Roman's faint-end science is riding on a stray-light model built at Ames

ROSALIA does two jobs that observatories usually keep apart. One runs forward: predict where photons scattering inside the optical system will land as glints, so an astronomer can move an observation and keep the glint off the target ^[8][9]. The other runs backward: model the diffuse floor, which includes both scattered stray light and zodiacal emission from sunlight bouncing off interplanetary dust, then subtract it to leave the faint edges of galaxies visible ^[10][11][12]. The first decides what gets looked at. The second decides what the looking produced.

Both are the same team's model, built at Ames with collaborators at Goddard and IPAC/Caltech ^[6]. That is efficient and it is also a single point of dependence. A prediction good enough to steer a schedule away from glints is a prediction that will, in the subtraction step, be trusted to remove a background from the very pixels where the science signal is faintest ^[9][12]. Residual error in that background does not announce itself. It sits at galaxy edges looking like low surface brightness emission, which is the thing being measured ^[12].

The public evidence for the correction is a matched pair of images from the Wide Field Instrument, and both frames are simulated ^[19]. Simulation is the right place to build this, since you know the answer by construction. It is not where the model gets tested. The first real check arrives with real frames from a camera that images in optical and near-infrared light across wide fields ^[4], and by then the observing plans written against the prediction will already exist.

The coronagraph side has a cleaner shape and a harder gate. NASA describes the Roman Coronagraph as a demonstration of the most advanced technology flown for directly imaging planets around other stars, using masks and mirrors including two deformable mirrors to carve a dark zone beside a star ^[13][14]. Its baseline mode handles single stars only, because suppressing the extra starlight in a binary is not something current coronagraphs typically do ^[15]. Roughly half of Sun-like stars are in multi-star systems ^[16], so the baseline mode is pointed at about half the Sun-like population ^[20].

The Ames answer, Multi-Star Wavefront Control, is custom light-blocking masks plus software ^[17]. Through a collaboration with JPL, those masks are already on the flight instrument, beyond the baseline modes, and NASA says they could be used if the coronagraph team is granted additional observation time ^[18]. The metal has flown. Whether the capability exists in practice is now a scheduling decision made by people allocating hours, not an engineering one.