Source code for ase.calculators.aims

"""This module defines an ASE interface to FHI-aims.

Felix Hanke
Jonas Bjork
Simon P. Rittmeyer

import os

import numpy as np
import warnings
import time

from ase.units import Hartree
from import write_aims, read_aims
from import atomic_numbers
from ase.calculators.calculator import FileIOCalculator, Parameters, kpts2mp, \
    ReadError, PropertyNotImplementedError
from ase.utils import basestring

float_keys = [

exp_keys = [

string_keys = [

int_keys = [

bool_keys = [

list_keys = [

[docs]class Aims(FileIOCalculator): # was "command" before the refactoring to dynamical commands __command_default = 'aims.version.serial.x > aims.out' __outfilename_default = 'aims.out' implemented_properties = ['energy', 'forces', 'stress', 'dipole', 'magmom'] def __init__(self, restart=None, ignore_bad_restart_file=False, label=os.curdir, atoms=None, cubes=None, radmul=None, tier=None, aims_command=None, outfilename=None, **kwargs): """Construct the FHI-aims calculator. The keyword arguments (kwargs) can be one of the ASE standard keywords: 'xc', 'kpts' and 'smearing' or any of FHI-aims' native keywords. .. note:: The behavior of command/run_command has been refactored ase X.X.X It is now possible to independently specify the command to call FHI-aims and the outputfile into which stdout is directed. In general, we replaced <run_command> = <aims_command> + " > " + <outfilename That is,what used to be, e.g., >>> calc = Aims(run_command = "mpiexec -np 4 aims.x > aims.out") can now be achieved with the two arguments >>> calc = Aims(aims_command = "mpiexec -np 4 aims.x" >>> outfilename = "aims.out") Backward compatibility, however, is provided. Also, the command actually used to run FHI-aims is dynamically updated (i.e., the "command" member variable). That is, e.g., >>> calc = Aims() >>> print(calc.command) aims.version.serial.x > aims.out >>> calc.outfilename = "systemX.out" >>> print(calc.command) aims.version.serial.x > systemX.out >>> calc.aims_command = "mpiexec -np 4 aims.version.scalapack.mpi.x" >>> print(calc.command) mpiexec -np 4 aims.version.scalapack.mpi > systemX.out Arguments: cubes: AimsCube object Cube file specification. radmul: int Set radial multiplier for the basis set of all atomic species. tier: int or array of ints Set basis set tier for all atomic species. aims_command : str The full command as executed to run FHI-aims *without* the redirection to stdout. For instance "mpiexec -np 4 aims.x". Note that this is not the same as "command" or "run_command". .. note:: Added in ase X.X.X outfilename : str The file (incl. path) to which stdout is redirected. Defaults to "aims.out" .. note:: Added in ase X.X.X run_command : str, optional (default=None) Same as "command", see FileIOCalculator documentation. .. note:: Deprecated in ase X.X.X outfilename : str, optional (default=aims.out) File into which the stdout of the FHI aims run is piped into. Note that this will be only of any effect, if the <run_command> does not yet contain a '>' directive. plus_u : dict For DFT+U. Adds a +U term to one specific shell of the species. kwargs : dict Any of the base class arguments. """ # yes, we pop the key and run it through our legacy filters command = kwargs.pop('command', None) # Check for the "run_command" (deprecated keyword) # Consistently, the "command" argument should be used as suggested by the FileIO base class. # For legacy reasons, however, we here also accept "run_command" run_command = kwargs.pop('run_command', None) if run_command: # this warning is debatable... in my eyes it is more consistent to # use 'command' warnings.warn('Argument "run_command" is deprecated and will be replaced with "command". Alternatively, use "aims_command" and "outfile". See documentation for more details.') if command: warnings.warn('Caution! Argument "command" overwrites "run_command.') else: command=run_command # this is the fallback to the default value for empty init if np.all([i is None for i in (command, aims_command, outfilename)]): # we go for the FileIOCalculator default way (env variable) with the former default as fallback command = os.environ.get('ASE_AIMS_COMMAND', Aims.__command_default) # filter the command and set the member variables "aims_command" and "outfilename" self.__init_command(command=command, aims_command=aims_command, outfilename=outfilename) FileIOCalculator.__init__(self, restart, ignore_bad_restart_file, label, atoms, # well, this is not nice, but cannot work around it... command=self.command, **kwargs) self.cubes = cubes self.radmul = radmul self.tier = tier # handling the filtering for dynamical commands with properties, @property def command(self): return self.__command @command.setter def command(self, x): self.__update_command(command=x) @property def aims_command(self): return self.__aims_command @aims_command.setter def aims_command(self, x): self.__update_command(aims_command=x) @property def outfilename(self): return self.__outfilename @outfilename.setter def outfilename(self, x): self.__update_command(outfilename=x) def __init_command(self, command=None, aims_command=None, outfilename=None): """ Create the private variables for which properties are defines and set them accordingly. """ # new class variables due to dynamical command handling self.__aims_command = None self.__outfilename = None self.__command = None # filter the command and set the member variables "aims_command" and "outfilename" self.__update_command(command=command, aims_command=aims_command, outfilename=outfilename) # legacy handling of the (run_)command behavior a.k.a. a universal setter routine def __update_command(self, command=None, aims_command=None, outfilename=None): """ Abstracted generic setter routine for a dynamic behavior of "command". The command that is actually called on the command line and enters the base class, is <command> = <aims_command> > <outfilename>. This new scheme has been introduced in order to conveniently change the outfile name from the outside while automatically updating the <command> member variable. Obiously, changing <command> conflicts with changing <aims_command> and/or <outfilename>, which thus raises a <ValueError>. This should, however, not happen if this routine is not used outside the property definitions. Parameters ---------- command : str The full command as executed to run FHI-aims. This includes any potential mpiexec call, as well as the redirection of stdout. For instance "mpiexec -np 4 aims.x > aims.out". aims_command : str The full command as executed to run FHI-aims *without* the redirection to stdout. For instance "mpiexec -np 4 aims.x" outfilename : str The file (incl. path) to which stdout is redirected. """ # disentangle the command if given if command: if aims_command: raise ValueError('Cannot specify "command" and "aims_command" simultaneously.') if outfilename: raise ValueError('Cannot specify "command" and "outfilename" simultaneously.') # check if the redirection of stdout is included command_spl = command.split('>') if len(command_spl) > 1: self.__aims_command = command_spl[0].strip() self.__outfilename = command_spl[-1].strip() else: # this should not happen if used correctly # but just to ensure legacy behavior of how "run_command" was handled self.__aims_command = command.strip() self.__outfilename = Aims.__outfilename_default else: if aims_command is not None: self.__aims_command = aims_command elif outfilename is None: # nothing to do here, empty call with 3x None return if outfilename is not None: self.__outfilename = outfilename else: # default to 'aims.out' if not self.outfilename: self.__outfilename = Aims.__outfilename_default self.__command = '{0:s} > {1:s}'.format(self.aims_command, self.outfilename) def set_atoms(self, atoms): self.atoms = atoms def set_label(self, label, update_outfilename=False): self.label = label = label self.prefix = '' # change outfile name to "<label.out>" if update_outfilename: self.outfilename="{}.out".format(os.path.basename(label)) self.out = os.path.join(label, self.outfilename) def check_state(self, atoms): system_changes = FileIOCalculator.check_state(self, atoms) # Ignore unit cell for molecules: if not atoms.pbc.any() and 'cell' in system_changes: system_changes.remove('cell') return system_changes def set(self, **kwargs): xc = kwargs.get('xc') if xc: kwargs['xc'] = {'LDA': 'pw-lda', 'PBE': 'pbe'}.get(xc, xc) changed_parameters = FileIOCalculator.set(self, **kwargs) if changed_parameters: self.reset() return changed_parameters def write_input(self, atoms, properties=None, system_changes=None, ghosts=None, scaled=False): FileIOCalculator.write_input(self, atoms, properties, system_changes) have_lattice_vectors = atoms.pbc.any() have_k_grid = ('k_grid' in self.parameters or 'kpts' in self.parameters) if have_lattice_vectors and not have_k_grid: raise RuntimeError('Found lattice vectors but no k-grid!') if not have_lattice_vectors and have_k_grid: raise RuntimeError('Found k-grid but no lattice vectors!') write_aims(os.path.join(, ''), atoms, scaled, ghosts) self.write_control(atoms, os.path.join(, '')) self.write_species(atoms, os.path.join(, '')) self.parameters.write(os.path.join(, 'parameters.ase')) def prepare_input_files(self): """ Wrapper function to prepare input filesi, e.g., to a run on a remote machine """ if self.atoms is None: raise ValueError('No atoms object attached') self.write_input(self.atoms) def write_control(self, atoms, filename): lim = '#' + '='*79 output = open(filename, 'w') output.write(lim + '\n') for line in ['FHI-aims file: ' + filename, 'Created using the Atomic Simulation Environment (ASE)', time.asctime(), '', 'List of parameters used to initialize the calculator:', ]: output.write('# ' + line + '\n') for p, v in self.parameters.items(): s = '# {} : {}\n'.format(p, v) output.write(s) output.write(lim + '\n') assert not ('kpts' in self.parameters and 'k_grid' in self.parameters) assert not ('smearing' in self.parameters and 'occupation_type' in self.parameters) for key, value in self.parameters.items(): if key == 'kpts': mp = kpts2mp(atoms, self.parameters.kpts) output.write('%-35s%d %d %d\n' % (('k_grid',) + tuple(mp))) dk = 0.5 - 0.5 / np.array(mp) output.write('%-35s%f %f %f\n' % (('k_offset',) + tuple(dk))) elif key == 'species_dir' or key == 'run_command': continue elif key == 'plus_u': continue elif key == 'smearing': name = self.parameters.smearing[0].lower() if name == 'fermi-dirac': name = 'fermi' width = self.parameters.smearing[1] output.write('%-35s%s %f' % ('occupation_type', name, width)) if name == 'methfessel-paxton': order = self.parameters.smearing[2] output.write(' %d' % order) output.write('\n' % order) elif key == 'output': for output_type in value: output.write('%-35s%s\n' % (key, output_type)) elif key == 'vdw_correction_hirshfeld' and value: output.write('%-35s\n' % key) elif key in bool_keys: output.write('%-35s.%s.\n' % (key, repr(bool(value)).lower())) elif isinstance(value, (tuple, list)): output.write('%-35s%s\n' % (key, ' '.join(str(x) for x in value))) elif isinstance(value, basestring): output.write('%-35s%s\n' % (key, value)) else: output.write('%-35s%r\n' % (key, value)) if self.cubes: self.cubes.write(output) output.write(lim + '\n\n') output.close() def read(self, label):, label) geometry = os.path.join(, '') control = os.path.join(, '') for filename in [geometry, control, self.out]: if not os.path.isfile(filename): raise ReadError self.atoms = read_aims(geometry) self.parameters =, 'parameters.ase')) self.read_results() def read_results(self): converged = self.read_convergence() if not converged: os.system('tail -20 ' + self.out) raise RuntimeError('FHI-aims did not converge!\n' + 'The last lines of output are printed above ' + 'and should give an indication why.') self.read_energy() if ('compute_forces' in self.parameters or 'sc_accuracy_forces' in self.parameters): self.read_forces() if ('compute_numerical_stress' in self.parameters or 'compute_analytical_stress' in self.parameters): self.read_stress() if ('dipole' in self.parameters.get('output', []) and not self.atoms.pbc.any()): self.read_dipole() def write_species(self, atoms, filename=''): self.ctrlname = filename species_path = self.parameters.get('species_dir') if species_path is None: species_path = os.environ.get('AIMS_SPECIES_DIR') if species_path is None: raise RuntimeError( 'Missing species directory! Use species_dir ' + 'parameter or set $AIMS_SPECIES_DIR environment variable.') control = open(filename, 'a') symbols = atoms.get_chemical_symbols() symbols2 = [] for n, symbol in enumerate(symbols): if symbol not in symbols2: symbols2.append(symbol) if self.tier is not None: if isinstance(self.tier, int): self.tierlist = np.ones(len(symbols2), 'int') * self.tier elif isinstance(self.tier, list): assert len(self.tier) == len(symbols2) self.tierlist = self.tier for i, symbol in enumerate(symbols2): fd = os.path.join(species_path, '%02i_%s_default' % (atomic_numbers[symbol], symbol)) reached_tiers = False for line in open(fd, 'r'): if self.tier is not None: if 'First tier' in line: reached_tiers = True self.targettier = self.tierlist[i] self.foundtarget = False self.do_uncomment = True if reached_tiers: line = self.format_tiers(line) control.write(line) if self.tier is not None and not self.foundtarget: raise RuntimeError( "Basis tier %i not found for element %s" % (self.targettier, symbol)) if self.parameters.get('plus_u') is not None: if symbol in self.parameters.plus_u.keys(): control.write('plus_u %s \n' % self.parameters.plus_u[symbol]) control.close() if self.radmul is not None: self.set_radial_multiplier() def format_tiers(self, line): if 'meV' in line: assert line[0] == '#' if 'tier' in line and 'Further' not in line: tier = line.split(" tier")[0] tier = tier.split('"')[-1] current_tier = self.translate_tier(tier) if current_tier == self.targettier: self.foundtarget = True elif current_tier > self.targettier: self.do_uncomment = False else: self.do_uncomment = False return line elif self.do_uncomment and line[0] == '#': return line[1:] elif not self.do_uncomment and line[0] != '#': return '#' + line else: return line def translate_tier(self, tier): if tier.lower() == 'first': return 1 elif tier.lower() == 'second': return 2 elif tier.lower() == 'third': return 3 elif tier.lower() == 'fourth': return 4 else: return -1 def set_radial_multiplier(self): assert isinstance(self.radmul, int) newctrl = self.ctrlname +'.new' fin = open(self.ctrlname, 'r') fout = open(newctrl, 'w') newline = " radial_multiplier %i\n" % self.radmul for line in fin: if ' radial_multiplier' in line: fout.write(newline) else: fout.write(line) fin.close() fout.close() os.rename(newctrl, self.ctrlname) def get_dipole_moment(self, atoms): if ('dipole' not in self.parameters.get('output', []) or atoms.pbc.any()): raise PropertyNotImplementedError return FileIOCalculator.get_dipole_moment(self, atoms) def get_stress(self, atoms): if ('compute_numerical_stress' not in self.parameters and 'compute_analytical_stress' not in self.parameters): raise PropertyNotImplementedError return FileIOCalculator.get_stress(self, atoms) def get_forces(self, atoms): if ('compute_forces' not in self.parameters and 'sc_accuracy_forces' not in self.parameters): raise PropertyNotImplementedError return FileIOCalculator.get_forces(self, atoms) def read_dipole(self): "Method that reads the electric dipole moment from the output file." for line in open(self.out, 'r'): if line.rfind('Total dipole moment [eAng]') > -1: dipolemoment = np.array([float(f) for f in line.split()[6:9]]) self.results['dipole'] = dipolemoment def read_energy(self): for line in open(self.out, 'r'): if line.rfind('Total energy corrected') > -1: E0 = float(line.split()[5]) elif line.rfind('Total energy uncorrected') > -1: F = float(line.split()[5]) self.results['free_energy'] = F self.results['energy'] = E0 def read_forces(self): """Method that reads forces from the output file. If 'all' is switched on, the forces for all ionic steps in the output file will be returned, in other case only the forces for the last ionic configuration are returned.""" lines = open(self.out, 'r').readlines() forces = np.zeros([len(self.atoms), 3]) for n, line in enumerate(lines): if line.rfind('Total atomic forces') > -1: for iatom in range(len(self.atoms)): data = lines[n + iatom + 1].split() for iforce in range(3): forces[iatom, iforce] = float(data[2 + iforce]) self.results['forces'] = forces def read_stress(self): lines = open(self.out, 'r').readlines() stress = None for n, line in enumerate(lines): if (line.rfind('| Analytical stress tensor') > -1 or line.rfind('Numerical stress tensor') > -1): stress = [] for i in [n + 5, n + 6, n + 7]: data = lines[i].split() stress += [float(data[2]), float(data[3]), float(data[4])] # rearrange in 6-component form and return self.results['stress'] = np.array([stress[0], stress[4], stress[8], stress[5], stress[2], stress[1]]) def read_convergence(self): converged = False lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('Have a nice day') > -1: converged = True return converged def get_number_of_iterations(self): return self.read_number_of_iterations() def read_number_of_iterations(self): niter = None lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('| Number of self-consistency cycles') > -1: niter = int(line.split(':')[-1].strip()) return niter def get_electronic_temperature(self): return self.read_electronic_temperature() def read_electronic_temperature(self): width = None lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('Occupation type:') > -1: width = float(line.split('=')[-1].strip().split()[0]) return width def get_number_of_electrons(self): return self.read_number_of_electrons() def read_number_of_electrons(self): nelect = None lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('The structure contains') > -1: nelect = float(line.split()[-2].strip()) return nelect def get_number_of_bands(self): return self.read_number_of_bands() def read_number_of_bands(self): nband = None lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('Number of Kohn-Sham states') > -1: nband = int(line.split(':')[-1].strip()) return nband def get_k_point_weights(self): return self.read_kpts(mode='k_point_weights') def get_bz_k_points(self): raise NotImplementedError def get_ibz_k_points(self): return self.read_kpts(mode='ibz_k_points') def get_spin_polarized(self): return self.read_number_of_spins() def get_number_of_spins(self): return 1 + self.get_spin_polarized() def get_magnetic_moment(self, atoms=None): return self.read_magnetic_moment() def read_number_of_spins(self): spinpol = None lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('| Number of spin channels') > -1: spinpol = int(line.split(':')[-1].strip()) - 1 return spinpol def read_magnetic_moment(self): magmom = None if not self.get_spin_polarized(): magmom = 0.0 else: # only for spinpolarized system Magnetisation is printed for line in open(self.out, 'r').readlines(): if line.find('N_up - N_down') != -1: # last one magmom = float(line.split(':')[-1].strip()) return magmom def get_fermi_level(self): return self.read_fermi() def get_eigenvalues(self, kpt=0, spin=0): return self.read_eigenvalues(kpt, spin, 'eigenvalues') def get_occupations(self, kpt=0, spin=0): return self.read_eigenvalues(kpt, spin, 'occupations') def read_fermi(self): E_f = None lines = open(self.out, 'r').readlines() for n, line in enumerate(lines): if line.rfind('| Chemical potential (Fermi level) in eV') > -1: E_f = float(line.split(':')[-1].strip()) return E_f def read_kpts(self, mode='ibz_k_points'): """ Returns list of kpts weights or kpts coordinates. """ values = [] assert mode in ['ibz_k_points', 'k_point_weights'] lines = open(self.out, 'r').readlines() kpts = None kptsstart = None for n, line in enumerate(lines): if line.rfind('| Number of k-points') > -1: kpts = int(line.split(':')[-1].strip()) for n, line in enumerate(lines): if line.rfind('K-points in task') > -1: kptsstart = n # last occurrence of ( assert not kpts is None assert not kptsstart is None text = lines[kptsstart + 1:] values = [] for line in text[:kpts]: if mode == 'ibz_k_points': b = [float(c.strip()) for c in line.split()[4:7]] else: b = float(line.split()[-1]) values.append(b) if len(values) == 0: values = None return np.array(values) def read_eigenvalues(self, kpt=0, spin=0, mode='eigenvalues'): """ Returns list of last eigenvalues, occupations for given kpt and spin. """ values = [] assert mode in ['eigenvalues', 'occupations'] lines = open(self.out, 'r').readlines() # number of kpts kpts = None for n, line in enumerate(lines): if line.rfind('| Number of k-points') > -1: kpts = int(line.split(':')[-1].strip()) break assert not kpts is None assert kpt + 1 <= kpts # find last (eigenvalues) eigvalstart = None for n, line in enumerate(lines): # eigenvalues come after Preliminary charge convergence reached if line.rfind('Preliminary charge convergence reached') > -1: eigvalstart = n break assert not eigvalstart is None lines = lines[eigvalstart:] for n, line in enumerate(lines): if line.rfind('Writing Kohn-Sham eigenvalues') > -1: eigvalstart = n break assert not eigvalstart is None text = lines[eigvalstart + 1:] # remove first 1 line # find the requested k-point nbands = self.read_number_of_bands() sppol = self.get_spin_polarized() beg = ((nbands + 4 + int(sppol) * 1) * kpt * (sppol + 1) + 3 + sppol * 2 + kpt * sppol) if self.get_spin_polarized(): if spin == 0: beg = beg end = beg + nbands else: beg = beg + nbands + 5 end = beg + nbands else: end = beg + nbands values = [] for line in text[beg:end]: # aims prints stars for large values ... line = line.replace('**************', ' 10000') line = line.replace('***************', ' 10000') line = line.replace('****************', ' 10000') b = [float(c.strip()) for c in line.split()[1:]] values.append(b) if mode == 'eigenvalues': values = [Hartree * v[1] for v in values] else: values = [v[0] for v in values] if len(values) == 0: values = None return np.array(values)
[docs]class AimsCube: "Object to ensure the output of cube files, can be attached to Aims object" def __init__(self, origin=(0, 0, 0), edges=[(0.1, 0.0, 0.0), (0.0, 0.1, 0.0), (0.0, 0.0, 0.1)], points=(50, 50, 50), plots=None): """parameters: origin, edges, points: Same as in the FHI-aims output plots: what to print, same names as in FHI-aims """ = 'AimsCube' self.origin = origin self.edges = edges self.points = points self.plots = plots def ncubes(self): """returns the number of cube files to output """ if self.plots: number = len(self.plots) else: number = 0 return number def set(self, **kwargs): """ set any of the parameters ... """ # NOT IMPLEMENTED AT THE MOMENT! def move_to_base_name(self, basename): """ when output tracking is on or the base namem is not standard, this routine will rename add the base to the cube file output for easier tracking """ for plot in self.plots: found = False cube = plot.split() if (cube[0] == 'total_density' or cube[0] == 'spin_density' or cube[0] == 'delta_density'): found = True old_name = cube[0] + '.cube' new_name = basename + '.' + old_name if cube[0] == 'eigenstate' or cube[0] == 'eigenstate_density': found = True state = int(cube[1]) s_state = cube[1] for i in [10, 100, 1000, 10000]: if state < i: s_state = '0' + s_state old_name = cube[0] + '_' + s_state + '_spin_1.cube' new_name = basename + '.' + old_name if found: os.system('mv ' + old_name + ' ' + new_name) def add_plot(self, name): """ in case you forgot one ... """ self.plots += [name] def write(self, file): """ write the necessary output to the already opened """ file.write('output cube ' + self.plots[0] + '\n') file.write(' cube origin ') for ival in self.origin: file.write(str(ival) + ' ') file.write('\n') for i in range(3): file.write(' cube edge ' + str(self.points[i]) + ' ') for ival in self.edges[i]: file.write(str(ival) + ' ') file.write('\n') if self.ncubes() > 1: for i in range(self.ncubes() - 1): file.write('output cube ' + self.plots[i + 1] + '\n')