Deep Space Tuesdays · your week in deep space

From The Decisional Pass

Last Tuesday I told you the eclipse was a scheduled exam — a rare chance to check a physics model against an answer key published by celestial mechanics.

The exam happened. And the most interesting result didn't come from anyone under the shadow. It came from a pair of spacecraft 60,000 km above Earth that were busy making their own eclipse and got interrupted by the real one.

Eight minutes and forty seconds

ESA's Proba-3 is two spacecraft pretending to be one instrument. The Occulter flies 150 metres in front of the Coronagraph, carrying a 1.4-metre disc that throws a 5-centimetre shadow onto the other spacecraft's aperture — and the pair hold that alignment autonomously, to within a few millimetres, for six hours at a time. The instrument sitting in that shadow, ASPIICS, gets an uninterrupted view of the Sun's inner corona for those six hours — the thing a ground observer gets for two minutes, a few times a decade.

On the morning of 12 August the pair made their 65th artificial eclipse. That's the image above.

Then the Moon wandered into the shot.

Over about three hours, our Moon crossed ASPIICS's field of view, and for 8 minutes 40 seconds it covered the solar disc completely — while the two spacecraft went right on holding their own eclipse. A double eclipse. It lasted far longer than anything on the ground, because Proba-3 was closer to the Moon than Earth was; maximum totality down here was 2 minutes 18 seconds, just off the coast of Iceland.

Here's the part that made me sit up, and it's pure instrument engineering.

The enemy of every coronagraph is stray light. The corona is roughly a million times fainter than the disc you're hiding, so the limiting error is never sensitivity — it's the sliver of sunlight that diffracts around the edge of your occulting disc and scatters into the optics. Proba-3's whole design is an attack on that problem: move the occulter 150 metres out instead of mounting it inside the instrument, and the diffracted fringe gets far fainter. But it never goes to zero. You correct for it. You model it. You live with a term you can't independently measure.

For 8 minutes and 40 seconds, that term went away.

The Moon's shadow is slightly larger than the Occulter's disc, so during the overlap ASPIICS was, in ESA's words, essentially free of the parasitic light produced by diffraction on the Occulter's edge. Which means the team got two things at once: the cleanest corona image the instrument will ever take, and — more valuable — a null measurement. Everything ASPIICS recorded that it shouldn't have during those nine minutes is instrument error, full stop. That's a calibration you cannot buy, cannot schedule, and cannot reproduce on a test bench, because no vacuum chamber on Earth can reproduce a shadow cast from a third of a million kilometres away.

ESA calls it once-in-a-lifetime for the mission, and they mean it literally. Before Proba-3's mission ends, the Moon will cross ASPIICS's field of view twice more — during the ground eclipses of 2 August 2027 and 22 July 2028 — but neither will cover the full solar disc. Two partials. That's the rest of the schedule.

I've spent enough time around flight hardware to know how rare it is to get a free, unambiguous measurement of your own error. Usually you pay for those in months of test time and still argue about what the residual means. Here the solar system handed one over as a coincidence.

The answer key

The models, then. Last week I pointed at the teams who published forecasts of the corona ahead of totality — Predictive Science Inc. and KU Leuven's Coconut model, with a third prediction from an Indian group led by the Raman Research Institute.

Predictive Science has now put the comparison online, and I think their choice of how to do it is worth a minute. They did not publish a score. They published a blink comparator.

Their page defaults to the last simulation output produced before totality — the "last-data-assimilated" frame at 08:00 UT, posted publicly before the shadow reached mainland Spain — rendered from Earth's perspective at 18:30 UT, the moment of totality in Palencia. On top of that you can overlay real photographs and instrument data from the day: images from the IAC-led NATE project, from photographers who shot totality, from the Mauna Loa K-Cor coronagraph, from ESA's Proba-2 in the extreme ultraviolet, and from SDO. Then you fade between them with a slider, or hold the b key to blink prediction against reality, or switch to a colour filter that renders the simulation purple, the observation green, and the overlap white.

No verdict. Just: here is what we said, here is what happened, look for yourself.

That is a genuinely brave way to publish a forecast, and it's the format I wish more modelling groups used. A single skill score compresses a rich 2D disagreement — this streamer is in the right place, that one is 10 degrees off, the southern hole is too open — into one number that hides where the physics failed. The blink shows you which feature moved.

The cheapest prediction of the lot, meanwhile, came from Proba-3 itself. About two weeks before the eclipse, the mission photographed the corona from what was effectively the other side; the Sun takes roughly two weeks to rotate halfway around, so mirroring that image forecasts the view at totality. ESA's assessment of that trick, in their own caption: rather accurate. A fourteen-day-old photograph, flipped left to right, competing respectably with supercomputer MHD.

On the ground, for the record: 1 minute 21 seconds of totality at the Observatorio Astrofísico de Javalambre, where ESA ran its broadcast, and an estimated ten to twelve thousand people at ESA's free public event in León. Carole Mundell, ESA's Director of Science, watching it from Javalambre: "The detail was incredible. This eerie, nighttime light is what it feels like, but it's not nighttime."

Elsewhere: a break in a spectrum

The other big deep-space result of the week landed the same day as the eclipse, and it also turns on a measurement rather than a picture.

Writing in Nature on 12 August, Rohan Naidu of MIT and colleagues describe MoM-BH*-1, a source seen 660 million years after the Big Bang that they argue is a new kind of object: a supermassive black hole wrapped in a dense, dust-free envelope of hydrogen roughly the size of the solar system. The team is calling it a black hole star.

The evidence is a Balmer break — a cliff in the spectrum where light drops away below a particular wavelength, produced when dense gas absorbs photons. Every stellar population makes one. This one is deeper than any ever recorded, by a wide margin. Naidu's summary: the break "rules out 'ordinary' stars as the source." As he put it, you have something that looks a bit like a star but is 100 billion times brighter — and you cannot power that with nuclear fusion. Their model is a black hole of about 100,000 solar masses feeding inside an envelope perhaps 10 to 100 AU across.

If that holds, it's a candidate explanation for the "little red dots" — the small ruddy sources Webb finds nearly everywhere in the early universe and essentially nowhere today, and one of the most-argued puzzles of the mission so far. It would also be direct evidence for super-Eddington accretion: a way to grow a billion-solar-mass black hole fast enough to exist at cosmic dawn, which is a problem theorists have been stuck on for years.

One object. The paper is careful about that, and so am I. But note the shape of the argument — it's the same shape as the eclipse story. Nobody resolved this thing. They found a feature in a curve that no ordinary explanation could produce, and followed it.

That's most of what deep space actually is: not the picture, but the residual.

— Elio

References

• ESA, "A double eclipse for Proba-3" — 14 August 2026
• ESA, "Proba-3's eclipse hours before totality in Europe" — 14 August 2026
• ESA, "A look back at the 2026 total solar eclipse" — 14 August 2026
• ESA, "Proba-3 Mission" — mission parameters, accessed 18 August 2026
• ESA Proba-3 blog, "A double eclipse" — 17 August 2026
• Predictive Science Inc., "Comparisons — 2026 Total Solar Eclipse" — accessed 18 August 2026
• Raman Research Institute, "Prediction of the Solar Coronal Structure During the 12 August 2026 Total Solar Eclipse" — accessed 18 August 2026
• Naidu, R. P., et al., "A gas-enshrouded and gas-reddened black hole at cosmic dawn", Nature — 12 August 2026
• MIT News, "Astronomers discover a brand-new type of astrophysical object: A black hole star" — 12 August 2026
• Sci.News, "Astronomers Find Black Hole Star in Early Universe" — 12 August 2026

Forwarded this? Get The Decisional Pass in your inbox: