A new preprint reports that the Hubble Space Telescope's STIS instrument can produce remarkably steady optical and near-infrared measurements when it scans an exoplanet host star across the detector. After correcting for instrumental patterns, the broad-band data left residual scatter of 27 to 39 parts per million, a level that could support detailed transit measurements.
The same analysis found a strong rise in the apparent transit radius toward longer wavelengths. In the scanned observations, the radius in the reddest bin near 1 micrometre was more than 40% larger than in the bluest bin around 0.6 micrometres. That pattern also appeared in each of three archival, fixed-pointing observations of 55 Cnc e, although its size varied from visit to visit.
A focused test of scanning mode
The work was an instrument-performance study centred on 55 Cnc and its transiting planet, 55 Cnc e. It combined 20 short scans taken outside transit to test repeatability with 42 scans collected across five Hubble orbits during a transit. For comparison, the researchers used three separate archival transits observed in saturated stare mode, in which the telescope kept the target fixed on the detector. Those stare exposures lasted 36 seconds and reached about 31 times the detector's saturation level.
The adopted reduction used L.A.COSMIC to remove cosmic rays, Python defringe in the stistools package to reduce interference patterns, and an exoplanet-style procedure to remove trends. The visits were fitted separately, with the first orbit and the first exposure of every orbit left out because they carried more systematic noise.
Scans were repeatable, with a small orbit offset
The geometry of the scan mattered. Forward scans returned to a repeatable starting position, while reverse, round-trip scans showed offsets that depended on scan rate. The report therefore recommends using forward-only scans. The second visit used a scan angle of 90.065 degrees, although it is not yet clear whether the measured angle difference is typical.
Within an individual orbit, the measured fluxes agreed to about 0.01%, or 100 parts per million. The final seven to eight scans came even closer, agreeing within 0.005%, or 50 parts per million. The two orbits still sat about 0.04% apart, and their detailed patterns were similar without being identical.
Before the data were detrended, the differences between scan-mode orbits were smaller than those typically seen in pointed, saturated exposures. The comparison is not a clean head-to-head test of every observing choice, however, because the longer scan exposures and the averaging over more detector pixels may help explain part of the difference.
A promising signal with important caveats
The scanned white-light fits gave transit depths of 448 and 453 parts per million. The corresponding planet-to-star radius ratios were 0.02117 and 0.02129, with a reported uncertainty of 0.00027 on each ratio.
The narrower red interval was substantially noisier. Its detrending residuals were 104 and 86 parts per million, though defringing reduced those residuals by roughly 15% to 20%. Lower throughput and stronger systematic trends made the redder, narrower measurements harder to use.
The transit-depth estimates also varied more between visits than expected for a nominally shared system. The three stare-mode visits differed by more than three standard deviations, and the fitted transit centre in stare-mode visit 2 was about 26 minutes earlier than the ephemeris. The analysis leaves incomplete transit coverage and instrumental effects as possible explanations.
The pattern is not yet enough to identify its physical source. The authors regard the positive long-wavelength slope as persistent across visits and data reductions, but they also stress that its magnitude changes and that its physical origin remains unknown.
A useful instrument demonstration, not a final answer
The evidence is limited to one stellar system, two scan-mode visits, three archival stare-mode transit visits and calibration observations. Transit coverage was incomplete and uneven because Earth occultation interrupted the observations, while narrow and red bins retained larger noise and systematic trends. The study therefore demonstrates performance for these observations rather than establishing how STIS scanning will work for every exoplanet host.
The results do not settle whether the long-wavelength increase is astrophysical or instrumental, or whether the visit-to-visit transit-depth differences reflect true temporal variability. Better-covered independent observations are needed to test both questions.
The supplied document is a preprint identified as Instrument Science Report STIS 2026-05, version 1 of arXiv:2608.25184, dated 25 August 2026.
Paper data and sources
Original title: Spatially Scanned STIS Spectra of the Exoplanet Host Star 55 Cnc
Authors: D. E. Welty, J. D. Lothringer, D. K. Sing et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text