Source code for pytaser.generator

import itertools
from tqdm import tqdm
import warnings
import numpy as np
from multiprocessing import Pool, cpu_count

from pymatgen.electronic_structure.core import Spin
from pymatgen.electronic_structure.dos import FermiDos, f0
from pymatgen.ext.matproj import MPRester
from pymatgen.io.vasp import optics
from pymatgen.io.vasp.inputs import UnknownPotcarWarning
from pymatgen.io.vasp.outputs import Vasprun, Waveder

from pytaser.kpoints import get_kpoint_weights
from pytaser.tas import Tas


warnings.filterwarnings("ignore", category=RuntimeWarning)


[docs]def gaussian(x, width, center=0.0, height=None): """ Returns the values of gaussian(x) where x is array-like. Args: x: Input array. width: Width of the gaussian. center: Center of the gaussian. height: height of the gaussian. If height is None, a normalized gaussian is returned. """ x = np.asarray(x) if height is None: height = 1.0 / (width * np.sqrt(2 * np.pi)) return height * np.exp(-(((x - center) / width) ** 2) / 2.0)
[docs]def set_bandgap(bandstructure, dos, bandgap): """ Shifts all bands of a material to correct the DFT-underestimated bandgap according to the input experimental bandgap. Args: bandstructure: PMG bandstructure object dos: PMG dos object bandgap: experimental bandgap from literature (float) Returns: new bandstructure and dos objects for the same material, but with corrected bandgap. """ from copy import deepcopy if abs(dos.efermi - bandstructure.efermi) > 0.001: raise ValueError( "DOS and band structure are not from the same calculation" ) if bandstructure.is_metal(): raise ValueError("System is a metal, cannot change bandgap") scissor = bandgap - bandstructure.get_band_gap()["energy"] midgap = ( bandstructure.get_cbm()["energy"] + bandstructure.get_vbm()["energy"] ) / 2 new_bandstructure = deepcopy(bandstructure) for spin, spin_energies in bandstructure.bands.items(): vb_mask = spin_energies <= midgap new_bandstructure.bands[spin][vb_mask] -= scissor / 2 new_bandstructure.bands[spin][~vb_mask] += scissor / 2 fermi_idx = np.argmin(np.abs(dos.energies - midgap)) de = np.diff(dos.energies).mean() shift = int(scissor / 2 // de) new_dos = deepcopy(dos) for spin in dos.densities.keys(): dens = np.zeros_like(dos.energies) if shift > 0: dens[: fermi_idx - shift] = dos.densities[spin][shift:fermi_idx] dens[fermi_idx + shift :] = dos.densities[spin][fermi_idx:-shift] else: dens[abs(shift) : fermi_idx] = dos.densities[spin][ : fermi_idx + shift ] dens[fermi_idx:+shift] = dos.densities[spin][fermi_idx - shift :] new_dos.densities[spin] = dens new_bandstructure.efermi = midgap new_dos.efermi = midgap return new_bandstructure, new_dos
[docs]def jdos(bs, f, i, occs, energies, kweights, gaussian_width, spin=Spin.up): """ Args: bs: bandstructure object f: final band i: initial band occs: occupancies over all bands. energies: energy mesh (eV) kweights: k-point weights gaussian_width: width of gaussian plot. spin: Which spin channel to include. Returns: Cumulative JDOS value for a specific i->f transition, with consideration of partial occupancy and spin polarisation. """ jdos = np.zeros(len(energies)) for k in range(len(bs.bands[spin][i])): final_occ = occs[f][k] init_energy = bs.bands[spin][i][k] final_energy = bs.bands[spin][f][k] init_occ = occs[i][k] k_weight = kweights[k] factor = k_weight * (init_occ - final_occ) jdos += factor * gaussian( energies, gaussian_width, center=final_energy - init_energy ) return jdos
def _calculate_oscillator_strength(args): """Calculates the oscillator strength of a single band-band transition""" ( ib, jb, ik, occs, dfc, eigs_shifted, norm_kweights, rspin, spin, sigma, ) = args ispin = 0 if spin == Spin.up else 1 init_energy = eigs_shifted[ib, ik, ispin] final_energy = eigs_shifted[jb, ik, ispin] if final_energy > init_energy: init_occ = occs[ib][ik] final_occ = occs[jb][ik] A = ( sum( dfc.cder[ib, jb, ik, ispin, idir] * np.conjugate(dfc.cder[ib, jb, ik, ispin, idir]) for idir in range(3) ) / 3 ) decel = dfc.eigs[jb, ik, ispin] - dfc.eigs[ib, ik, ispin] matrix_el_wout_occ_factor = np.abs(A) * norm_kweights[ik] * rspin tdm = ( np.abs(A) * rspin * decel ) # kweight and occ factor already accounted for with JDOS abs_occ_factor = init_occ * (1 - final_occ) em_occ_factor = (1 - init_occ) * final_occ both_occ_factor = init_occ - final_occ abs_matrix_el = abs_occ_factor * matrix_el_wout_occ_factor em_matrix_el = em_occ_factor * matrix_el_wout_occ_factor both_matrix_el = both_occ_factor * matrix_el_wout_occ_factor smeared_wout_matrix_el = optics.get_delta( x0=decel, sigma=sigma, nx=dfc.nedos, dx=dfc.deltae, ismear=dfc.ismear, ) absorption = smeared_wout_matrix_el * abs_matrix_el emission = smeared_wout_matrix_el * em_matrix_el both = smeared_wout_matrix_el * both_matrix_el return absorption, emission, both, tdm, ib, jb, ik return 0, 0, 0, 0, 0, 0, 0
[docs]def occ_dependent_alpha( dfc, occs, spin=Spin.up, sigma=None, cshift=None, processes=None ): """Calculate the expected optical absorption given the groundstate orbital derivatives and eigenvalues (via dfc) and specified band occupancies. Templated from pymatgen.io.vasp.optics.epsilon_imag(). Args: dfc: DielectricFunctionCalculator object occs: Array of band occupancies with shape (nspin, nbands, nkpoints) spin: Which spin channel to include. sigma: Smearing width (in eV) for broadening of the dielectric function (see https://www.vasp.at/wiki/index.php/SIGMA). If not set, uses the value of SIGMA from the underlying VASP WAVEDER calculation. cshift: Complex shift in the Kramers-Kronig transformation of the dielectric function (see https://www.vasp.at/wiki/index.php/CSHIFT). If not set, uses the value of CSHIFT from the underlying VASP WAVEDER calculation. processes: Number of processes to use for multiprocessing. If not set, defaults to one less than the number of CPUs available. Returns: (alpha_dict, tdm_array) where alpha_dict is a dictionary of band-to-band absorption, stimulated emission and summed contributions to the total overall absorption coefficient under illumination in cm^-1, and tdm_array is an array of shape (nbands, nbands, nkpoints) with the transition dipole matrix elements for each band-to-band transition. """ if sigma is None: sigma = dfc.sigma if cshift is None: cshift = dfc.cshift egrid = np.arange(0, dfc.nedos * dfc.deltae, dfc.deltae) dielectric_dict = { key: np.zeros_like(egrid, dtype=np.complex128) for key in ["absorption", "emission", "both"] } # array of shape equal to cder but without the last two dimensions (ispin, idir) tdm_array = np.zeros_like(dfc.cder[:, :, :, 0, 0]) # ib, jb, ik, ispin norm_kweights = np.array(dfc.kweights) / np.sum(dfc.kweights) eigs_shifted = dfc.eigs - dfc.efermi rspin = ( 3 - dfc.cder.shape[3] ) # 2 for ISPIN = 1, 1 for ISPIN = 2 (spin-polarised) min_band0, max_band0 = np.min(np.where(dfc.cder)[0]), np.max( np.where(dfc.cder)[0] ) min_band1, max_band1 = np.min(np.where(dfc.cder)[1]), np.max( np.where(dfc.cder)[1] ) _, _, nk, _ = dfc.cder.shape[:4] iter_idx = [ range(min_band0, max_band0 + 1), range(min_band1, max_band1 + 1), range(nk), ] num_ = (max_band0 - min_band0) * (max_band1 - min_band1) * nk spin_string = "up" if spin == Spin.up else "down" light_dark_string = ( "under illumination" if any( occs[b][k] not in [0, 1] for b in range(min_band0, max_band0 + 1) for k in range(nk) ) else "dark" ) args = [] for ib, jb, ik in itertools.product(*iter_idx): args.append( ( ib, jb, ik, occs, dfc, eigs_shifted, norm_kweights, rspin, spin, sigma, ) ) if processes is None: processes = cpu_count() - 1 with Pool(processes) as pool: results = pool.map( _calculate_oscillator_strength, tqdm( args, total=num_, desc=f"Calculating oscillator strengths (spin {spin_string}, {light_dark_string})", ), ) for absorption, emission, both, tdm, ib, jb, ik in results: dielectric_dict["absorption"] += absorption dielectric_dict["emission"] += emission dielectric_dict["both"] += both tdm_array[ib, jb, ik] = tdm tdm_array = ( tdm_array.real ) # real part of A is the TDM (imag part is zero after taking the # complex conjugate) alpha_dict = {} for key, dielectric in dielectric_dict.items(): eps_in = dielectric * optics.edeps * np.pi / dfc.volume eps = optics.kramers_kronig( eps_in, nedos=dfc.nedos, deltae=dfc.deltae, cshift=cshift ) eps += 1.0 + 0.0j dielectric_dict[key] = eps # complex dielectric function # convert to alpha: n = np.sqrt(eps) # complex refractive index alpha = ( n.imag * egrid * 4 * np.pi / 1.23984212e-4 ) # absorption coefficient in cm^-1 alpha_dict[key] = alpha return alpha_dict, tdm_array
[docs]def get_cbm_vbm_index(bs): """ Args: bs: bandstructure object Returns: Valence and Conduction band as an index number for different spins. """ vbm_index = {} cbm_index = {} for spin, spin_bands in bs.bands.items(): vbm_index[spin] = np.where(np.all(spin_bands <= bs.efermi, axis=1))[ 0 ].max() cbm_index[spin] = vbm_index[spin] + 1 return vbm_index, cbm_index
[docs]class TASGenerator: """ Class to generate a TAS spectrum (decomposed and cumulative) from a bandstructure and dos object. Args: bs: Pymatgen-based bandstructure object kpoint_weights: kpoint weights either found by the function or inputted. dos: Pymatgen-based dos object dfc: Pymatgen-based DielectricFunctionCalculator object (for computing oscillator strengths) Attributes: bs: Pymatgen bandstructure object kpoint_weights: k-point weights (degeneracies). dos: Pymatgen-based dos object dfc: Pymatgen-based DielectricFunctionCalculator object (for computing oscillator strengths) bg_centre: Energy (eV) of the bandgap centre. vb: Spin dict detailing the valence band maxima. cb: Spin dict detailing the conduction band minima """ def __init__(self, bs, kpoint_weights, dos, dfc=None): self.bs = bs self.kpoint_weights = kpoint_weights self.dos = FermiDos(dos) self.bg_centre = (bs.get_cbm()["energy"] + bs.get_vbm()["energy"]) / 2 self.dfc = dfc if self.bs.is_metal(): raise ValueError("System is metallic, cannot compute TAS") self.vb = get_cbm_vbm_index(self.bs)[0] self.cb = get_cbm_vbm_index(self.bs)[1]
[docs] @classmethod def from_vasp_outputs(cls, vasprun_file, waveder_file=None): """Create a TASGenerator object from VASP output files.""" warnings.filterwarnings("ignore", category=UnknownPotcarWarning) warnings.filterwarnings( "ignore", message="No POTCAR file with matching TITEL fields" ) vr = Vasprun(vasprun_file) if waveder_file: waveder = Waveder.from_binary(waveder_file) # check if LVEL was set to True in vasprun file: if not vr.incar.get("LVEL", False): lvel_error_message = ( "LVEL must be set to True in the INCAR for the VASP optics calculation to output the full " "band-band orbital derivatives and thus allow PyTASer to parse the WAVEDER and compute oscillator " "strengths. Please rerun the VASP calculation with LVEL=True (if you use the WAVECAR from the " "previous calculation this should only require 1 or 2 electronic steps!" ) if vr.incar.get("ISYM", 2) not in [-1, 0]: isym_error_message = "ISYM must be set to 0 and " raise ValueError(isym_error_message + lvel_error_message) else: raise ValueError(lvel_error_message) dfc = optics.DielectricFunctionCalculator.from_vasp_objects( vr, waveder ) else: dfc = None return cls( vr.get_band_structure(), vr.actual_kpoints_weights, vr.complete_dos, dfc, )
[docs] def band_occupancies(self, temp, conc, dark=True): """ Gives band occupancies. Args: temp: Temperature of material we wish to investigate (affects the FD distribution) conc: Carrier concentration of holes and electrons (both are the same). Inversely proportional to pump-probe time delay. dark: Bool; dark = True indicates pump is on. Returns: A dictionary of {Spin: occs} for all bands across all k-points. """ # Calculate the quasi-Fermi levels q_fermi_e = self.dos.get_fermi( -conc, temp ) # quasi-electron fermi level q_fermi_h = self.dos.get_fermi(conc, temp) # quasi-hole fermi level occs = {} if dark: for spin, spin_bands in self.bs.bands.items(): hole_mask = spin_bands < self.bg_centre elec_mask = spin_bands > self.bg_centre # fully occupied hole mask, completely empty electron mask spin_occs = np.zeros_like(spin_bands) spin_occs[hole_mask] = 1 spin_occs[elec_mask] = 0 occs[spin] = spin_occs else: for spin, spin_bands in self.bs.bands.items(): # Calculate the occupancies at the initial and final energy according to # the Fermi-Dirac distribution electron_occs = f0(spin_bands, q_fermi_e, temp) # q-e-f hole_occs = f0(spin_bands, q_fermi_h, temp) # q-h-f hole_mask = spin_bands < self.bg_centre elec_mask = spin_bands > self.bg_centre # set occupancies equal to computed by f0 spin_occs = np.zeros_like(hole_occs) spin_occs[hole_mask] = hole_occs[hole_mask] spin_occs[elec_mask] = electron_occs[elec_mask] occs[spin] = spin_occs return occs
[docs] def generate_tas( self, temp, conc, energy_min=0, energy_max=5, gaussian_width=0.1, cshift=None, step=0.01, light_occs=None, dark_occs=None, processes=None, ): """ Generates TAS spectra based on inputted occupancies, and a specified energy mesh. If the TASGenerator has not been generated from VASP outputs (and thus does not have a dfc attribute), then the output TAS is generated using the change in joint density of states (JDOS) under illumination, with no consideration of oscillator strengths. Otherwise, the output TAS is generated considering all contributions to the predicted TAS spectrum. Args: temp: Temperature (K) of material we wish to investigate (affects the FD distribution) conc: Carrier concentration (cm^-3) of holes and electrons (both are equivalent). Inversely proportional to pump-probe time delay. energy_min: Minimum band transition energy to consider for energy mesh (eV) energy_max: Maximum band transition energy to consider for energy mesh (eV) gaussian_width: Width of gaussian curve cshift: Complex shift in the Kramers-Kronig transformation of the dielectric function (see https://www.vasp.at/wiki/index.php/CSHIFT). If not set, uses the value of CSHIFT from the underlying VASP WAVEDER calculation. (only relevant if the TASGenerator has been generated from VASP outputs) step: Interval between energy points in the energy mesh. light_occs: Optional input parameter for occupancies of material under light, otherwise automatically calculated based on input temperature (temp) and carrier concentration (conc) [dict] dark_occs: Optional input parameter for occupancies of material in dark, otherwise automatically calculated based on input temperature (temp) and carrier concentration (conc) [dict] processes: Number of processes to use for multiprocessing. If not set, defaults to one less than the number of CPUs available. Returns: TAS class containing the following inputs; - tas_total: overall deltaT TAS spectrum for a material under the specified conditions. - jdos_diff_if: JDOS difference (from dark to light) across the energy mesh for a specific band transition i (initial) -> f (final) [dict] - jdos_light_total: overall JDOS (pump-on) for a material under the specified conditions - jdos_light_if: JDOS (pump-on) across the energy mesh for a specific band transition i (initial) -> f (final) [dict] - jdos_dark_total: overall JDOS (pump-off) for a material under the specified conditions - jdos_dark_if: JDOS (pump-off) across the energy mesh for a specific band transition i (initial) -> f (final) [dict] - energy_mesh_ev: Energy mesh of spectra in eV, with an interval of 'step'. - bandgap: Bandgap of the system, in eV, rounded to 2 decimal points - temp: Temperature of the system, in K - conc: Carrier concentration of the system, in cm^-3 - alpha_dark: Absorption coefficient of the material in the dark, in cm^-1 (only calculated if the TASGenerator has been generated from VASP outputs) - alpha_light_dict: Dictionary of band-to-band absorption, stimulated emission and summed contributions to the total overall absorption coefficient under illumination, in cm^-1 (only calculated if the TASGenerator has been generated from VASP outputs) - weighted_jdos_diff_if: JDOS difference (from dark to light) across the energy mesh for a specific band transition i (initial) -> f (final), weighted by the oscillator strength of the transition [dict] - weighted_jdos_light_if: JDOS (pump-on) across the energy mesh for a specific band transition i (initial) -> f (final), weighted by the oscillator strength of the transition [dict] - weighted_jdos_dark_if: JDOS (pump-off) across the energy mesh for a specific band transition i (initial) -> f (final), weighted by the oscillator strength of the transition [dict] """ occs_light = light_occs occs_dark = dark_occs if light_occs is None: occs_light = self.band_occupancies(temp, conc, dark=False) if dark_occs is None: occs_dark = self.band_occupancies(temp, conc) bandgap = round(self.bs.get_band_gap()["energy"], 2) energy_mesh_ev = np.arange(energy_min, energy_max, step) jdos_light_if = {} jdos_dark_if = {} jdos_diff_if = {} weighted_jdos_light_if = {} weighted_jdos_dark_if = {} weighted_jdos_diff_if = {} tas_total = np.zeros(len(energy_mesh_ev)) jdos_dark_total = np.zeros(len(energy_mesh_ev)) jdos_light_total = np.zeros(len(energy_mesh_ev)) if self.dfc is not None: egrid = np.arange( 0, self.dfc.nedos * self.dfc.deltae, self.dfc.deltae ) alpha_dark = np.zeros_like(egrid, dtype=np.complex128) alpha_light_dict = { key: np.zeros_like(egrid, dtype=np.complex128) for key in ["absorption", "emission", "both"] } for spin, spin_bands in self.bs.bands.items(): if self.dfc is not None: alpha_dark_dict, tdm_array = occ_dependent_alpha( self.dfc, occs_dark[spin], sigma=gaussian_width, cshift=cshift, processes=processes, ) alpha_dark += alpha_dark_dict[ "both" ] # stimulated emission should be # zero in the dark calculated_alpha_light_dict = occ_dependent_alpha( self.dfc, occs_light[spin], sigma=gaussian_width, cshift=cshift, processes=processes, )[0] for key, array in alpha_light_dict.items(): alpha_light_dict[key] += calculated_alpha_light_dict[key] for i in range(len(spin_bands)): for f in range(len(spin_bands)): if f > i: jd_light = jdos( self.bs, f, i, occs_light[spin], energy_mesh_ev, self.kpoint_weights, gaussian_width, spin=spin, ) jd_dark = jdos( self.bs, f, i, occs_dark[spin], energy_mesh_ev, self.kpoint_weights, gaussian_width, spin=spin, ) jdos_diff = jd_light - jd_dark jdos_dark_total += jd_dark jdos_light_total += jd_light tas_total += jdos_diff new_i = i - self.vb[spin] new_f = f - self.vb[spin] if self.bs.is_spin_polarized: spin_str = "up" if spin == Spin.up else "down" key = (new_i, new_f, spin_str) else: key = (new_i, new_f) jdos_dark_if[key] = jd_dark jdos_light_if[key] = jd_light jdos_diff_if[key] = jdos_diff if self.dfc is not None: weighted_jd_light = jdos( self.bs, f, i, occs_light[spin], energy_mesh_ev, np.array(self.kpoint_weights) * tdm_array[i, f, :], gaussian_width, spin=spin, ) weighted_jd_dark = jdos( self.bs, f, i, occs_dark[spin], energy_mesh_ev, np.array(self.kpoint_weights) * tdm_array[i, f, :], gaussian_width, spin=spin, ) weighted_jdos_diff = ( weighted_jd_light - weighted_jd_dark ) weighted_jdos_light_if[key] = weighted_jd_light weighted_jdos_dark_if[key] = weighted_jd_dark weighted_jdos_diff_if[key] = weighted_jdos_diff # need to interpolate alpha arrays onto JDOS energy mesh: if self.dfc is not None: alpha_dark = np.interp(energy_mesh_ev, egrid, alpha_dark) for key, array in alpha_light_dict.items(): alpha_light_dict[key] = np.interp(energy_mesh_ev, egrid, array) tas_total = ( alpha_light_dict["absorption"] - alpha_light_dict["emission"] - alpha_dark ) return Tas( tas_total, jdos_diff_if, jdos_light_total, jdos_light_if, jdos_dark_total, jdos_dark_if, energy_mesh_ev, bandgap, temp, conc, alpha_dark if self.dfc is not None else None, alpha_light_dict if self.dfc is not None else None, weighted_jdos_light_if if self.dfc is not None else None, weighted_jdos_dark_if if self.dfc is not None else None, weighted_jdos_diff_if if self.dfc is not None else None, )
[docs] @classmethod def from_mpid(cls, mpid, bg=None, api_key=None, mpr=None): """ Import the desired bandstructure and dos objects from the legacy Materials Project database. Args: mpid: The Materials Project ID of the desired material. bg: The experimental bandgap (eV) of the material. If None, the band gap of the MP calculation will be used. api_key: The user's Materials Project API key. mpr: An MPRester object if already generated by user. Returns: A TASGenerator object. """ if mpr is None: if api_key is None: mpr = MPRester() else: mpr = MPRester(api_key=api_key) mp_dos = mpr.get_dos_by_material_id(mpid) mp_bs = mpr.get_bandstructure_by_material_id(mpid, line_mode=False) if bg is not None: mp_bs, mp_dos = set_bandgap(mp_bs, mp_dos, bg) kweights = get_kpoint_weights(mp_bs) return TASGenerator(mp_bs, kweights, mp_dos)