Overview
- Northwestern researchers published a Nature paper on Nov. 4 introducing FLIP (Physics-based Learning for Imaging Planets) and reported 2–15× sensitivity gains when applying it to three archival JWST coronagraphic datasets.
- FLIP builds a physics-based digital twin of JWST and uses physics-driven simulation plus iterative optimization similar to machine-learning training to match and subtract the telescope’s speckle patterns.
- Applying FLIP to archived data made the giant planet 51 Eridani b visible in JWST images for the first time despite the planet having been known from ground-based telescopes.
- Authors and outside experts say FLIP’s largest gains occur close to stars and could allow engineers to rethink some stability trade-offs for future missions, but they stress the method needs broader validation before changing design requirements.
- The result highlights limits and opportunities: JWST remains best for Saturn-sized or larger worlds at wide separations, Earth-like imaging will still need optics 100–1,000× more stable, and FLIP could unlock more discoveries in existing data while shifting where effort is spent between hardware and computation.