Before It Can Map Earth's Hydrogen Halo, a NASA Instrument Must Calibrate on Starlight

From the Earth-Sun L1 point, a spacecraft stares back at Earth. Its target is not the bright blue disk but the geocorona, the hydrogen exosphere the mission is named for. The GeoCoronal Imager is built to capture ultraviolet emission from exospheric hydrogen at the Lyman-alpha line, a 1216-angstrom wavelength. The Carruthers Geocorona Observatory, NASA's first mission dedicated to investigating the fundamental nature of Earth's exosphere, launched in September 2025 and began science operations at L1 in January 2026. Its primary payload, the GeoCoronal Imager (GCI), consists of two co-aligned broadband photometric imagers, the Narrow Field Imager (NFI) and the Wide Field Imager (WFI), each with a six-position filter wheel.

A raw image is not yet a measurement

The detectors do not record physical hydrogen fluxes directly. They record digital numbers, shaped by the instrument's responsivity: how efficiently the instrument converts incoming photons at each wavelength into digital signal. That sensitivity can change after launch because of thermal variability, contamination, or gradual optical degradation. Ground tests in a vacuum chamber at the Centre Spatial de Liège in Belgium in late 2023 provided a starting point, but the mission needs to check and refine the calibration on orbit.

To do that, Carruthers will point at stars. Stars are convenient calibration lamps because many hot, young, massive stars shine brightly in the ultraviolet, and decades of space telescopes, including the International Ultraviolet Explorer, Hubble, and others, have built up libraries of their spectra. The catch is that the GCI is a broadband instrument. It gathers light across a wide range of wavelengths, and both the incoming starlight and the instrument's sensitivity vary across that range. A single star can yield an average sensitivity, but not a detailed sensitivity curve.

Turning 858 stars into a calibration toolbox

The team compiled a larger ultraviolet stellar dataset from several sources, including Hubble's CALSPEC flux standards, SOLSTICE and SPICAM measurements, and refined archival spectra from the International Ultraviolet Explorer. After correcting known artifacts, discarding unreliable data, and applying visibility and signal-to-noise cuts, they tied the final collection to the CALSPEC absolute scale, which carries about a 1 percent systematic uncertainty. The final calibration catalog contains 858 nonvariable stars, including 90 strong Lyman-alpha emitters.

From that catalog, the researchers use an algorithm to choose the best 30 stars to observe. The number 30 matches the mission's minimum operational requirement. The selection aims to minimize coherence, a measure of how similar the stars' spectra are. Stars with more distinct spectral shapes give the inversion more independent information, especially in the wavelengths where the instrument is expected to be most sensitive. This makes the target list objective and repeatable rather than hand-picked.

Mathematically, the problem becomes a large, underdetermined set of equations: 30 stellar measurements must constrain a sensitivity curve sampled at roughly 2,800 wavelengths. The team regularizes the solution, applying mathematical penalties that keep the retrieved curve smooth and close to laboratory expectations. This is the same general strategy used in source separation, where mixed signals are untangled using prior knowledge.

What the simulated observations recovered—and what they do not prove yet

To test the workflow, the researchers generated synthetic measurements of the 30 selected stars. They deliberately made the ground truth different from the laboratory prior. For the open, LyaN, and LyaX bandpass filters, they scaled the expected sensitivity down by 10 percent, a conservative stand-in for throughput loss. For the longpass filters, they also scaled the prior down by 10 percent and shifted the filter edge by 0, minus 5, or plus 5 angstroms to mimic thermal drift.

Across 50 trials for each channel and filter combination, the retrieval tracked the ground truth well, except where sensitivity approaches zero. In those low-signal regions, the solution becomes noisy because the algorithm is trying to fit data that contain little information. The mission's nominal science filter is the narrow Lyman-alpha channel, called LyaN. For that filter, the mean retrieved passband errors were 6.9 percent plus or minus 0.27 percent for the NFI and 4.7 percent plus or minus 1.68 percent for the WFI. The paper reports passband error rates of less than 7 percent for both primary Lyman-alpha channels.

These are simulation results, not on-orbit measurements. The validation assumes that any star in the calibration catalog is available at any time, bypassing the real scheduling constraints of power, thermal limits, and spacecraft pointing. It also assumes perfect subtraction of background light. Real observations will have to contend with those complications, and with the remaining uncertainties in stellar spectra.

The longpass filters bring their own caveat. The algorithm does not pin down the exact size of a thermal edge shift with high precision, though the study says that discrepancy is negligible for the mission's main scientific goals. A distinct dip near 1380 angstroms in the SrF2 filter is also left as a possible target for future refinement.

Once the GCI's sensitivity is known across its passbands, raw digital numbers can be converted into physical fluxes. That conversion is necessary before the geocoronal Lyman-alpha signal—already cleaned of instrument and background effects in companion steps—can be interpreted as physical emission. The Carruthers mission is designed to watch the geocorona with high cadence and global coverage, turning those calibrated measurements into maps of exospheric hydrogen. The new preprint suggests the star-based calibration plan can meet its precision requirements at Lyman-alpha, the line the mission cares about most. But the deeper message is about the quiet work that makes discovery possible. Before a spacecraft can map the breath of Earth's atmosphere, it must first learn exactly how it sees the stars.