"""
This module implements an interface to the critic2 Bader analysis code.
For most Bader analysis purposes, users are referred to
pymatgen.command_line.bader_caller instead, this module is for advanced
usage requiring identification of critical points in the charge density.
This module depends on a compiled critic2 executable available in the path.
Please follow the instructions at https://github.com/aoterodelaroza/critic2
to compile.
New users are *strongly* encouraged to read the critic2 manual first.
In brief,
* critic2 searches for critical points in charge density
* a critical point can be one of four types: nucleus, bond, ring
or cage
* it does this by seeding locations for likely critical points
and then searching in these regions
* there are two lists of critical points in the output, a list
of non-equivalent points (with in-depth information about the
field at those points), and a full list of points generated
by the appropriate symmetry operations
* connectivity between these points is also provided when
appropriate (e.g. the two nucleus critical points linked to
a bond critical point)
* critic2 can do many other things besides
If you use this module, please cite the following:
A. Otero-de-la-Roza, E. R. Johnson and V. Luaña,
Comput. Phys. Commun. 185, 1007-1018 (2014)
(http://dx.doi.org/10.1016/j.cpc.2013.10.026)
A. Otero-de-la-Roza, M. A. Blanco, A. Martín Pendás and
V. Luaña, Comput. Phys. Commun. 180, 157–166 (2009)
(http://dx.doi.org/10.1016/j.cpc.2008.07.018)
"""
import os
import subprocess
import warnings
import numpy as np
import glob
from scipy.spatial import KDTree
from pymatgen.io.vasp.outputs import Chgcar
from pymatgen.analysis.graphs import StructureGraph
from pymatgen.core.periodic_table import DummySpecie
from monty.os.path import which
from monty.dev import requires
from monty.json import MSONable
from monty.tempfile import ScratchDir
from enum import Enum
import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
__author__ = "Matthew Horton"
__maintainer__ = "Matthew Horton"
__email__ = "mkhorton@lbl.gov"
__status__ = "Production"
__date__ = "July 2017"
[docs]class Critic2Caller:
"""
Class to call critic2 and store standard output for further processing.
"""
@requires(which("critic2"), "Critic2Caller requires the executable critic to be in the path. "
"Please follow the instructions at https://github.com/aoterodelaroza/critic2.")
def __init__(self, structure, chgcar=None, chgcar_ref=None,
user_input_settings=None, write_cml=False):
"""
Run Critic2 in automatic mode on a supplied structure, charge
density (chgcar) and reference charge density (chgcar_ref).
The reason for a separate reference field is that in
VASP, the CHGCAR charge density only contains valence
electrons and may be missing substantial charge at
nuclei leading to misleading results. Thus, a reference
field is commonly constructed from the sum of AECCAR0
and AECCAR2 which is the total charge density, but then
the valence charge density is used for the final analysis.
If chgcar_ref is not supplied, chgcar will be used as the
reference field. If chgcar is not supplied, the promolecular
charge density will be used as the reference field -- this can
often still give useful results if only topological information
is wanted.
User settings is a dictionary that can contain:
* GRADEPS, float (field units), gradient norm threshold
* CPEPS, float (Bohr units in crystals), minimum distance between
critical points for them to be equivalent
* NUCEPS, same as CPEPS but specifically for nucleus critical
points (critic2 default is depedent on grid dimensions)
* NUCEPSH, same as NUCEPS but specifically for hydrogen nuclei
since associated charge density can be significantly displaced
from hydrogen nucleus
* EPSDEGEN, float (field units), discard critical point if any
element of the diagonal of the Hessian is below this value,
useful for discarding points in vacuum regions
* DISCARD, float (field units), discard critical points with field
value below this value, useful for discarding points in vacuum
regions
* SEED, list of strings, strategies for seeding points, default
is ['WS 1', 'PAIR 10'] which seeds critical points by
sub-dividing the Wigner-Seitz cell and between every atom pair
closer than 10 Bohr, see critic2 manual for more options
:param structure: Structure to analyze
:param chgcar: Charge density to use for analysis. If None, will
use promolecular density.
:param chgcar_ref: Reference charge density. If None, will use
chgcar as reference.
:param user_input_settings (dict):
:param write_cml (bool): Useful for debug, if True will write all
critical points to a file 'table.cml' in the working directory
useful for visualization
"""
settings = {'CPEPS': 0.1, 'SEED': ["WS", "PAIR DIST 10"]}
if user_input_settings:
settings.update(user_input_settings)
# Load crystal structure
input_script = ["crystal POSCAR"]
# Load data to use as reference field
if chgcar_ref:
input_script += ["load VASPCHG CHGCAR_ref id chgcar_ref",
"reference chgcar_ref"]
# Load data to use for analysis
if chgcar:
input_script += ["load VASPCHG CHGCAR_int id chgcar",
"integrable chgcar"]
# Command to run automatic analysis
auto = "auto "
for k, v in settings.items():
if isinstance(v, list):
for item in v:
auto += '{} {} '.format(k, item)
else:
auto += '{} {} '.format(k, v)
input_script += [auto]
if write_cml:
input_script += ["cpreport ../table.cml cell border graph"]
input_script = "\n".join(input_script)
with ScratchDir(".") as temp_dir:
os.chdir(temp_dir)
with open('input_script.cri', 'w') as f:
f.write(input_script)
structure.to(filename='POSCAR')
if chgcar:
chgcar.write_file('CHGCAR_int')
if chgcar_ref:
chgcar_ref.write_file('CHGCAR_ref')
args = ["critic2", "input_script.cri"]
rs = subprocess.Popen(args,
stdout=subprocess.PIPE,
stdin=subprocess.PIPE, close_fds=True)
stdout, stderr = rs.communicate()
stdout = stdout.decode()
if stderr:
stderr = stderr.decode()
warnings.warn(stderr)
if rs.returncode != 0:
raise RuntimeError("critic2 exited with return code {}.".format(rs.returncode))
self._stdout = stdout
self._stderr = stderr
self.output = Critic2Output(structure, stdout)
[docs] @classmethod
def from_path(cls, path, suffix=''):
"""
Convenience method to run critic2 analysis on a folder containing
typical VASP output files.
This method will:
1. Look for files CHGCAR, AECAR0, AECAR2, POTCAR or their gzipped
counterparts.
2. If AECCAR* files are present, constructs a temporary reference
file as AECCAR0 + AECCAR2.
3. Runs critic2 analysis twice: once for charge, and a second time
for the charge difference (magnetization density).
:param path: path to folder to search in
:param suffix: specific suffix to look for (e.g. '.relax1' for
'CHGCAR.relax1.gz')
:return:
"""
def _get_filepath(filename, warning, path=path, suffix=suffix):
paths = glob.glob(os.path.join(path, filename + suffix + '*'))
if not paths:
warnings.warn(warning)
return None
if len(paths) > 1:
# using reverse=True because, if multiple files are present,
# they likely have suffixes 'static', 'relax', 'relax2', etc.
# and this would give 'static' over 'relax2' over 'relax'
# however, better to use 'suffix' kwarg to avoid this!
paths.sort(reverse=True)
warnings.warn('Multiple files detected, using {}'.format(os.path.basename(path)))
path = paths[0]
return path
chgcar_path = _get_filepath('CHGCAR', 'Could not find CHGCAR!')
chgcar = Chgcar.from_file(chgcar_path)
aeccar0_path = _get_filepath('AECCAR0', 'Could not find AECCAR0, interpret Bader results with caution.')
aeccar0 = Chgcar.from_file(aeccar0_path) if aeccar0_path else None
aeccar2_path = _get_filepath('AECCAR2', 'Could not find AECCAR2, interpret Bader results with caution.')
aeccar2 = Chgcar.from_file(aeccar2_path) if aeccar2_path else None
chgcar_ref = aeccar0.linear_add(aeccar2) if (aeccar0 and aeccar2) else None
return cls(chgcar.structure, chgcar, chgcar_ref)
[docs]class CriticalPointType(Enum):
"""
Enum type for the different varieties of critical point.
"""
nucleus = "nucleus" # (3, -3)
bond = "bond" # (3, -1)
ring = "ring" # (3, 1)
cage = "cage" # (3, 3)
[docs]class CriticalPoint(MSONable):
"""
Access information about a critical point and the field values at that point.
"""
def __init__(self, index, type, frac_coords, point_group,
multiplicity, field, field_gradient,
coords=None, field_hessian=None):
"""
Class to characterise a critical point from a topological
analysis of electron charge density.
Note this class is usually associated with a Structure, so
has information on multiplicity/point group symmetry.
:param index: index of point
:param type: type of point, given as a string
:param coords: Cartesian co-ordinates in Angstroms
:param frac_coords: fractional co-ordinates
:param point_group: point group associated with critical point
:param multiplicity: number of equivalent critical points
:param field: value of field at point (f)
:param field_gradient: gradient of field at point (grad f)
:param field_hessian: hessian of field at point (del^2 f)
"""
self.index = index
self._type = type
self.coords = coords
self.frac_coords = frac_coords
self.point_group = point_group
self.multiplicity = multiplicity
self.field = field
self.field_gradient = field_gradient
self.field_hessian = field_hessian
@property
def type(self):
"""
Returns: Instance of CriticalPointType
"""
return CriticalPointType(self._type)
def __str__(self):
return "Critical Point: {} ({})".format(self.type.name, self.frac_coords)
@property
def laplacian(self):
"""
Returns: The Laplacian of the field at the critical point
"""
return np.trace(self.field_hessian)
@property
def ellipticity(self):
"""
Most meaningful for bond critical points,
can be physically interpreted as e.g. degree
of pi-bonding in organic molecules. Consult
literature for more information.
Returns: The ellpiticity of the field at the critical point
"""
eig, _ = np.linalg.eig(self.field_hessian)
eig.sort()
return eig[0]/eig[1] - 1
[docs]class Critic2Output(MSONable):
"""
Class to process the standard output from critic2.
"""
def __init__(self, structure, critic2_stdout):
"""
This class is used to store results from the Critic2Caller.
To explore the bond graph, use the "structure_graph"
method, which returns a user-friendly StructureGraph
class with bonding information. By default, this returns
a StructureGraph with edge weights as bond lengths, but
can optionally return a graph with edge weights as any
property supported by the `CriticalPoint` class, such as
bond ellipticity.
This class also provides an interface to explore just the
non-symmetrically-equivalent critical points via the
`critical_points` attribute, and also all critical
points (via nodes dict) and connections between them
(via edges dict). The user should be familiar with critic2
before trying to understand these.
Indexes of nucleus critical points in the nodes dict are the
same as the corresponding sites in structure, with indices of
other critical points arbitrarily assigned.
:param structure: associated Structure
:param critic2_stdout: stdout from running critic2 in automatic
mode
"""
self.structure = structure
self._critic2_stdout = critic2_stdout
self.nodes = {}
self.edges = {}
self._parse_stdout(critic2_stdout)
[docs] def structure_graph(self, edge_weight=None, edge_weight_units=None,
include_critical_points=("bond", "ring", "cage")):
"""
A StructureGraph object describing bonding information
in the crystal. Lazily constructed.
Args:
edge_weight: a value to store on the Graph edges,
by default this is "bond_length" but other supported
values are any of the attributes of CriticalPoint
edge_weight_units: optional metadata for
book-keeping (e.g. Å for "bond_length" edge weight)
include_critical_points: add DummySpecie for
the critical points themselves, a list of
"nucleus", "bond", "ring", "cage", set to None
to disable
Returns: a StructureGraph
"""
structure = self.structure.copy()
if include_critical_points:
# atoms themselves don't have field information
# so set to 0
for prop in ("ellipticity", "laplacian", "field"):
structure.add_site_property(prop, [0]*len(structure))
for idx, node in self.nodes.items():
cp = self.critical_points[node["unique_idx"]]
if cp.type.value in include_critical_points:
specie = DummySpecie("{}cp".format(cp.type.value[0]), oxidation_state=None)
structure.append(specie, node["frac_coords"],
properties={"ellipticity": cp.ellipticity,
"laplacian": cp.laplacian,
"field": cp.field})
sg = StructureGraph.with_empty_graph(structure, name="bonds",
edge_weight_name=edge_weight,
edge_weight_units=edge_weight_units)
edges = self.edges.copy()
idx_to_delete = []
# check for duplicate bonds
for idx, edge in edges.items():
unique_idx = self.nodes[idx]['unique_idx']
# only check edges representing bonds, not rings
if self.critical_points[unique_idx].type == CriticalPointType.bond:
if idx not in idx_to_delete:
for idx2, edge2 in edges.items():
if idx != idx2 and edge == edge2:
idx_to_delete.append(idx2)
warnings.warn("Duplicate edge detected, try re-running "
"critic2 with custom parameters to fix this. "
"Mostly harmless unless user is also "
"interested in rings/cages.")
logger.debug("Duplicate edge between points {} (unique point {})"
"and {} ({}).".format(idx, self.nodes[idx]['unique_idx'],
idx2, self.nodes[idx2]['unique_idx']))
# and remove any duplicate bonds present
for idx in idx_to_delete:
del edges[idx]
for idx, edge in edges.items():
unique_idx = self.nodes[idx]['unique_idx']
# only add edges representing bonds, not rings
if self.critical_points[unique_idx].type == CriticalPointType.bond:
from_idx = edge['from_idx']
to_idx = edge['to_idx']
from_lvec = edge['from_lvec']
to_lvec = edge['to_lvec']
relative_lvec = np.subtract(to_lvec, from_lvec)
if edge_weight == "bond_length":
weight = self.structure.get_distance(from_idx, to_idx, jimage=relative_lvec)
elif edge_weight:
weight = getattr(self.critical_points[unique_idx],
edge_weight, None)
else:
weight = None
sg.add_edge(from_idx, to_idx,
from_jimage=from_lvec, to_jimage=to_lvec,
weight=weight)
return sg
[docs] def get_critical_point_for_site(self, n):
"""
Args:
n: Site index n
Returns: A CriticalPoint instance
"""
return self.critical_points[self.nodes[n]['unique_idx']]
def _parse_stdout(self, stdout):
stdout = stdout.split("\n")
# NOTE WE ARE USING 0-BASED INDEXING:
# This is different from critic2 which
# uses 1-based indexing, so all parsed
# indices have 1 subtracted.
# Parsing happens in two stages:
# 1. We construct a list of unique critical points
# (i.e. non-equivalent by the symmetry of the crystal)
# and the properties of the field at those points
# 2. We construct a list of nodes and edges describing
# all critical points in the crystal
# Steps 1. and 2. are essentially indepdendent, except
# that the critical points in 2. have a pointer to their
# associated unique critical point in 1. so that more
# information on that point can be retrieved if necessary.
unique_critical_points = []
# parse unique critical points
for i, line in enumerate(stdout):
if "mult name f |grad| lap" in line:
start_i = i + 1
elif "* Analysis of system bonds" in line:
end_i = i - 2
# if start_i and end_i haven't been found, we
# need to re-evaluate assumptions in this parser!
for i, line in enumerate(stdout):
if i >= start_i and i <= end_i:
l = line.replace("(", "").replace(")", "").split()
unique_idx = int(l[0]) - 1
point_group = l[1]
# type = l[2] # type from definition of critical point e.g. (3, -3)
type = l[3] # type from name, e.g. nucleus
frac_coords = [float(l[4]), float(l[5]), float(l[6])]
multiplicity = float(l[7])
# name = float(l[8])
field = float(l[9])
field_gradient = float(l[10])
# laplacian = float(l[11])
point = CriticalPoint(unique_idx, type, frac_coords, point_group,
multiplicity, field, field_gradient)
unique_critical_points.append(point)
# TODO: may be other useful information to parse here too
for i, line in enumerate(stdout):
if '+ Critical point no.' in line:
unique_idx = int(line.split()[4]) - 1
elif "Hessian:" in line:
l1 = list(map(float, stdout[i + 1].split()))
l2 = list(map(float, stdout[i + 2].split()))
l3 = list(map(float, stdout[i + 3].split()))
hessian = [[l1[0], l1[1], l1[2]],
[l2[0], l2[1], l2[2]],
[l3[0], l3[1], l3[2]]]
unique_critical_points[unique_idx].field_hessian = hessian
self.critical_points = unique_critical_points
# parse graph connecting critical points
for i, line in enumerate(stdout):
if "#cp ncp typ position " in line:
start_i = i + 1
elif "* Attractor connectivity matrix" in line:
end_i = i - 2
# if start_i and end_i haven't been found, we
# need to re-evaluate assumptions in this parser!
# Order of nuclei provided by critic2 doesn't
# necessarily match order of sites in Structure.
# We perform a mapping from one to the other,
# and re-index all nodes accordingly.
node_mapping = {} # critic2_index:structure_index
# ensure frac coords are in [0,1] range
frac_coords = np.array(self.structure.frac_coords) % 1
kd = KDTree(frac_coords)
for i, line in enumerate(stdout):
if i >= start_i and i <= end_i:
l = line.split()
if l[2] == "n":
critic2_idx = int(l[0]) - 1
frac_coord = np.array([float(l[3]), float(l[4]), float(l[5])]) % 1
node_mapping[critic2_idx] = kd.query(frac_coord)[1]
if len(node_mapping) != len(self.structure):
warnings.warn("Check that all sites in input structure have "
"been detected by critic2.")
def _remap(critic2_idx):
return node_mapping.get(critic2_idx, critic2_idx)
for i, line in enumerate(stdout):
if i >= start_i and i <= end_i:
l = line.replace("(", "").replace(")", "").split()
idx = _remap(int(l[0]) - 1)
unique_idx = int(l[1]) - 1
frac_coords = [float(l[3]), float(l[4]), float(l[5])]
self._add_node(idx, unique_idx, frac_coords)
if len(l) > 6:
from_idx = _remap(int(l[6]) - 1)
to_idx = _remap(int(l[10]) - 1)
self._add_edge(idx, from_idx=from_idx, from_lvec=(int(l[7]), int(l[8]), int(l[9])),
to_idx=to_idx, to_lvec=(int(l[11]), int(l[12]), int(l[13])))
self._map = node_mapping
def _add_node(self, idx, unique_idx, frac_coords):
"""
Add information about a node describing a critical point.
:param idx: unique index
:param unique_idx: index of unique CriticalPoint,
used to look up more information of point (field etc.)
:param frac_coord: fractional co-ordinates of point
:return:
"""
self.nodes[idx] = {'unique_idx': unique_idx, 'frac_coords': frac_coords}
def _add_edge(self, idx, from_idx, from_lvec, to_idx, to_lvec):
"""
Add information about an edge linking two critical points.
This actually describes two edges:
from_idx ------ idx ------ to_idx
However, in practice, from_idx and to_idx will typically be
atom nuclei, with the center node (idx) referring to a bond
critical point. Thus, it will be more convenient to model
this as a single edge linking nuclei with the properties
of the bond critical point stored as an edge attribute.
:param idx: index of node
:param from_idx: from index of node
:param from_lvec: vector of lattice image the from node is in
as tuple of ints
:param to_idx: to index of node
:param to_lvec: vector of lattice image the to node is in as
tuple of ints
:return:
"""
self.edges[idx] = {'from_idx': from_idx, 'from_lvec': from_lvec,
'to_idx': to_idx, 'to_lvec': to_lvec}