from icet import ClusterSpace
from icet.core.sublattices import Sublattices
import numpy as np
from scipy.integrate import cumulative_trapezoid
from ase import Atoms
from ase.units import kB
from mchammer import DataContainer
import operator
from collections import Counter
from math import factorial
from functools import reduce
def _lambda_function_forward(n_steps: int, step: int) -> float:
"""
Returns the current lambda value for a backward calculation.
Parameters
----------
n_steps
Total number of steps..
step
Current step.
"""
x = (step + 1) / (n_steps)
lam = x**5*(70*x**4 - 315*x**3 + 540*x**2 - 420*x + 126)
# Due to numerical precision lambda may be slightly outside the interval [0.0, 1.0]
lam = np.clip(lam, 0.0, 1.0)
return lam
def _lambda_function_backward(n_steps: int, step: int) -> float:
"""
Returns the current lambda value for a backward calculation.
Parameters
----------
n_steps
Total number of steps.
step
Current step.
"""
return 1 - _lambda_function_forward(n_steps, step)
def _stirling(n: int) -> float:
""" Stirling approximation of log(n!). """
return n * np.log(n) - n + 0.5 * np.log(2*np.pi*n)
def _lognpermutations(array: list[int]) -> float:
"""If _npermutations becomes too large we resort to calculating the
logarithm directly using Stirling's approximation, i.e., we
calculate
.. math::
log(n!) - (log(a1!) + log(a2!) + log(a3!) + ... + log(ak!)
where we denote in the code
.. math::
v n = log(n!)
den = (log(a1!) + log(a2!) + log(a3!) + ... + log(ak!)
"""
n = _stirling(len(array))
# Gets the number of each unique atom type
mults = Counter(array).values()
den = reduce(operator.add,
(_stirling(v) for v in mults),
1)
return n - den
def _npermutations(array: list[int]) -> float:
"""
Returns the total number of ways we can permutate the array which is given by
.. math::
n! / (a1! * a2! * a3! * ... * ak!)
where we denote in the code
.. math::
den = (a1! * a2! * a3! * ... * ak!)
"""
n = factorial(len(array))
# Gets the number of each unique atom type
a = Counter(array).values()
den = reduce(operator.mul,
(factorial(v) for v in a),
1)
return n / den
def _ideal_mixing_entropy(swap_sublattice_probabilities: list[int],
atoms_on_sublattices: np.ndarray,
boltzmann_constant: float) -> float:
"""
Calculates the ideal mixing entropy for the supercell.
Parameters
----------
swap_sublattice_probabilites
Sublattices that are active during the MC simulation.
atoms_on_sublattices
Sorted atomic numbers in sublattices lists.
boltzmann_constant
Value of Boltzmann constant in the units used in the MC simulation.
"""
log_multiplicity = []
sublattices = np.array(swap_sublattice_probabilities) > 0
for aos in atoms_on_sublattices[sublattices]:
try:
log_multiplicity.append(np.log(_npermutations(aos)))
except Exception:
log_multiplicity.append(_lognpermutations(aos))
return boltzmann_constant * np.sum(log_multiplicity)
def _get_atoms_on_sublattice(structure: Atoms,
sublattices: Sublattices) -> np.ndarray:
"""Sorts the atom symbols in the structure to lists involving
specific sublattices, for example,
[[5, 5, 3, 3, 3], [11, 11, 10, 10, 10 10, 10]]
Parameters
----------
structure
Atomic structure.
sublattices
Sublattices for the supercell obtained from the cluster space.
"""
_atoms_on_sublattices = []
for sl in sublattices:
atoms_on_sublattice = []
for an in sl.atomic_numbers:
natoms = np.where(structure.numbers == an)[0].size
atoms_on_sublattice.extend([an] * natoms)
_atoms_on_sublattices.append(atoms_on_sublattice)
_atoms_on_sublattices = np.array(_atoms_on_sublattices,
dtype=object)
return _atoms_on_sublattices
[docs]
def get_free_energy_thermodynamic_integration(
dc: DataContainer,
cluster_space: ClusterSpace,
forward: bool,
max_temperature: float = np.inf,
sublattice_probabilities: list[float] = None,
boltzmann_constant: float = kB,
) -> (np.ndarray, np.ndarray):
r"""Returns the free energy calculated via thermodynamic integration
using the :class:`ThermodynamicIntegrationEnsemble
<mchammer.ensembles.ThermodynamicIntegrationEnsemble>`.
The temperature dependence of the free energy can be extracted
from the thermodynamic integration as
.. math::
F(T) = \frac{F_0(\lambda)}{\lambda} + \frac{T_0}{\lambda} S_\text{B}
where :math:`S_\text{B}` is the Boltzmann entropy,
.. math::
T = \frac{T_0}{\lambda}
and
.. math::
F_0(\lambda) = \int_0^\lambda \left\langle\frac{\mathrm{d}H(\lambda)}
{\mathrm{d}\lambda}\right\rangle_{H} \mathrm{d}\lambda
Parameters
----------
dc
Data container from the thermodynamic integration simulation.
cluster_space
Cluster space used to construct the cluster expansion used for the simulation.
forward
Whether or not the thermodynamic integration was carried out forward or backward.
max_temperature
Largest temperature to extract from the thermodynamic integration.
sublattice_probababilites
Sublattice probabilties that were provided to the thermodynamic integration
simulation.
boltzmann_constant
Boltzmann constant in the units used for the thermodynamic integration
simulation.
"""
lambdas, potentials = dc.get('lambda', 'potential')
if not forward:
# We want to integrate from high to low temperature even though
# the simulation was from low to high.
potentials = potentials[::-1]
lambdas = lambdas[::-1]
sublattices = cluster_space.get_sublattices(dc.structure)
if sublattice_probabilities is None:
sublattice_probabilities = [True] * len(sublattices)
atoms_on_sublattices = _get_atoms_on_sublattice(dc.structure,
sublattices)
ideal_mixing_entropy = _ideal_mixing_entropy(sublattice_probabilities,
atoms_on_sublattices,
boltzmann_constant)
temperature = dc._ensemble_parameters['temperature']
with np.errstate(divide='ignore'):
# First value of lambdas will be zero (we ignore this warning)
temperatures = (1 / lambdas) * temperature
temp_inds = np.where(temperatures <= max_temperature)[0]
free_energy_change = cumulative_trapezoid(potentials, x=lambdas, initial=0)
with np.errstate(invalid='ignore', divide='ignore'):
# First value of lambdas will be zero (we ignore this warning)
free_energy = (free_energy_change / lambdas -
temperatures * ideal_mixing_entropy)
free_energy[np.isnan(free_energy)] = 0
return (temperatures[temp_inds], free_energy[temp_inds])
[docs]
def get_free_energy_temperature_integration(
dc: DataContainer,
cluster_space: ClusterSpace,
forward: bool,
temperature_reference: float,
free_energy_reference: float = None,
sublattice_probabilities: list[float] = None,
max_temperature: float = np.inf,
boltzmann_constant: float = kB,
) -> (np.ndarray, np.ndarray):
r""" Returns the free energy calculated using temperature integration and
the corresponding temperature.
.. math::
\frac{A(T_{2})}{T_{2}} = \frac{A(T_{1})}{T_{1}}
- \int_{T_{1}}^{T_{2}}\frac{U(T)}{T^2}\mathrm{d}T
Parameters
----------
dc
Data container from a canonical annealing simulation.
The first (last for :attr:`forward=False`) temperature in the
data container has to be at the same temperature as
:attr:`temperature_reference`.
cluster_space
Cluster space used to construct the cluster expansion
that was used in the simulations.
forward
If ``True`` the canonical annealing simulation was carried out
from high to low temperature, otherwise the opposite is assumed.
temperature_reference
Temperature at which :attr:`free_energy_reference` was calculated
free_energy_reference
Reference free energy. If set to ``None`` it will be assumeed that the
free energy at :attr:`temperature_reference` can be approximated by
:math:`T S_B` where :math:`S_B` is the ideal mixing entropy.
sublattice_probababilites
Sublattice probabilties that were provided to the canonical annealing
simulation.
max_temperature
Largest temperature to extract from the temperature integration.
boltzmann_constant
Boltzmann constant in the units used for the thermodynamic integration
simulation.
"""
temperatures, potentials = dc.get('temperature', 'potential')
if not forward:
# We want to integrate from high to low temperature even though
# the simulation was from low to high.
potentials = potentials[::-1]
temperatures = temperatures[::-1]
if not np.isclose(temperatures[0], temperature_reference):
raise ValueError(
'The first or the last temperature in the temperature list'
' has to equal the reference temperature.'
f' reference_temperature: {temperature_reference}'
f' first canonical temperature: {temperatures[0]}')
temp_inds = np.where(temperatures <= max_temperature)[0]
if free_energy_reference is None:
sublattices = cluster_space.get_sublattices(dc.structure)
if sublattice_probabilities is None:
sublattice_probabilities = [True] * len(sublattices)
atoms_on_sublattices = _get_atoms_on_sublattice(dc.structure,
sublattices)
ideal_mixing_entropy = _ideal_mixing_entropy(sublattice_probabilities,
atoms_on_sublattices,
boltzmann_constant)
free_energy_reference = -ideal_mixing_entropy * temperature_reference
reference = free_energy_reference / temperature_reference
integral = cumulative_trapezoid(potentials / temperatures**2,
x=temperatures, initial=0)
free_energy = (reference - integral) * temperatures
return (temperatures[temp_inds], free_energy[temp_inds])
