Real time optimisation of laser-matter interaction experiments using Machine Learning

该项目的目标是探索机器学习在激光物质实验的实时控制和优化中的有效性。最初的目标将是最大限度地增加通过高次谐波产生（HHG）过程产生的X射线光子的数量，其中强烈的红外飞秒激光脉冲与气体样品相互作用。学生将需要建立一个实验系统，其中计算机在计算机代码的控制下实时改变实验参数，同时记录用X射线探测器测量的X射线通量（例如，脉冲强度通过电动波片与偏振器组合进行计算机控制，气体密度通过电子阀等进行计算机控制）。将探讨各种计算方法，包括贝叶斯优化，遗传算法和神经网络。HHG优化为该项目提供了一个有用的起点，因为这是一个重要且长期存在的问题，在科学文献中已经以多种方式解决，并将提供有用的基准。这将为探索其他实验的控制和优化提供基础，例如，两个光脉冲与分子样品的相互作用，其中产物是使用飞行时间光谱学和速度图成像测量的离子和电子，并且人们想要最大化（或最小化）特定的电离或碎裂通道。

The goal of this project is to explore the effectiveness of Machine Learning for the real-time control and optimisation of laser-matter experiments. The initial goal will be to maximise the number of x-ray photons produced via the process of high harmonic generation (HHG), where an intense, infra-red femtosecond laser pulse interacts with a gas sample. The student will need to set up an experimental system, where a computer changes the experimental parameters in real-time under control of a computer code while recording the x-ray flux measured with an x-ray detector (for example, the pulse intensity is computer controlled via a motorised waveplate in combination with a polariser, the gas density is computer controlled via an electronic valve etc). A variety of computational approaches will be explored, including Bayesian Optimisation, Genetic Algorithms and Neural Networks. HHG optimisation provides a useful starting point for this project since this is an important and long-standing problem that has been tackled in numerous ways in the scientific literature and will provide useful benchmarking. This would provide a foundation for exploring the control and optimisation of other experiments, e.g., the interaction of two light pulses with a molecular sample, where the products are ions and electrons measured using time-of-flight spectroscopy and velocity map imaging and one wants to maximise (or minimise) a particular ionisation or fragmentation channel.