Memory-dependent electronic friction for nonadiabatic dynamics at metal surfaces

arXiv:2608.12572 · cond-mat.mtrl-sci, physics.chem-ph · Submitted 2026-08-12 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "Memory-dependent electronic friction for nonadiabatic dynamics at metal surfaces".

Mira: Here's the gist of the paper. Electronic excitation induced by nuclear motion is a key energy dissipation channel in chemical dynamics at metal surfaces Here, nonadiabatic effects can be treated via molecular dynamics with electronic friction, where they act as frictional drag and fluctuation force contributions A theoretical formalism is presented to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction Importantly, memory-dependent evaluation of electronic friction removes the need to define a single effective Markovian friction coefficient from the structured frequency-dependent electronic response Theoretical Framework for Memory Effects The paper introduces a theoretical formalism to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles.

Kai: This approach builds on mixed quantum-classical formalisms to derive a consistent derivation based on a local linear-response formulation where the frequency-dependent friction kernel is evaluated from electronic-structure quantities at a single nuclear configuration.

Mira: The memory kernel in the time domain is expressed as Kµν(t − t′; x(t)) = Λµν(t − t′; x(t))θ(t − t′) and in the frequency domain as Kµν(ω; x) = Λµν(ω; x)/2 = -π/2ω ReΓµν(ω; x).

Kai: The spectral dissipation function Λµν is determined by Γµν, the interaction-weighted electron-phonon density of states.

Paper summary: Kai: We've just touched on how this paper tackles energy dissipation in chemical dynamics on metal surfaces. The core idea they are pushing is that when we look at nonadiabatic effects, treating electronic friction through molecular dynamics needs to account for memory.

Mira: They claim they developed a theoretical formalism to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles, moving away from the standard Markovian approach where memory is simply ignored.

Kai: What this means is that instead of assuming the response is instantaneous, they are showing how it depends on the past configurations and the strength of the excitation, which shows up in these memory kernels.

Lev: From a hardware standpoint, if we're trying to build a system where we measure these dynamics, this formalism tells us what kind of complexity we are facing when modeling that friction.

Mira: They applied this formalism by looking at models like Newns Anderson Hamiltonian and also within Kohn Sham density functional theory. This lets them test the idea against different levels of theoretical sophistication.

Kai: The key finding they highlight is that memory effects can actually enhance vibrational energy loss and reduce translational energy loss for nitric oxide scattering on gold one eleven, which points to a stronger coupling than expected from simpler models.

Mira: They also used classical path approximations to show that this memory-dependent frictional energy dissipation is stronger than the strictly pseudo spectral density Markovian friction for both high and low initial states in the GHM model.

Lev: So, it’s telling us that if you are running simulations on real hardware, you need to consider these memory terms because they can significantly alter the energy budget of the system.

Kai: This isn't just a small correction; it alters how we partition electronic energy loss and increases mode anisotropy between molecular degrees of freedom in their analysis.

Mira: They also used full-dimensional, tensorial electronic friction kernels calculated from Kohn Sham density functional theory to revisit the NO on Au one eleven scattering problem, which is a big validation step.

Lev: The limitation they point out is that the Markov approximation doesn't reproduce this non-Markovian behavior when comparing different Markovian approximations, which means we aren't at a simple fixed point yet.

Kai: This suggests that while the Markov approximation might be useful for some applications, it’s not sufficient for studying hyperthermal scattering dynamics.

Mira: It also opens the door to thinking that memory effects could influence temperature dependence of electron-phonon coupling constants across a broad range of materials.

Conclusion: Kai: So looking at the full picture, this work by Xuexun Lu, Connor L. Box, Nils Hertl, etc., on memory-dependent electronic friction really shows that the way we model energy dissipation in chemical dynamics on metal surfaces is more nuanced than we thought.

Mira: The title itself points to the main point: that memory effects matter when you're dealing with nonadiabatic dynamics at metal surfaces, and they provide a first-principles method to calculate these complex friction terms.

Kai: In simple terms, what it means for the listener is that if you are modeling how molecules interact with metals, you can't treat the electronic friction as a simple constant drag force. You have to account for its history and its configuration dependence.

Mira: The authors are giving us a tool to move toward more realistic simulations where we don't just rely on the Markov approximation when studying things like NO scattering on gold one eleven.

Kai: It suggests that these effects aren't just noise; they actively shape the energy partitioning between different parts of the system, leading to different outcomes depending on what you start with.

Mira: So, for the listener, it means that if you're trying to build something precise, like a simulation of a molecule on a surface, ignoring memory is going to lead to an incomplete picture of the energy flow.

Kai: It’s about moving from a simple instantaneous drag coefficient to understanding the underlying structure of how electronic interactions evolve over time.

Mira: Ultimately, this work gives us the practical framework to incorporate these memory effects into first-principles calculations in a way that should be useful for studying many different materials and conditions.

Lev: For those of us building hardware, it means that when we design sensors or simulators, we need to anticipate that the friction isn't just a simple constant; it has this history you have to model.

Kai: It’s about getting better fidelity in our models by acknowledging this complexity instead of trying to force too much reality into a simplified Markovian framework.

Department of Chemistry, University of Warwick · Yusuf Hamied Department of Chemistry, University of Cambridge · Department of Physics, University of Warwick · University of Vienna Faculty of Physics, University of Vienna Institute for Physical Chemistry Georg-August University Göttingen Max Planck Institute for Multidisciplinary Sciences

cond-mat.mtrl-sci, physics.chem-ph

Submitted: 2026-08-12

Updated: 2026-11-08

Comments: 23 pages, 7 figures

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 88/100

The gist: Electronic excitation induced by nuclear motion is a key energy dissipation channel in chemical dynamics at metal surfaces Here…

Key concepts

Electronic Friction
This describes how electronic excitation caused by nuclear motion acts as a frictional drag force in chemical dynamics at metal surfaces. It represents the dissipation of energy from nuclear motion into electronic degrees of freedom, which is crucial for understanding nonadiabatic processes.
Memory Kernel
The memory kernel mathematically describes the time-dependent friction effect, showing how past configurations and electronic states influence the current frictional drag. Unlike simple models, these kernels capture the history of the system's interaction with its environment.
Markovian Limit
This is a simplified approximation where the system's response to nuclear motion depends only on its instantaneous configuration, ignoring past history. The paper shows that this approximation is insufficient for hyperthermal scattering, as memory effects are necessary to accurately describe the physics.

Terminology

Summary

Electronic excitation induced by nuclear motion is a key energy dissipation channel in chemical dynamics at metal surfaces Here, nonadiabatic effects can be treated via molecular dynamics with electronic friction, where they act as frictional drag and fluctuation force contributions A theoretical formalism is presented to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction Importantly, memory-dependent evaluation of electronic friction removes the need to define a single effective Markovian friction coefficient from the structured frequency-dependent electronic response

Theoretical Framework for Memory Effects

The paper introduces a theoretical formalism to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles. This approach builds on mixed quantum-classical formalisms to derive a consistent derivation based on a local linear-response formulation where the frequency-dependent friction kernel is evaluated from electronic-structure quantities at a single nuclear configuration. p. 1

Where the paper says it

In this work, we introduce a practically feasible approach to evaluate memory-dependent electronic friction kernels from first principles. Building on previously established mixed quantum-classical formalisms [13, 17, 19], we provide a consistent derivation based on a local linear-response formulation in which the frequency-dependent friction kernel can be evaluated from electronic-structure quantities at a single nuclear configuration.

The memory kernel in the time domain is expressed as Kµν(t − t′; x(t)) = Λµν(t − t′; x(t))θ(t − t′) and in the frequency domain as Kµν(ω; x) = Λµν(ω; x)/2 = -π/2ω ReΓµν(ω; x). p. 3
Where the paper says it

All quantities in either frequency space ω or time space within this section are understood to be evaluated at, and to retain their dependence on, configuration x(t). Following the NEGF formalism [19], the friction kernel in frequency domain, Kµν(ω), is expressed as: Kµν(ω ; x) := Λµν(ω ; x)/2 = − π 2ω ReΓµν(ω ; x).

The spectral dissipation function Λµν is determined by Γµν, the interaction-weighted electron-phonon density of states. p. 3
Where the paper says it

Following the NEGF formalism [19], the friction kernel in frequency domain, Kµν(ω), is expressed as: Kµν(ω ; x) := Λµν(ω ; x)/2 = − π 2ω ReΓµν(ω ; x). (17) The spectral dissipation function Λµν is determined by Γµν, the interaction-weighted electron-phonon density of states (for spinless electrons).

Modeling and Approximations

The study investigates friction kernels for Newns–Anderson Hamiltonian models as well as within Kohn–Sham density functional theory. The Markovian limit is achieved when the spectrum ωJµν(ω; x) varies negligibly over the frequency scale of nuclear motion, reducing the memory kernel to a local in time and Markovian friction tensor ηµν(x). p. 2

Where the paper says it

(7) is exact. If the spectrum ωJµν(ω; x, x ′ ) varies negligibly over the frequency scale of nuclear motion, known as the constant coupling limit, the cosine transform reduces to a Dirac delta distribution δ(t − t ′ ), such that the memory kernel becomes local in time and reduces to a Markovian friction tensor, ηµν(x(t)): Kµνt − t ′ ; x(t), x(t ′ ) −→ ηµν(x(t)) δ(t − t ′ ), (8) In this limit, the generalised Langevin equation (Eq.

In the linear-coupling regime, electronic degrees of freedom are linearly coupled to nuclear displacements by operators Mˆ µ(x(t)). p. 2
Where the paper says it

(14) By applying the Feynman-Vernon influence-functional or Keldysh nonequilibrium Green’s function (NEGF) framework [17, 19, 49] to the linearized Hamiltonian Eq. (10) in the quasi-adiabatic limit, the electronic degrees of freedom are linearly coupled to the nuclear displacements by the operators {Mˆ µ(x(t))}.

The local position based memory electronic friction kernel is illustrated as Kµν(t − t′; x(t)):= Λµν(t − t′; x(t))θ(t − t′). p. 3
Where the paper says it

Consequently, friction kernels for electronic Hamiltonian models and full ab initio calculations based on Kohn-Sham DFT will subsequently be formulated within the local linear-coupling approximation, in which the electron–nuclear coupling matrices are evaluated at a single reference configuration, while the non-Markovian character of the dynamics is retained through the frequency dependence of the resulting friction kernel. Local position-dependent friction kernel Within the electronic Hamiltonian linearisation regime discussed previously, the local position based memory electronic friction kernel for mode µ and ν is illustrated as Kµν(t − t ′ ; x(t)) := Λµν(t − t ′ ; x(t))θ(t − t ′ ) (16) where Λµν(t − t ′ ; x(t)) represents a symmetrized-in-time correlation function satisfying Λµν(t ′−t; x(t)) = Λµν(t− t ′ ; x(t)) and θ is the Heaviside step function that ensures causality.

Results and Physical Implications

The results show that memory effects can lead to an increase or a reduction of energy dissipation, depending on the configurations that are dynamically sampled. For NO on Au(111), memory effects enhance vibrational energy loss and reduce translational energy loss, further emphasising the anisotropic nature of electronic friction. p. 1

Where the paper says it

Using the classical path approximation, we quantify the effect of memory on nonadiabatic energy dissipation during scattering dynamics. In NO on Au(111), memory effects enhance vibrational energy loss and reduce translational energy loss, further emphasising the anisotropic nature of electronic friction.

The analysis using classical-path approximation shows that memory-dependent frictional energy dissipation is stronger than the strictly PSD Markovian friction for both high and low initial states in the GHM model. p. 7
Where the paper says it

This is a heuristic choice of Markovian friction tensor with a magnitude representative of the friction spectrum at finite frequencies to aid our later analysis of the effect of including memory in hyperthermal scattering dynamics. Memory friction energy dissipation via classical-path approximation In the general non-Markovian MDEF dynamics discussed in Sec.

Computational Validation

The calculation of full-dimensional, tensorial electronic friction kernels from Kohn–Sham DFT enables revisiting the NO/Au(111) scattering problem. The comparison between different Markovian approximations shows that neither can reproduce the non-Markovian behaviour. p. 1

Where the paper says it

In NO on Au(111), memory effects enhance vibrational energy loss and reduce translational energy loss, further emphasising the anisotropic nature of electronic friction. We compare against two types of Markovian approximations of the friction kernel; neither can reproduce the non-Markovian behaviour.

The CPA analysis of the full-dimensional ab initio trajectories of NO scattering on Au(111) showed that memory effects alter the partitioning of electronic energy loss and increase mode anisotropy between molecular degrees of freedom. p. 0
Where the paper says it

For hyperthermal atomic and diatomic scattering, memory effects arising from frequency and configuration dependence of electronic friction affect energy exchange between adsorbate and metal electrons. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction.

p. 0
Where the paper says it

For hyperthermal atomic and diatomic scattering, memory effects arising from frequency and configuration dependence of electronic friction affect energy exchange between adsorbate and metal electrons. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction.

The inclusion of memory friction effects may significantly increase the vibrational energy loss observed for NO/Au(111). p. 0
Where the paper says it

For hyperthermal atomic and diatomic scattering, memory effects arising from frequency and configuration dependence of electronic friction affect energy exchange between adsorbate and metal electrons. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction.

Conclusion and Future Directions

The authors conclude that while the Markov approximation to MDEF is not valid for the study of hyperthermal scattering, it remains unclear whether inclusion of memory is able to (somewhat) extend the realm of applicability of molecular dynamics with electronic friction. The findings suggest that memory effects may also affect the temperature dependence of electron-phonon coupling constants for a broad range of materials. p. 0

Where the paper says it

Associated with this are memory effects that conceptually arise from the coarse-graining of electronic degrees of freedom into a bath with a broad spectrum of frequencies [13, 30], which are coupled to the nuclear motion. The resulting non-Markovian system–bath dynamics include the dependence of the system-bath coupling on the configuration of the system, but also the variation of system-bath coupling with the excitation frequency of the electrons.

The work provides a practical route for incorporating memory effects into realistic first-principles electronic friction calculations. p. 1
Where the paper says it

[44] were able to reach better agreement with experiment, suggesting that inclusion of memory effects may be able to extend the applicability regime of electronic friction. In this work, we introduce a practically feasible approach to evaluate memory-dependent electronic friction kernels from first principles.

Pruning redundant reasoning tokens

The paper systematically develops a formalism to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles. The key finding is that the inclusion of memory effects in molecular dynamics with electronic friction can significantly alter the partitioning of electronic energy loss and increase mode anisotropy between molecular degrees of freedom. p. 0

Where the paper says it

For hyperthermal atomic and diatomic scattering, memory effects arising from frequency and configuration dependence of electronic friction affect energy exchange between adsorbate and metal electrons. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction.

How the training signal is built

The theoretical framework is based on a local linearisation of the coupling between electronic excitations and nuclear motion. The memory effect arises from the fact that the electron-nuclear response depends on the configuration of the atoms and on the strength of the perturbation, i. p. 0

Where the paper says it

Seminal experiments demonstrated nonadiabatic chemicurrents in gas-surface scattering [3], activated diffusion during inelastic electron tunelling spectroscopy of hydrogen (H) on copper [4, 5], and vibrationally and translationally inelastic scattering of nitric oxide (NO) [6, 7], carbon monoxide (CO) [8, 9] and H atoms from metal surfaces [10–12]. When the electronic excitations can be treated as a weakly coupled bath, their effect on the nuclei can be described using molecular dynamics with electronic friction (MDEF) [13–15].

e., the energy of the excited electrons. p. 1
Where the paper says it

Markovian MDEF based on ab initio data (directly or through surrogate models) has been successfully applied to simulate the translational energy loss in H atom scattering experiments from metal surfaces [11, 12, 39], laser-induced desorption [40–42], and electron-hole-pair-driven vibrational relaxation rates [16, 37, 43]. In the case of the vibrationally inelastic scattering of NO from Au(111), it was shown that Markovian MDEF significantly underestimates vibrational energy loss compared to experiment for highly excited initial vibrational states [44].

The conventional Markovian approximation corresponds to retaining only the static, zero-frequency limit of this response. p. 9
Where the paper says it

Figure 2 shows that the frequencydependent electronic friction spectra K(ω ; x) (solid) exhibit a peaked structure, where, over a wide energy range, the electronic friction first rises above the Markovian, zero-frequency limit (horizontal dashed lines) before decaying to zero. The corresponding kernel in the timedomain corresponds to underdamped system-bath dynamics.

Where the gains come from

The gains come from evaluating friction kernels for Newns–Anderson Hamiltonian models as well as within Kohn–Sham density functional theory. The analysis shows that memory effects enhance vibrational energy loss and reduce translational energy loss in the case of nitric oxide scattering on Au(111). p. 0

Where the paper says it

For hyperthermal atomic and diatomic scattering, memory effects arising from frequency and configuration dependence of electronic friction affect energy exchange between adsorbate and metal electrons. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction.

p. 0
Where the paper says it

For hyperthermal atomic and diatomic scattering, memory effects arising from frequency and configuration dependence of electronic friction affect energy exchange between adsorbate and metal electrons. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction.

The CPA analysis of the full-dimensional ab initio trajectories of NO scattering on Au(111) showed that memory effects alter the partitioning of electronic energy loss and increase mode anisotropy between molecular degrees of freedom.

The gains come from evaluating friction kernels for Newns–Anderson Hamiltonian models as well as within Kohn–Sham density functional theory. The analysis shows that memory effects enhance vibrational energy loss and reduce translational energy loss in the case of nitric oxide scattering on Au(111).

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI can perform nonadiabatic dynamics simulations by incorporating memory effects into molecular dynamics with electronic friction (MDEF) using a theoretical formalism derived from first principles for tensor and configuration-dependent friction kernels. This allows the AI to move beyond the Markov approximation, which ignores memory effects, leading to a more accurate description of energy dissipation in hyperthermal scattering.

  2. The AI system can accurately predict vibrational energy loss for NO on Au(111) by using the local CPA memory kernel and comparing results against Markovian approximations. This capability is crucial because the paper shows that memory effects enhance vibrational energy loss and reduce translational energy loss, further emphasising the anisotropic nature of electronic friction.

  3. The AI can distinguish between different Markovian approximations of the friction kernel to determine which one is most appropriate for a given physical scenario. This is achieved by analyzing how different approximations affect energy dissipation, as Markovian MDEF treatment may significantly affect nonadiabatic energy dissipation in hyperthermal scattering.

  4. The AI system can analyze the frequency-dependent electronic friction spectrum K(w; x) to identify regions where the Markovian approximation breaks down. This involves checking if the friction spectrum indicates a breakdown of strict positive semidefiniteness by looking for negative eigenvalues, which happens outside the threshold frequency omega star.

  5. The AI can calculate mode-resolved energy losses for NO/Au(111) scattering by using the memory kernel to show how dissipation is partitioned between different molecular degrees of freedom. This analysis reveals that memory effects increase mode anisotropy and significantly enhance vibrational energy loss compared to both the zero-frequency limit and the frequency-averaged Markovian case.

Abstract

Electronic excitation induced by nuclear motion is a key energy dissipation channel in chemical dynamics at metal surfaces. Here, nonadiabatic effects can be treated via molecular dynamics with electronic friction, where they act as frictional drag and fluctuation force contributions. Commonly, the Markov approximation is imposed, so memory effects are ignored. A theoretical formalism is presented to evaluate tensorial and configuration-dependent electronic friction memory kernels from first principles. We evaluate friction kernels for Newns--Anderson Hamiltonian models as well as within Kohn--Sham density functional theory and analyse their mathematical properties and configuration dependence. These friction kernels capture memory effects that arise from frequency and configuration dependence and are used to study energy loss in hyperthermal atomic and diatomic scattering studied via the classical path approximation. Memory effects lead to an increase of vibrational and a reduction of translational energy loss in the case of nitric oxide scattering on Au(111), leading to an increase of directional anisotropy of friction. Importantly, memory-dependent evaluation of electronic friction removes the need to define a single effective Markovian friction coefficient from the structured frequency-dependent electronic response.

Sources

Related papers