An ab initio approach to energy alignment and charge-state prediction of adsorbates on ultrathin insulators

Technical Summary: An ab initio approach to energy alignment and charge-state prediction of adsorbates on ultrathin insulators

Problem Statement
The manipulation of individual adsorbate spin states on ultrathin oxide layers supported by metal substrates is a critical capability for bottom-up quantum-coherent systems, such as atomic-scale qubits and qudits. Electron spin resonance (ESR) requires unpaired spin density on the adsorbate, often achieved via charge transfer from the substrate. However, predicting the final charge and spin state is challenging due to complex phenomena including electronic bandgap narrowing, charge transfer, Fermi-level pinning, and the re-ordering of adsorbate orbitals. While experimental techniques like scanning tunneling microscopy (STM) and spectroscopy (STS) have revealed these effects, a predictive ab initio framework capable of capturing them without the prohibitive computational cost of full adsorbate/oxide/metal simulations using high-level methods (such as quasiparticle GW) remains an open challenge. Existing models often fail to reconcile work function shifts observed in local probes (STM/STS, ~1 eV) with those from non-local techniques (UPS/IPES, ~0.3 eV) and struggle to accurately predict orbital re-ordering after charge transfer.

Methodology
The authors present a theoretical approach that balances computational efficiency with quantitative accuracy by separating key physical mechanisms rather than simulating the entire interface at the highest level of theory. The framework integrates Density Functional Theory (DFT) and many-body perturbation theory (GW):

1. Ionization Potentials (IP) and Electron Affinities (EA): The energy levels of isolated neutral and charged adsorbates are determined using the GW approximation, specifically the COHSEX (Coulomb-hole and screened-exchange) variant. This method relates the HOMO and LUMO positions to the cost of removing or adding an electron, respectively, providing accurate chemical potentials relative to the vacuum level.
2. Substrate Work Function and Pauli Pushback: DFT calculations within the Generalized Gradient Approximation (GGA) are used to determine the work function of the metal substrate and the reduction in work function ($\Delta\phi_{PB}$) caused by the Pauli repulsion ("push-back") when an ultrathin oxide layer is deposited.
3. Bandgap Polarization: The screening effect of the substrate on the adsorbate's electronic bandgap is quantified. The effective dielectric constant lowers the LUMO and raises the HOMO, reducing the fundamental gap ($E_g \to E'_g = E_g - 2P$). This polarization ($P$) is calculated using GW for the molecule adsorbed on the oxide layer at equilibrium height. For systems where full GW simulations of the metal/oxide/adsorbate are computationally infeasible, the authors estimate the metal's contribution to polarization by adding a constant term derived from benchmark systems (e.g., benzene on MgO/Ag).
4. Charge Transfer and Fermi-Level Pinning: The model evaluates whether the renormalized LUMO or HOMO aligns with the substrate's Fermi level. If pinning occurs, charge transfer is induced, creating an interface dipole. The resulting work function shift ($\Delta\phi$) is calculated from the local variations of the electrostatic potential at a specific probe height (e.g., 4 Å), distinguishing between local (STM) and spatially averaged (UPS) measurements.
5. Orbital Re-ordering and Reorganization: Following charge transfer, the molecular orbitals are re-organized. The model accounts for inner reorganization energy ($\lambda$) and specific orbital shifts (e.g., $d_{z^2}$ orbital shifts in transition metals due to crystal field effects) to predict the final Singly Occupied (SOMO) and Singly Unoccupied (SUMO) molecular orbital energies.

Key Results and Benchmarks
The method was benchmarked against six representative systems: Benzene/NaCl/Cu, Pentacene/MgO/Ag, FePc/MgO/Ag, PTCDA/NaCl/Ag, TCNE/MgO/Ag, and Ti/MgO/Ag.

• Benzene on NaCl/Cu: The model correctly predicts a neutral character for benzene, as the renormalized IP and EA do not approach the Fermi energy, consistent with experimental findings.
• FePc on MgO/Ag: The approach successfully predicts charge transfer from the substrate to the molecule ($[FePc]^0 \to [FePc]^{-1}$), driven by the pinning of the LUMO. It captures the re-ordering of $d$-orbitals and predicts a screened Hubbard $U$ ($2.16$ eV) and charging energy ($\epsilon \approx -1.4$ eV) that aligns reasonably well with experimental STS data, though the charge extraction barrier shows some discrepancy attributed to local potential variations.
• $\pi$-Conjugated Molecules (Pentacene, PTCDA, TCNE): The framework accurately predicts the formation of single anions for these molecules. For Pentacene, it identifies charge transfer driven by the proximity of the renormalized LUMO to the Fermi level. For PTCDA and TCNE, it correctly identifies LUMO pinning and the resulting negative charging, with predicted SOMO/SUMO energies matching experimental STS measurements.
• Ti on MgO/Ag: The study highlights a limitation of the physisorption model. While the isolated atom model predicts no charge transfer, explicit DFT simulations of the full Ti/MgO/Ag system reveal chemisorption and charge transfer ($[Ti]^0 \to [Ti]^{+1}$) due to hybridization between Ti orbitals and the oxide layer. This suggests that single-atom adsorbates may require full-system simulations beyond the physisorption picture.

Significance and Claims
The paper claims to provide a reliable, fully ab initio method for determining the charge state and frontier orbital alignment of molecular and atomic adsorbates on oxide/metal substrates. Its primary contributions are:

1. Physical Transparency: By separating mechanisms (Pauli pushback, polarization, pinning), the method offers clearer interpretation of the physical processes governing adsorption.
2. Computational Efficiency: It avoids the high cost of full interface GW simulations while retaining quantitative reliability, making it suitable for high-throughput screening of molecular qubits and organic electronic interfaces.
3. Accuracy: The use of COHSEX GW on top of GGA densities is shown to accurately predict substrate-induced polarization and the correct re-ordering of molecular orbitals, which are often underestimated by standard DFT or hybrid functionals.

The authors conclude that while standard DFT is sufficient for work functions, the GW-based approach is essential for reliable IP and EA predictions. They emphasize that their method is generally applicable but note that for specific cases like single transition metal atoms (Ti), the breakdown of the physisorption picture necessitates full-system simulations to capture chemisorption effects.