Particle accelerators are devices of primary importance in a large range of applications such as fundamental particle physics, nuclear physics, light sources, imaging, neutron sources, and transmutation of nuclear waste. They are also used every day for cargo inspection, medical diagnostics, and radiotherapy worldwide. Electron is the easiest particle to produce and manipulate, resulting in unequaled energy over cost ratio. However, there is an urgent and growing need to reduce the footprint of accelerators in order to lower their cost and environmental impact, from the future high-energy colliders to the portable relativistic electron source for industrial and societal applications. The radical new vision we propose will revolutionize the use of accelerators in terms of footprint, beam time delivery, and electron beam properties (stability, reproducibility, monochromaticity, femtosecond-scale bunch duration), which is today only a dream for a wide range of users. We propose developing a new structure sustaining the accelerating wave pushing up the particle energy, which will enable democratizing the access to femtosecond-scale electron bunch for ultrafast phenomena studies. This light and compact accelerator, for which we propose breaking through the current technological barriers, will open the way toward compact accelerators with an energy gain gradient of more than 100 MeV/m and enlarge time access in the medical environment (preclinical and clinical phase studies).

The development of ERLs has been recognized as one of the five main pillars of accelerators R&D in support of the European Strategy for Particle Physics (ESPP). The international panel in charge of the ERL Roadmap definition recognized PERLE project as “a central part of the roadmap for the development of energy-recovery linacs”, with milestones to be achieved by the next ESPP in 2026. PERLE project is aiming at the construction of a novel ERL facility for the development and application of the energy recovery technique in multi-turn configuration, high current and large energy regime. It will operate in a 3-turns mode, first at 250 MeV, then upgraded to 500 MeV with 20mA beam current. Such challenging parameters make PERLE a unique multi-turn ERL facility operating at an unexplored operational power regime (10MW), studying and validating a broad range of accelerator phenomena, paving the way for the future larger scale ERLs. PERLE will be the necessary demonstrator for the future HEP machine (LHeC / FCC-eh), with which it shares the same technological choices and beam parameters. Furthermore, PERLE opens a new frontier for the physics of “the electromagnetic probe”. It will be the first ERL dedicated to Nuclear Physics for studying the eN interaction with radioactive nuclei. Here we will report on the project status, introduce the main ongoing achievements and describe the staged strategy we will adopt toward the construction of PERLE machine at its nominal performances.

Spatio-temporal couplings (STCs) [1] can have a detrimental effect on the intensity at focus of ultrashort femtosecond lasers. The laser spatio-temporal intensity profile control is a key issue for stable operation of laser wakefield acceleration (LWFA) [2]. Thus, it is necessary to measure and correct STCs. Techniques such as INSIGHT [3] or TERMITES [4] allow reconstructing the full spatial phase for different spectral components of the laser pulse using a phase-retrieval iterative algorithm. However this requires a computing time of the order of several minutes, making it inappropriate for single-shot online monitoring of lasers running at repetition rates of several hertz. We propose a method to characterize STCs in real-time using a multispectral camera [5] coupled with wavefront and temporal measurements and a machine learning algorithm. We will present the sensitivity characterization of the STCs measurement, which has been tested at 10 Hz for the optimization of a large optical compressor. Finally, we will discuss the status of the reinforcement learning implementation for full laser field reconstruction.

Accelerator physics simulators accurately predict the propagation of a beam in a particle accelerator, taking into account the particle interactions (a.k.a. space charge) inside the beam. A precise estimation of the space charge is required to understand the potential errors causing the difference between simulations and reality. Unfortunately, the space charge is computationally expensive, needing the simulation of a few dozen thousand particles to obtain an accurate prediction. This paper presents a Machine Learning-based approximation of the simulator output, a.k.a. surrogate model. Such an inexpensive surrogate model can support multiple experiments in parallel, allowing the wide exploration of the simulator control parameters. While the state of the art is limited to considering a few such parameters with a restricted range, the proposed approach, LinacNet, scales up to one hundred parameters with wide domains. LinacNet uses a large-size particle cloud to represent the beam and estimates the particle behavior using a dedicated neural network architecture reflecting the architecture of a Linac and its different physical regimes.