Fast, efficient piston correction of deployable space telescopes using machine learning.

Area of Science:

• Optics and Astronomy
• Machine Learning Applications

Background:

• Segmented telescope alignment is critical for monolithic primary mirror performance.
• Diffraction-limited imaging demands high Strehl ratios (e.g., >0.8) and low wavefront errors (<32 nm RMS at 450 nm).

Purpose of the Study:

• To develop and implement a fast, two-step piston sensing technique for telescope alignment.
• To utilize a machine learning model for direct piston error retrieval from Point Spread Function (PSF) images.

Main Methods:

• A machine learning model was implemented for a four-petal telescope.
• The model processed synthetic misalignments within ±300 nm (±2λ/3).
• Performance was evaluated based on Strehl ratio improvement and signal-to-noise ratio (SNR).

Main Results:

• The machine learning model improved the mean Strehl ratio from a degraded state to 0.95 after one iteration and 0.99 after two.
• A signal-to-noise ratio (SNR) greater than 40 was sufficient for phasing, achieving a Strehl ratio of at least 0.97.
• The technique enables rapid, accurate correction of piston errors in segmented mirrors.

Conclusions:

• The developed machine learning-based piston sensing technique is effective for achieving diffraction-limited imaging with segmented telescopes.
• This method offers a significant advancement in mirror alignment, crucial for high-performance optical systems.
• The technique demonstrates the potential of AI in real-time optical system correction and optimization.