Hot Jupiter TOI-2535 b’s transits vanished; unseen neighbour proposed

Across seven years of NASA TESS data, the planet’s path over its star climbed from near the centre to the edge and then missed it, a drift of 0.16 stellar radii a year that its discoverers call the fastest tilt change recorded for a planet around a single star.

Peer reviewed · Observational study · Companion inferred from modelling

A hot Jupiter about 1,100 light-years away has stopped crossing the face of its star as seen from Earth. Samuel Yee and Erik Petigura of the University of California, Los Angeles, and 26 colleagues reported in The Astrophysical Journal Letters on 9 November 2026 that the transits of TOI-2535 b shrank over seven years of NASA TESS data and were gone by March 2026. They attribute the change to the pull of a neighbouring giant planet on a tilted orbit. Nobody has detected that neighbour. It is inferred from models, and several different orbits fit the data.

Evidence at a glance

Is this a big deal?

It is a solid observation with a provisional explanation. The vanishing act is measured, and the authors describe TOI-2535 b as the first confirmed planet around a single star whose transits have been directly watched to stop. The unseen neighbour explains the data better than the alternatives the team tested, yet it still has to be found.

What scientists found

A transit is the small dip in starlight when a planet passes between its star and us. TOI-2535 b is a hot Jupiter, a gas giant on a tight orbit. It circles a star a little smaller and cooler than the Sun every 8.236 days, at about 11 million km, roughly a fifth of Mercury’s distance from the Sun. Its radius is 0.84 times Jupiter’s.

Astronomers describe where a planet crosses its star with a number called the impact parameter: 0 is a path through the centre of the disc and 1 is the edge. In TESS data from 2019 and 2021 the value was about 0.2. By 2023 it was about 0.58, and by March 2025 about 0.87. The last transit anyone caught was on 4 May 2025, from a 0.4-metre telescope in Chile.

TESS looked again in early 2026 and saw no dip. Ground telescopes in Poland and Texas tried on 13 and 22 March 2026 and also came up empty. A straight-line fit gives a drift of 0.16 ± 0.01 per year, which the paper calls the “fastest rate of inclination change observed for an exoplanet orbiting a single star”.

The timing misbehaved too. Transits came as much as 6 hours ahead of or behind a regular schedule, about 3% of the orbital period. The authors say that is the largest timing swing known for a hot Jupiter with an orbit shorter than 10 days.

How they found it

The candidate first turned up in a search of TESS light curves by a collaboration called WINE (Warm gIaNts with tEss), and it caused trouble from the start. One early transit fell in a gap in the data, so the TESS catalogue lists a period of 16.4453 days, twice the real value. Observers on the ground then tried 15 times to catch a transit and succeeded 6 times, because the planet kept arriving off schedule.

Fifteen spectra from the NEID instrument on the WIYN 3.5-metre telescope at Kitt Peak, Arizona, showed the star being tugged by about 100 metres per second. That is a planet’s pull, far too weak for a hidden companion star.

The team then tested causes for the drift. A stretched orbit turning in its own plane would need an eccentricity above 0.692, which the data rule out. The star’s equatorial bulge would have to be far larger than expected for an old, slowly spinning star. A distant object would need more than 8.7 Jupiter masses, and the velocities show nothing that heavy.

That left a nearby planet. The team tuned 300,000 simulated two-planet systems to the data and ran the 100 best for a million years. About two-thirds fell apart. The survivors share a pattern: a second planet of roughly 0.3 Jupiter masses or more (Saturn is 0.30), on an orbit tilted at least 5° from TOI-2535 b’s, with the two orbital periods near a 2-to-1 or 3-to-2 ratio.

In those models the neighbour’s gravity swings TOI-2535 b’s orbital plane around once every 30 to 50 years. From Earth, the swing shows up as the transit path sliding up the star and off it. SciQuest’s arithmetic from the paper’s numbers puts the change in viewing angle at about 0.5° a year, or roughly 3.4° over the seven years.

Why it matters to you

Nothing here changes daily life. Its value is a view of orbits moving on a human timescale. Petigura offered a comparison close to home in UCLA’s release: if the Moon’s orbit lined up exactly with Earth’s path around the Sun, we would get an eclipse every month. The Moon’s orbit is tilted by about 5°, so most months the shadows miss.

TOI-2535 b needed only a few degrees to slip from view. The comparison stops at the geometry, though. It explains why a small tilt hides a crossing, and says nothing about why this planet’s tilt is changing so fast.

What the evidence does not show

It does not show the neighbour. The paper states that the current data “do not uniquely identify a system architecture”. The single best-scoring model put the neighbour where it would itself transit. No such transits are seen, so the authors discarded it. Versions with the neighbour inside TOI-2535 b’s orbit and outside it both remain in play.

The fits are imperfect as well. Even the best models miss some transit times by up to 10 minutes, and about 30 metres per second of scatter in the star’s velocity stays unexplained. The authors say there may be more than one unseen planet. TOI-2535 b’s own mass is unsettled: a one-planet fit gives about one Jupiter mass, and the paper calls that fit “only illustrative”.

Treat the return date with the same care. UCLA’s release says the planet should transit again in 30 to 50 years. That range matches the swing period in the favoured models, yet one of the paper’s five example solutions takes about 72 years.

How this fits what we already knew

Planets orbiting two stars are known to drift in and out of transit, because the stars themselves move. Around single stars it is rare. The paper cites a 2021 search of 561 Kepler systems that found long-term changes in transit shape in only 15. A few planets, such as TOI-216 b and K2-146 c, have been caught moving further into view. The orbital swing reported for TOI-216 takes 150 years, against decades here.

The proposed pairing is unusual too. Few hot Jupiters have close planetary neighbours, and nearly all the known neighbours are small. The authors argue that two giants this close cannot have arrived by the route often invoked for hot Jupiters, in which a planet is thrown onto a stretched orbit that later shrinks. They favour a slow inward drift through the star’s birth disc of gas. A disc should flatten any tilt, however, so something must have tipped the orbits later, and that step is unexplained. Our report on two families of multi-planet systems covers how such arrangements are sorted.

Independent view

SciQuest found no on-record comment from a researcher outside the team as of 10 November 2026 and was unable to request one during this automated run. We will add one if we obtain it.

How other headlines framed it

No independent outlet had covered the paper when we checked, so the comparison is between the study and its own publicity. UCLA’s release, which Phys.org reposted, leads with a neighbour that makes the planet wobble like a top, stated as fact. The paper’s title mentions only the vanishing transits, and its text says the perturbing planet could not be uniquely pinned down. Lead author Yee keeps to the observation in the release, saying the transit “disappeared while we were watching”.

The release also rounds the orbit to 8.25 days and the distance to about 1,000 light-years. The paper measures 8.236 days, and the Gaia parallax it lists works out to about 1,100.

What happens next

Timing transits is no longer possible, since there are none. The authors call for frequent radial-velocity measurements instead. Their models predict that the neighbour tugs the star by 30 to 120 metres per second, which a larger set of spectra should pick out. That would turn the inferred planet into a detected one, or show that the real arrangement is something else. TESS is still scanning the sky, and the paper notes that its first four years of data would not have revealed this drift securely.

In one sentence

TOI-2535 b’s transits slid off the edge of its star between 2019 and 2026, and the leading explanation is the pull of a tilted, Saturn-mass-or-larger neighbour that no telescope has yet detected.

Read next

• Two families of multi-planet systems tied to their host stars: how planetary systems are grouped by what orbits in them.
• JWST maps WASP-17b’s atmosphere: what transits reveal about a hot Jupiter while it still crosses its star.
• New starlight-removal method recovers 51 Eridani b: another way to find giant planets that do not transit.

Sources and disclosure

1. Yee S. W., Petigura E. A., Brahm R., Collins K. A., Srdoc G., Latham D. W. and 22 others. “A TESS Disappearing Act: The Vanishing Transits of TOI-2535 b.” The Astrophysical Journal Letters 1011, L60 (2026). doi:10.3847/2041-8213/aeb4a4. Open access, CC BY 4.0. Received 2 September 2026, accepted 13 October 2026, published 9 November 2026.
2. UCLA Newsroom (Holly Ober), 9 November 2026: First known “disappearing planet” has a neighbor that makes it wobble like a top.
3. NASA Space Science Data Coordinated Archive planetary fact sheets for Mercury, the Moon, Jupiter and Saturn (orbit sizes, the Moon’s 5.145° orbital tilt, planet masses).

Disclosure: This article was drafted with AI assistance and checked against the sources above. SciQuest read the full open-access paper and checked UCLA’s release against it; short quotes come from the paper and, for Samuel Yee, from the release. Calculations by SciQuest from values in the paper: distance 1,000 ÷ 2.929 milliarcseconds = 341 parsecs, or about 1,100 light-years; orbital distance 18.74 × 0.877 solar radii = about 11.4 million km, against Mercury’s 57.9 million km; timing swing 6 hours ÷ 197.7 hours = 3%; tilt drift 0.16 ÷ 18.74 = 0.0085 radians, about 0.5° a year, assuming a circular orbit seen nearly edge-on; Saturn’s mass 568.32 ÷ 1,898.13 = 0.30 Jupiter masses. Calendar dates of the transits were converted by SciQuest from the Julian dates in the paper. The graphic is a SciQuest original redrawn from the paper’s published values; no third-party images are used. Spotted an error? See our corrections policy.