A numerical model for the Habitable Worlds Observatory suggests that JWST-like mirror motion would make its 10−10 raw-contrast goal difficult to reach, even when the simulation includes active wavefront correction. With the JWST-derived motion left unscaled, only one modeled case reached the threshold: a 100 Hz loop using an out-of-band wavefront sensor, and only for stars of magnitude 4 or brighter. It is a result from a simulation, not a performance test of an operating observatory.
The work asks a tightly defined engineering question: what contrast stability could a future space telescope with JWST-like stability achieve if it used a modern coronagraph and deformable mirrors to correct telescope drifts? To explore it, the researchers combined JWST in-flight primary-mirror telemetry with a numerical model of a JWST-like segmented telescope, an Apodized Pupil Lyot Coronagraph and an unobstructed off-axis segmented aperture. In the model, deformable mirrors supplied active wavefront control, meaning the simulated optical surfaces adjusted to correct changes in the shape of the incoming light. The comparison included different wavefront sensors, control frequencies, stellar magnitudes and two versions of the JWST-derived coefficients, one original and one divided by 10.
How the telescope motion was modeled
The inputs came from JWST NIRCam weak-lens time-series observations in program 1445. The dataset contained 11,752 frames measuring 18 mirror segments. During preprocessing, the 200 frames with the highest standard deviation were replaced, and a central segment was treated as static. The frames were inputs to the simulation rather than a set of independent experimental subjects.
The analysis used singular value decomposition, or SVD, to represent the telemetry as a set of modes, or recurring patterns. It produced 57 modes, 51 of which made a significant contribution after the central segment was held static. The model then used a power spectral density, or PSD, to describe how the motion was distributed across frequency. Above 0.02 Hz, the simulations replaced the PSD floor with a power law whose exponent was minus 2.
The optical model also made favorable choices. It used an unobstructed off-axis segmented aperture and an Apodized Pupil Lyot Coronagraph, or APLC, a starlight-suppressing optical system. The APLC was designed for raw contrast better than 10−10 across a 10% bandpass, with a dark zone extending from 3.4 to 12 times wavelength divided by aperture diameter. Beyond 4.5 of those units, the design specified 25% core throughput. The main setup used a 6.6-metre aperture, a 575-nanometre central wavelength, 10% wavefront-sensor transmission, one simulated wavelength and a two-frame total loop delay.
The result changes with the assumptions
Once the JWST coefficients and their PSDs were left at their original levels, the target proved elusive. With 10% end-to-end throughput, only the 100 Hz out-of-band WFS loop reached 10−10, and only for stars of magnitude 4 or brighter. All other modeled sensor and frequency combinations in that comparison fell short.
The model became more favorable in a deliberately hypothetical scaling exercise. Dividing the segment-motion coefficients by 10 is equivalent in the simulation to dividing the corresponding PSDs by 100. Under that change, delta contrast fell below 10−10 for stars brighter than magnitude 8 across all simulations. The paper reports this as a delta-contrast result, separate from the full raw-contrast result discussed next.
The conclusion gives a separate raw-contrast result. It reports raw contrast below 10−10 for both wavefront-sensor configurations when measured segment-motion amplitudes were reduced by a factor of 10, equivalent to a 100-fold PSD reduction. That result applied to simulated G2V stars brighter than magnitude 11, with G2V denoting Sun-like stars. It remains a modeled estimate for the chosen telescope and coronagraph, not a demonstrated operating result.
The vibration warning
The sharpest warning came from the vibration test. In a conservative simulation, 3 nanometres of total vibration power was distributed across 51 modes. Even with a 2,000 Hz control loop, the achievable contrast was reported to be more than 100 times worse. The study did not establish whether that exact modal coupling would occur in a future observatory, so this is best read as a stress test of the modeled control scheme.
The frequency treatment leaves another important gap. Above 0.02 Hz, the main extrapolation used a power law with exponent minus 2, but it did not capture JWST reaction-wheel disturbances observed at approximately 10 to 35 Hz. The analysis also assumed an unobstructed off-axis architecture, perfect correction of segment piston, tip and tilt, negligible detector noise and a simplified out-of-band wavefront-sensor model. Those are favorable assumptions, so the numbers describe a conditional simulation rather than a direct operating forecast.
Taken together, the simulations describe a conditional engineering benchmark. JWST-level segment stability did not consistently reach the target in the unscaled cases, while the tenfold-scaled case reached the reported raw-contrast threshold for both sensor configurations for G2V stars brighter than magnitude 11, under the stated assumptions. The vibration scenario shows why that conclusion remains sensitive to disturbances missing from the main frequency treatment. The manuscript is an arXiv preprint, version 1, dated 28 August 2026. The JWST time-series data from observation program 1445 are publicly available through MAST.
Paper data and sources
Original title: JWST telemetry combined with active coronagraphy: raw contrast predictions for exoplanet imaging with the Habitable Worlds Observatory
Authors: Raphaël Pourcelot, Laurent A. Pueyo, Emiel H. Por et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-28
DOI: Not available
Original paper · Full text