Mitigation of Emittance Dilution Due to Wakefields in Accelerator Cavities Using Advanced Diagnostics with Machine Learning Techniques

The preservation of low-emittance electron beams will continue to be a challenge and an objective in rflinac-driven accelerators where off-axis steering can lead to both transverse long-range wakefields (LRWs) and short-range wakefields (SRWs) that dilute the emittance [1]. Earlier experimentalists using normal conducting S-band and L-band accelerators have mitigated these effects by steering the beam optimally through the cavities while watching downstream imaging screens [2] or streak camera images [3]. One can even tune the wakefields to cancel some of the effects in a few normal conducting L-band structures by not centering the beam on a screen or BPM after each structure in simulations [4]. Since the transverse wakefields depend on 1/a where a is the cavity bore radius, it was somewhat surprising to identify both LRWs including higher order modes (HOMs) [5] and SRWs [6] in the superconducting rf TESLA-type cavities with their larger 35-mm radii. These cavities are used in major accelerator facilities (FLASH and the European XFEL), the under-construction LCLS-II XFEL, the Superconducting Test Facility in Japan, and proposed for the conceptual International Linear Collider (ILC) in Japan. Recent tests at Fermilab showed that near-resonance conditions of an HOM frequency with a beam harmonic resulted in submacropulse centroid oscillations at 100 kHz that diluted macropulse-averaged beam size [5]. More importantly, the same off-axis steering resulted in the generation of SRWs whose submicropulse transverse head-tail kicks produced projected beam size dilutions of 40% and greater in the sampled distributions, an effect at least 5x larger than that of the HOMs [6]. Such effects would also dilute the emittance values, and they would be a particular problem for ultra-low emittance preservation. These effects were seen after only two TESLA-type cavities with beam injected at 4.5 MeV and a final energy of 41 MeV. The transverse wakes depend on charge, beam offset, and the SQRT of bunch length, but inversely on beam energy. Thus, the emittance dilution threat is highest at the lower energies in the first accelerator cavities after the gun such as occurs in the LCLS-II injector with <1 MeV into the cryomodule. This same principle applies to all accelerators in labs around the country at Fermilab, SLAC, and Argonne. In the case of the SC rf cavities, HOM couplers provide online signals of dipolar modes dependent on beam offset and downstream streak camera images or a rf transverse deflecting cavity (TDC) plus screen provide submicropulse information. Such a scenario of multiple options for steering and tracking beam effects should be a prime application for machine learning techniques with extensions to virtual diagnostics [7].