Polarization-induced renormalization of molecular levels at metallic and semiconducting surfaces

J. M. García‐Lastra, C. Rostgaard, Ángel Rubio, Kristian S. Thygesen

TL;DR

This paper investigates how the electronic energy levels of a benzene molecule change when it sits on different metal and semiconductor surfaces. Using advanced quantum mechanical calculations, they found that the molecule's energy gap shrinks due to the surface's electrical response, and this effect can be predicted by understanding the surface's electronic properties. This work is crucial for designing future molecular electronic devices.

What Problem Does This Paper Address?

Understanding how molecules behave when placed on surfaces is fundamental to fields like catalysis, molecular electronics, and sensors. A key aspect is how the electronic energy levels of the molecule align with those of the surface. Traditional methods, like Density Functional Theory (DFT) in its common Kohn-Sham formulation, treat the interaction as a static, mean-field effect. However, experiments show that the dynamic electrical response of the surface (polarization) significantly alters these energy levels, particularly the gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO). This "polarization-induced renormalization" is not captured by standard DFT. This paper aims to provide a microscopic understanding of this phenomenon and how it depends on the nature of the surface.

Methods

The researchers employed a combination of theoretical techniques:

  • First-principles ${G}_{0}{W}_{0}$ calculations: This is a sophisticated many-body perturbation theory method that goes beyond standard DFT to accurately describe excited-state properties, like electron addition and removal energies (quasiparticle energies). They used this to calculate the electronic structure of a benzene molecule physisorbed on various metallic and semiconducting surfaces.
  • DFT calculations (LDA and PBE0): As a baseline and for comparison, they also performed calculations using the Local Density Approximation (LDA) and the PBE0 hybrid functional.
  • Model GW calculations: To gain deeper insight and test their findings in a simplified setting, they developed and used model Hamiltonians representing metal-molecule and semiconductor-molecule interfaces. These models allowed them to systematically vary parameters like interaction strength and substrate electronic properties.
  • Classical Image Charge Model: They also tested how well a simplified classical model, which describes the interaction of a charge with its "image" in a polarizable medium, could explain the observed level shifts.

The ${G}_{0}{W}_{0}$ approach is standard for high-accuracy electronic structure calculations, but applying it to complex surface-molecule systems is computationally demanding. The model GW calculations and the microscopic theory developed in this paper offer a more interpretable framework.

Datasets

Not discussed in the paper. The study uses theoretical models and calculations based on the electronic structure of various materials (benzene, Li, Al, Ti, Rh, Pt, TiO2, BaO, MgO, CaO, NaCl, Graphite).

Key Insights and Ideas

  • Dynamic Polarization is Key: The dynamic electrical response of the substrate (polarization) is crucial for accurately describing molecular energy levels at interfaces, and this is not captured by standard DFT.
  • Image Charge Effect: The primary mechanism for this renormalization is the "image charge" effect, where the molecule's charge state induces a polarization in the substrate, which in turn interacts back with the molecule.
  • Surface Properties Dictate Renormalization: The extent of the energy gap reduction scales predictably with the substrate's electronic structure:
  • For metals, it scales with the density of states at the Fermi level.
  • For semiconductors, it scales with the band gap.
  • Microscopic Explanation: The paper derives a theoretical framework that links the level shifts directly to the substrate's joint density of states and Coulomb interaction matrix elements, providing a microscopic understanding of the observed trends.
  • DFT Error Cancellation: Surprisingly, standard DFT (LDA and PBE0) can yield results for occupied molecular levels that are in good agreement with the more accurate ${G}_{0}{W}_{0}$ results, even though DFT misses the polarization effects. This is attributed to a fortuitous cancellation of errors: DFT overestimates occupied levels due to self-interaction errors, and this is compensated by the missing image charge correction.

Key Results

  • HOMO-LUMO Gap Reduction: The HOMO-LUMO gap of benzene is consistently reduced upon physisorption, with the magnitude of the reduction varying significantly depending on the substrate.
  • Classical Model Agreement: A classical image charge model can qualitatively capture the dependence of the gap on distance and substrate properties, but its parameters are sensitive to microscopic details.
  • Scaling Trends Verified: The derived theoretical scaling relations (gap reduction with substrate DOS for metals and band gap for semiconductors) are confirmed by both the first-principles ${G}_{0}{W}_{0}$ calculations and the model GW calculations.
  • Good Agreement for Occupied Levels: LDA and PBE0 Kohn-Sham eigenvalues for the occupied molecular orbitals of benzene show remarkably good agreement with the ${G}_{0}{W}_{0}$ quasiparticle energies, especially on metallic surfaces, due to error cancellation.

Limitations and Future Work

  • Physisorption Focus: The study primarily focuses on physisorption, where the molecule is weakly bound to the surface. Chemisorption, involving stronger chemical bonds and significant charge transfer, would likely exhibit different polarization effects.
  • Benzene Specificity: While benzene is a common test case, the specific details of the molecular orbitals and their interactions might differ for other molecules.
  • Model Simplifications: The model GW calculations involve simplifications (e.g., neglecting hybridization, using simplified interaction terms) that, while useful for understanding trends, might not capture all nuances of real systems.
  • Future Work: The authors suggest that their findings have significant implications for designing molecular electronic devices, where precise control over energy level alignment is critical. Further work could explore chemisorption, different molecular systems, and the impact of surface defects or roughness on polarization effects. The derived microscopic theory can be a valuable tool for predicting and understanding level shifts in a wide range of surface-molecule systems.

Summary updated

Abstract

On the basis of first-principles ${G}_{0}{W}_{0}$ calculations we systematically study how the electronic levels of a benzene molecule are renormalized by substrate polarization when physisorbed on different metallic and semiconducting surfaces. The polarization-induced reduction in the energy gap between occupied and unoccupied molecular levels is found to scale with the substrate density of states at the Fermi level (for metals) and substrate band gap (for semiconductors). These conclusions are further supported by self-consistent GW calculations on simple lattice models. By expressing the electron self-energy in terms of the substrate's joint density of states we relate the level shift to the surface electronic structure, thus providing a microscopic explanation of the trends in the GW and ${G}_{0}{W}_{0}$ calculations. While image charge effects are not captured by semilocal and hybrid exchange-correlation functionals, we find that error cancellations lead to remarkably good agreement between the ${G}_{0}{W}_{0}$ and Kohn-Sham energies for the occupied orbitals of the adsorbed molecule.
BibTeX
@inproceedings{garcialastra2009polarization,
  author = {J. M. García‐Lastra and C. Rostgaard and Ángel Rubio and Kristian S. Thygesen},
  title = {{Polarization-induced renormalization of molecular levels at metallic and semiconducting surfaces}},
  booktitle = {Physical},
  year = {2009},
  doi = {10.1103/physrevb.80.245427},
}
Figure