def get_energy_edges(self):
""" Evaluate what minimum and maximum energies are for wang-landau sampling and split into bins
where no bin has a zero count
"""
energy_store = []
N = 10000
ii = 0
z1 = copy.deepcopy(self)
while ii < N:
z2, r1, r2 = MC.IonMove(z1)
delG = MC.Energy_Difference(z1, z2, r1, r2)
z2.energy += delG
energy_store.append(z2.energy)
z1 = copy.deepcopy(z2)
del (z2)
ii += 1
# Find unique values and then keep reducing bin size until all bins have at least one count
tmp1 = list(set(energy_store))
bins = 80
tmp2 = np.histogram(tmp1, bins=bins)
while 0 in tmp2[0]:
bins -= 1
tmp2 = np.histogram(tmp1, bins=bins)
print("# of bins: ", bins)
self.energy_edges = tmp2[1]
return tmp1
def wang_landau(self):
""" Perform wang_landau simulation
Arguments:
self {class} -- Lattice class
"""
# Possible energies
energy_edges = self.energy_edges
energy_edges = energy_edges[:-1]
energy_store = []
hist = np.zeros(len(energy_edges))
entropy = np.zeros(len(energy_edges))
ln_factor = 1
ii = 1
while ln_factor > 1e-3 and ii < 80000:
# Perform Monte Carlo Move on lattice
z1, r1, r2 = MC.IonMove(self)
delG = MC.Energy_Difference(self, z1, r1, r2)
z1.energy += delG
energy_store.append(z1.energy)
new_edge = np.digitize(z1.energy, energy_edges) - 1
old_edge = np.digitize(self.energy, energy_edges) - 1
P = np.exp(entropy[old_edge] - entropy[new_edge])
if P > np.random.rand():
old_edge = new_edge
self = copy.deepcopy(z1)
entropy[old_edge] += ln_factor
hist[old_edge] += 1
del (z1)
if is_flat(hist):
hist[:] = 0
ln_factor /= 2
print(ln_factor, ii)
ii += 1
print("ln(f): ", ln_factor, "iter #: ", ii)
return hist, entropy, energy_store