Advancing laser-driven ion acceleration : optimising with machine learning and investigating sources of instability

Dolier, Ewan James

Thesis

Advancing laser-driven ion acceleration : optimising with machine learning and investigating sources of instability

Download PDF

Creator

Dolier, Ewan James

Rights statement

Strathclyde Thesis Copyright

Awarding institution

University of Strathclyde

Date of award

2024

Thesis identifier

T16957

Person Identifier (Local)

201950027

Qualification Level

Doctoral (Postgraduate)

Qualification Name

Doctor of Philosophy (PhD)

Department, School or Faculty

Department of Physics

Abstract

This thesis reports on numerical and experimental investigations of proton acceleration driven by intense laser pulse (∼1021 Wcm−2 ) interactions with foil targets. The resultant beam of protons has unique properties compared to those produced in conventional accelerators. As a result, these novel accelerator sources are expected to have an important impact on both research and societal applications. For this to be realised, key properties such as the maximum proton energy and laser-to-proton energy conversion efficiency must be improved, in addition to the beam reproducibility and stability from one laser shot to another. Progress towards this goal is presented in two main investigations, the first of which involves the development of methods to automatically optimise properties of laser-driven proton beams in numerical simulations, advancing beyond conventional grid-search optimisation. Optimal values for laser energy, pulse duration, target foil thickness, and pre-plasma density scale length are identified with ∼200 fewer data-points, corresponding to a reduction of ∼48 days in simulation time, by employing a newly developed code called BISHOP with an integrated machine learning (ML) model. This four parameter optimisation is made feasible because of this technique, and is found to double the maximum energy of protons produced in the target normal sheath acceleration (TNSA) regime, compared to optimising for only the laser energy and pulse duration. The ML model also uncovered novel optimal pre-plasma conditions that increase laser energy coupling to fast electrons in the pre-plasma, whilst mitigating their overall divergence upon propagation through the target foil, thus increasing the sheath field strength on the target normal axis, and, as a result, the maximum energy of TNSA protons. A second, numerical and experimental investigation, demonstrates that the onset of relativistic self-induced transparency (RSIT) enhances proton energies beyond those that are achieved solely via TNSA, but with less stability when RSIT is induced at an optimal interaction time that maximises proton energies. This sensitivity is significant when factoring in shot-to-shot fluctuations in laser energy and pulse duration, demonstrated to occur in experiments at high power laser facilities. This exacerbates the known sensitivity of optimised RSIT enhanced proton acceleration to target foil thickness and laser temporal-intensity contrast. Early onset of RSIT deoptimises this regime in terms of proton energy, but makes it less susceptible to fluctuations in laser pulse parameters, whilst still enabling proton energy enhancement compared to the TNSA regime. Together, these investigations contribute to the development of laser-driven proton sources towards applications, by identifying new pathways to improve proton beam properties towards required specifications, whilst increasing understanding of how to produce these properties consistently

Advisor / supervisor