def get_z(x, chemical_rules=False):
descriptors = ['X', 'average_ionic_radius']
a, b, c = mend_to_name(x[0], x[1], x[2])
# name = a + b + c
# conglomerate = [get_pymatgen_descriptor(name, d) for d in descriptors]
# means = [np.mean(k) for k in conglomerate]
# stds = [np.std(k) for k in conglomerate]
# ranges = [np.ptp(k) for k in conglomerate]
# z = means + stds + ranges
z = []
for d in descriptors[:1]:
ab_attrs = [getattr(Element(el), d) for el in (a, b)]
c_attrs = get_pymatgen_descriptor(c, d)
for attrs_set in [ab_attrs, c_attrs]:
z.append(np.mean(attrs_set))
z.append(np.ptp(attrs_set))
z.append(np.std(attrs_set))
z += [Element(a).max_oxidation_state, Element(a).min_oxidation_state]
z += [Element(b).max_oxidation_state, Element(b).min_oxidation_state]
# Chemical rules
rx = np.mean(get_pymatgen_descriptor(c, 'average_ionic_radius'))
ra = Element(a).average_ionic_radius
rb = Element(b).average_ionic_radius
gs_dev = abs(1 - (ra + rx) / (2**0.5 * (rb + rx)))
if chemical_rules:
z.append(gs_dev)
return z
def test_descriptor_ionic_radii(self):
ionic_radii = get_pymatgen_descriptor(self.nacl_formula_2, "ionic_radii")
self.assertEqual(ionic_radii, [1.16, 1.67])
with self.assertRaises(ValueError):
get_pymatgen_descriptor(self.nacl_formula_1, "ionic_radii")
ionic_radii = get_pymatgen_descriptor(self.fe2o3_formula_1, "ionic_radii")
self.assertEqual(ionic_radii, [0.785, 0.785, 1.26, 1.26, 1.26])
def test_descriptor_ionic_radii(self):
ionic_radii = get_pymatgen_descriptor(self.nacl_formula_2,
"ionic_radii")
self.assertEqual(ionic_radii, [1.16, 1.67])
with self.assertRaises(ValueError):
get_pymatgen_descriptor(self.nacl_formula_1, "ionic_radii")
ionic_radii = get_pymatgen_descriptor(self.fe2o3_formula_1,
"ionic_radii")
self.assertEqual(ionic_radii, [0.785, 0.785, 1.26, 1.26, 1.26])
def predict(self, structure):
"""
Given a structure, returns back the predicted volume.
Args:
structure (Structure): structure w/unknown volume
Returns:
a float value of the predicted volume
"""
if not structure.is_ordered:
raise ValueError("VolumePredictor requires ordered structures!")
smallest_ratio = None # ratio of observed vs expected bond distance
ionic_mix = min(
np.std(get_pymatgen_descriptor(structure.composition, 'X')) *
self.ionic_factor, 1)
for site in structure:
el1 = site.specie
if el1.atomic_radius:
r1 = el1.average_ionic_radius * ionic_mix + \
el1.atomic_radius * (
1 - ionic_mix) if el1.average_ionic_radius else el1.atomic_radius
neighbors = structure.get_neighbors(
site, el1.atomic_radius + self.cutoff)
for site2, dist in neighbors:
el2 = site2.specie
if el2.atomic_radius:
r2 = el2.average_ionic_radius * ionic_mix + \
el2.atomic_radius * (
1 - ionic_mix) if el2.average_ionic_radius else el2.atomic_radius
expected_dist = float(r1 + r2)
if not smallest_ratio or dist / expected_dist \
< smallest_ratio:
smallest_ratio = dist / expected_dist
else:
warnings.warn("VolumePredictor: no atomic radius data "
"for {}".format(el2))
else:
warnings.warn("VolumePredictor: no atomic radius data for "
"{}".format(el1))
if not smallest_ratio:
raise ValueError("Could not find any bonds in this material!")
volume_factor = (1 / smallest_ratio)**3
return structure.volume * volume_factor
def predict(self, structure):
"""
Given a structure, returns back the predicted volume.
Args:
structure (Structure): structure w/unknown volume
Returns:
a float value of the predicted volume
"""
if not structure.is_ordered:
raise ValueError("VolumePredictor requires ordered structures!")
smallest_ratio = None # ratio of observed vs expected bond distance
ionic_mix = min(np.std(get_pymatgen_descriptor(structure.composition, "X")) * self.ionic_factor, 1)
for site in structure:
el1 = site.specie
if el1.atomic_radius:
r1 = (
el1.average_ionic_radius * ionic_mix + el1.atomic_radius * (1 - ionic_mix)
if el1.average_ionic_radius
else el1.atomic_radius
)
neighbors = structure.get_neighbors(site, el1.atomic_radius + self.cutoff)
for site2, dist in neighbors:
el2 = site2.specie
if el2.atomic_radius:
r2 = (
el2.average_ionic_radius * ionic_mix + el2.atomic_radius * (1 - ionic_mix)
if el2.average_ionic_radius
else el2.atomic_radius
)
expected_dist = float(r1 + r2)
if not smallest_ratio or dist / expected_dist < smallest_ratio:
smallest_ratio = dist / expected_dist
else:
warnings.warn("VolumePredictor: no atomic radius data " "for {}".format(el2))
else:
warnings.warn("VolumePredictor: no atomic radius data for " "{}".format(el1))
if not smallest_ratio:
raise ValueError("Could not find any bonds in this material!")
volume_factor = (1 / smallest_ratio) ** 3
return structure.volume * volume_factor
def add_pmgdescriptor_column(self, descriptor, stat_function, stat_name):
"""
Args:
descriptor: (str) name of descriptor - must match the name in the source library
stat_function: function to approximate the descriptor. For example, numpy.mean, numpy.std, etc.
stat_name: (str) name of stat function to append to new descriptor column name
Returns: dataframe with appended descriptor column
"""
try:
self.df[descriptor + self.separator + stat_name] = self.df[self.formula_colname]. \
map(lambda x: stat_function(get_pymatgen_descriptor(Composition(x), descriptor)))
except ValueError:
self.df.loc[descriptor + self.separator + stat_name] = None
except AttributeError:
raise ValueError('Invalid pymatgen Element attribute!')
return self.df
def add_pmgdescriptor_column(self, descriptor, stat_function, stat_name):
"""
Args:
descriptor: (str) name of descriptor - must match the name in the source library
stat_function: function to approximate the descriptor. For example, numpy.mean, numpy.std, etc.
stat_name: (str) name of stat function to append to new descriptor column name
Returns: dataframe with appended descriptor column
"""
try:
self.df[descriptor + self.separator + stat_name] = self.df[self.formula_colname]. \
map(lambda x: stat_function(get_pymatgen_descriptor(Composition(x), descriptor)))
except ValueError:
self.df.loc[descriptor + self.separator + stat_name] = None
except AttributeError:
raise ValueError('Invalid pymatgen Element attribute!')
return self.df
'Li', 'Be', 'B', 'Na', 'Mg', 'Al', 'Si', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr',
'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Rb', 'Sr', 'Y',
'Zr', 'Nb', 'Mo', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te',
'Cs', 'Ba', 'La', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg',
'Tl', 'Pb', 'Bi'
]
c_names = ['O3', 'O2N', 'ON2', 'N3', 'O2F', 'OFN', 'O2S']
# Atomic
ab_atomic = [Element(name).Z for name in ab_names]
c_atomic = list(range(7))
# Mendeleev number
ab_mend = [int(Element(name).mendeleev_no) for name in ab_names]
c_mend = [
int(np.sum(get_pymatgen_descriptor(anion, 'mendeleev_no')))
for anion in c_names
]
# Mendeleev rank
ab_mendrank = [sorted(ab_mend).index(i) for i in ab_mend]
c_mendrank = [4, 3, 2, 1, 6, 5, 0]
# Corrected and relevant perovskite data
perovskites = pd.read_csv('unc.csv')
# dim = [(0, 51), (0, 51), (0, 6)]
# todo: use categorical anions
dim = [ab_mend, ab_mend, c_names]
# Defining search space with exclusions
def test_descriptor_ionic_radii_from_composition(self):
cscl = Composition({Specie("Cs", 1): 1, Specie("Cl", -1): 1})
ionic_radii = get_pymatgen_descriptor(cscl, "ionic_radii")
self.assertEqual(ionic_radii, [1.81, 1.67])
ionic_radii_2 = get_pymatgen_descriptor("Cs+1Cl-1", "ionic_radii")
self.assertEqual(ionic_radii, ionic_radii_2)
def test_descriptor_ionic_radii_from_composition(self):
cscl = Composition({Specie("Cs", 1): 1, Specie("Cl", -1): 1})
ionic_radii = get_pymatgen_descriptor(cscl, "ionic_radii")
self.assertEqual(ionic_radii, [1.81, 1.67])
ionic_radii_2 = get_pymatgen_descriptor("Cs+1Cl-1", "ionic_radii")
self.assertEqual(ionic_radii, ionic_radii_2)
