def read_vasp_xml(filename='vasprun.xml', index=-1):
"""Parse vasprun.xml file.
Reads unit cell, atom positions, energies, forces, and constraints
from vasprun.xml file
"""
import xml.etree.ElementTree as ET
from ase.constraints import FixAtoms, FixScaled
from ase.calculators.singlepoint import (SinglePointDFTCalculator,
SinglePointKPoint)
from ase.units import GPa
from collections import OrderedDict
tree = ET.iterparse(filename, events=['start', 'end'])
atoms_init = None
calculation = []
ibz_kpts = None
kpt_weights = None
parameters = OrderedDict()
try:
for event, elem in tree:
if event == 'end':
if elem.tag == 'kpoints':
for subelem in elem.iter(tag='generation'):
kpts_params = OrderedDict()
parameters['kpoints_generation'] = kpts_params
for par in subelem.iter():
if par.tag in ['v', 'i']:
parname = par.attrib['name'].lower()
kpts_params[parname] = __get_xml_parameter(par)
kpts = elem.findall("varray[@name='kpointlist']/v")
ibz_kpts = np.zeros((len(kpts), 3))
for i, kpt in enumerate(kpts):
ibz_kpts[i] = [float(val) for val in kpt.text.split()]
kpt_weights = elem.findall('varray[@name="weights"]/v')
kpt_weights = [float(val.text) for val in kpt_weights]
elif elem.tag == 'parameters':
for par in elem.iter():
if par.tag in ['v', 'i']:
parname = par.attrib['name'].lower()
parameters[parname] = __get_xml_parameter(par)
elif elem.tag == 'atominfo':
species = []
for entry in elem.find("array[@name='atoms']/set"):
species.append(entry[0].text.strip())
natoms = len(species)
elif (elem.tag == 'structure'
and elem.attrib.get('name') == 'initialpos'):
cell_init = np.zeros((3, 3), dtype=float)
for i, v in enumerate(
elem.find("crystal/varray[@name='basis']")):
cell_init[i] = np.array(
[float(val) for val in v.text.split()])
scpos_init = np.zeros((natoms, 3), dtype=float)
for i, v in enumerate(
elem.find("varray[@name='positions']")):
scpos_init[i] = np.array(
[float(val) for val in v.text.split()])
constraints = []
fixed_indices = []
for i, entry in enumerate(
elem.findall("varray[@name='selective']/v")):
flags = (np.array(
entry.text.split() == np.array(['F', 'F', 'F'])))
if flags.all():
fixed_indices.append(i)
elif flags.any():
constraints.append(FixScaled(cell_init, i, flags))
if fixed_indices:
constraints.append(FixAtoms(fixed_indices))
atoms_init = Atoms(species,
cell=cell_init,
scaled_positions=scpos_init,
constraint=constraints,
pbc=True)
elif elem.tag == 'dipole':
dblock = elem.find('v[@name="dipole"]')
if dblock is not None:
dipole = np.array(
[float(val) for val in dblock.text.split()])
elif event == 'start' and elem.tag == 'calculation':
calculation.append(elem)
except ET.ParseError as parse_error:
if atoms_init is None:
raise parse_error
if calculation and calculation[-1].find("energy") is None:
calculation = calculation[:-1]
if not calculation:
yield atoms_init
if calculation:
if isinstance(index, int):
steps = [calculation[index]]
else:
steps = calculation[index]
else:
steps = []
for step in steps:
# Workaround for VASP bug, e_0_energy contains the wrong value
# in calculation/energy, but calculation/scstep/energy does not
# include classical VDW corrections. So, first calculate
# e_0_energy - e_fr_energy from calculation/scstep/energy, then
# apply that correction to e_fr_energy from calculation/energy.
lastscf = step.findall('scstep/energy')[-1]
dipoles = step.findall('scstep/dipole')
if dipoles:
lastdipole = dipoles[-1]
else:
lastdipole = None
de = (float(lastscf.find('i[@name="e_0_energy"]').text) -
float(lastscf.find('i[@name="e_fr_energy"]').text))
free_energy = float(step.find('energy/i[@name="e_fr_energy"]').text)
energy = free_energy + de
cell = np.zeros((3, 3), dtype=float)
for i, vector in enumerate(
step.find('structure/crystal/varray[@name="basis"]')):
cell[i] = np.array([float(val) for val in vector.text.split()])
scpos = np.zeros((natoms, 3), dtype=float)
for i, vector in enumerate(
step.find('structure/varray[@name="positions"]')):
scpos[i] = np.array([float(val) for val in vector.text.split()])
forces = None
fblocks = step.find('varray[@name="forces"]')
if fblocks is not None:
forces = np.zeros((natoms, 3), dtype=float)
for i, vector in enumerate(fblocks):
forces[i] = np.array(
[float(val) for val in vector.text.split()])
stress = None
sblocks = step.find('varray[@name="stress"]')
if sblocks is not None:
stress = np.zeros((3, 3), dtype=float)
for i, vector in enumerate(sblocks):
stress[i] = np.array(
[float(val) for val in vector.text.split()])
stress *= -0.1 * GPa
stress = stress.reshape(9)[[0, 4, 8, 5, 2, 1]]
dipole = None
if lastdipole is not None:
dblock = lastdipole.find('v[@name="dipole"]')
if dblock is not None:
dipole = np.zeros((1, 3), dtype=float)
dipole = np.array([float(val) for val in dblock.text.split()])
dblock = step.find('dipole/v[@name="dipole"]')
if dblock is not None:
dipole = np.zeros((1, 3), dtype=float)
dipole = np.array([float(val) for val in dblock.text.split()])
efermi = step.find('dos/i[@name="efermi"]')
if efermi is not None:
efermi = float(efermi.text)
kpoints = []
for ikpt in range(1, len(ibz_kpts) + 1):
kblocks = step.findall(
'eigenvalues/array/set/set/set[@comment="kpoint %d"]' % ikpt)
if kblocks is not None:
for spin, kpoint in enumerate(kblocks):
eigenvals = kpoint.findall('r')
eps_n = np.zeros(len(eigenvals))
f_n = np.zeros(len(eigenvals))
for j, val in enumerate(eigenvals):
val = val.text.split()
eps_n[j] = float(val[0])
f_n[j] = float(val[1])
if len(kblocks) == 1:
f_n *= 2
kpoints.append(
SinglePointKPoint(kpt_weights[ikpt - 1], spin, ikpt,
eps_n, f_n))
if len(kpoints) == 0:
kpoints = None
atoms = atoms_init.copy()
atoms.set_cell(cell)
atoms.set_scaled_positions(scpos)
atoms.calc = SinglePointDFTCalculator(atoms,
energy=energy,
forces=forces,
stress=stress,
free_energy=free_energy,
ibzkpts=ibz_kpts,
efermi=efermi,
dipole=dipole)
atoms.calc.name = 'vasp'
atoms.calc.kpts = kpoints
atoms.calc.parameters = parameters
yield atoms
def read_vasp(filename='CONTCAR'):
"""Import POSCAR/CONTCAR type file.
Reads unitcell, atom positions and constraints from the POSCAR/CONTCAR
file and tries to read atom types from POSCAR/CONTCAR header, if this fails
the atom types are read from OUTCAR or POTCAR file.
"""
from ase.constraints import FixAtoms, FixScaled
from ase.data import chemical_symbols
fd = filename
# The first line is in principle a comment line, however in VASP
# 4.x a common convention is to have it contain the atom symbols,
# eg. "Ag Ge" in the same order as later in the file (and POTCAR
# for the full vasp run). In the VASP 5.x format this information
# is found on the fifth line. Thus we save the first line and use
# it in case we later detect that we're reading a VASP 4.x format
# file.
line1 = fd.readline()
lattice_constant = float(fd.readline().split()[0])
# Now the lattice vectors
a = []
for ii in range(3):
s = fd.readline().split()
floatvect = float(s[0]), float(s[1]), float(s[2])
a.append(floatvect)
basis_vectors = np.array(a) * lattice_constant
# Number of atoms. Again this must be in the same order as
# in the first line
# or in the POTCAR or OUTCAR file
atom_symbols = []
numofatoms = fd.readline().split()
# Check whether we have a VASP 4.x or 5.x format file. If the
# format is 5.x, use the fifth line to provide information about
# the atomic symbols.
vasp5 = False
try:
int(numofatoms[0])
except ValueError:
vasp5 = True
atomtypes = numofatoms
numofatoms = fd.readline().split()
# check for comments in numofatoms line and get rid of them if necessary
commentcheck = np.array(['!' in s for s in numofatoms])
if commentcheck.any():
# only keep the elements up to the first including a '!':
numofatoms = numofatoms[:np.arange(len(numofatoms))[commentcheck][0]]
if not vasp5:
# Split the comment line (first in the file) into words and
# try to compose a list of chemical symbols
from ase.formula import Formula
atomtypes = []
for word in line1.split():
word_without_delims = re.sub(r"-|_|,|\.|=|[0-9]|^", "", word)
if len(word_without_delims) < 1:
continue
try:
atomtypes.extend(list(Formula(word_without_delims)))
except ValueError:
# print(atomtype, e, 'is comment')
pass
# Now the list of chemical symbols atomtypes must be formed.
# For example: atomtypes = ['Pd', 'C', 'O']
numsyms = len(numofatoms)
if len(atomtypes) < numsyms:
# First line in POSCAR/CONTCAR didn't contain enough symbols.
# Sometimes the first line in POSCAR/CONTCAR is of the form
# "CoP3_In-3.pos". Check for this case and extract atom types
if len(atomtypes) == 1 and '_' in atomtypes[0]:
atomtypes = get_atomtypes_from_formula(atomtypes[0])
else:
atomtypes = atomtypes_outpot(fd.name, numsyms)
else:
try:
for atype in atomtypes[:numsyms]:
if atype not in chemical_symbols:
raise KeyError
except KeyError:
atomtypes = atomtypes_outpot(fd.name, numsyms)
for i, num in enumerate(numofatoms):
numofatoms[i] = int(num)
[atom_symbols.append(atomtypes[i]) for na in range(numofatoms[i])]
# Check if Selective dynamics is switched on
sdyn = fd.readline()
selective_dynamics = sdyn[0].lower() == 's'
# Check if atom coordinates are cartesian or direct
if selective_dynamics:
ac_type = fd.readline()
else:
ac_type = sdyn
cartesian = ac_type[0].lower() == 'c' or ac_type[0].lower() == 'k'
tot_natoms = sum(numofatoms)
atoms_pos = np.empty((tot_natoms, 3))
if selective_dynamics:
selective_flags = np.empty((tot_natoms, 3), dtype=bool)
for atom in range(tot_natoms):
ac = fd.readline().split()
atoms_pos[atom] = (float(ac[0]), float(ac[1]), float(ac[2]))
if selective_dynamics:
curflag = []
for flag in ac[3:6]:
curflag.append(flag == 'F')
selective_flags[atom] = curflag
if cartesian:
atoms_pos *= lattice_constant
atoms = Atoms(symbols=atom_symbols, cell=basis_vectors, pbc=True)
if cartesian:
atoms.set_positions(atoms_pos)
else:
atoms.set_scaled_positions(atoms_pos)
if selective_dynamics:
constraints = []
indices = []
for ind, sflags in enumerate(selective_flags):
if sflags.any() and not sflags.all():
constraints.append(FixScaled(atoms.get_cell(), ind, sflags))
elif sflags.all():
indices.append(ind)
if indices:
constraints.append(FixAtoms(indices))
if constraints:
atoms.set_constraint(constraints)
return atoms