Accelerating $GW$-Based Energy Level Alignment Calculations for Molecule-Metal Interfaces Using a Substrate Screening Approach

2019 
The physics of electronic energy level alignment at interfaces formed between molecules and metals can in general be accurately captured by the \emph{ab initio} $GW$ approach. However, the computational cost of such $GW$ calculations for typical interfaces is significant, given their large system size and chemical complexity. In the past, approximate self-energy corrections, such as those constructed from image-charge models together with gas-phase molecular level corrections, have been used to compute level alignment with good accuracy. However, these approaches often neglect dynamical effects of the polarizability and require the definition of an image plane. In this work, we propose a new approximation to enable more efficient $GW$-quality calculations of interfaces, where we greatly simplify the calculation of the non-interacting polarizability, a primary bottleneck for large heterogeneous systems. This is achieved by first computing the non-interacting polarizability of each individual component of the interface, e.g., the molecule and the metal, without the use of large supercells; and then using folding and spatial truncation techniques to efficiently combine these quantities. Overall this approach significantly reduces the computational cost for conventional $GW$ calculations of level alignment without sacrificing the accuracy. Moreover, this approach captures both dynamical and nonlocal polarization effects without the need to invoke a classical image-charge expression or to define an image plane. We demonstrate our approach by considering a model system of benzene at relatively low coverage on aluminum (111) surface. Although developed for such interfaces, the method can be readily extended to other heterogeneous interfaces.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    56
    References
    0
    Citations
    NaN
    KQI
    []