SkyBlink

Case Files

Guided investigations, worked end-to-end on real archive data — the way a “Cosmic Detective” case should run: evidence first, claim second, error bars always.

CASE 1 · THE HERO

The Visitor From Between the Stars — comet 3I/ATLAS

3I/ATLAS is only the third interstellar object ever seen passing through our Solar System. SPHEREx caught it twice: during its August 2025 visit and again in December 2025 — and between those visits it crossed 79.2° of sky. IRSA publishes a dedicated observation table giving the comet's position for every one of its 259 SPHEREx exposures; SkyBlink's case pipeline takes the comet's own coordinates at each visit and renders the field around them from the regular archive. The two visits are 79.2° apart — completely different star fields — so they are shown side by side, not blinked against each other.

Case data: real SPHEREx frames at the comet's per-visit positions (IRSA 3I/ATLAS table).

  1. Find it. In each visit's frame, look for the soft, slightly extended glow that is not in the star charts — comets are fuzzy; stars are not. The comet is faint and extended (low signal-to-noise), and in the August frame it may not be visible by eye at all.
  2. Check it. Run the Two-Epoch Check in reverse: the comet should fail the “known star at the new position” test — nothing in Gaia sits where it went.
  3. Place it. The two positions come from IRSA's observation table, not from our own measurement: RA 244.37°, Dec −16.47° on 2025-08-13 and RA 167.21°, Dec +5.15° on 2025-12-14 — two points 79.2° apart on one interstellar trajectory.
CASE 2 · THE METHOD

The Two-Epoch Check

Most things that “move” in survey images are lies: cosmic-ray hits, detector ghosts, saturation trails. SkyBlink's named vetting method — applied to every candidate the blink turns up — has three tests, and a candidate must pass all of them:

  1. Both epochs. The candidate appears in at least two independent epochs, shifted by an amount consistent with one steady motion — not present in one frame only.
  2. Not a known star. ESA Gaia DR3, proper-motion propagated to the epoch, has no star at the candidate's new position. If Gaia says a star lives there, it is a star.
  3. Survives subtraction. In the difference view the candidate leaves a clean dipole — dark at the old position, bright at the new one — not a one-sided spike.

Anything less is logged as an artefact, not a discovery. The full method, with the pipeline decisions behind it, is on the Methods page.

CASE 3 · THE MEASUREMENT

The Star That Barely Moves — Barnard's Star, measured

Barnard's Star owns the largest proper motion of any star: 10.4 arcseconds per year. That is about 3.4 of SkyBlink's rendered pixels (3.1″ each) a year, or 1.7 native SPHEREx pixels (6.15″ each) — so this case measures it properly. We centroid the star in each of SkyBlink's five frames and compare with the position ESA Gaia predicts for that date:

ObservedReleaseDetectorCentre distanceMeasured − Gaia
2025-08-27DR1D50.23°2.4″
2025-09-30DR1D41.06°7.4″
2026-03-17DR1D20.29°1.3″
2026-05-04DR1D31.09°6.0″
2026-08-21QR3D20.04°2.8″

Barnard's Star lies within 1.3–7.4″ of Gaia's predicted position in all five frames (0.4–2.4 rendered pixels of 3.1″; 0.2–1.2 native SPHEREx pixels of 6.15″). The two largest offsets come from the two frames whose centre sits about 1.06° and 1.09° from the star; the other three frames are within 0.3°. With only five frames we cannot separate the effects of detector, position on the detector, and wavelength band, so we treat that pattern as a hypothesis, not a finding. Either way, it is why the Two-Epoch Check uses a tolerance, not a hair trigger. Gaia reference: source 4472832130942575872, μ = (−801.6, +10362.4) mas/yr.

Blink Barnard's Star yourself →