Data-driven modeling and control of an X-ray bimorph adaptive mirror
Gautam Gunjala1, Antoine Wojdyla2, Kenneth A Goldberg2
1Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, Berkeley, California, USA.
Journal of Synchrotron Radiation
|January 5, 2023
Summary
A novel data-driven approach enables precise control of adaptive X-ray optics. This method uses neural networks to manage mirror shapes without continuous feedback, achieving sub-2 nm RMS stability for advanced synchrotron and X-ray laser applications.
Area of Science:
- Optics and Photonics
- Materials Science
- Control Systems Engineering
Background:
- Adaptive X-ray optics are crucial for high-coherent-flux synchrotron and X-ray free-electron laser beamlines.
- Dynamic phase control and aberration compensation are essential for preserving wavefront quality.
- Challenges include the inability to continuously probe the wavefront, necessitating feedback-independent control methods.
Purpose of the Study:
- To demonstrate a data-driven approach for controlling adaptive X-ray optics with piezo-bimorph actuators.
- To develop a control method that requires little to no feedback for wavefront preservation.
- To achieve precise shape control and stability in adaptive X-ray mirrors.
Main Methods:
- A data-driven approach approximating non-linear system dynamics with a discrete-time model.
- Utilizing random mirror shapes and interferometric measurements as training data for a neural network.
- Incorporating prior states and voltage inputs into the model to account for shape-change trajectory.
Main Results:
- Achieved shape control and stability below 2 nm RMS.
- Demonstrated open-loop shape control across diverse states using a trained model and ex situ metrology.
- The control algorithm achieved shape error magnitudes within diffraction-limited performance.
Conclusions:
- The data-driven neural network approach effectively models and controls adaptive X-ray optics, accommodating system nonlinearities like drift, creep, and hysteresis.
- This method enables accurate, feedback-independent wavefront control essential for advanced X-ray applications.
- The demonstrated technique offers a path towards enhanced performance and stability in synchrotron and X-ray free-electron laser beamlines.
