Calibrating tiny stellar motions: RETINA’s progress toward Earth-mass planet detection

Finding an Earth-mass planet by watching its star move requires more than a sensitive camera. Researchers must also know precisely how the camera’s pixels and telescope optics alter the apparent positions of stars. In a preprint posted on November 6, 2026, Fabien Malbet and colleagues describe RETINA, a collaboration developing detectors and calibration methods for future high-precision space astrometry. The work reports laboratory measurements and instrument-development plans—not an exoplanet discovery or a completed demonstration of the proposed missions’ performance.

Measuring a star’s response to a planet

Astrometry measures the positions and movements of objects on the sky. A planet’s gravitational pull makes its host star move slightly; measuring that motion can reveal the planet’s mass and orbit.

The preprint gives an illustrative signal of about 0.3 microarcseconds for an Earth-mass planet orbiting a Sun-like star at a distance of 10 parsecs. A microarcsecond is one millionth of an arcsecond, an already small unit of angle. Detecting such a signal requires extremely tight control of measurement errors.

The scientific ambition includes planets in habitable zones—the regions around stars discussed as potentially suitable for habitable worlds. Measuring a planet’s mass and orbit would not, by itself, establish that it is habitable or hosts life.

The paper discusses technologies intended for the proposed Theia mission and an astrometric instrument concept for NASA’s Habitable Worlds Observatory, called HWO-AGATE. These are future applications, not instruments whose performance in space is demonstrated here.

What the detector tests establish

RETINA brings together IPAG, CEA-Saclay and detector manufacturer Pyxalis. Its aim is to increase the technological maturity of components needed for precise astrometry.

A central development target is the GIGAPYX 220-megapixel detector, with pixels spaced 4.4 micrometres apart. Multiple detectors would form a large focal plane: the surface on which the telescope forms its images. The design omits microlenses, tiny lenses associated with individual pixels, to avoid the response distortions they can introduce.

The reported independent detector-characterization measurements, however, were made on the earlier GIGAPYX-4600, described in the paper as a 46-megapixel detector. They should not be read as a completed validation of the 220-megapixel model.

The authors report electronic readout noise as low as 12 electrons, departures from a proportional response to illumination below 1%, and pixel-to-pixel sensitivity variation below 1.3%. The defective-pixel fraction was 0.0004%.

Not every result matched the manufacturer’s specification. Dark current—electrical signal generated even without light—was measured at 36 ± 20 electrons per second at 35°C, slightly above the stated specification. The authors consider it acceptable and expect improvement in future revisions; that improvement is not a reported result.

Using moving light fringes to map pixels

Good detector specifications alone do not establish that star positions can be measured accurately enough. Small differences in pixel geometry or sensitivity within a pixel can shift the measured centre of a star’s image.

RETINA uses Young’s fringe projection to investigate these effects. Light from a helium-neon laser travels through two fibres and overlaps on the detector, producing alternating bright and dark interference bands. By moving these bands and recording each pixel’s response, researchers estimate offsets in the pixels’ effective sensing positions.

The paper gives an overall laboratory precision of 0.0005 pixel for the current technique, with a target of 0.00005 pixel. That target is 50 micropixels, where a micropixel means one millionth of a pixel. For the proposed 4.4-micrometre pixels, it corresponds to a distance of approximately 200 picometres on the detector—not an angular measurement on the sky.

More specifically, repeated measurements in a 70-by-70-pixel window on the GIGAPYX-4600 achieved precision of 0.0006 pixel horizontally and 0.001 pixel vertically. The authors suspect that imperfect calibration of electrical coupling between neighbouring pixels limits these results. This explanation remains their interpretation.

Scaling the method is another unresolved task. Treating the interference bands as straight is an approximation that breaks down over large detectors. The team is developing a model that accounts for their hyperbolic shape.

Correcting the telescope as well as the camera

Telescope optics can also shift apparent star positions in ways that vary across an image. RETINA therefore combines pixel calibration with a second calibration layer based on reference stars in the observed field.

A laboratory bench projects an artificial star field from a high-resolution micro-OLED display. Controlled changes to the optical setup allow researchers to test mathematical descriptions of image distortion.

The team also describes software that jointly estimates star positions, telescope pointing, image scale and optical distortion. Image scale relates distance on the detector to angular distance on the sky. Fitting these quantities together is intended to separate actual stellar motion from instrumental effects.

The paper reports preliminary laboratory evidence that this approach is feasible, with simulations continuing. It does not provide an end-to-end demonstration that the full system has reached the required sub-microarcsecond performance.

A proposed instrument, with requirements still to meet

The HWO-AGATE concept places four 220-megapixel detectors around a central 14-megapixel detector. The central sensor would rapidly read a selected region containing the bright target star, while the surrounding sensors would record fainter reference stars. The proposed field of view is 3 by 4 arcminutes.

Its error budget—the allocation of allowable measurement error among different sources—requires a single-measurement error no greater than 0.5 microarcseconds. The design aims for a signal-to-noise ratio of at least six after 100 revisits. These are performance requirements, not reported observations.

Remaining steps include validating calibration on 220-megapixel detectors, testing multi-detector configurations and reaching the targeted technology-maturity milestone by the HWO Mission Consolidation Review in 2029.

The preprint also identifies dark-matter studies as a motivation for precise astrometry, but reports no new dark-matter measurement. Its contribution is a review of detector and calibration progress, alongside the work still needed to turn those components into a functioning astronomical measurement system.

Source and status

This article is based exclusively on the supplied preprint, Advances in High-Precision Astrometry for Exoplanet Detection for Astrometric Missions, by Fabien Malbet and colleagues, version 1: https://arxiv.org/abs/2611.09657v1.