def get_features(self, formula):
'''
Parameters
----------
formula: string
put a valid chemical fomula as a string. Example( 'NaCl')
Return
----------
features: np.array()
This is an 1xN length array containing feature values for use in the
machine learning model.
'''
try:
fractional_composition = mg.Composition(formula).fractional_composition.as_dict()
element_composition = mg.Composition(formula).element_composition.as_dict()
avg_feature = np.zeros(len(self.element_df.iloc[0]))
sum_feature = np.zeros(len(self.element_df.iloc[0]))
for key in fractional_composition:
try:
avg_feature += self.element_df.loc[key].values * fractional_composition[key]
diff_feature = self.element_df.loc[list(fractional_composition.keys())].max()-self.element_df.loc[list(fractional_composition.keys())].min()
except:
print('The element:', key, 'from formula', formula,'is not currently supported in our database')
return np.array([np.nan]*len(self.element_df.iloc[0])*4)
max_feature = self.element_df.loc[list(fractional_composition.keys())].max()
min_feature = self.element_df.loc[list(fractional_composition.keys())].min()
features = pd.DataFrame(np.concatenate([avg_feature, diff_feature, np.array(max_feature), np.array(min_feature)]))
features = np.concatenate([avg_feature, diff_feature, np.array(max_feature), np.array(min_feature)])
return features.transpose()
except:
print('There was an error with the Formula: '+ formula + ', this is a general exception with an unkown error')
return [np.nan]*len(self.element_df.iloc[0])*4
def get_ABE(self, formula, A_site, B_site, verbose=False):
"""
Estimate average metal-oxygen bond energy for complex perovskite oxide from simple oxide thermo data
Formula from Sammells et al. (1992), Solid State Ionics 52, 111-123.
Parameters:
-----------
formula: oxide formula
A_site: list of A-site elements
B_site: list of B-site elements
verbose: if True, print info about which simple oxides used in calculation
"""
#validated on compounds in Sammells 1992 - all but CaTi0.7Al0.3O3 agree
#validated on (La,Sr)(Cr,Co,Fe)O3 compounds in https://pubs.acs.org/doi/suppl/10.1021/acs.jpcc.6b10571/suppl_file/jp6b10571_si_001.pdf
#works if Co3O4 specified in oxide_dict
comp = mg.Composition(formula)
cd = comp.get_el_amt_dict()
metals = [x for x in cd.keys() if x != 'O']
abe = 0
if verbose == True:
print('Oxides used for ABE calculation:')
for metal in metals:
amt = cd[metal]
met_mg = mg.Element(metal)
try: #oxide_dict specifies which oxide to use
oxide = self.oxide_dict[metal]
oxide_mg = mg.Composition(oxide)
m = oxide_mg.get(metal)
n = oxide_mg.get('O')
obe = self.oxide_obe(oxide)
except KeyError: #if no oxide indicated in oxide_dict
"placeholder - for now, take the lowest common oxidation state with a corresponding stable oxide"
i = 0
while i != -1:
ox = met_mg.common_oxidation_states[i]
oxide, m, n = self.oxide_formula(metal, ox, return_mn=True)
try:
obe = self.oxide_obe(oxide)
#print(obe)
i = -1
except LookupError as err:
i += 1 #try the next oxidation state
if verbose == True:
print(oxide)
#print('m: {}, n: {}'.format(m,n))
if metal in A_site:
abe += amt * obe / (12 * m)
elif metal in B_site:
abe += amt * obe / (6 * m)
else:
raise KeyError(
'{} is not assigned to A or B site'.format(metal))
#print(abe)
return abe
def get_features(self, formula):
try:
fractional_composition = mg.Composition(
formula).fractional_composition.as_dict() #显示化学式归一成分
element_composition = mg.Composition(
formula).element_composition.as_dict() #显示化学式成分
avg_feature = np.zeros(len(self.element_df.iloc[0]))
std_feature = np.zeros(len(self.element_df.iloc[0]))
for key in fractional_composition:
try:
avg_feature += self.element_df.loc[
key].values * fractional_composition[key]
#element_df.loc[key].values为元素表中相应属性值，
#fractional_composition[key]为化学成分中元素所对应的成分
#element_df.loc[key].values * fractional_composition[key]=原子属性值在化学式式中所占有的比例值
diff_feature = self.element_df.loc[list(
fractional_composition.keys())].max(
) - self.element_df.loc[list(
fractional_composition.keys())].min()
#找出化学式中每种原子的每种属性的最大值和最小值，然后相减
except Exception as e:
print('The element:', key, 'from formula', formula,
'is not currently supported in our database')
return np.array([np.nan] * len(self.element_df.iloc[0]) *
5)
max_feature = self.element_df.loc[list(
fractional_composition.keys())].max()
min_feature = self.element_df.loc[list(
fractional_composition.keys())].min()
std_feature = self.element_df.loc[list(
fractional_composition.keys())].std(ddof=0)
# 把相关的信息拼接成
features = pd.DataFrame(
np.concatenate([
avg_feature, diff_feature,
np.array(max_feature),
np.array(min_feature),
np.array(std_feature)
]))
features = np.concatenate([
avg_feature, diff_feature,
np.array(max_feature),
np.array(min_feature),
np.array(std_feature)
])
return features.transpose()
except:
print('There was an error with the Formula: ' + formula +
', this is a general exception with an unkown error')
return [np.nan] * len(self.element_df.iloc[0]) * 5
def run(mpfile, **kwargs):
import pymatgen
import pandas as pd
from mpcontribs.users.swf.rest.rester import SwfRester
# load data from google sheet
google_sheet = mpfile.document[mp_level01_titles[0]].pop("google_sheet")
google_sheet += "/export?format=xlsx"
df_dct = pd.read_excel(google_sheet, sheet_name=None)
# rename sheet columns
elements = ["Fe", "V", "Co"]
df_dct["IP Energy Product"].columns = ["IP_Energy_product"] + elements
df_dct["total"].columns = elements
df_dct["MOKE"].columns = elements + ["thickness", "MOKE_IP_Hc"]
df_dct["VSM"].columns = elements + ["thickness", "VSM_IP_Hc"]
df_dct["formula"].columns = elements
df_dct["Kondorsky"].columns = ["angle", "Kondorsky_Model", "Experiment"]
# round all compositions to 100%
for sheet, df in df_dct.items():
if sheet != "Kondorsky":
for idx, row in df.iterrows():
df.loc[idx:idx, elements] = round_to_100_percent(row[elements])
row5 = df_dct["formula"].iloc[0]
formula5 = get_composition_from_string(
pymatgen.Composition(10 * row5).formula.replace(" ", ""))
dct = dict((k, clean_value(v, "%")) for k, v in row5.to_dict().items())
mpfile.add_hierarchical_data({"data": dct}, identifier=formula5)
mpfile.add_data_table(formula5,
df_dct["Kondorsky"],
name="Angular Dependence of Switching Field")
for sheet, df in df_dct.items():
if sheet == "formula" or sheet == "Kondorsky" or sheet == "total":
continue
for idx, row in df.iterrows():
composition = pymatgen.Composition(row[elements] * 10)
formula = get_composition_from_string(
composition.formula.replace(" ", ""))
dct = dict((k, clean_value(v, "%"))
for k, v in row[elements].to_dict().items())
mpfile.add_hierarchical_data({"data": dct}, identifier=formula)
columns = [x for x in row.index if x not in elements]
if columns:
data = row[columns].round(decimals=1)
dct = dict(
(k, clean_value(v)) for k, v in data.to_dict().items())
mpfile.add_hierarchical_data({"data": dct}, identifier=formula)
def run(mpfile, **kwargs):
import pymatgen
import pandas as pd
from mpcontribs.users.swf.rest.rester import SwfRester
# load data from google sheet
google_sheet = mpfile.document[mp_level01_titles[0]].pop('google_sheet')
google_sheet += '/export?format=xlsx'
df_dct = pd.read_excel(google_sheet, sheet_name=None)
# rename sheet columns
elements = ['Fe', 'V', 'Co']
df_dct['IP Energy Product'].columns = ['IP_Energy_product'] + elements
df_dct['total'].columns = elements
df_dct['MOKE'].columns = elements + ['thickness', 'MOKE_IP_Hc']
df_dct['VSM'].columns = elements + ['thickness', 'VSM_IP_Hc']
df_dct['formula'].columns = elements
df_dct['Kondorsky'].columns = ['angle', 'Kondorsky_Model', 'Experiment']
# round all compositions to 100%
for sheet, df in df_dct.items():
if sheet != 'Kondorsky':
for idx, row in df.iterrows():
df.loc[idx:idx, elements] = round_to_100_percent(row[elements])
row5 = df_dct['formula'].iloc[0]
formula5 = get_composition_from_string(
pymatgen.Composition(10 * row5).formula.replace(' ', ''))
dct = dict((k, clean_value(v, '%')) for k, v in row5.to_dict().items())
mpfile.add_hierarchical_data({'data': dct}, identifier=formula5)
mpfile.add_data_table(formula5,
df_dct['Kondorsky'],
name='Angular Dependence of Switching Field')
for sheet, df in df_dct.items():
if sheet == 'formula' or sheet == 'Kondorsky' or sheet == 'total':
continue
for idx, row in df.iterrows():
composition = pymatgen.Composition(row[elements] * 10)
formula = get_composition_from_string(
composition.formula.replace(' ', ''))
dct = dict((k, clean_value(v, '%'))
for k, v in row[elements].to_dict().items())
mpfile.add_hierarchical_data({'data': dct}, identifier=formula)
columns = [x for x in row.index if x not in elements]
if columns:
data = row[columns].round(decimals=1)
dct = dict(
(k, clean_value(v)) for k, v in data.to_dict().items())
mpfile.add_hierarchical_data({'data': dct}, identifier=formula)
def get_class(self, formula):
output = ''
try:
dc = pg.Composition(formula, strict=False).as_dict().keys()
except Exception as ce:
print("Exception when parsing " + str(formula) + ". Error: " +
str(ce))
# Trying with some tricks
c_with_replacements = re.sub(r'[+-][ZXYzxy]', '', formula)
try:
print("Trying to parse " + str(c_with_replacements))
dc = pg.Composition(c_with_replacements,
strict=False).as_dict().keys()
except Exception as ce:
print("Exception when parsing " + str(c_with_replacements) +
". Error: " + str(ce))
# We give up... skipping this record
return output
input_formula = list(dc)
# print(" Input Formula: " + str(input_formula))
for composition in self.composition_map:
and_compounds = []
if 'and_compounds' in composition:
and_compounds = composition['and_compounds']
or_compounds = []
if 'or_compounds' in composition:
or_compounds = composition['or_compounds']
output_class = composition['name']
if len(and_compounds) > 0:
if all(elem in input_formula for elem in and_compounds):
output = output_class
break
elif len(or_compounds) > 0:
if any(elem in input_formula for elem in or_compounds):
output = output_class
break
if output == '':
output = "Alloy"
return output
def check_neutrality_4(formula):
charge_neutral_count = 0
comp = mg.Composition(formula)
reduce_formula = comp.get_el_amt_dict()
list_of_elements = list(reduce_formula.keys())
max_num = max(reduce_formula.values())
for i, ele_a in enumerate(list_of_elements):
for ox_a in smact.Element(ele_a).oxidation_states:
for j, ele_b in enumerate(list_of_elements[i+1:]):
for ox_b in smact.Element(ele_b).oxidation_states:
for k, ele_c in enumerate(list_of_elements[i+j+2:]):
for ox_c in smact.Element(ele_c).oxidation_states:
for m, ele_d in enumerate(list_of_elements[i+j+k+3:]):
for ox_d in smact.Element(ele_d).oxidation_states:
# Checks if the combination is charge neutral before printing it out! #
cn_e, cn_r = neutral_ratios([ox_a, ox_b, ox_c, ox_d], threshold = int(max_num))
if cn_e:
for num in cn_r:
if tuple(reduce_formula.values()) == num:
return True
# print(cn_e)
# print(cn_r)
# print(ox_a, ox_b, ox_c)
# charge_neutral_count = charge_neutral_count + 1
# print('{0:3s} {1:3s} {2:3s}'.format(ele_a, ele_b, ele_c))
print('Number of combinations = {0}'.format(charge_neutral_count))
print("--- {0} seconds to run ---".format(time.time() - start_time))
return False
def ox_states_from_binary_formula(self,formula,anion=None,anion_ox_state=None):
"""
Determine oxidation states from binary formula.
Could also use mg.Composition.oxi_state_guesses(), but the logic used is more complex.
Args:
formula: chemical formula
anion: Element symbol of anion. If None, search for common anion
anion_ox_state: oxidation state of anion. If None, assume common oxidation state
"""
comp = mg.Composition(formula)
if len(comp.elements) != 2:
raise ValueError('Formula must be binary')
# determine anion
if anion is None:
anion = np.intersect1d([e.name for e in comp.elements],self.common_anions)
if len(anion) > 1:
raise ValueError('Found multiple possible anions in formula. Please specify anion')
elif len(anion)==0:
raise ValueError('No common anions found in formula. Please specify anion')
else:
anion = anion[0]
metal = np.setdiff1d(comp.elements,mg.Element(anion))[0].name
#get common oxidation state for anion
if anion_ox_state is None:
anion_ox_state = [ox for ox in mg.Element(anion).common_oxidation_states if ox < 0]
if len(anion_ox_state) > 1:
raise Exception(f"Multiple common oxidation states for {anion}. Please specify anion_ox_state")
else:
anion_ox_state = anion_ox_state[0]
metal_ox_state = -comp.get(anion)*anion_ox_state/comp.get(metal)
return {metal:metal_ox_state,anion:anion_ox_state}
def check_electronegativity_2(formula):
pauling_count = 0
comp = mg.Composition(formula)
reduce_formula = comp.get_el_amt_dict()
list_of_elements = list(reduce_formula.keys())
max_num = max(reduce_formula.values())
for i, ele_a in enumerate(list_of_elements):
paul_a = smact.Element(ele_a).pauling_eneg
for ox_a in smact.Element(ele_a).oxidation_states:
for j, ele_b in enumerate(list_of_elements[i + 1:]):
paul_b = smact.Element(ele_b).pauling_eneg
for ox_b in smact.Element(ele_b).oxidation_states:
# Puts elements, oxidation states and electronegativites into lists for convenience #
elements = [ele_a, ele_b]
oxidation_states = [ox_a, ox_b]
pauling_electro = [paul_a, paul_b]
# Checks if the electronegativity makes sense and if the combination is charge neutral #
electroneg_makes_sense = pauling_test(
oxidation_states, pauling_electro, elements)
cn_e, cn_r = neutral_ratios([ox_a, ox_b],
threshold=int(max_num))
if cn_e:
if electroneg_makes_sense:
pauling_count = pauling_count + 1
for num in cn_r:
if tuple(reduce_formula.values()) == num:
# print('{0:2s}{1:3d} {2:2s}{3:3d} {4:2s}{5:3d}'.format(ele_a, ox_a, ele_b,
# ox_b, ele_c, ox_c))
return True
print('Number of combinations = {0}'.format(pauling_count))
print("--- {0} seconds to run ---".format(time.time() - start_time))
return False
def check_neutrality_2(formula):
charge_neutral_count = 0
comp = mg.Composition(formula)
reduce_formula = comp.get_el_amt_dict()
list_of_elements = list(reduce_formula.keys())
# stoichs = np.array(list(reduce_formula.values())).astype(np.int32)[:,np.newaxis]
max_num = max(reduce_formula.values())
for i, ele_a in enumerate(list_of_elements):
for ox_a in smact.Element(ele_a).oxidation_states:
for j, ele_b in enumerate(list_of_elements[i+1:]):
for ox_b in smact.Element(ele_b).oxidation_states:
# Checks if the combination is charge neutral before printing it out! #
# neutral_ratios(oxidations, stoichs=False, threshold=5) #speed up by providing stoichs
# cn_e, cn_r = neutral_ratios([ox_a, ox_b], threshold = int(max_num))
cn_e, cn_r = neutral_ratios([ox_a, ox_b], threshold = int(max_num))
if cn_e:
for num in cn_r:
if tuple(reduce_formula.values()) == num:
return True
# print(cn_e)
# print(cn_r)
# print(ox_a, ox_b, ox_c)
# charge_neutral_count = charge_neutral_count + 1
# print('{0:3s} {1:3s} {2:3s}'.format(ele_a, ele_b, ele_c))
print('Number of combinations = {0}'.format(charge_neutral_count))
print("--- {0} seconds to run ---".format(time.time() - start_time))
return False
def get_features(self, formula):
try:
fractional_composition = mg.Composition(
formula).fractional_composition.as_dict()
element_composition = mg.Composition(
formula).element_composition.as_dict()
avg_feature = np.zeros(len(self.element_df.iloc[0]))
std_feature = np.zeros(len(self.element_df.iloc[0]))
for key in fractional_composition:
try:
avg_feature += self.element_df.loc[
key].values * fractional_composition[key]
diff_feature = self.element_df.loc[list(
fractional_composition.keys())].max(
) - self.element_df.loc[list(
fractional_composition.keys())].min()
except Exception as e:
print('The element:', key, 'from formula', formula,
'is not currently supported in our database')
return np.array([np.nan] * len(self.element_df.iloc[0]) *
5)
max_feature = self.element_df.loc[list(
fractional_composition.keys())].max()
min_feature = self.element_df.loc[list(
fractional_composition.keys())].min()
std_feature = self.element_df.loc[list(
fractional_composition.keys())].std(ddof=0)
features = pd.DataFrame(
np.concatenate([
avg_feature, diff_feature,
np.array(max_feature),
np.array(min_feature),
np.array(std_feature)
]))
features = np.concatenate([
avg_feature, diff_feature,
np.array(max_feature),
np.array(min_feature),
np.array(std_feature)
])
return features.transpose()
except:
print('There was an error with the Formula: ' + formula +
', this is a general exception with an unkown error')
return [np.nan] * len(self.element_df.iloc[0]) * 5
def get_elem_of_interest(formula_str):
"""Get the element of interest from the chemical formula"""
# get all the elements
elem_lst = mg.Composition(formula_str).elements
# grab the second last element from the right
elem_of_interest = elem_lst[-2]
# return the element symbol in string format
return elem_of_interest.symbol
def check_electronegativity_8(formula):
pauling_count = 0
comp = mg.Composition(formula)
reduce_formula = comp.get_el_amt_dict()
list_of_elements = list(reduce_formula.keys())
max_num = max(reduce_formula.values())
# for element in reduce_formula.keys():
# print(len(smact.Element(element).oxidation_states), end = ",")
for i, ele_a in enumerate(list_of_elements):
paul_a = smact.Element(ele_a).pauling_eneg
for ox_a in smact.Element(ele_a).oxidation_states:
for j, ele_b in enumerate(list_of_elements[i+1:]):
paul_b = smact.Element(ele_b).pauling_eneg
for ox_b in smact.Element(ele_b).oxidation_states:
for k, ele_c in enumerate(list_of_elements[i+j+2:]):
paul_c = smact.Element(ele_c).pauling_eneg
for ox_c in smact.Element(ele_c).oxidation_states:
for m, ele_d in enumerate(list_of_elements[i+j+k+3:]):
paul_d = smact.Element(ele_d).pauling_eneg
for ox_d in smact.Element(ele_d).oxidation_states:
for n, ele_e in enumerate(list_of_elements[i+j+k+m+4:]):
paul_e = smact.Element(ele_e).pauling_eneg
for ox_e in smact.Element(ele_e).oxidation_states:
for p, ele_f in enumerate(list_of_elements[i+j+k+m+n+5:]):
paul_f = smact.Element(ele_f).pauling_eneg
for ox_f in smact.Element(ele_f).oxidation_states:
for q, ele_g in enumerate(list_of_elements[i+j+k+m+n+p+6:]):
paul_g = smact.Element(ele_g).pauling_eneg
for ox_g in smact.Element(ele_g).oxidation_states:
for s, ele_h in enumerate(list_of_elements[i+j+k+m+n+p+q+7:]):
paul_h = smact.Element(ele_h).pauling_eneg
for ox_h in smact.Element(ele_h).oxidation_states:
# Puts elements, oxidation states and electronegativites into lists for convenience #
elements = [ele_a, ele_b, ele_c, ele_d, ele_e, ele_f, ele_g, ele_h]
oxidation_states = [ox_a, ox_b, ox_c, ox_d, ox_e, ox_f, ox_g, ox_h]
pauling_electro = [paul_a, paul_b, paul_c, paul_d, paul_e, paul_f, paul_g, paul_h]
if None in pauling_electro:
print("No pauling electronegativity data")
return False
# Checks if the electronegativity makes sense and if the combination is charge neutral #
electroneg_makes_sense = pauling_test(oxidation_states, pauling_electro, elements)
cn_e, cn_r = neutral_ratios([ox_a, ox_b, ox_c, ox_d, ox_e, ox_f, ox_g, ox_h], threshold = int(max_num))
if cn_e:
if electroneg_makes_sense:
pauling_count = pauling_count + 1
for num in cn_r:
if tuple(reduce_formula.values()) == num:
# print('{0:2s}{1:3d} {2:2s}{3:3d} {4:2s}{5:3d}'.format(ele_a, ox_a, ele_b,
# ox_b, ele_c, ox_c))
return True
# print('Number of combinations = {0}'.format(pauling_count))
# print("--- {0} seconds to run ---".format(time.time() - start_time))
return False
def check_electronegativity_4(formula):
pauling_count = 0
comp = mg.Composition(formula)
reduce_formula = comp.get_el_amt_dict()
list_of_elements = list(reduce_formula.keys())
stoichs = list(
np.array(list(reduce_formula.values())).astype(np.int32)[:,
np.newaxis])
max_num = max(reduce_formula.values())
for i, ele_a in enumerate(list_of_elements):
paul_a = smact.Element(ele_a).pauling_eneg
for ox_a in smact.Element(ele_a).oxidation_states:
for j, ele_b in enumerate(list_of_elements[i + 1:]):
paul_b = smact.Element(ele_b).pauling_eneg
for ox_b in smact.Element(ele_b).oxidation_states:
for k, ele_c in enumerate(list_of_elements[i + j + 2:]):
paul_c = smact.Element(ele_c).pauling_eneg
for ox_c in smact.Element(ele_c).oxidation_states:
for m, ele_d in enumerate(
list_of_elements[i + j + k + 3:]):
paul_d = smact.Element(ele_d).pauling_eneg
for ox_d in smact.Element(
ele_d).oxidation_states:
# Puts elements, oxidation states and electronegativites into lists for convenience #
elements = [ele_a, ele_b, ele_c, ele_d]
oxidation_states = [ox_a, ox_b, ox_c, ox_d]
pauling_electro = [
paul_a, paul_b, paul_c, paul_d
]
if None in pauling_electro:
print(
"No pauling electronegativity data"
)
return False
# Checks if the electronegativity makes sense and if the combination is charge neutral #
electroneg_makes_sense = pauling_test(
oxidation_states, pauling_electro,
elements)
cn_e, cn_r = neutral_ratios(
[ox_a, ox_b, ox_c, ox_d],
stoichs=stoichs,
threshold=int(max_num))
if cn_e:
if electroneg_makes_sense:
pauling_count = pauling_count + 1
for num in cn_r:
if tuple(reduce_formula.values(
)) == num:
# print('{0:2s}{1:3d} {2:2s}{3:3d} {4:2s}{5:3d}'.format(ele_a, ox_a, ele_b,
# ox_b, ele_c, ox_c))
return True
# print('Number of combinations = {0}'.format(pauling_count))
# print("--- {0} seconds to run ---".format(time.time() - start_time))
return False
def transform(self, X_df):
formulas = X_df.formula.values
input = np.zeros(shape=(len(formulas), len(elements)),
dtype=np.float32)
for i, formula in enumerate(formulas):
comp = mg.Composition(formula).as_dict()
for k in comp.keys():
input[i][elements.index(k)] = comp[k]
return input
def get_fH(self,formula, phase='solid', data_type='exp',silent=True,exclude_phases=[]):
"""
Get average experimental formation enthalpy for formula and phase
Parameters:
-----------
formula: chemical formula string
phase: phase string. Can be 'solid', 'liquid', 'gas', or a specific solid phase (e.g. 'monoclinic'). If 'solid', returns average across all solid phases
"""
#first check for corrected/saved data in fH_dict
try:
fH,msg = self.fH_dict[(formula,phase,data_type,','.join(exclude_phases))]
if silent==False:
#print('already calculated')
print(msg)
#if no entry exists, look up in MP
except KeyError:
results = self.mp.get_data(formula,data_type=data_type)
if data_type=='exp':
#results = self.mp.get_exp_thermo_data(formula)
if phase=='solid':
phase_results = [r for r in results if r.type=='fH' and r.phaseinfo not in ['liquid','gas']+exclude_phases]
else:
phase_results = [r for r in results if r.type=='fH' and r.phaseinfo==phase]
phases = np.unique([r.phaseinfo for r in phase_results])
fH = [r.value for r in phase_results]
elif data_type=='vasp':
if phase in ('liquid','gas'):
raise ValueError('VASP data only valid for solid phases')
elif phase=='solid':
#get entry with lowest energy above hull
srt_results = sorted(results,key=lambda x: x['e_above_hull'])
phase_results = srt_results[0:1]
else:
phase_results = [r for r in results if r['spacegroup']['crystal_system']==phase]
phases = np.unique([r['spacegroup']['crystal_system'] for r in phase_results])
n_atoms = mg.Composition(formula).num_atoms
#DFT formation energies given in eV per atom - need to convert to kJ/mol
fH = [r['formation_energy_per_atom']*n_atoms*96.485 for r in phase_results]
if len(fH)==0:
raise LookupError('No {} data for {} in {} phase'.format(data_type,formula,phase))
maxdiff = np.max(fH) - np.min(fH)
if maxdiff > 15:
warnings.warn('Max discrepancy of {} in formation enthalpies for {} exceeds limit'.format(maxdiff,formula))
fH = np.mean(fH)
msg = 'Formation enthalpy for {} in {} phase includes {} data from phases: {}'.format(formula,phase,data_type,', '.join(phases))
if silent==False:
print(msg)
#store value and info message for future lookup
self.fH_dict[(formula,phase,data_type,','.join(exclude_phases))] = (fH,msg)
return fH
def _contains_element(self, comp):
"""
Returns 1 if comp contains that element, and 0 if not.
Uses ints because sklearn and numpy like number classes better than bools. Could even be
something crazy like "contains {element}" and "does not contain {element}" if you really
wanted.
"""
comp = pymatgen.Composition(comp)
count = comp[self.element]
return int(count != 0)
def normalize_and_alphabetize_formula(formula):
'''Normalizes composition labels. Used to enable matching / groupby on compositions.'''
if formula:
try:
comp = mg.Composition(formula)
weights = [comp.get_atomic_fraction(ele) for ele in comp.elements]
normalized_weights = [round(w / max(weights), 3) for w in weights]
normalized_comp = "".join([
str(x) + str(y)
for x, y in zip(comp.elements, normalized_weights)
])
return mg.Composition(normalized_comp).alphabetical_formula
except:
print("INVALID: ", formula)
return None
else:
return None
def __init__(self,composition,radius_type='ionic_radius',normalize_formula=False):
self.cations = [el.name for el in composition.elements if el.name!='O']
self.radius_type = radius_type
if normalize_formula==True:
#scale to single total unit of cations
tot_cat_amt = sum([composition[c] for c in self.cations])
composition = mg.Composition({el:amt/tot_cat_amt for el,amt in composition.get_el_amt_dict().items()})
self.composition = composition
self.metal_composition = mg.Composition({c:self.composition[c] for c in self.cations})
#checks
if len(self.cations)==0:
raise Exception('No cations in composition')
# if self.composition['O']!=self.composition['Ba'] + self.composition['Ca'] + self.composition['Al']*3/2:
# raise Exception('Oxygen amount does not match BaO, CaO, Al2O3 stoichiometry')
if self.radius_type not in ('crystal_radius','ionic_radius'):
raise Exception(f'Invalid radius type {self.radius_type}. Options are crystal_radius and ionic_radius')
def featurize(self, comp):
"""
Args:
comp: (Composition)
pymatgen Composition object
Returns:
HOMO_character: (str) orbital symbol ('s', 'p', 'd', or 'f')
HOMO_element: (str) symbol of element for H**O
HOMO_energy: (float in eV) absolute energy of H**O
LUMO_character: (str) orbital symbol ('s', 'p', 'd', or 'f')
LUMO_element: (str) symbol of element for LUMO
LUMO_energy: (float in eV) absolute energy of LUMO
gap_AO: (float in eV)
the estimated bandgap from H**O and LUMO energeis
"""
integer_comp, factor = comp.get_integer_formula_and_factor()
# warning message if composition is dilute and truncated
if not (len(mg.Composition(comp).elements) ==
len(mg.Composition(integer_comp).elements)):
warn('AtomicOrbitals: {} truncated to {}'.format(comp,
integer_comp))
homo_lumo = MolecularOrbitals(integer_comp).band_edges
feat = collections.OrderedDict()
for edge in ['H**O', 'LUMO']:
if homo_lumo[edge] is not None:
feat['{}_character'.format(edge)] = homo_lumo[edge][1][-1]
feat['{}_element'.format(edge)] = homo_lumo[edge][0]
feat['{}_energy'.format(edge)] = homo_lumo[edge][2]
else:
#if LUMO is None
feat['{}_character'.format(edge)] = 'na'
feat['{}_element'.format(edge)] = 'na'
#unclear what this value should be. Arbitrarily set to 0. Don't want NaN for modeling
feat['{}_energy'.format(edge)] = 0
feat['gap_AO'] = feat['LUMO_energy'] - feat['HOMO_energy']
return list(feat.values())
def calculate_density(formula):
'''Calculates densisty based on Rule of Mixtures (ROM).'''
comp = mg.Composition(formula)
weights = [comp.get_atomic_fraction(e) for e in comp.elements]
vols = np.array([e.molar_volume for e in comp.elements])
atomic_masses = np.array([e.atomic_mass for e in comp.elements])
val = np.sum(weights * atomic_masses) / np.sum(weights * vols)
return round(val, 1)
def _contains_all_elements(self, compositions):
elements = list()
df_trans = pd.DataFrame()
for comp in compositions.values:
comp = pymatgen.Composition(comp)
for element in comp.elements:
if element not in elements:
elements.append(element)
for element in elements:
self.element = element
self.new_column_name = "has_" + str(self.element)
has_element = compositions.apply(self._contains_element)
df_trans[self.new_column_name] = has_element
return df_trans
def calculate_youngs_modulus(formula):
'''Calculates Young Modulus based on Rule of Mixtures (ROM).'''
comp = mg.Composition(formula)
weights = np.array([comp.get_atomic_fraction(e) for e in comp.elements])
vols = np.array([e.molar_volume for e in comp.elements])
ym_vals = np.array([e.youngs_modulus for e in comp.elements])
if None in ym_vals:
return ''
val = np.sum(weights * vols * ym_vals) / np.sum(weights * vols)
return int(round(val, 0))
def load_data(file_path): if not os.path.exists(file_path): myfile = open(file_path, "w") myfile.close() file = open(file_path, "r") all_formula = [] for ind, form in enumerate(file.readlines()): if form == "reduced_cell_formula\n": continue comp = mg.Composition(form) reduce_formula = comp.get_el_amt_dict() for key in reduce_formula.keys(): reduce_formula[key] = int(reduce_formula[key]) all_formula.append(reduce_formula) return all_formula
def MX_bond_energy(self,formula,data_type='exp',ordered_formula=False,silent=True,exclude_phases=[]):
"""
Get metal-anion bond energy per mole of metal for binary ionic compound
Parameters:
-----------
formula: chemical formula string
ordered_formula: if true, assume that first element in formula is metal, and second is anion (i.e. MmXn)
exclude_phases: phases to exclude from aggregate over all solid phases
"""
comp = mg.Composition(formula)
formula = comp.reduced_formula
try:
#look up compound if already calculated
abe,msg = self.calc_MX_bond_energy[(formula,data_type,','.join(exclude_phases))]
if silent==False:
#print('already calculated')
print(msg)
except KeyError:
if len(comp.elements) != 2:
raise Exception("Formula is not a binary compound")
if ordered_formula is False:
anions = [el.name for el in comp.elements if el.name in self.common_anions]
if len(anions) == 0:
raise Exception('No common anions found in formula. Use ordered formula to indicate metal and anion')
elif len(anions) > 1:
raise Exception('Multiple anions found in formula. Use ordered formula to indicate metal and anion')
else:
anion = anions[0]
metal = [el.name for el in comp.elements if el.name!=anion][0]
elif ordered_formula is True:
metal = comp.elements[0].name
anion = comp.elements[1].name
m = comp.get_el_amt_dict()[metal]
n = comp.get_el_amt_dict()[anion]
fH = self.get_fH(formula,data_type=data_type,silent=silent,exclude_phases=exclude_phases) #oxide formation enthalpy
H_sub = self.get_fH(metal, phase='gas',silent=silent,exclude_phases=[]) #metal sublimation enthalpy - must be exp data (no vasp data for gas)
#look up info messages from get_fH to store in dict
msg = self.fH_dict[formula,'solid',data_type,','.join(exclude_phases)][1] + '\n'
msg += self.fH_dict[metal,'gas','exp',''][1]
DX2 = self.dissocation_energy[anion] #anion dissociation energy
abe = (fH - m*H_sub - (n/2)*DX2)/m #M-O bond energy per mole of M
self.calc_MX_bond_energy[(formula,data_type,','.join(exclude_phases))] = (abe,msg)
return abe
def __call__(self, formulas):
if isinstance(formulas, str):
formulas = [formulas]
reduce_forms = []
for i, formula in enumerate(formulas):
comp = mg.Composition(formula)
reduce_form = comp.get_el_amt_dict()
reduce_forms.append(reduce_form)
one_hot_vec = []
for ind, formula in enumerate(reduce_forms):
vec = np.zeros((1, len(self.Periodic_table)))
keys = formula.keys()
for symbols in keys:
symbols_index = list(self.Periodic_table).index(symbols)
vec[0, symbols_index] = float(formula[symbols])
one_hot_vec.append(vec)
one_hot_vec = np.concatenate(one_hot_vec, axis=0)
return one_hot_vec
def create_inequalities(compounds, eoi, control_element = None):
# Assumes that the anion is placed last
inequalities = {}
formulas = {}
if control_element is None:
control_element = eoi[-1]
for compound in compounds:
comp = pmg.Composition(compound['pretty_formula'])
dict = comp.get_el_amt_dict()
lhs = []
for k in list(dict.keys()):
lhs.append([k, dict[k]])
formulas[compound['pretty_formula']] = lhs
for compound in compounds:
per_comp = []
c_formula = formulas[compound['pretty_formula']]
# Basic inequalities
per_comp.append([c_formula, '<', compound['final_energy']])
for items in c_formula:
if items[0] != control_element:
per_comp.append([[items], '>', compound['final_energy']])
per_comp.append([[items], '<', 0])
# Generates inequalities outside the first one
for o_comp in compounds:
if o_comp != compound:
o_formula = formulas[o_comp['pretty_formula']]
compul = [x[1] for x in c_formula if x[0] == control_element][0]
o_compul = [x[1] for x in o_formula if x[0] == control_element][0]
f = lcm(compul, o_compul)
constants = [[x[0], x[1]*f/compul] for x in c_formula]
o_constants = [[x[0], x[1]*f/o_compul] for x in o_formula]
f_constants = []
for elm in [x for x in eoi if x!= control_element]:
el = [x[1] for x in constants if x[0] == elm]
sel = [x[1] for x in o_constants if x[0] == elm]
if len(el) == 0:
el = [0]
if len(sel) == 0:
sel = [0]
f_constants.append([elm, el[0]-sel[0]])
per_comp.append([o_comp['pretty_formula'], f_constants, '>',
f*(compound['final_energy']/compul-o_comp['final_energy']/o_compul)])
inequalities[compound['pretty_formula']] = per_comp
return inequalities
def check_neutrality_8(formula):
charge_neutral_count = 0
comp = mg.Composition(formula)
reduce_formula = comp.get_el_amt_dict()
list_of_elements = list(reduce_formula.keys())
stoichs = list(
np.array(list(reduce_formula.values())).astype(np.int32)[:,np.newaxis]
)
max_num = max(reduce_formula.values())
for element in reduce_formula.keys():
print(len(smact.Element(element).oxidation_states), end = ",")
for i, ele_a in enumerate(list_of_elements):
for ox_a in smact.Element(ele_a).oxidation_states:
for j, ele_b in enumerate(list_of_elements[i+1:]):
for ox_b in smact.Element(ele_b).oxidation_states:
for k, ele_c in enumerate(list_of_elements[i+j+2:]):
for ox_c in smact.Element(ele_c).oxidation_states:
for m, ele_d in enumerate(list_of_elements[i+j+k+3:]):
for ox_d in smact.Element(ele_d).oxidation_states:
for n, ele_e in enumerate(list_of_elements[i+j+k+m+4:]):
for ox_e in smact.Element(ele_e).oxidation_states:
for p, ele_f in enumerate(list_of_elements[i+j+k+m+n+5:]):
for ox_f in smact.Element(ele_f).oxidation_states:
for q, ele_g in enumerate(list_of_elements[i+j+k+m+n+p+6:]):
for ox_g in smact.Element(ele_g).oxidation_states:
for s, ele_h in enumerate(list_of_elements[i+j+k+m+n+p+q+7:]):
for ox_h in smact.Element(ele_h).oxidation_states:
# Checks if the combination is charge neutral before printing it out! #
cn_e, cn_r = neutral_ratios([ox_a, ox_b, ox_c, ox_d, ox_e, ox_f, ox_g, ox_h], stoichs = stoichs, threshold = int(max_num))
if cn_e:
for num in cn_r:
if tuple(reduce_formula.values()) == num:
return True
# print(cn_e)
# print(cn_r)
# print(ox_a, ox_b, ox_c)
# charge_neutral_count = charge_neutral_count + 1
# print('{0:3s} {1:3s} {2:3s}'.format(ele_a, ele_b, ele_c))
print('Number of combinations = {0}'.format(charge_neutral_count))
print("--- {0} seconds to run ---".format(time.time() - start_time))
return False
def get_comp_from_coords(coords, tern_axes=['Ca', 'Al', 'Ba'], scale=1):
if len(coords) == 2:
a, b = coords
c = scale - a - b
coords = (a, b, c)
# else:
# a,b,c = coords
oxides = {
'Ba': 'BaO',
'Ca': 'CaO',
'Al': 'Al2O3',
'B': 'B2O3',
'Mg': 'MgO',
'Sr': 'SrO'
}
formula = ''.join(
['({}){}'.format(oxides[m], amt) for m, amt in zip(tern_axes, coords)])
return mg.Composition(formula)
def __call__(self, formulas):
if isinstance(formulas, str):
formulas = [formulas]
reduce_forms = []
for i, formula in enumerate(formulas):
comp = mg.Composition(formula)
reduce_form = comp.get_el_amt_dict()
reduce_forms.append(reduce_form)
atom2vec = []
for ind, formula in enumerate(reduce_forms):
matrix = 0
keys = formula.keys()
for symbols in keys:
symbols_index = list(self.atom).index(symbols)
matrix += self.atom_fec[symbols_index] * float(
formula[symbols])
atom2vec.append(matrix[np.newaxis, :])
atom2vecs = np.concatenate(atom2vec, axis=0)
return atom2vecs