X-ray multiphoton ionization dynamics: Quantum-state-resolved calculations and machine-learning approach

Abstract

X-ray free-electron lasers (XFELs) offer unique opportunities for unraveling ultrafast dynamics in matter and for imaging biomolecules with almost atomic resolution. For these attractive applications of XFELs, deepening our understanding of the interaction of high-intensity X rays with individual atoms is essential. In a single intense X-ray pulse, atoms are highly ionized by multiple sequences of one-photon ionization events and accompanying decay processes. These X-ray multiphoton ionization dynamics are commonly simulated by a rate-equation approach for electronic configurations. However, the configuration-based rate-equation approach does not include individual quantum states and, thereby, is unsuitable for studying the electron-cloud alignment of the produced atomic ions. It has been well-known that atomic photoionization can align an electron cloud, initially being perfectly spherically symmetric. But it is not clear how the alignment of the electron cloud evolves during X-ray multiphoton ionization dynamics. However, such a study requires a computationally expensive description of individual quantum states and quantum-state-resolved atomic transitions and includes solving rate equations in a generally extremely large space of states. In this thesis, I present a comprehensive framework for quantum-state-resolved calculations of X-ray multiphoton ionization dynamics of atoms and apply machine learning for handling the high computational cost. To this end, a quantum-state-resolved electronic-structure framework for isolated atoms and atomic ions is introduced in the first part of the thesis. This framework uses first-order many-body perturbation theory, which improves accuracy of transition energies. In addition, I employ quantum-state-resolved electronic-structure calculations to study how much the electron cloud of argon ions can be aligned through a single X-ray-induced atomic transition. A nonnegligible degree of alignment is observed. Combining the quantum-state-resolved electronic-structure framework with a Monte Carlo rate-equation method in a follow-up study enables me to calculate quantum-state-resolved X-ray multiphoton ionization dynamics. Results for neon atoms demonstrate that state-resolved calculations provide similar charge-state distributions, but more precise information about resonant excitations and electron and photon spectra than the common configuration-based calculations. Moreover, calculated time-resolved spectra of electrons and photons present detailed insight into ultrafast dynamics of state-resolved X-ray multiphoton ionization. However, performing such state-resolved calculations for atoms much heavier than neon is very costly due to extremely time-consuming state-resolved calculations of a very large number of atomic transition parameters. Because of this limitation, I next present a strategy that embeds machine-learning models for predicting atomic transition parameters into the state-resolved calculation of X-ray multiphoton ionization dynamics. As potential machine-learning models, I discuss feedforward neural networks and random forest regressors, which exhibit a similarly acceptable, but limited accuracy. In addition, fully calculated and machine-learning-based charge-state distributions and electron and photon spectra are compared for argon atoms. The comparison demonstrates that the machine-learning strategy works in principle and that the performance, in terms of charge-state distributions and electron and photon spectra, is good. Lastly, I apply the state-resolved X-ray multiphoton ionization dynamics calculations to explore the possibility to align the electron cloud of argon ions through a linearly polarized XFEL pulse. The induced X-ray multiphoton ionization dynamics generate ions in a wide range of charge states with nonzero orbital- and spin-angular momentum. While the electron-cloud alignment is suppressed with progressing ionization dynamics when averaging over all individual quantum states, the simulations clearly demonstrate nonnegligible electron-cloud alignment for orbital-angular-momentum- and charge-resolved states. Overall, this thesis contributes to a deeper understanding of the interaction of XFEL pulses with atoms by providing more accurate state-resolved information, complemented by insight into electron-cloud alignment dynamics. It also establishes a first step toward computationally efficient calculations of X-ray multiphoton ionization dynamics for easily examining a variety of atoms and XFEL beam parameters in the future.