def test_phi_at_x(self):
energy = np.array([0.1, 0.2, 0.3, 0.4, 0.5])
coordinates = np.array([-2.0, -1.0, 0.0, 1.0, 2.0])
with patch(
'pyscses.grid.index_of_grid_at_x') as mock_index_of_grid_at_x:
mock_index_of_grid_at_x.side_effect = [0, 2, 3]
self.assertEqual(
phi_at_x(phi=energy, coordinates=coordinates, x=-1.5), 0.1)
self.assertEqual(
phi_at_x(phi=energy, coordinates=coordinates, x=-0.1), 0.3)
self.assertEqual(
phi_at_x(phi=energy, coordinates=coordinates, x=0.6), 0.4)
def create_space_charge_region(self, grid, pos_or_neg_scr, scr_limit):
"""Calculate the space charge region. The space charge region is defined as the region when the electrostatic potential is greater than a predefined limit.
Args:
grid (:obj:`pyscses.Grid`): Grid object - contains properties of the grid including the x coordinates and the volumes. Used to access the x coordinates.
pos_or_neg_scr (str): 'positive' - for a positive space charge potential.
'negative' - for a negative space charge potential.
scr_limit (float): The minimum electrostatic potential that the electrostatic potential must exceed to be included in the space charge region.
Returns:
list: List of x coordinates for sites within the space charge region.
"""
space_charge_region = []
self.phi_on_mobile_defect_grid = [
phi_at_x(self.phi, self.grid.x, x) for x in grid.x
]
x_and_phi = np.column_stack((grid.x, self.phi_on_mobile_defect_grid))
for i in range(len(x_and_phi)):
if pos_or_neg_scr == 'positive':
if x_and_phi[i, 1] - x_and_phi[0, 1] > scr_limit:
space_charge_region.append(x_and_phi[i, 0])
if pos_or_neg_scr == 'negative':
if x_and_phi[i, 1] - x_and_phi[0, 1] < scr_limit:
space_charge_region.append(x_and_phi[i, 0])
return space_charge_region
def calculate_probabilities(self,
grid: Grid,
phi: np.ndarray,
temp: float) -> np.ndarray:
"""
Calculates the probability of a site being occupied by its corresponding defect.
Args:
grid (Grid): Grid object - contains properties of the grid including the x coordinates and the volumes. Used to access the x coordinates.
phi (array): electrostatic potential on a one-dimensional grid.
temp (float): Absolute temperature.
Returns:
array: probabilities of defects occupying each site using their grid points
"""
# TODO: This is Jacob's fixed code, but it is inefficient and slow.
probability = np.zeros_like(grid.x)
for i,j in enumerate(grid.x):
prob = []
for site in self.sites:
if j == site.x:
prob.append(site.probabilities_as_list(phi_at_x(phi, grid.x, site.x), temp))
if len(prob) == 0:
probability[i] = 0
else:
probability[i] = np.mean(prob)
return probability
def solve(self, approximation):
"""
Self-consistent solving of the Poisson-Boltzmann equation. Iterates until the convergence is less than the convergence limit. The outputs are stored as Calculation attributes.
Calculation.phi (array): Electrostatic potential on a one-dimensional grid.
Calculation.rho (array): Charge density on a one-dimensional grid.
Calculation.niter (int): Number of iterations performed to reach convergence.
Args:
approximation (str): Approximation used for the defect behaviour.
'mott-schottky' - Some defects immobile / fixed to bulk mole fractions.
'gouy-chapman' - All defects mobile / able to redistribute.
"""
poisson_solver = MatrixSolver(
self.grid,
self.dielectric,
self.temp,
boundary_conditions=self.boundary_conditions)
phi = np.zeros_like(self.grid.x)
rho = np.zeros_like(self.grid.x)
conv = 1
niter = 0
while conv > self.convergence:
predicted_phi, rho = poisson_solver.solve(phi)
if approximation == 'gouy-chapman':
average_phi = self.calculate_average(self.grid,
self.bulk_x_min,
self.bulk_x_max,
predicted_phi)
predicted_phi -= average_phi
if approximation == 'mott-schottky':
vo_predicted_phi = [
phi_at_x(predicted_phi, self.grid.x, x)
for x in self.grid.subgrid(self.site_labels[0]).x
]
average_vo_predicted_phi = self.calculate_average(
self.grid.subgrid(self.site_labels[0]), self.bulk_x_min,
self.bulk_x_max, vo_predicted_phi)
predicted_phi -= average_vo_predicted_phi
phi = self.alpha * predicted_phi + (1.0 - self.alpha) * phi
conv = sum((predicted_phi - phi)**2) / len(self.grid.x)
prob = self.grid.set_of_sites.calculate_probabilities(
self.grid, phi, self.temp)
niter += 1
# if niter % 500 == 0.0:
# if niter == 1:
# print(conv)
# print(phi, rho)
# stop
self.phi = phi
self.rho = self.grid.rho(phi, self.temp)
self.niter = niter
def calculate_defect_density( self, grid, phi, temp ):
"""
Calculates the defect density at each site.
Args:
grid (object): Grid object - contains properties of the grid including the x coordinates and the volumes. Used to access the x coordinates.
phi (array): electrostatic potential on a one-dimensional grid.
temp (float): Absolute temperature.
Returns:
array: defect density for each site using their grid points
"""
defect_density = np.zeros_like( grid.x )
for site in self.sites:
i = index_of_grid_at_x( grid.x, site.x )
defect_density[ i ] += np.asarray( site.probabilities( phi_at_x( phi, grid.x, site.x ), temp ) ) / grid.volumes[ i ]
return defect_density
def calculate_probabilities(self, grid, phi, temp):
"""
Calculates the probability of a site being occupied by its corresponding defect.
Args:
grid (object): Grid object - contains properties of the grid including the x coordinates and the volumes. Used to access the x coordinates.
phi (array): electrostatic potential on a one-dimensional grid.
temp (float): Absolute temperature.
Returns:
probability (array): probabilities of defects occupying each site using their grid points
"""
probability = np.zeros_like(grid.x)
for site in self.sites:
probability[index_of_grid_at_x(grid.x, site.x)] = np.asarray(
site.probabilities(phi_at_x(phi, grid.x, site.x), temp))
return probability
def subgrid_calculate_defect_density(self,
sub_grid: Grid,
full_grid: Grid,
phi: np.ndarray,
temp: float) -> np.ndarray:
"""
Calculates the defect density at each site for a given subset of sites.
Args:
subgrid (Grid): Grid object - contains properties of the grid including the x coordinates and the volumes. Used to access the x coordinates. For a given subset of sites.
full_grid (Grid): Grid object - contains properties of the grid including the x coordinates and the volumes. Used to access the x coordinates. For all sites.
phi (array): electrostatic potential on a one-dimensional grid.
temp (float): Absolute temperature.
Returns:
array: defect density for each site using their grid points
"""
defect_density = np.zeros_like( sub_grid.x )
for site in self.sites:
i = index_of_grid_at_x( sub_grid.x, site.x )
defect_density[ i ] += np.asarray( site.probabilities_as_list( phi_at_x( phi, full_grid.x, site.x ), temp ) ) / sub_grid.volumes[ i ]
return defect_density
def solve(self, approximation: str, verbose: bool = False) -> None:
"""
Self-consistent solving of the Poisson-Boltzmann equation. Iterates until the convergence is less than the convergence limit. The outputs are stored as Calculation attributes.
Calculation.phi (array): Electrostatic potential on a one-dimensional grid.
Calculation.rho (array): Charge density on a one-dimensional grid.
Calculation.niter (int): Number of iterations performed to reach convergence.
Args:
approximation (str): Approximation used for the defect behaviour.
'mott-schottky' - Some defects immobile / fixed to bulk mole fractions.
'gouy-chapman' - All defects mobile / able to redistribute.
verbose (optional, bool): Verbose output. Default is False.
"""
poisson_solver = MatrixSolver(
grid=self.grid,
dielectric=self.dielectric,
temp=self.temp,
boundary_conditions=self.boundary_conditions)
phi = np.zeros_like(self.grid.x)
rho = np.zeros_like(self.grid.x)
conv = 1.0
niter = 0
while conv > self.convergence:
predicted_phi, rho = poisson_solver.solve(phi)
if approximation == 'gouy-chapman':
average_phi = self.calculate_average(
grid=self.grid,
min_cutoff=self.bulk_x_min,
max_cutoff=self.bulk_x_max,
sc_property=predicted_phi)
predicted_phi -= average_phi
elif approximation == 'mott-schottky':
subgrid = self.grid.subgrid(self.site_labels[0])
predicted_phi_subgrid = np.array([
phi_at_x(phi=predicted_phi, coordinates=self.grid.x, x=x)
for x in subgrid.x
])
average_predicted_phi = self.calculate_average(
grid=subgrid,
min_cutoff=self.bulk_x_min,
max_cutoff=self.bulk_x_max,
sc_property=predicted_phi_subgrid)
predicted_phi -= average_predicted_phi
# TODO: This is inefficient. Jacob has some ideas for how to improve things.
# TODO: This definition of convergence should not be averaged over all sites.
phi = self.alpha * predicted_phi + (1.0 - self.alpha) * phi
conv = sum((predicted_phi - phi)**2) / len(self.grid.x)
# prob = self.grid.set_of_sites.calculate_probabilities( self.grid, phi, self.temp) # Jacob: Does this do anything?
niter += 1
if verbose:
if niter % 500 == 0:
print(
f'Iteration: {niter} -> Convergence: {conv} / {self.convergence}'
)
if verbose:
print(
f'Converged at iteration {niter} -> Convergence: {conv} / {self.convergence}'
)
self.phi = phi
self.rho = self.grid.rho(phi, self.temp)
self.niter = niter