Description
Interactions between matter and high intensity lasers ($a_0\gg1$) produce brilliant sources of high energy X-rays via the Non-linear Inverse Compton Scattering process, which can go on to produce large numbers of Positrons through the Non-linear Breit-Wheeler and Bethe-Heitler processes [1,2]. This can produce a collection of electron-positron pairs dense enough to be considered a plasma, similar to the coronas of stellar remnants such as pulsars [3]. In this study, we use an active machine learning method, specifically an iterative Gaussian Process Regression model [4], to produce surrogate models of this positron production as well as the underlying X-ray emission across a multi-dimensional parameter space. These models are trained using a set of full-PIC simulations modelling the interaction between a Laser pulse in the intensity range $\sim10^{22}$-$10^{24}\textrm{Wcm}^{-2}$ and a $50\mu\textrm{m}$ gold target. We investigate the accuracy of the resulting models compared to a test dataset as well as how they compare with theoretical expectations.
[1] Ridgers, CP. et al (2012) ‘Dense electron-positron
plasmas and bursts of gamma-rays from laser-generated
quantum electrodynamic plasmas’, Physics of Plasmas
[2] Chen, H. et al (2010) ‘Relativistic
Quasimonoenergetic Positron Jets from Intense Laser-
Solid Interactions’, Physical Review Letters
[3] Sarri, G. et al. (2015) ‘Overview of laser-driven
generation of electron–positron beams’, Journal of
Plasma Physics
[4] Williams CK, Rasmussen CE. (2006) Gaussian
processes for machine learning. Cambridge, MA: MIT
press.
