Meta-Analysis of High Intensity Laser-Proton Acceleration Experiments

We present a comprehensive meta-analysis of laser-proton acceleration compiling more than 1,000 data points from over sixty scientific papers. This will be the largest publicly available dataset, ensuring organization and accessibility for ongoing and future research. Features collected include wavelength, intensity, pulse duration, spot size, angle of incidence, target thickness, and the resulting maximum proton energy for each experiment. We analyzed this dataset using a variety of statistical methods including correlation matrices, principal component analysis, and regression to uncover underlying patterns, outliers, and insights into factors that strongly impact the maximum proton energy. We compare existing theoretical models using this dataset including those developed by Fuchs et al. [Nature Phys 2, 48–54 (2006)] and Schreiber et al. [PRL 97, 045005 (2006)]. We find that the Schrieber model provides a reasonable upper bound for maximum proton energy based on just the laser power and conversion efficiency. Additionally, we evaluate different machine learning methods including neural networks. This comprehensive approach provides valuable insights for the optimization of laser proton acceleration experiments and the development of predictive models to help direct future experimental efforts.