Photoemission Orbital Tomography

• Photoemission orbital tomography is a technique that maps molecular orbitals by interpreting angular photoelectron distributions as momentum-space fingerprints.
• It employs Fourier transforms, phase retrieval, and iterative algorithms to reconstruct both 2D and 3D orbital images from spectral data.
• The method benchmarks electronic structure theories, resolves excitonic dynamics, and reveals hybridization and structural effects at organic/metal interfaces.

Photoemission orbital tomography (POT) is an angle-resolved photoemission spectroscopy technique in which the angular distribution of photoelectrons, measured as a function of binding energy, is interpreted as the momentum-space signature of molecular orbitals. In the established formulation for well-oriented organic layers, the method yields an orbital-resolved picture of the electronic density of states at organic/metal interfaces; in later developments it was extended to direct real-space reconstruction, time-resolved access to transiently occupied states, sparse-data three-dimensional reconstruction, and excitonic wave functions (Haags et al., 2022, Bennecke et al., 25 Feb 2025, Wallauer et al., 2020, Kern et al., 2023). Although frontier orbitals are not quantum-mechanical observables in a strict sense, POT provides experimentally accessible momentum-space images that can be related to orbital-like quantities under controlled approximations (Wallauer et al., 2020).

1. Physical basis and photoemission image formation

The central approximation in POT is the plane wave approximation for the photoemission final state. In that limit, the measured intensity for orbital $i$ is written as

$$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$

so that a momentum map or $k$-map corresponds to the squared modulus of the Fourier transform of the initial-state orbital, modulated by a polarization factor (Haags et al., 2022, Brandstetter et al., 2020). This approximation has been emphasized as particularly suitable for $\pi$-orbitals in large, planar molecules composed of light atoms, and it was later shown to be well fulfilled also for $\sigma$-orbitals over a broad energy range (Brandstetter et al., 2020, Haags et al., 2021).

A momentum map is therefore not a generic angular pattern but an orbital fingerprint. Distinct nodal structures, rings, and lobe arrangements encode the symmetry and spatial character of the emitting state. For extended or dispersive overlayers, the same logic applies to Bloch-like states: the photoemission intensity behaves essentially like that of the isolated molecule, modulated by the band dispersion due to intermolecular interactions (Ules et al., 2015).

The matrix-element structure remains essential. In the broader ARPES context, the transition amplitude is written as

$$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$

and the observed intensity patterns depend sensitively on light polarization, beam orientation, photon energy, and the symmetry of the initial state (Wang et al., 2012). A recurrent point in the literature is that asymmetries in the measured intensity are not incidental artifacts: they can contain decisive orbital information, and symmetrization can destroy that information (Wang et al., 2012). This is one of the main reasons POT analyses retain the full unsymmetrized momentum distribution.

2. Reconstruction strategies and computational methodology

Two methodological lines coexist in POT. One line uses forward simulation: density-functional wave functions are Fourier transformed, polarization factors are applied, and the resulting momentum maps are compared directly to experiment. The other line treats POT as an inverse problem and reconstructs real-space orbitals or orbital coefficients from the measured momentum data.

For forward simulation and fitting, kMap.py provides a widely used workflow. It computes simulated momentum maps numerically from a fast Fourier transform of real-space molecular orbital distributions, allows variation of final-state kinetic energy, molecular orientation, and polarization state, and supports direct visual comparison or automatic optimization against experiment (Brandstetter et al., 2020). In this framework, molecular orientation is typically parameterized by Euler angles, while orbital deconvolution is formulated as a least-squares problem for the energy-dependent weights of several candidate orbitals (Brandstetter et al., 2020).

For direct inversion, iterative phase retrieval algorithms were adapted from coherent diffraction imaging. A detailed formulation includes background subtraction by maximization of mutual information, followed by hybrid input-output and error-reduction cycles with shrinkwrap support updates (Kliuiev et al., 2018). An important refinement is a two-step procedure using a tight-centered object support, which removes translational ambiguity and yields artifact-free centered reconstructions (Kliuiev et al., 2018). The resulting two-dimensional orbital image can be interpreted as a superposition of an in-focus distribution at $z=0$ and out-of-focus distributions at other $z=\mathrm{const}$ planes; this suggests that full three-dimensional reconstruction is possible when the axial resolution is sufficiently high (Kliuiev et al., 2018).

A different inversion strategy replaces unconstrained phase retrieval by a tight-binding or linear-combination-of-atomic-orbitals ansatz. In this approach, the photoemission intensity is written in terms of orbital coefficients and an operator $F_{\mathbf{k}}$, which naturally incorporates incoherent superpositions from multiple molecular orientations. This enables reconstruction of three-dimensional orbitals from a single two-dimensional PMM even for multi-orientation systems, and it also permits simultaneous optimization of molecular structure and orbital coefficients (Nozaki et al., 2023). The associated PhaseLift formulation promotes positive-semidefinite, rank-one solutions and remains usable in the presence of experimental or theoretical uncertainties (Nozaki et al., 2023).

Robust sparse PhaseLift pushes this logic further by explicitly providing atomic positions and basis functions, reconstructing three-dimensional orbital phases from a single PMM, and discriminating adsorption-induced molecular deformations with an accuracy of $0.05$ $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$0 (Niki et al., 2023). A complementary minimalist iterative projection algorithm reformulates three-dimensional POT as cyclic projections onto measurement, support, sparsity, symmetry, and low-pass constraint sets. For pentacene frontier orbitals, this strategy demonstrated full orbital reconstruction from only four simulated photoemission momentum measurements, without interpolation of densely sampled three-dimensional datasets (Dinh et al., 2024).

3. Orbital assignment, pDOS extraction, and benchmarking of electronic structure theory

A major achievement of static POT is the conversion of momentum-resolved photoemission data into orbital-resolved partial densities of states. In bisanthene on Cu(110), POT identified an unprecedented number of $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$1 $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$2 and $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$3 $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$4 orbitals and used momentum-map deconvolution to extract the corresponding experimental orbital-projected density of states (Haags et al., 2022). Subsequent work broadened the accessible window further: $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$5 $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$6-orbitals and $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$7 $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$8-orbitals were identified in a wide valence-energy range for chemical analysis (Haags et al., 2021), and a later study reported a comprehensive experimental identification of $$I_i(k_x,k_y) \propto |\mathbf{A}\cdot \mathbf{k}|^2 \, |\tilde{\Psi}_i(\mathbf{k})|^2,$$9 $k$0 and $k$1 $k$2 orbitals, i.e. $k$3 non-degenerate molecular orbitals, over a binding-energy range larger than $k$4 eV (Haags et al., 9 Jan 2025).

The underlying decomposition is typically written as

$k$5

with the weights $k$6 representing orbital-resolved pDOS (Haags et al., 2022, Haags et al., 9 Jan 2025). Because the momentum fingerprints of different orbitals are highly distinctive, the fit can remain informative even in spectral regions with strong overlap.

This orbital resolution turns POT into a stringent benchmark for theory. For bisanthene/Cu(110), four exchange-correlation functionals from GGA, global hybrid, and range-separated hybrid families were compared orbital by orbital. In that benchmark, the overlap of simulated orbitals from different functionals exceeded $k$7, indicating that pDOS extraction is not very sensitive to the functional used for the momentum maps themselves, while the energy alignment is strongly functional dependent (Haags et al., 2022). HSE gave the best agreement with experiment for both $k$8 and $k$9 states, with errors $\pi$0 eV in one benchmark and mean absolute errors well below $\pi$1 eV for most orbitals in a later wide-range study (Haags et al., 2022, Haags et al., 9 Jan 2025).

The same benchmarking role appears in first-principles spectral theories. Koopmans-compliant functionals were shown to produce molecular photoemission spectra and momentum maps of Dyson orbitals in excellent agreement with experimental ultraviolet photoemission spectroscopy and orbital tomography data, and KIPZ was highlighted as particularly important for correct orbital ordering and momentum maps (Nguyen et al., 2014). In the bisanthene case, the remarkable agreement between experimental and Kohn-Sham orbital energies over more than $\pi$2 eV was taken to suggest that Kohn-Sham orbitals can approximate Dyson orbitals much better than previously thought (Haags et al., 9 Jan 2025). This suggests that POT is not merely an assignment tool but a direct test of how electronic-structure approximations represent charged excitations.

4. Ultrafast POT and the separation of coherent and incoherent pathways

Time-resolved photoemission orbital tomography (tr-POT) extends the static framework to transiently occupied unoccupied states. By combining femtosecond pump-probe spectroscopy with momentum microscopy, the method records a four-dimensional dataset $\pi$3, allowing the momentum-space distribution of excited electrons to be followed in time (Wallauer et al., 2020). In the first implementation on PTCDA/Cu(001)-2O, a $\pi$4 eV pump pulse and a $\pi$5 eV high-harmonic probe pulse were used to image the LUMO in momentum space, establish a LUMO lifetime $\pi$6 fs, and show that the orbital pattern remained constant during decay (Wallauer et al., 2020).

The same experiment demonstrated that excitation pathways can be separated by momentum mapping and polarization control. By resolving contributions from differently oriented molecules and varying the excitation geometry, it distinguished direct HOMO$\pi$7LUMO excitation from substrate-mediated population pathways. Under some conditions, the population transfer displayed a fast maximum at $\pi$8 fs, attributed to persistent coherent HOMO-LUMO polarization, while a four-level optical Bloch analysis yielded a very slow decoherence $\pi$9 fs for the molecule-internal channel (Wallauer et al., 2020).

A later tr-POT study on CuPc/Cu(001)-2O established in a more explicit way how coherent and incoherent excitation pathways appear in the momentum patterns themselves. At early time delays, during pump-probe overlap, a coherent two-photon photoemission contribution from the projected HOMO dominates; at later delays, the incoherent population of the LUMO dominates (Adamkiewicz et al., 2023). The measured intensity was modeled as

$\sigma$0

with the temporal weights obtained from density-matrix simulations of a three-level system (Adamkiewicz et al., 2023). Variation of pump photon energy provides an especially clean discriminator: the coherent HOMO contribution shifts linearly with $\sigma$1, whereas the incoherent LUMO contribution remains at fixed kinetic energy (Adamkiewicz et al., 2023). For CuPc/Cu(001)-2O, the LUMO population lifetime extracted from optical Bloch fits was $\sigma$2 fs (Adamkiewicz et al., 2023).

tr-POT also adds orientational selectivity. With s-polarized pump light at normal incidence, the electric field lies in the molecular plane and can be aligned along one of the two possible diagonals of the CuPc molecules, producing asymmetric LUMO momentum maps and selective excitation of molecules with a specific orientation (Adamkiewicz et al., 2023). A plausible implication is that tr-POT can probe domain-selective ultrafast dynamics in systems where structural anisotropy is electronically active.

5. Hybridization, chemical specificity, and structural sensitivity

The relation between momentum-space patterns and real-space orbitals makes POT sensitive not only to electronic state assignment but also to hybridization, bonding, and structure. In pentacene monolayers, a comparison between Ag(110) and Cu(110) showed two distinct regimes: on Ag(110), the orbitals remain essentially isolated-molecule like, whereas on Cu(110) strong substrate-enhanced dispersion and orbital modification occur (Ules et al., 2015). On Cu(110), the LUMO is fully occupied, disperses by over $\sigma$3 eV, and its minor lobes shift in $\sigma$4-space in a way that corresponds to a real-space orbital expansion by approximately $\sigma$5 (Ules et al., 2015). POT thereby provided direct evidence for adsorption-induced orbital distortion.

Chemical analysis becomes especially powerful when $\sigma$6-orbitals are accessible. Whereas delocalized frontier $\sigma$7-orbitals are not always sensitive to local bond modifications at molecular peripheries, $\sigma$8-orbitals are directly associated with local bonding. In bisanthene formed by dehalogenation and cyclodehydrogenation on Cu(110), the uppermost two $\sigma$9-orbitals were sufficiently sensitive to edge termination that their momentum maps distinguished hydrogen-passivated from Cu-metalated products (Haags et al., 2021). The same study emphasized that the plane wave approximation remains well fulfilled for $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$0-orbitals and that POT can therefore serve as a detailed probe of surface chemical reactions (Haags et al., 2021).

POT also provides structural information in thicker films. For $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$1-sexithiophene on Cu(110)-p($$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$2)O, band dispersion and HOMO momentum maps were analyzed from monolayer to eight layers (Stettner et al., 6 Mar 2026). The periodicity of an intermolecular band changed with film thickness, revealing an increase of the intralayer distance between molecules, while the HOMO momentum distribution disclosed a decrease of molecular tilt angle from $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$3° at $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$4 ML to $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$5° at $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$6 ML (Stettner et al., 6 Mar 2026). The corresponding intermolecular spacing increased from $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$7 $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$8 at $$M_{if}=\langle f|\mathbf{A}\cdot \mathbf{r}|i\rangle,$$9 ML to $z=0$0 $z=0$1 at $z=0$2 ML, showing purely from electronic-structure data that the surface-templated monolayer relaxes toward the bulk crystal structure (Stettner et al., 6 Mar 2026). This suggests that POT is not confined to submonolayer orbital imaging but can trace subtle structural relaxation in multilayer organic semiconductors.

6. Three-dimensional POT and table-top implementation

Three-dimensional POT seeks reconstruction of the full orbital, rather than a projected two-dimensional distribution. Earlier direct approaches required densely sampled, well-calibrated photon-energy-dependent datasets because each photon energy probes a different hemispherical shell in momentum space (Dinh et al., 2024). The minimalist iterative projection framework changed this requirement by operating directly on sparse measured shells and enforcing measurement consistency, support, symmetry, realness, and sparsity. For pentacene frontier orbitals, the full orbital was reconstructed from only four simulated momentum measurements, and the method remained robust under intensity-calibration errors of up to $z=0$3; for the best-gap solutions, reconstruction errors were typically $z=0$4 (Dinh et al., 2024).

A table-top experimental realization followed with ultrafast momentum microscopy and a spectrally tunable high-harmonic source. In that setup, a Yb-fiber amplifier delivered $z=0$5 fs pulses at $z=0$6 kHz, high harmonics from $z=0$7 to $z=0$8 eV were generated in argon, and single harmonics were selected by a grazing-incidence off-plane diffraction-grating monochromator (Bennecke et al., 25 Feb 2025). On PTCDA/Ag(110), three-dimensional images of both HOMO and LUMO were reconstructed from between four and ten photon energies, with real-space resolution of approximately $z=0$9 $z=\mathrm{const}$0, energy resolution between $z=\mathrm{const}$1 and $z=\mathrm{const}$2 meV, and momentum resolution of about $z=\mathrm{const}$3 to $z=\mathrm{const}$4 $z=\mathrm{const}$5 (Bennecke et al., 25 Feb 2025). The same work reported that four photon energies required about eight hours of data collection, which is short enough to make time-resolved three-dimensional POT a practical target (Bennecke et al., 25 Feb 2025).

These developments separate two conceptual routes to three-dimensional information. One route uses direct phase retrieval from photon-energy-dependent data and can be simulation-free in principle (Dinh et al., 2024, Bennecke et al., 25 Feb 2025). The other uses a model space, typically LCAO or tight binding, and infers three-dimensional orbital coefficients and structure from a reduced measurement set (Nozaki et al., 2023, Niki et al., 2023). Both routes reduce the former dependence on dense synchrotron scans.

7. Excitonic and many-body extensions

Recent work has generalized POT from one-electron orbital imaging to correlated electron-hole states. For organic molecules, exciton photoemission orbital tomography extends the standard orbital picture while respecting both excitonic entanglement and energy conservation (Kern et al., 2023). In the Tamm-Dancoff form, the exciton wave function is

$z=\mathrm{const}$6

and the corresponding exPOT intensity becomes

$z=\mathrm{const}$7

The coherent sum over unoccupied orbitals preserves orbital intuition, while the incoherent sum over final hole states enforces energy conservation (Kern et al., 2023). Natural transition orbitals provide a compact interpretation of the same structure, and real-time TDDFT simulations validated the predicted photoelectron angular distributions for TCNQ, porphin, and PTCDA (Kern et al., 2023).

For periodic systems, exPOT was formulated within $z=\mathrm{const}$8+Bethe-Salpeter theory. In this case the measured signal is a coherent sum of Fourier-transformed conduction-band Bloch states weighted by BSE eigenvectors, so that the ARPES pattern encodes the correlated nature of the exciton rather than a simple occupation change (Kaidisch et al., 18 Nov 2025). In monolayer hexagonal boron nitride, the formalism predicted strong dependence of the photoemission pattern on pump polarization for some excitons, and it naturally extends to finite center-of-mass momentum, making it suitable for momentum-dark excitons (Kaidisch et al., 18 Nov 2025).

An experimental realization of excitonic trPOT was reported for $z=\mathrm{const}$9-sexithiophene thin films. There, femtosecond time-resolved POT directly imaged the momentum-space distribution of the exciton and reconstructed the real-space wave function, including spatial extent and internal phase structure (Theilen et al., 28 Nov 2025). The reconstructed wave function was found to be coherently delocalized across approximately three molecular units and to contract by about $F_{\mathbf{k}}$0 within $F_{\mathbf{k}}$1 fs, which was interpreted as self-trapping driven by exciton-phonon coupling (Theilen et al., 28 Nov 2025). This shifts POT from orbital imaging toward direct access to correlated many-body wave functions.

A related extension uses angle-resolved resonant photoemission to access local spin and orbital character in itinerant ferromagnets. In bcc Fe, resonant diffraction patterns were shown to provide real-space imaging of pure spin-flip and entangled spin-flip–orbital-flip excitations, suggesting a route toward magnetic tomography with element and site specificity (Pieve, 2015). This suggests that the conceptual scope of photoemission tomography now includes not only occupied and unoccupied molecular orbitals, but also excitons and local spin-orbital excitations in correlated materials.