Asymmetry and coverage dependence in two-pulse correlation measurements of CO photodesorption from Pd(111): Insights from theory

TL;DR

This work tackles the mechanism behind CO photodesorption from Pd(111) under two-pulse laser excitation, addressing the observed asymmetry in desorption probability $P_{\rm des}$ and its coverage dependence. The authors develop a coupled theoretical framework combining a multicoverage neural-network PES for ML-based MD with two parameterizations of the two-temperature model, augmented by a temperature-dependent electronic friction derived from ab initio data. They show that including $C_e(T_e)$ and $G(T_e)$ in 2TM (2TM-2) is essential to capture the positive/negative delay asymmetry, and that a $T_e$-dependent friction improves the zero-delay desorption yield, moving theory closer to experiment. However, the characteristic subpicosecond peak seen experimentally remains unresolved, indicating missing physics such as nonthermal electronic distributions or rapid electronic structure changes at extreme $T_e$; TDDFT or nonthermal models may be required for full explanation. Overall, the study highlights the critical role of temperature-dependent electronic structure and coupling in extreme non-equilibrium surface dynamics and provides a pathway for more accurate simulations of 2PC experiments on metal surfaces.

Abstract

Two-pulse correlation experiments performed using pulses of different intensities on Pd(111) with different CO coverages showed that the CO photodesorption probability depends on whether the strong or the weak pulse arrives first to the surface, being this difference particularly large for the low-covered surface. Motivated by these experiments, we perform molecular dynamics simulations using a multicoverage potential energy surface that was previously constructed with the embedded atom neural network method. The process is modeled by combining the two-temperature model (2TM)--to describe the laser-excited electrons and phonons--and Langevin dynamics with electronic time-dependent temperature $T_\textrm{e}$--to model the coupling of the nuclei degrees of freedom with the laser-excited electrons. We show that improving the energy balance description in 2TM--by including \textit{ab-initio} $T_\textrm{e}$-dependent electronic heat capacity and electron-phonon coupling constant--is key to reproduce the asymmetry of the photodesorption probability $P_{\rm des}$ between positive and negative time delays. Furthermore, we also explore possible reasons for the usual underestimation of $P_{\rm des}$ at zero delay given by state-of-the-art calculations. In particular, we improve the description of the energy exchange between CO and the metal surface at high $T_\textrm{e}$ by including in the simulations a $T_\textrm{e}$-dependent friction coefficient. The prediction for $P_{\rm des}$ at zero delay in this case, increases by an order of magnitude, reducing its discrepancy with the experimental value. Altogether, our results hint to the importance of accounting for the temperature dependence of the electronic structure and properties in describing the extreme conditions generated in 2PC experiments.

Figures (6)

• Figure 1: Electronic temperature dependence predicted by different models present in the literature for the palladium (a) electron heat capacity $C_\mathrm{e}$ and (b) electron-phonon energy exchange coupling factor $G$. In blue, we show $C_\text{e,1}$ and $G_1$ adjusted to experimental data for temperatures below 1500 K, extracted from Refs. Muzas2024 and Szymanski2007. In red, we show $C_\text{e,2}$ and $G_2$ extracted from DFT-based calculations in Ref. Li2022. Illustrative 2TM temperature profiles obtained with said models in the case of two pulse irradiation with 38 and 93 J/m$^2$ of absorbed laser fluence and 2.5 ps of delay for (c) electronic temperature $T_\textrm{e}$ and (d) phononic temperature $T_\textrm{l}$.
• Figure 2: $T_\textrm{e}$-dependence of the atom-resolved electronic friction coefficient for CO on Pd(111). Dotted (dashed) lines corresponds to calculations where the CO is adsorbed at bridge (three-fold hollow) positions with a coverage of 0.5 ML. Their average is shown in solid lines.
• Figure 3: CO desorption probability as a function of time delay $\delta t$ between the high and low intensity pulses obtained in our $(T_\mathrm{e},T_\mathrm{l})$-MDEF simulations based on 2TM-1 and $\eta(0)$ for the adsorbates friction coefficients (red squares and lines) and in two-pulse correlation measurements from Ref. Hong2016 (black squares and lines) normalized to $\delta t=5$ ps. Positive (negative) delays correspond to the low intensity pulse arriving first (second). Left: the initial coverage is 0.75 ML in both experiments and simulations. Right: the initial coverage is 0.24 ML in experiments and 0.33 ML in our simulations.
• Figure 4: Same as Fig. \ref{['fig:2pc_camillone']} but for our $(T_\mathrm{e},T_\mathrm{l})$-MDEF simulations based on the electronic temperature calculated with 2TM-2 and the friction coefficients at 0 K for C and O ($\eta(0)$).
• Figure 5: Positive-to-negative delay comparison of $T_\textrm{ads}(t)$ (solid lines) and $T_\textrm{e}(t)$ (dashed lines) for 0.75 ML (left panel) and 0.33 ML (right panel). Results obtained with 2TM-2 and MDEF-2 for time delays $\delta t=\!5$ ps (blue lines) and $\delta t=\!-5$ ps (red lines). Insets: Difference between $T_\textrm{e}(t)$ at $\delta t=\!5$ and $\delta t=\!-5$ ps as obtained for the laser pulses employed at 0.75 ML (green, left inset) and 0.33 ML (orange, right inset).