ARPES simulation (PAOFLOW.arpes)#

PAOFLOW.arpes simulates angle-resolved photoemission from the PAO Hamiltonian: dipole matrix elements of the PAO Bloch states with a plane-wave final state, the spectral function with an optional band- and k-resolved self-energy (electron-phonon, from PAOFLOW.elphon.self_energy), the Fermi factor and the experimental resolution. The formalism follows chinook (R. P. Day, B. Zwartsenberg, I. S. Elfimov and A. Damascelli, npj Quantum Materials 4, 54 (2019), MIT licence; see THIRD_PARTY_NOTICES.md), re-implemented in the real-harmonic PAO basis and vectorised over k-points and bands. Please cite chinook when publishing results obtained with this module. The design note is arpes-chinook-plan.md.

Workflow#

from PAOFLOW import PAOFLOW
from PAOFLOW.arpes import ArpesExperiment, ConstantSelfEnergy

pf = PAOFLOW.PAOFLOW(savedir='si.save', outputdir='output')
pf.projections(basispath='BASIS_PS', configuration='standard')
pf.projectability(pthr=0.95)
pf.pao_hamiltonian()

experiment = ArpesExperiment(
    photon_ev=21.2, work_function_ev=4.6, inner_potential_ev=12.0, temperature_k=30.0,
    polarization=(1.0, 0.0, 0.0), energy=(-3.0, 0.2, 400),
    kx=(-1.5, 1.5, 151), ky=(-1.5, 1.5, 151), resolution=(0.02, 0.01),
)
arpes_map = pf.arpes(experiment, self_energy=ConstantSelfEnergy(0.03))
arpes_map.save('output/arpes.npz')
pf.plotter().plot_arpes(arpes_map, 'arpes_band_map.png', cut='band_map', value=0.0)

ArpesExperiment fields (eV, 1/Å, radians): photon_ev, work_function_ev, inner_potential_ev (free-electron \(k_z\) from the photon energy; None keeps kz), temperature_k, polarization (Cartesian in the sample frame, complex for circular light), energy=(min, max, n), kx/ky=(min, max, n) or an explicit kpoints (N, 3) (a path), azimuth, resolution=(dE, dk) FWHM, escape_depth (slab depths), radial ('pao', 'slater', 'hydrogenic' or a RadialIntegrals), phase_shifts, fixed_radial, position_phase, kz_at_binding_energy, window_margin_ev, kblock.

The result, IntensityMap, holds kpoints, w, intensity (broadened), raw, the window bands with their energies and matrix-element weights, the mesh grid/axes, and the settings in metadata; as_grid(), energy_distribution_curve(ik), momentum_distribution_curve(E), save(path).

Formalism#

For \(|n\mathbf k\rangle = \sum_i c_{i,n\mathbf k}|\phi_i,\mathbf k\rangle\) (cell-periodic gauge, BlochHamiltonian: \(H(\mathbf k) = \sum_n W_n H(R_n)e^{-2\pi i\mathbf k\cdot n}\) on the Wigner–Seitz images of the FFT cells) and a photoelectron \(\mathbf k_f\) with \(|\mathbf k_f|^2 = 2m(h\nu + \varepsilon_{n\mathbf k} - W)/\hbar^2\), \(\mathbf k_{f\parallel} = \mathbf k_\parallel\),

\[M^\alpha_{n\mathbf k} = \sum_i c_{i,n\mathbf k}\,\sigma_i\,e^{-d_i/2\lambda}\, e^{-i(\mathbf k_f-\mathbf k)\cdot\boldsymbol\tau_i}\,s_i\sum_{l'=l_i\pm1} B^{l'}_i(|\mathbf k_f|) \sum_{m'} G^\alpha_{l_i m_i,l'm'}\,Y_{l'm'}(\hat{\mathbf k}_f), \qquad B^{l'}_i(k) = i^{l'}\!\int_0^\infty\! dr\,\chi_i(r)\,r^2 j_{l'}(kr),\]

with \(\chi_i = rR_i\) the PAO radial table of the orbital (arrays['basis'][i]['wfc']), \(G\) the real Gaunt coefficients of the dipole operator (\(x, y, z\) are real \(l = 1\) harmonics; hamiltonian/_two_center.real_gaunt_coefficient), \(Y_{l'm'}\) the real harmonics, \(s_i\) the sign of QE’s harmonic relative to the standard one (harmonics.qe_harmonic), \(\sigma_i\) a cross-section factor and \(d_i/\lambda\) the depth over the escape depth. The gauge phase is chinook’s atomic-gauge convention made explicit; it is unity for \(\mathbf k_f = \mathbf k\) and off by default (position_phase=False). The intensity is

\[I(\mathbf k,\omega) = \sum_n\Big|\sum_\alpha\epsilon_\alpha M^\alpha_{n\mathbf k}\Big|^2 A_{n\mathbf k}(\omega)\,f(\omega,T)\otimes g_{\Delta E}\otimes g_{\Delta k},\qquad A_{n\mathbf k}(\omega) = -\frac1\pi\mathrm{Im}\frac{1}{\omega-\varepsilon_{n\mathbf k}-\Sigma_{n\mathbf k}(\omega)+i0^+}.\]

Module map: basis.py (Orbital, PhotoemissionBasis.from_data_controller, with_atomic_radial, with_depths), atomic_radial.py (Slater and hydrogenic fallbacks, ported from chinook with the physical Bohr radius and a corrected normalisation), radial.py (radial_integral, RadialIntegrals tables with phase shifts and fixed values), harmonics.py (QE order, dipole Gaunt tables), kinematics.py (\(|k_f|\), \(k_z\), meshes, emission angles), bloch.py, matrix_elements.py, spectral.py (spectral_function, assemble_intensity, ConstantSelfEnergy, GridSelfEnergy, kramers_kronig_self_energy, gaussian_broadening), experiment.py (ArpesExperiment, simulate, IntensityMap).

Self-energies#

The self-energy protocol is a constant, an array (N, nb, nw) aligned with the window bands, or a callable f(w, energies, kpoints, bands) evaluated per k-block. Sources:

  • ConstantSelfEnergy(gamma); GridSelfEnergy(w, sigma) for a tabulated \(\Sigma(\omega)\); kramers_kronig_self_energy(w, im_sigma) builds the real part of a model;

  • PAOFLOW.elphon.self_energy.eliashberg_self_energy(omega, a2f, w, T): the isotropic Engelsberg–Schrieffer form from \(\alpha^2F\) of eliashberg_dense_q (or the per-state \(\alpha^2F_{n\mathbf k}\) of the anisotropic coupling), with the kink and the mass enhancement \(\lambda = -\mathrm{Re}\,\Sigma'(0)\) (mass_enhancement);

  • PAOFLOW.elphon.self_energy.fan_migdal_self_energy(...): the band-resolved Fan–Migdal \(\Sigma_{n\mathbf k}(\omega, T)\) at arbitrary k-points (the ARPES path or mesh) from the dense-q vertex of prepare_dense_vertex, intermediate states diagonalised at every \(\mathbf k+\mathbf q\) (elph_bloch.bloch_states_at), Lorentzian width eta_ev, full dense q-grid, MPI over q; TabulatedSelfEnergy.from_result(result, kpoints) adapts it to the protocol (k-points matched by position, bands by index).

Examples#

One generic driver, arpes/arpes.py of every system folder (examples/elphon/<System>, a copy of src/PAOFLOW/gen/elphon_templates/arpes/arpes.py built on PAOFLOW.elphon.workflows.arpes), with the per-system settings in the ARPES dictionary of system.py and --self-energy {isotropic, anisotropic, fan-migdal}: Pb (metal: the kink from the isotropic α²F of the superconductivity workflow, optionally the state-resolved form with adaptive widths), MgB₂ (the σ-band E₂g kink from the anisotropic Fermi-surface coupling, one self-energy per irreducible state) and Si (semiconductor: band-resolved Fan–Migdal linewidths versus temperature along the full Γ-X line and the dressed band maps); each README has the numbers.

Slabs#

build_slab(bloch, basis, nlayers, axis) stacks nlayers cells of the bulk Hamiltonian along a lattice vector with open boundaries (hoppings from the Wigner–Seitz images of HRs, zero beyond their range) and returns the two-dimensional BlochHamiltonian of the slab and the basis of every layer with the depth below the top surface, for the escape-depth attenuation of ArpesExperiment(escape_depth=...). The stacking vector should be the surface normal (build a supercell first for a general Miller index). The self-energy of slab states is not computed by the dense-q driver; use the isotropic form or the bulk one at the projected momentum.

Metals: the state-resolved Fan–Migdal self-energy on an affordable q-grid is noisy with a fixed width far below the level spacing \(|v_F|\Delta q\) of the grid (Pb: 5 meV against 0.3 eV at 18³), and the adaptive widths (adaptive_smearing=, group velocities by finite differences at \(\mathbf k+\mathbf q\)) that smooth it are then far larger than the phonon energies; the kink of a metal is therefore resolved by the isotropic or anisotropic Eliashberg form, the band-resolved form being the tool for semiconductors and for the energy-dependent linewidths over wide windows (the Si example).

Spinor (fully relativistic) PAO bases are supported: each |l j m_j⟩ PAO function is expanded on the real harmonics and the two spin components (harmonics.spinor_components, from clebsch_jm_matrix and calc_ylmg_complex_0), the radial table of its j is used, and the matrix elements get a spin index (Orbital.components); a closed shell gives the same emission in both representations (test).

Not yet covered: the Debye–Waller term; the self-energy of slab states.

Validation#

tests/unit/arpes (26 tests): the QE harmonic order against calc_ylmg, orthonormality of the real harmonics, the dipole Gaunt sum rule \(\sum_{\alpha l'm'}G^2 = 1\) and the parity selection rule, the radial integral of hydrogen 1s against quadrature, table interpolation, phase shifts and fixed values, Slater’s rules (C 2p: 3.25, Na 3s: 2.2) and normalisation, the Bloch Hamiltonian against the FFT grid, isotropic emission of an \(s\) orbital and of a closed \(p\) shell, the gauge phase and the escape-depth attenuation, spinor components, kinematics (free-electron \(k_z\), azimuth, evanescent states), the spectral function normalisation, the intensity assembly against a literal state loop, Kramers–Kronig of a Lorentzian, the resolution broadening, mesh and path simulations with the inner potential, and the polarisation selection rule at normal emission, the slab of a cubic \(s\) band against the open chain, the spinor p shell against the scalar one, and parity with chinook on its square-lattice Sb 5p tutorial model (data/chinook_square_lattice.npz, produced by data/chinook_square_lattice_reference.py with chinook 1.1.3 in a throwaway environment): eigenvalues to 1e-8 eV and the matrix-element weights of the linear polarisations to 2e-4 (chinook rounds its hopping vectors to 1e-4 Å), up to one global constant; for circular light chinook’s spherical components label the helicities the other way round, so its (x + iy) weights equal PAOFLOW’s (x − iy) ones. tests/unit/elphon/test_self_energy.py: the Einstein-model mass enhancement and \(\mathrm{Im}\,\Sigma\) plateau at \(T = 0\) and finite \(T\), the Fan–Migdal kernel against a literal sum and its on-shell identity with the SERTA sum with a Lorentzian delta, and the tabulated adapter.