Smoothing starlight before the search for faint planets

Why blocked starlight still leaks through

How do you photograph a faint planet beside a much brighter star? A coronagraph suppresses the star’s central image, but imperfections in the surrounding optics scatter light into speckles. Those speckles can obscure a planet or resemble one.

A report by Anton Neisius and colleagues describes work on MITHiC, a visible-light optical testbed at the Laboratoire d’Astrophysique de Marseille. Available as an arXiv preprint with a listed SPIE conference-proceedings reference, it presents laboratory measurements and software-based mirror emulation—not astronomical observations or a demonstrated planet detection.

The central result is a reduction in measured wavefront error from 34.8 to 2.4 nanometers root mean square after 15 correction cycles. This measures variations in the light’s optical path across the beam. It is not a planet-to-star brightness ratio.

Measuring the error, then reversing it

MITHiC uses a fiber-fed laser at a wavelength of 670.7 nanometers to imitate a point-like star. Its equipment includes a Lyot coronagraph, a science camera, a spatial light modulator and a wavefront sensor called ZELDA.

The modulator adjusts the phase of light—the position within its oscillation—across a finely sampled grid. ZELDA uses a small phase-shifting mask and a camera to reconstruct optical-path differences across the beam. The control loop applies an opposing phase pattern to the modulator, measures what remains and repeats.

This arrangement avoids the large interaction-matrix calibration commonly used to translate sensor measurements into mirror commands. It does not eliminate calibration: the researchers must still establish how positions on the sensor correspond to positions on the modulator. They report that omitting this geometric correction makes the loop diverge.

The team also filters out spatial variations too fine for useful correction. A preliminary step brings the wavefront into the sensor’s approximately linear operating range before the full correction begins.

A flatter wavefront is not yet a dark hole

The 2.4-nanometer result concerns the ordinary bench aberrations. When the researchers introduced a pattern representing misaligned segments of a telescope mirror, the residual was higher, at 8.6 nanometers. Sharp steps between segments introduce fine-scale structure that the current correction leaves behind.

The paper also examines image brightness averaged around rings centered on the artificial star. With 50 science-camera frames, the authors report that the corrected profile reaches their estimated photon-plus-read-noise regime beyond roughly 11 diffraction units. One diffraction unit, written as wavelength divided by aperture diameter, expresses angular separation relative to the instrument’s resolution scale.

Closer to the center, residual speckles remain dominant. Importantly, the noise calculation describes uncertainty in an annular average. It should not be read as a demonstrated sensitivity to a localized planet, whose detection would require an appropriate analysis of the image and its residual structure.

A focal-plane “dark hole” is a selected image region where controlled interference suppresses stellar light. Implementing the algorithms that create such a region is future work in this report, rather than an outcome already demonstrated.

Trying out mirrors without replacing hardware

The researchers use the modulator to emulate a deformable mirror with a limited number of actuators. Software restricts its commands to combinations of smooth, Gaussian-shaped actuator responses. These are experiments on the optical bench with an imposed mathematical model, not measurements of a newly installed physical mirror.

For an emulated 10-by-10 actuator grid, measured wavefront error fell from 35.5 to 7.6 nanometers. The authors estimate a fitting floor of about 6.1 nanometers: structure that the chosen actuator patterns cannot reproduce. Covering less of the beam leaves its outer region poorly corrected.

The apparent floor initially decreased during correction. The authors attribute this to sensor reconstruction bias at larger starting errors, not to the mirror model overcoming its own physical limitations.

Keeping the correction stable

During a 12-hour run, temperature changes correlated with pointing-related wavefront terms and defocus, at approximately 20–30 nanometers per degree Celsius. These measurements motivate temperature-based predictive correction, but do not demonstrate sustained dark-hole stabilization by that method.

The roadmap includes focal-plane control algorithms and an additional deformable mirror for independent amplitude and phase control. The present contribution establishes measured starting conditions and a way to explore actuator limitations. Whether those limits predict achievable dark-hole contrast remains a question for the next experiments.