High power lasers are on the verge of creating a new quantum electrodynamic (QED) plasma state in the laboratory. Multi-petawatt laser facilities coming online can produce focused light intensities exceeding 1023 Wcm-2. At these extreme intensities, the plasma dynamics are dominated by feedback between strong field QED and classical plasma processes. The plasma electrons accelerated in the laser-plasma interaction of this type radiate a significant fraction of their energy as an extremely bright flash of gamma-ray photons, produced by synchrotron emission in the ultraintense laser fields. This can result in other strong field phenomena such as radiation reaction, relativistic anomalous opacity and multi-photon electron–positron pair generation. In addition to the exploration of high field plasma physics, there is also interest in developing this approach as a compact source of ultra-bright gamma-rays for a range of other fundamental researches, including particle physics, laboratory astrophysics and nuclear physics.
Maximizing the usefulness of the gamma-ray source requires an improved understanding of the laser-plasma interaction dynamics resulting in radiation emission and optimization of the radiation beam output, as a function of various laser and target parameters. Improved understanding of the conditions under which efficient absorption of laser energy into radiation occurs is also important for the development of other high power laser-dense-plasma interaction topics, such as laser-driven ion acceleration and high harmonic generation, given that copious gamma-ray generation in the laser field changes the laser-plasma interaction dynamics.
To efficiently explore the conditions for which the generation of synchrotron radiation is optimized, researchers from the University of Strathclyde applied machine learning approaches in simulations of the laser-plasma interactions with particle-in-cell modelling (using the BISHOP code developed at Strathclyde). They used Gaussian process regression on multidimensional parameter scans to provide new understanding of the role of each parameter on the synchrotron emission. It was found that the energy conversion efficiency from the laser pulse to synchrotron radiation is maximized for foil targets sufficiently thin as to become relativistically transparent to the intense laser light during the interaction. In this scenario, the relativistic mass increase of the electrons quivering in the laser fields is sufficient to cause the plasma to become transparent, resulting in partial transmission of the laser pulse. The machine learning algorithm was given control of the choice of initial conditions for several hundred simulations, and directed to search for conditions that maximized various aspects of the synchrotron emission, including the overall yield, divergence and number of photons within the highest energy part of the spectrum. The results indicate that optimum emission is achieved when the laser pulse is focused to the shortest and smallest possible spot close to the target surface, and at an oblique angle of incidence.
Identification of the optimum laser pulse parameters enabled additional, more detailed modelling of the interaction physics under those conditions, which showed that the majority of the synchrotron emission originates from electrons pulled from the surface of the target by the laser field. These energetic electron bunches are accelerated into the laser focal spot whilst the laser radiation pressure forms a channel in the target. The intense fields experienced by these highly relativistic electrons results in copious gamma-ray production in their direction of motion. By irradiating the target at an oblique angle with a p-polarized laser pulse, the propagation of the laser pulse along the target surface on one side of the focal spot enhances laser energy coupling to electrons and subsequent synchrotron emission. Control of the gamma-ray production was also demonstrated by variation of the laser polarization, due to its influence on the motion of the electrons.
Gaussian process regression was also used for optimization with multiple objectives. In particular, the simultaneous minimization of bremsstrahlung emission (produced by relativistic electrons propagating within the target) and maximization of synchrotron emission, to find conditions where the synchrotron emission can be best used or investigated in isolation. Foils with nanometre-scale thickness, irradiated at a large angle of incidence with respect to target normal, were identified as optimum.
This work provides multiple new insights into the physics of gamma-ray generation in laser-foil interactions, and especially the role of the laser angle of incidence onto the target. The results will inform the design of experiments to explore the new QED-plasma regime. Optimization of the radiation in experiments will require consideration of additional parameters that influence the interaction, such as the temporal- and spatial-intensity contrast of the laser pulse. The work presented here may be further extended by considering more complex targets and related phenomena, such as the generation of electron–positron pairs from the high-energy photons in the intense laser fields.
The research is published in High Power Laser Science and Engineering: Goodman, J., King, M., Dolier, E. J., Wilson, R., Gray, R. J. and McKenna, P. Optimization and control of synchrotron emission in ultraintense laser–solid interactions using machine learning[J]. High Power Laser Science and Engineering, 2023, 11(3): 03000e34
