseEnsemble):
"""Instances of this class allow one to carry out simulated annealing in the
canonical ensemble, i.e. the temperature is varied in pre-defined fashion
while the composition is kept fixed.
See :class:`CanonicalEnsemble <mchammer.ensembles.CanonicalEnsemble>` for
more information about the standard canonical ensemble.
The canonical annealing ensemble can be useful, for example, for
finding ground states or generating low energy configurations.
The temperature control scheme is selected via the :attr:`cooling_function`
keyword argument, while the initial and final temperature are set via the
:attr:`T_start` and :attr:`T_stop` arguments.
Several pre-defined temperature control schemes are available including
``'linear'`` and ``'exponential'``.
In the latter case the temperature varies logarithmically as a function
of the MC step, emulating the exponential temperature dependence of the
atomic exchange rate encountered in many materials.
It is also possible to provide a user defined cooling function via the
keyword argument.
This function must comply with the following function header::
def cooling_function(step, T_start, T_stop, n_steps):
T = ... # compute temperature
return T
Here :attr:`step` refers to the current MC trial step.
Parameters
----------
structure
Atomic configuration to be used in the Monte Carlo simulation.
It also defines the initial occupation vector.
calculator
Calculator to be used for calculating the potential changes
that enter the evaluation of the Metropolis criterion.
T_start
Temperature from which the annealing is started.
T_stop
Final temperature for annealing.
n_steps
Number of steps to take in the annealing simulation.
cooling_function
to use the predefined cooling functions provide a string
``'linear'`` or ``'exponential'``, otherwise provide a function.
boltzmann_constant
Boltzmann constant :math:`k_B` in appropriate units, i.e., units that
are consistent with the underlying cluster expansion and the
temperature units.
By default eV/K.
user_tag
Human-readable tag for the ensemble.
random_seed
Seed for the random number generator used in the Monte Carlo simulation.
dc_filename
Name of file the data container associated with the ensemble will be
written to.
If the file exists it will be read, the data container will be
appended, and the file will be updated/overwritten.
data_container_write_period
Period in seconds at which the data container is written to file.
Writing periodically to file provides both a way to examine the
progress of the simulation and to back up the data.
By default 600 s.
ensemble_data_write_interval
interval at which data is written to the data container.
This includes for example the current value of the calculator (i.e.,
usually the energy) as well as ensembles specific fields such as
temperature or the number of atoms of different species.
By default the number of sites in :attr:`structure`.
trajectory_write_interval
interval at which the current occupation vector of the atomic
configuration is written to the data container.
By default the number of sites in :attr:`structure`.
sublattice_probabilities
Probability for picking a sublattice when doing a random swap.
This should be as long as the number of sublattices and should
sum up to 1.
neighbor_sites_to_avoid
Sites that must not be occupied simultaneously, keyed by site index.
A site counts as occupied when it is not held by a vacancy, and sites
that are absent from the mapping are unconstrained.
The mapping must be symmetric and the initial configuration must
already satisfy it.
The constraint is enforced by rejecting trial moves that would violate
it, which keeps the acceptance criterion exact.
A restrictive mapping can nevertheless leave the allowed configurations
disconnected under the available trial moves, in which case only the
reachable part of them is sampled.
The mapping itself is not written to the data container, only a digest
of it, so a restart has to be given the same mapping again.
"""
def __init__(self,
structure: Atoms,
calculator: ClusterExpansionCalculator,
T_start: float,
T_stop: float,
n_steps: int,
cooling_function: str = 'exponential',
user_tag: str | None = None,
boltzmann_constant: float = kB,
random_seed: int | None = None,
dc_filename: str | None = None,
data_container_write_period: float = 600,
ensemble_data_write_interval: int | None = None,
trajectory_write_interval: int | None = None,
sublattice_probabilities: list[float] | None = None,
neighbor_sites_to_avoid: dict[int, list[int]] | None = None) -> None:
self._ensemble_parameters = dict(n_steps=n_steps)
# add species count to ensemble parameters
for sl in calculator.sublattices:
for symbol in sl.chemical_symbols:
key = 'n_atoms_{}'.format(symbol)
count = structure.get_chemical_symbols().count(symbol)
self._ensemble_parameters[key] = count
super().__init__(
structure=structure, calculator=calculator, user_tag=user_tag,
random_seed=random_seed,
dc_filename=dc_filename,
data_container_class=DataContainer,
data_container_write_period=data_container_write_period,
ensemble_data_write_interval=ensemble_data_write_interval,
trajectory_write_interval=trajectory_write_interval,
boltzmann_constant=boltzmann_constant,
neighbor_sites_to_avoid=neighbor_sites_to_avoid)
self._temperature = T_start
self._T_start = T_start
self._T_stop = T_stop
self._n_steps = n_steps
self._ground_state_candidate = self.configuration.structure
self._ground_state_candidate_potential = calculator.calculate_total(
occupations=self.configuration.occupations)
# setup cooling function
if isinstance(cooling_function, str):
available = sorted(available_cooling_functions.keys())
if cooling_function not in available:
raise ValueError(
'Select from the available cooling_functions {}'.format(available))
self._cooling_function = available_cooling_functions[cooling_function]
elif callable(cooling_function):
self._cooling_function = cooling_function
else:
raise TypeError('cooling_function must be either str or a function')
if sublattice_probabilities is None:
self._swap_sublattice_probabilities = self._get_swap_sublattice_probabilities()
else:
self._swap_sublattice_probabilities = sublattice_probabilities
@property
def temperature(self) -> float:
""" Current temperature. """
return self._temperature
@property
def T_start(self) -> float:
""" Starting temperature. """
return self._T_start
@property
def T_stop(self) -> float:
""" Final temperature. """
return self._T_stop
@property
def n_steps(self) -> int:
""" Number of steps to carry out. """
return self._n_steps
@property
def estimated_ground_state(self) -> Atoms:
""" Structure with lowest observed potential during run. """
return self._ground_state_candidate.copy()
@property
def estimated_ground_state_potential(self) -> float:
""" Lowest observed potential during run. """
return self._ground_state_candidate_potential
[docs]
def run(self) -> None:
""" Runs the annealing simulation. """
if self.step >= self.n_steps:
raise Exception('Annealing has already finished')
super().run(self.n_steps - self.step)
def _do_trial_step(self) -> int:
""" Carries out one Monte Carlo trial step. """
self._temperature = self._cooling_function(
self.step, self.T_start, self.T_stop, self.n_steps)
sublattice_index = self.get_random_sublattice_index(self._swap_sublattice_probabilities)
return self.do_canonical_swap(sublattice_index=sublattice_index)
def _get_ensemble_data(self) -> dict:
"""Returns the data associated with the ensemble.
For the CanonicalAnnealing this specifically includes the temperature.
"""
data = super()._get_ensemble_data()
data['temperature'] = self.temperature
if data['potential'] < self._ground_state_candidate_potential:
self._ground_state_candidate_potential = data['potential']
self._ground_state_candidate = self.configuration.structure
return data
def _cooling_linear(step, T_start, T_stop, n_steps):
return T_start + (T_stop-T_start) * (step + 1) / n_steps
def _cooling_exponential(step, T_start, T_stop, n_steps):
return T_start - (T_start - T_stop) * np.log(step+1) / np.log(n_steps)
available_cooling_functions = dict(linear=_cooling_linear, exponential=_cooling_exponential)
