PhD Thesis Defense | Alfredo Serrano Jiménez

Theoretical description of femtosecond laser-induced desorption dynamics using ab initio and machine learning methods: pure CO and mixed CO+O adlayers on Pd(111)

Supervisors: Maite Alducin Ochoa and Joseba Iñaki Juaristi Oliden

DOWNLOAD FULL THESIS

Summary

In this thesis, femtosecond laser-induced desorption of CO from Pd(111) at various coverages, as well as photodesorption and photooxidation of CO coadsorbed with oxygen (0.25 ML O + 0.50 ML CO/Pd(111)), are examined. Using ab initio molecular dynamics with electronic friction (AIMDEF) and neural network-based potential energy surfaces (NN PESs), the research addresses the interplay of electronic and phononic excitations. Multiple factors—including coverage, laser fluence, and isotope effects—are investigated. Simulations underscore the cooperation of electrons and phonons in aforementioned processes, and reveal their relative roles on desorption. Oxygen coadsorption modifies interadsorbate energy exchange, enabling the formation of CO₂. By incorporating NN PESs, computational efficiency is significantly increased, allowing long-timescale dynamics and more comprehensive statistical analyses. These findings provide a deeper understanding of CO desorption and oxidation on Pd(111), offering insights for catalytic system design and advancing theoretical modeling and potential applications in surface chemistry. Future studies can refine these methodologies and insights.

00"]

Left: Sketch illustrating the construction of a multicoverage neural network potential energy surface for CO/Pd(111). Right: CO desorption probability against the absorbed fluence for different initial coverages. Reddish, bluish, and greenish full circles stand for (Te, Tl) − MDEF theoretical results for 0.75, 0.60, and 0.33 ML coverages, respectively, with varying simulation cell size. Darker red, blue, and green triangles correspond to the experimental results by Hong et al. (2016) for 0.75, 0.64, and 0.24ML coverages, respectively. Each set of data points has been fitted to a log-linear function (solid lines).[/caption]