Dynamic and momentum apertures are a critical parameter of storage rings, limiting the flux of x-ray sources and the luminosity of colliders. Traditional optimizers must track thousands of particles for each lattice, severely limiting the design search. MultipointBAX instead tracks single particles, each selected based on a neural-network model of the stability maps. On a fourth-generation light-source design, MultipointBAX matches a genetic-algorithm Pareto front with over two orders of magnitude fewer tracking simulations, a gain that should extend to future light sources, colliders and other large facilities, enabling more complex and higher-performance designs.

This paper introduces a novel concept to reduce the longitudinal emittance in a low momentum compaction factor (low-alpha) storage ring, enabling the stable storage of electron bunches shorter than 100 nm. This design strategy can be applied to any quasi-isochronous storage ring to achieve very high radiation power through the longitudinal coherence of the emitted radiation. This paper addresses the nonlinear dynamics and aperture optimization for this type of storage ring. An optimal design example based on our analysis is presented. Single-particle tracking results demonstrate that an electron beam with an equilibrium root-mean-square bunch length of approximately 80 nm can be maintained in this ring, featuring a dynamic aperture larger than 1 mm.

Inverse Compton scattering produces X-rays by colliding a laser pulse with relativistic electrons. Conventional head-on geometries require low beam energies, causing wide emission cones that limit brightness. A shallow, nearly co-propagating crossing angle lets a higher-energy beam deliver the same photon energy with far tighter collimation, yet this regime had not been explored in practice. The authors derive the geometric scaling laws of shallow-angle scattering, confirm them experimentally at a 5.8° crossing angle, and use the validated framework to evaluate future high-brightness radiation sources across diverse accelerator platforms.

We propose creative applications of axially magnetized permanent magnet rings for low-energy (keV) electron optics with an emphasis on preventing apparent emittance growth in a magnetic field, ease of integration, and minimization of parasitic aberrations. Extensive analysis of tolerances, and preliminary magnetic field measurements indicate that ‘extraordinary performance’ can be achieved using off-the-shelf ‘fridge’ magnets. Our design philosophy offers a different perspective on the otherwise arduous challenges of mechanical alignment of accelerator components.

Modern accelerators generate machine states that are high-dimensional and tightly coupled, making single channel monitoring an incomplete picture of true operating conditions. Building on graph neural network embeddings of the CEBAF injector, this work shows that 14 months of operational history collapses into a small number of persistent, physically interpretable regimes within a 16-dimensional learned space. The embedding supports stability baselining, outlier screening, and historical case-based retrieval, and a controlled beam study confirms it tracks deliberate reconfiguration coherently — demonstrating a practical operational reference space for holistic machine state monitoring.