[Paper Review] Ab initio calculation of electron-phonon linewidths and molecular dynamics with electronic friction at metal surfaces with numeric atom-centered orbitals

Connor L. Box, Wojciech G. Stark+1 more|arXiv (Cornell University)|Nov 30, 2021

Advanced Chemical Physics Studies4 citations

TL;DR

This paper presents an all-electron, numeric atom-centered orbital (NAO) implementation of electron-phonon coupling and electronic friction within the FHI-aims DFT code, enabling scalable and efficient ab initio calculations of vibrational linewidths and nonadiabatic dynamics at metal surfaces. The method shows faster, more monotonic convergence than plane-wave pseudopotential approaches and reveals that many prior electron-phonon calculations may be underconverged, especially for systems like CO on Cu(100).

ABSTRACT

Molecular motion at metallic surfaces is affected by nonadiabatic effects and electron-phonon coupling. The ensuing energy dissipation and dynamical steering effects are not captured by classical molecular dynamics simulations, but can be described with the molecular dynamics with electronic friction method and linear response calculations based on density functional theory. Herein, we present an implementation of electron-phonon response based on an all-electron numeric atomic orbital description in the electronic structure code FHI-aims. After providing details of the underlying approximations and numerical considerations, we present significant scalability and performance improvements of the new code compared to a previous implementation [Phys. Rev. B 94, 115432 (2016)]. We compare convergence behaviour and results of our simulations for exemplary systems such as CO on Cu(100), H$_2$ adsorption on Cu(111), and CO on Ru(0001) against existing plane wave implementations. We find that our all-electron calculations exhibit faster and more monotonic convergence behaviour than conventional plane-wave-based electron-phonon calculations. Our findings suggest that many electron-phonon linewidth calculations in literature may be underconverged. Finally, we showcase the capabilities of the new code by studying the contribution of interband and intraband excitations to vibrational linewidth broadening of aperiodic motion in previously unfeasibly large periodic surface models.

Motivation & Objective

• To develop a scalable, all-electron ab initio method for calculating electron-phonon coupling and electronic friction using numeric atom-centered orbitals (NAO) in FHI-aims.
• To address the challenge of slow and oscillatory convergence in plane-wave pseudopotential implementations for metal-organic interfaces, particularly for systems with low-density-of-states like Cu.
• To enable large-scale simulations of vibrational linewidths and nonadiabatic dynamics in complex, aperiodic surface systems beyond the reach of conventional plane-wave codes.
• To provide a robust and efficient alternative to plane-wave-based electron-phonon calculations with improved convergence behavior and accuracy for realistic surface models.

Proposed method

• Implementation of linear response theory for electron-phonon coupling within the all-electron FHI-aims code using numeric atom-centered orbitals (NAO) and density functional theory (DFT).
• Calculation of the electronic friction tensor via the imaginary part of the frequency-dependent polarization response, using the delta function approximation with a broadening parameter σ = 0.05 eV.
• Use of the electronic friction tensor in molecular dynamics with electronic friction (MDEF) and projection onto normal mode displacements to compute vibrational linewidths.
• Employment of distributed matrix parallelism via the ELPA library for high scalability on high-performance architectures.
• Application of Wannier-based interpolation and real-space interpolation techniques to extend the method to non-periodic and aperiodic surface motions.
• Validation against plane-wave pseudopotential results from Quantum ESPRESSO and experimental data for benchmark systems such as CO/Cu(100), H₂/Cu(111), and CO/Ru(0001).

Experimental results

Research questions

• RQ1 How does the convergence behavior of electron-phonon linewidths differ between all-electron NAO and plane-wave pseudopotential implementations for metal-organic surface systems?
• RQ2 To what extent do existing plane-wave-based electron-phonon calculations suffer from insufficient convergence, particularly for systems like CO on Cu(100)?
• RQ3 Can the NAO-based implementation in FHI-aims achieve accurate and efficient calculation of vibrational linewidths and electronic friction in large, complex surface unit cells?
• RQ4 What is the relative contribution of interband and intraband excitations to vibrational linewidth broadening in aperiodic surface motion?
• RQ5 Why do discrepancies persist between NAO and plane-wave results in certain components of the electronic friction tensor, particularly those perpendicular to the surface?

Key findings

• The all-electron NAO implementation in FHI-aims exhibits significantly faster and more monotonic convergence of vibrational linewidths with respect to substrate layer thickness compared to plane-wave pseudopotential methods.
• For CO on Cu(100), convergence of electronic friction components requires up to 20 substrate layers, indicating that many prior calculations may be underconverged due to insufficient layer thickness.
• Discrepancies between NAO and plane-wave results persist for components of the electronic friction tensor perpendicular to the surface, likely due to differences in the treatment of the electronic response in pseudopotential approximations.
• The NAO-based method achieves quantitative agreement with literature and experimental data for vibrational lifetimes when averaging over inter- and intraband contributions in the 'instant-dephasing' limit.
• The new implementation enables the study of vibrational linewidths in aperiodic surface motion within periodic supercells of unprecedented size, previously unfeasible with plane-wave codes.
• The method is highly scalable and memory-efficient, making it suitable for large-scale ab initio simulations of nonadiabatic dynamics and machine-learning training data generation.