Ph.D position in Theoretical Atomic and Molecular Physics - Machine learning in ultrafast process in molecules driven by intense infrared laser pulses, University College London, UK (Deadline: 30.03.2026)

Attosecond science offers formidable tools to study correlated electron dynamics underlying ultra-fast phenomena. Ultra-short and ultra-intense laser pulses provide an impressive camera into the world of electron motion, with attoseconds and sub-femtoseconds being the natural time scales of electron motion. Quantum-mechanical techniques face hurdles in addressing ultra-fast phenomena. As a result, the key themes of this Ph.D project are currently beyond the reach of these techniques. We will address these challenges as follows:

• Objective 1. Deliver a powerful and versatile computational toolkit that comprises three-dimensional (3-d) semi-classical techniques. These techniques account for correlated electron dynamics and nuclear motion at the same time and for the magnetic-field component of the Lorentz force in three- and four-electron escape in atoms, diatomics and triatomics driven by ultra-intense, near-infrared (0.75-1.4 µm) and mid-infrared (2-8 µm) laser pulses}, i.e strongly driven systems. Using this toolkit, we will tackle Objectives 2.

• Objective 2. Study how correlated multi-electron dynamics couples with nuclear motion during Coulomb explosion of the nuclei, and steer electron capture to control the formation of Rydberg states in three- and four-active-electron driven triatomics, such as water.

Delivering the powerful computational toolkit in Objective 1 will involve high performance computing and building on and optimizing C++ codes currently employed in the group of Prof. A.Emmanouilidou. These C++ codes involve Monte-Carlo techniques used to obtain and identify multi-electron escape pathways. Some of these ionization pathways occur with a small probability. However, many such events are needed in order to perform a thorough analysis. To increase the number of multi-electron escape events we will explore the use of machine learning techniques in order to optimize the selection of initial conditions for the electrons involved. These will range from general purpose surrogate modelling approaches for efficiently precomputing initial conditions, to adaptive sampling approaches based on active learning and simulation-based inference while also taking measures for inspecting for and correcting the bias these sampling methods might introduce over naive sampling.

Prof. Agapi Emmanouilidou (first supervisor) is an expert in ultrafast science and multi-electron escape processes involving the interaction of atoms and molecules with laser pulses. Dr. N. Nikolaou (second supervisor) is an expert in Machine Learning / Artificial Intelligence (ML/AI) applications in Physics, Healthcare and Energy Systems. His research focuses on ML/AI model interpretability, causality, resource-efficient ML and ML/AI model uncertainty quantification. The last two areas are particularly relevant for this project.

The physics involved in this project is at the frontier of laser-matter interactions while the ML algorithms that will be developed will be useful for a wide range of other ultrafast phenomena.

Please contact Prof. Agapi Emmanouilidou by e-mail at ucapaem@ucl.ac.uk.