Machine learning control of an elliptically bent hard X-ray mirror
Sheikh T. Mashrafi, Ross Harder, Xianbo Shi, Deming Shu, Zhi Qiao, Max Wyman, Tim Mooney, Jayson Anton, Steven Kearney, Luca Rebuffi, Jun Qian, Bing Shi, Lahsen Assoufid
Abstract
This article showcases the high-resolution control of an elliptically bent hard X-ray mirror optics at the Advanced Photon Source. The mirror uses a compact laminar flexure bending mechanism to achieve elliptical shapes covering a large range of focal distances. An array of capacitive sensors are used as a surface profiler for in-situ monitoring of the mirror shape. Machine learning and control techniques were used to change the mirror shape and focus the incident X-ray at predefined focal planes. The mirror surface shape error can be controlled to be within 40 nm rms with high repeatability. This technique gives the capability to focus incident X-ray beam within a range of focal distances corresponding to shape deformation range of a mirror optics. This work would be beneficial for controlling similar adaptive optics for multiple adaptive optics systems.