Daylila

Space · Friday, 21 August 2026

01 · Briefing · what happened

A star found by staring may finally measure how fast our black hole spins

Space 7 min 25 sources

Astronomers have watched one speck of the Milky Way's centre since 2017. Out of that pile of six-minute exposures came a star faint enough to have hidden there all along. Its orbit is tight enough to feel the black hole turn.

8.7 years

S301's orbit

shortest known around our galaxy's black hole

8.5%

of light speed

its top speed, at closest approach

80-100 h

telescope time a year

spent on this one patch since 2017

19.3

how faint it is

on a scale where bigger means dimmer

At a glance

  • A team using the GRAVITY instrument in Chile found a new star, S301, circling the black hole at the centre of our galaxy.
  • Its orbit takes 8.7 years - the shortest ever measured there, beating a 12-year record.
  • At its closest it moves at 8.5% of the speed of light, near enough to feel the black hole dragging space around as it spins.
  • Nobody spotted it before because it is very faint. It only emerged once years of six-minute exposures had been stacked together.
  • Once they had its orbit, they found it again in old data from 2021 and 2017 that they had owned all along.
  • Elsewhere the same week: a comet studied by six months of watching, a rocky planet 23 times Earth's mass, and two warnings that satellite fleets are brightening the sky.

Forces in play

Patient observing High

Six years of watching one patch of sky produced S301; six months on one comet explained why it broke apart.

Brightening sky Building

About 15,000 satellites are up now, with paper plans for nearly two million, including data-centre fleets.

Launch tempo High

SpaceX flew its 100th mission of the year and passed 11,000 Starlinks; the White House wants 1,000 US launches a year by 2030.

Instrument reach Easing

The next generation should close gaps this one cannot. A bigger telescope able to split S301's light apart would finally read its make-up and its speed.

In play GRAVITY+ collaboration — found S301 and worked out its orbit Sagittarius A* — the black hole whose spin the star could reveal China — launching Chang'e-7 to hunt water ice at the lunar south pole SpaceX — 100 missions this year, past 11,000 satellites in orbit

How it unfolded

  1. 2017 monthly watching of the galactic centre begins; S301 is in the data, unseen
  2. Spring 2023 a faint dot appears near Sgr A* and is followed
  3. 2024-25 13 more positions pin the orbit down
  4. 19 Aug 2026 the discovery is published in Nature
  5. By 2035 enough measurements could fix the black hole's spin

Where this points

Watch whether the follow-up holds at ten measurements a year, and whether a deeper instrument can read S301's speed - both are needed for a spin measurement.

Full briefing

The star that only years of looking could find

Sagittarius A* is the black hole at the centre of our galaxy. It holds about 4.3 million times the mass of the Sun, and the paper puts it 8,277 parsecs away - roughly 27,000 light years [1]. Dozens of stars loop around it. Tracking them is how we know it is there.

On Wednesday a team reported a new star in Nature [1]. They found it with GRAVITY, an instrument that combines the light of the European Southern Observatory’s Very Large Telescope in Chile into one sharper view [1]. They call it S301.

S301 goes round once every 8.7 years. That is the shortest orbit ever measured at the galactic centre; the previous record was 12 years [1]. Its path is a long thin ellipse. At its closest it skims about 136 or 142 times the radius of the black hole’s point of no return [1]. The paper gives two numbers because the orbit can face either of two ways, and nothing yet tells them apart [1]. Either way that is roughly ten times closer than S2, the star that has carried most of this work until now [1]. At that instant it travels 25,000 to 25,600 kilometres a second, between 8.3% and 8.5% of the speed of light [1].

The closeness is the whole point. Einstein’s theory says a spinning black hole drags space around with it, and that effect fades very fast as you move away. No star found so far passes near enough to feel it. S301 does. The team estimates a fair chance of measuring that spin within about a decade [1]. Their forecast assumes roughly ten measurements a year out to 2035, plus ten more around the closest pass [1].

So why did nobody see it before? Because it is faint - magnitude 19.3 in infrared light, on a scale where bigger numbers mean dimmer [1]. The team has watched this patch monthly since 2017, in week-long runs from March to September, spending 80 to 100 hours a year on it [1]. The light arrives in six-minute chunks that are stacked together [1]. S301 surfaced in spring 2023. Once they had a rough orbit they went back through the archive and found it again in 2021, and weakly in 2017 [1]. It had been sitting in data they already owned. Those early positions are the least precise, and the paper says plainly why: fewer exposures were taken at those pointings [1].

One wrinkle matters more than it looks. Sgr A* itself flickers and is bright. Adding up more frames does not make it go away. The team had to model its light in each exposure and subtract it, then combine what was left and search that [1]. Even so, a separate instrument that splits light into its colours could not pick S301 out at all [1]. That is why its speed towards or away from us is still unknown [1].

The same week, the same currency

Time on a telescope was the price of almost every result this week.

A comet called 240P/NEAT split in two. David Jewitt’s group at UCLA worked out why by watching it from October 2025 to April 2026 [4]. That is two months before its closest pass by the Sun, and four months after [4]. They used a 2.56-metre telescope, small by modern standards [4]. Six months of patience produced dust-loss rates, sizes for both fragments and a separation speed [4]. It also ruled out tides, impacts and internal pressure [4]. What is left is the comet spinning itself apart [4]. The brighter piece is 400 to 600 metres across; the fainter one is harder to measure precisely because it is fainter [4]. (Sky & Telescope’s picture captions name the Astrophysical Journal, but its own citation and the paper’s identifier point to the Astronomical Journal.)

NASA has put online the transcript of a two-day symposium on Hubble held in Washington on 4 and 5 June [2]. Read as a ledger, it is a record of hours bought. One programme, ULLYSES, spent a thousand Hubble orbits on massive stars [2]. Another, Rocky Worlds, is spending 250 orbits watching small stars flare [2]. Giada Arney is project scientist for the proposed Habitable Worlds Observatory, and put the scale plainly. The faintest galaxies in the Hubble Ultra Deep Field are about as bright as the habitable planets it would hunt [2].

That is why the plan for it is so slow. To gather enough light, it would have to hold on a single planet for days to weeks [3]. That is why studying 25 worlds is expected to take a decade [3].

Elsewhere: astronomers reported GJ 523b, a rocky world 2.55 times Earth’s width but 23.5 times its mass, circling its star every 17.75 days. TESS flagged it; the 3.5-metre WIYN telescope at Kitt Peak measured the tug it puts on its star [5]. The system is about 170 million years old, which makes the density awkward - a core that heavy should have pulled in gas and become a small Jupiter. “This isn’t what we expected at all,” said lead author Max Kroft [5]. And Europe’s FLEX satellite was fuelled in French Guiana for a 15 September launch [6]. Its job is to measure the faint glow plants give off while photosynthesising, a signal invisible to the eye [6].

The sky that will not stay dark

Every one of those results depends on a background that stays quiet. This week two pieces of research argued that it will not.

Astronomers led by Aaron Boley of the University of British Columbia modelled the night sky under proposed orbital data centres. They posted the study on 3 August to arXiv, where physicists share work before formal review [7]. Companies have filed with US regulators for fleets of anywhere between 20,000 and one million satellites [7]. The favoured orbit runs over the poles along the day-night line, where sunlight never stops [7]. “People have lost the sense of the night sky because of the light-polluted cities, and now we’re rewriting that story with artificial satellites,” Boley said [7].

Roughly 15,000 satellites are up now, and there are plans on paper for nearly two million [8]. Hugh Lewis, a long-standing space-debris researcher, posted a separate study on runaway collisions [9]. His version accounts for satellites that dodge each other, and that get replaced when they die [9]. On disposal, Ars Technica ran the arithmetic on SpaceX’s proposed million-satellite fleet [10]. At a five-year life for the computer chips aboard, about 200,000 would be retired every year [10].

On the pad

China is expected to launch Chang’e-7 late on Sunday US Eastern time, which is Monday 24 August in China [11][12]. A Long March 5 was rolled out at Wenchang on 19 August [11]. The stack carries an orbiter, a lander, a rover and a hopping robot built to jump from sunlight into permanently shadowed craters near the south pole, hunting water ice [11][13]. Ice has been spotted from orbit; nobody has ever measured it on the ground [11]. Landing is set for later this year - Scientific American reports November [12]. Brown University’s James Head called it “the most ambitious, comprehensive, and complex robotic lunar mission ever attempted by any nation or entity” [13].

LandSpace landed the first stage of its ZhuQue-3 rocket in Gansu on 19 August local time, China’s first booster recovery on land using legs [14][15]. Video showed a fire in the tail afterwards [14]. It was the vehicle’s second flight; the first crashed on landing [14].

SpaceX’s 100th mission of the year was a Starlink flight from California [16]. It took the fleet past 11,000 satellites in orbit [18]. Days earlier it set a record with two Falcon 9s 38 minutes apart [17]. A Cape Canaveral flight carrying the year’s 2,000th Starlink was then stalled by a last-minute abort [19]. President Trump signed a memo on Thursday aimed at 1,000 US launches and re-entries a year by 2030, against 178 last year [20]. Muon Space raised $250 million, valuing it at $1.5 billion according to a person familiar with the round [21]. A new San Jose factory is meant to build 500 satellites a year by 2027 [21].

On the station, Anil Menon and Sophie Adenot spent six hours 23 minutes outside and removed a failed antenna, but ran out of time to fit the spare [22]. Adenot became the first French woman to walk in space [22]. Starship’s Ship 40, which splashed down intact in the Indian Ocean on 24 July, reached Christmas Island after 24 days under tow [23]. On Mars, Curiosity found a field of honeycomb cracks 4 to 8 centimetres across, a shape it had never met before [24]. Perseverance watched what appears to be the moon Phobos cross the Sun on 13 August, a day after the eclipse on Earth [25].

02 · Lesson · why it matters

Why the cure for a faint signal is time, not glass

Stare four times as long and the picture gets twice as clean - patience buys what money buys, until it meets an error time cannot wash out.

How it works

  1. A faint thing sends a steady trickle of light
  2. Random errors also arrive, but they cancel each other
  3. Stack more frames: the real signal adds, the errors partly wash out
  4. Four times the time buys twice the clarity - so patience substitutes for a bigger telescope
  5. But an error that repeats the same way never washes out
  6. So it has to be modelled and subtracted, not out-waited

The twist

Time buys clarity only against errors that are genuinely random, which is why a long careful look can end in a confident wrong answer rather than no answer.

Where you've seen this

Medical trials

more patients shrink chance error, but a biased way of choosing them stays exactly as wrong

Polling

a bigger sample tightens the margin; a skewed sample just gets more certain of a wrong number

Weighing something tiny

repeat the reading to beat the wobble, but a scale set wrong at zero is wrong every time

Listening for a quiet sound

record longer to lift a voice out of the hiss, unless a hum sits at the same pitch

The catch

The returns punish you: going from a rough picture to a good one is cheap, and going from good to excellent can cost a career's worth of nights.

Full lesson

The star was always there

S301 did not arrive in 2023. It was in the 2017 data and the 2021 data, sitting in files the team already owned. Nobody could see it, because on any single night it was buried under the noise of the instrument itself.

What changed was not the telescope. It was the pile.

The arithmetic behind this governs almost every faint measurement anyone makes. A real signal is steady. It points the same way in every frame, so a hundred frames give you a hundred times the signal. Random error is not steady. It points a different way each time, so adding frames makes it partly cancel itself. A hundred frames leave you about ten times the noise, not a hundred - the square root.

Signal grows faster than noise. That gap is the whole trick.

Four times the patience, twice the picture

Because the noise only grows as the square root, the ratio between the two improves as the square root too. Stare four times as long and the picture gets twice as clean. Stare a hundred times as long and it gets ten times cleaner.

That is why a modest telescope with time on its hands can outperform a giant with a short glance. The comet 240P was pulled apart by a 2.56-metre telescope - small, by the standards of the observatories that get on the news. It won because it looked for six months.

It is also why observing time is treated as currency. When the Hubble transcript records a thousand orbits on one programme and 250 on another, that is a budget, not a schedule. Hours are the thing being bought. It is why a future planet-hunting telescope would hold on one world for days to weeks, and why twenty-five worlds is a decade of work. The light arrives at the rate it arrives.

The returns get brutal

The square root cuts the other way too, and this is the half people forget.

Doubling your clarity always means quadrupling your time. The first useful result might take a night. The next step up takes four nights, then sixteen, then sixty-four. Somewhere on that ladder a scientist stops - not because the answer arrived, but because a life is only so long. The precision in a published paper usually marks where somebody ran out of nights, not where the physics stopped improving.

Some errors never wash out

All of this depends on one assumption: that the error is genuinely random.

Sgr A* is bright and it flickers. Add more frames and it does not fade, because it is really there in every one. The team could not out-wait it. They had to build a model of its light, subtract that model from each exposure, and only then combine what was left.

That is the difference between noise and a systematic error. Noise cancels. A systematic error repeats, and a repeated error survives averaging perfectly intact. Worse, the longer you look, the tighter your error bars get around the wrong answer. You do not end up uncertain. You end up confident and wrong.

A warm detector, a stray reflection, a setting slightly off - none of these announce themselves. They look exactly like a result.

Which is why a brightening sky matters

Now put the two halves together, and you can see why astronomers spent this week arguing about satellites.

A satellite streak in one frame is an annoyance you can throw away. Tens of thousands of them, arriving all night, every night, over every dark site on Earth - that stops being noise and starts being a floor. It does not cancel. It sets a level below which no amount of staring will take you.

The fleets are useful. They carry internet to places that had none, and emergency traffic when the ground breaks. That is real. But the cost lands on people who were never in the room - those whose whole method was patience, and whose patience is quietly worth less. Nobody voted to make the sky brighter. It is being settled in filings with a regulator, one fleet at a time.

What we cannot out-wait

Every one of us runs this same procedure on our own life. Try something a few times, average the results, decide what is true. More attempts, more confidence. That works against the coin-flip part - the luck, the bad day. It does nothing against the errors baked into where we stand: what we were never shown, what our vantage point cannot reach, the reflection off our own instrument.

Those do not fade with more looking. They go only when someone points them out, or when a different instrument in different hands sees the same thing from somewhere else.

So the confidence that comes from long attention is not evidence in itself. It is worth exactly as much as our list of what we thought to subtract first. And that list is written from inside the view we are trying to check.

03 · Lab · your turn

Spend the Season

Split twenty telescope nights between staring longer and subtracting a known error, and feel which one time can actually fix.

04 · Hope · carry this

The star was in the data all along. What finally revealed it was not a better machine but people who kept looking, and much of what we do not yet know is like that.

Across the beats