Explore the vast range of titles available.

Numerical optimization algorithms are already established tools to increase and stabilize the performance of particle accelerators. These algorithms have many advantages, are available out of the box, and can be adapted to a wide range of optimization problems in accelerator operation. The next boost in efficiency is expected to come from reinforcement learning algorithms that learn the optimal policy for a certain control problem and hence, once trained, can do without the time-consuming exploration phase needed for numerical optimizers. To investigate this approach, continuous model-free reinforcement learning with up to 16 degrees of freedom was developed and successfully tested at various facilities at CERN. The approach and algorithms used are discussed and the results obtained for trajectory steering at the AWAKE electron line and LINAC4 are presented. The necessary next steps, such as uncertainty aware model-based approaches, and the potential for future applications at particle accelerators are addressed.

To improve the performance-critical stability and brightness of the electron bunch at injection into the proton-driven plasma wakefield at the AWAKE CERN experiment, automation approaches based on unsupervised machine learning (ML) were developed and deployed. Numerical optimisers were tested together with different model-free reinforcement learning (RL) agents. In order to avoid any bias, RL agents have been trained also using a completely unsupervised state encoding using auto-encoders. To aid hyper-parameter selection, a full synthetic model of the beamline was constructed using a variational auto-encoder trained to generate surrogate data from equipment settings. This paper describes the novel approaches based on deep learning and RL to aid the automatic setup of a low energy line, as the one used to deliver beam to the AWAKE facility. The results obtained with the different ML approaches, including automatic unsupervised feature extraction from images using computer vision are presented. The prospects for operational deployment and wider applicability are discussed.

The Gamma Factory proof of principle experiment aims at colliding laser pulses with ultrarelativistic partially stripped ion beams at the CERN Super Proton Synchrotron. Its goals include the first demonstration of fast cooling of ultrarelativistic ion beams and opening up many possibilities for new physics measurements in various domains from atomic physics to particle physics. A high average-power, pulsed laser system delivering approximately 200 kW needs to be implemented for this aim. This is possible thanks to state-of-the-art optical systems that recently demonstrated similar performances in the laboratory environment. Challenges lie in the implementation of this kind of laser system in the harsh environment of hadronic machines including their robust and fully remote operation. The design of this laser system, involving a high quality factor enhancement cavity, is drawn and described in this article. Mitigation procedures are proposed to overcome limitations imposed by the occurrence of degenerate high-order mode at high average power in such optical resonators. We show that the operation at average power above 200 kW is feasible.

The flux of high-energy protons slow-extracted from the CERN Super Proton Synchrotron (SPS) is limited by the induced radioactivity caused by the beam loss intrinsic to the extraction process. Methods to substantially increase the efficiency of the extraction process are of great interest to fulfill requests for an increasing flux of 400 GeV protons to the present experiments, located in the North Area of the SPS, and also for potential future experiments with very high demanded protons on target. A crystal shadowing technique to significantly reduce the beam scattered and lost on the electrostatic extraction septum during the third-integer resonant slow extraction process has been developed and a prototype system tested with beam. The technique is based on the use of a thin, bent silicon crystal to coherently channel or volume reflect the portion of beam that would otherwise impinge the wire array of the electrostatic septum and instead eject it into the transfer line toward the production targets of the experiments. In this paper, the concept is described and applied to the SPS machine in order to specify the requirements of the prototype crystal shadowing system. Beam dynamics simulations of the prototype system are compared and benchmarked to the results obtained through beam tests, before being exploited to understand the characteristics of the present system and the potential performance reach of an optimized, future operational configuration. The remaining challenges faced to bring the system into operation, the optimization possibilities and other potential applications are discussed.

The powering of octupoles during third-integer resonant slow extraction has been studied and recently tested with the beam at the CERN Super Proton Synchrotron (SPS) in order to increase the extraction efficiency and reduce the induced radioactivity along the extraction straight. The octupoles distort the particle trajectories in phase space in such a way that the extracted separatrix is folded, which decreases the particle density impinging the wires of the extraction septum at the expense of increasing the extracted beam emittance. During experimental SPS machine studies a reduction of over 40% in the specific (per extracted proton) beam loss measured at the extraction septum was achieved. In this paper, the prerequisite studies needed to safely deploy the new extraction scheme in a limited time-frame are described, the experimental results are presented and an outlook given toward the next steps to bring slow extraction with octupoles into routine operation.

Slow extraction of a quasicontinuous flux of high-energy protons is an important requirement for many high-energy physics experiments. This extraction type is associated with an unavoidable beam loss due to scattering on the thin septum element. The energy deposition of scattering products and resulting activation place performance limits on existing and planned high-power, high-energy fixed-target proton facilities. In the400GeV/cSuper Proton Synchrotron (SPS) at CERN, a diffuser (or prescatterer), comprising an array of dense wires or ribbon located upstream of the electrostatic septum, has been designed to reduce absolute losses on the septum wires. As part of a concerted effort to investigate loss reduction techniques in the SPS in view of new physics experiments, the diffuser concept was explored in numerical simulation and analytically. A prototype device has been designed, built, installed, and tested in the SPS to prove the feasibility and quantify the performance reach. In this paper, the diffuser concept is briefly recalled and design considerations for the SPS use case are presented, with the analytical considerations and simulation studies for the optimization of the material and geometry. The device design is described, and the experimental results with a beam are presented and analyzed. The results are discussed, and an outlook is given for the operational feasibility and maximum obtainable performance gain. Conclusions are drawn on the implications for the application of the concept.

We use beam position measurements over the first part of the AWAKE electron beamline, together with beamline modeling, to deduce the beam average momentum and to predict the beam position in the second part of the beamline. Results show that using only the first five beam position monitors leads to much larger differences between predicted and measured positions at the last two monitors than when using the first eight beam position monitors. These last two positions can in principle be used with ballistic calculations to predict the parameters of closest approach of the electron bunch with the proton beam. In external injection experiments of the electron bunch into plasma wakefields driven by the proton bunch, only the first five beam position monitors measurements remain un-affected by the presence of the much higher charge proton bunch. Results with eight beam position monitors show the prediction method works in principle to determine electron and proton beams closest approach within the wakefields width (<1 mm), corresponding to injection of electrons into the wakefields. Using five beam position monitors is not sufficient.

The Advanced Proton Driven Plasma Wakefield Acceleration Experiment (AWAKE) is a project at CERN aiming to accelerate an electron bunch in a plasma wakefield driven by a proton bunch. The plasma is induced in a 10 m long rubidium vapor cell using a pulsed Ti:Sapphire laser, with the wakefield formed by a proton bunch from the CERN Super Proton Synchrotron (SPS). A 16 MeV electron bunch is simultaneously injected into the plasma cell to be accelerated by the wakefield to energies in the GeV range over this short distance. After successful runs with the proton and laser beams, the electron beam line was installed and commissioned at the end of 2017 to produce and inject a suitable electron bunch into the plasma cell. To achieve the goals of the experiment, it is important to have reliable beam instrumentation measuring the various parameters of the proton, electron and laser beams. This contribution presents the status of the beam instrumentation in AWAKE and reports on the performance achieved during the AWAKE runs in 2017.

FLASH radiation therapy calls for the delivery of fast bursted spills of particles with dose delivery times of the order of milliseconds. The requirements overlap with fundamental physics experimental requests that are being studied at CERN, albeit at very different energy scales. In this contribution, a scheme for extracting millisecond bursts from synchrotrons is explored by controlling a third-integer resonant and chromatic extraction with RF phase displacement acceleration. The scheme would be implementable in existing medical and experimental synchrotron facilities. Using a model of the CERN Proton Synchrotron, both single-burst and multi-burst extractions are simulated. Results show that 80 - 90% of the total beam intensity is extracted in a single burst of 40 - 60 ms. This would correspond to a ~10 ms burst in a typical medical synchrotron, namely the one outlined in the Proton Ion Medical Machine Study. A set of 3 consecutive bursts of 30 ms was simulated in the Proton Synchrotron with optimised machine parameters.

To improve the performance-critical stability and brightness of the electron bunch at injection into the proton-driven plasma wakefield at the AWAKE CERN experiment, automation approaches based on unsupervised Machine Learning (ML) were developed and deployed. Numerical optimisers were tested together with different model-free reinforcement learning agents. In order to avoid any bias, reinforcement learning agents have been trained also using a completely unsupervised state encoding using auto-encoders. To aid hyper-parameter selection, a full synthetic model of the beamline was constructed using a variational auto-encoder trained to generate surrogate data from equipment settings. This paper describes the novel approaches based on deep learning and reinforcement learning to aid the automatic setup of a low energy line, as the one used to deliver beam to the AWAKE facility. The results obtained with the different ML approaches, including automatic unsupervised feature extraction from images using computer vision are presented. The prospects for operational deployment and wider applicability are discussed.