Vanishing transits of hot Jupiter TOI-2535 b from orbital precession

TESS watched a hot Jupiter's transits shrink and stop between 2019 and 2026, because a hidden second planet tilts its orbit. We rebuilt the fade from the paper's own measurements with fibub's transit maths.

Planets do not simply disappear, but their shadows can. A paper published in The Astrophysical Journal Letters in November 2026 describes TOI-2535 b, a giant planet that the TESS space telescope saw crossing its star in 2019, then saw crossing less and less, until in early 2026 it stopped crossing at all. The team traces the cause to a second, unseen planet that slowly tilts the first one’s orbit. The paper is open access under CC BY 4.0. Below we explain the finding and then rebuild the fading transits with the same maths the transit fitter uses, with the exoplanet mass calculator for the orbit’s size.

What they found

TOI-2535 b is a hot Jupiter: a gas giant close to its star, going round once every 8.24 days. The star is an orange K dwarf about 1,000 light-years away. When TESS first looked, in 2019, the planet passed close to the middle of the star’s face. In each later visit the path had moved further towards the edge, so the dips in starlight got shorter. By early 2026 the planet missed the star altogether.

The team measured how fast the path moves. Astronomers describe the path with the impact parameter, written b: how far from the centre of the star the planet passes, in units of the star’s radius. A central crossing has b = 0; a crossing that just touches the edge has b close to 1. For TOI-2535 b, b grows by about 0.16 every year, the fastest change seen so far for a planet around a single star.

The transit times also wander, by up to about six hours early or late compared with a steady clock. Putting the timing, the changing transit length and radial-velocity measurements from the NEID spectrograph together, the team concludes that a second planet, at least as massive as Saturn and probably locked in a simple ratio of orbital periods with TOI-2535 b, is pulling on it from an orbit tilted by at least about 5 degrees.

Why it matters

A planet’s orbit around a star is not fixed in space if another planet is nearby on a tilted orbit. The two orbits pull on each other and slowly swing around, like a spinning top that wobbles. This is called nodal precession. From Earth, that swing changes the angle we see the orbit at, so a planet that transits today may not transit in twenty years.

That has two consequences. First, some planets we see transiting are only visible for part of the time, so counts of how common planets are need a small correction. Second, a changing transit is a way to find planets that never transit themselves: the hidden companion here was found by its effect, not by its own shadow.

How they did it

The team used seven years of TESS light curves, ground-based transit observations and NEID velocities. For each TESS visit they fitted the transit shape to get its length, depth and timing. A change in length at constant planet size points to a change in b. They then ran models of two planets pulling on each other to find which companions could produce both the timing shifts and the drift in b.

Checked with the tools

We do not have the paper’s light curves, but we do have its measurements, so every check below starts from the paper’s own numbers and says where each one comes from. From the paper: the period (8.236 days), the star’s mass (0.864 solar masses) and radius (0.877 solar radii) from its Table 1, the star’s wobble (about 100 m/s) and the seven impact parameters TESS measured in 2019, 2021 and 2023 (Table 2). The planet’s radius, 0.88 Jupiter radii, is the TESS catalogue value, because the paper’s fitted size ratio sits in an appendix we could not read. The only number we chose ourselves is how much darker the star’s edge is than its centre (limb darkening), which barely changes the timings.

Step 1: size the planet and its orbit. The exoplanet mass calculator turns a wobble of 100 m/s, an 8.236-day period and a 0.864 solar-mass star into a planet of at least 0.90 Jupiter masses: a hot Jupiter, as the paper says. Kepler’s third law then puts the orbit at 18.6 star radii (a/R★), and the planet covers a disc 0.103 times the star’s width (k = Rp/R★). That means the planet stops touching the star once b passes 1 + k ≈ 1.10. The same numbers give a star density of 1.8 g/cm³, normal for a K dwarf.

Step 2: read the path from the paper’s own measurements. Figure 1 plots the seven impact parameters in the paper’s Table 2. They stay near 0.2 in 2019 and 2021, then jump to 0.57 in 2023. A light curve only tells you how far from the centre the planet crosses, never which side. So there are two ways to draw a straight line through these points:

  • If the planet always crossed the same side, the best line rises by 0.13 ± 0.02 a year, and it fits badly (χ² of 8.9 for 5 degrees of freedom).
  • If the planet crossed the far side in 2019, passed straight over the centre around 2020.3, and has been moving outward since, the line rises by 0.20 ± 0.02 a year and fits the points almost exactly (χ² of 1.1).

The paper’s rate of 0.16 a year, which also uses later ground-based transits we do not have, sits between the two. The second reading is our own, not a claim of the paper, but it is what steady precession would do: the tilt of the orbit changes smoothly, so the path slides across the star’s face without stopping.

The measured path: the paper's b, read two waysThe measured path: the paper's b, read two ways-0.500.512020202220242026YearImpact parameter bPlanet no longer touches the star (b = 1 + k)Paper |b|2019 flippedSame side 0.13/yrCrossing 0.20/yrfibub.com
Figure 1. The seven TESS impact parameters measured in the paper (Table 2, grey, with error bars), and the same 2019 values placed on the far side of the star's centre (colour). Lines: weighted straight-line fits made for this article. Data: Yee et al. 2026, CC BY 4.0.

On that line, b reaches 1.10, the last grazing touch, at about 2025.9: just before the early-2026 TESS look that found no transit. At the paper’s rate of 0.16 a year from the 2023 value, the last touch would be in mid-2026, with a shallow graze of about 0.15% still there in early 2026, against 1.2% for the full transits in 2023. The same-side line would keep the transits going until 2027, which the 2026 non-detection rules out.

Step 3: rebuild the fade. Using the line through the paper’s points, we drew the transit for each year with the transit model inside the transit fitter (a limb-darkened star crossed by a dark disc). Figure 2 shows the shape: the dip loses its flat bottom and narrows.

The dip gets shorter and shallower each yearThe dip gets shorter and shallower each year00.51-3-2-10123Hours from mid-transitLight blocked (%)2020202220232025fibub.com
Figure 2. Model transits of TOI-2535 b, with the impact parameter following the line through the paper's values (Figure 1), using the paper's star size and mass. Made with the transit maths in the Transit fitter.

Figure 3 shows the transit length year by year. It is about 3.7 hours while the planet crosses near the centre in 2019 to 2021, 3.2 hours in 2023, and zero by about 2025.9.

Transit length, year by yearTransit length, year by year012342020202220242026YearTransit length (hours)Last transit ≈ 2025.9fibub.com
Figure 3. Transit length (first to last contact) from the same model. The line hits zero when the planet no longer touches the star.

Figure 4 shows something that surprises people: the depth hardly changes for years, then collapses. The length is the early warning; the depth only drops once the planet starts to graze the edge.

Transit depth, year by yearTransit depth, year by year00.512020202220242026YearLight blocked at mid-transit (%)Gonefibub.com
Figure 4. Transit depth at mid-transit from the same model.

Step 4: can the fitter measure the drift? We made five fake TESS visits from 2021 to 2024 from the model, each with three transits and noise of 1.2 parts per thousand per 10-minute point (our guess for a star of TESS magnitude 12.7, the value in the paper’s Table 1), and fitted each one blind with the Transit fitter. Figure 5 is the last visit.

The Transit fitter on one simulated visit (mid-2024)The Transit fitter on one simulated visit (mid-2024)00.51-3-2-10123Hours from mid-transitLight blocked (%)Simulated dataFitted modelfibub.com
Figure 5. Three simulated mid-2024 transits with TESS-like noise (grey), fitted by the Transit fitter (line). Simulated, not the real TESS data.

The fitter measures the transit length to about ±0.1 hours on every visit. Turning each length back into b (holding a/R★ at the star’s value) and drawing a straight line through the five points gives 0.18 ± 0.02 a year, against the 0.20 we put in. So a drift of this size is well within what TESS can measure, which supports the paper’s main finding.

Impact parameter from each visit, and the rate it givesImpact parameter from each visit, and the rate it gives00.250.50.75120212022202320242025YearImpact parameter bPlanet no longer touches the star (b = 1 + k)From the fitterStraight-line fitfibub.com
Figure 6. Impact parameter worked out from the transit length the fitter measured on each simulated visit, with a straight line through them. Simulated, not the real TESS data.
Check Paper or reports Our tool result
Planet mass hot Jupiter, K ≈ 100 m/s at least 0.90 Jupiter masses, from the wobble
Rate of change of b 0.16 a year 0.20 ± 0.02 a year from the paper’s TESS values if the path crossed the centre in 2020; 0.13 ± 0.02 if not (poor fit)
When the transits stop gone by early 2026 about 2025.9 on the better line; mid-2026 at 0.16 a year
Can TESS see the drift? yes 0.18 ± 0.02 a year recovered from simulated visits with 0.20 put in

One thing the fitter also showed: when the planet crosses near the centre, b on its own is poorly pinned down, because a central crossing of a smaller orbit looks much like an off-centre crossing of a larger one. You can see it in the paper’s own error bars: ±0.13 near the centre in 2019, ±0.05 off-centre in 2023.

Step 5: how long until it comes back? If the orbit swings to and fro by some angle either side of edge-on, and crosses the star’s edge at the paper’s rate of 0.16 a year, a simple swing model gives the years without transits. Figure 7 shows the answer for swings of up to 15 degrees. A swing of 5 degrees gives a gap of about 13 years; the 30 to 50 years quoted by the team need a swing of roughly 7.5 to 10 degrees. That fits with their lower limit of about 5 degrees on the tilt between the two orbits. This is a toy model, not the team’s full calculation.

Years without transits, for different swings of the orbitYears without transits, for different swings of the orbit02040608057.51012.515Swing of the orbit either side of edge-on (degrees)Years with no transits30 years50 yearsfibub.com
Figure 7. A simple tilting-orbit model held to the stated rate of 0.16 a year: how long the transits stay away for each size of swing. Dashed lines: the 30 to 50 years quoted by the team.

To go further, the radial velocity fits star wobbles like the NEID data, and the period finder looks for a repeating pattern in transit times, which is how timing wobbles like the six-hour shifts are usually studied.

How we made the graphs

Every chart above comes from one script, scripts/blog-figures/toi-2535b-vanishing-transits.mjs in fibub’s public code. It does not draw anything by hand: it calls the same functions that run inside the tools.

  • semiMajor, minMass and semiAmplitude from src/lib/sci/rvmass.ts, the maths behind the exoplanet mass calculator, turn the paper’s wobble (100 m/s), period and star mass into the planet’s mass and the size of its orbit.
  • fluxAt, durations and blocked from src/lib/sci/transit.ts, the model inside the transit fitter, draw each year’s dip, its length and its depth.
  • fitTransit from the same file is the fitter itself. We fed it simulated, noisy light curves and let it find the transit without being told the answer.

Each input carries a label in the script saying where it came from: the paper’s Table 1 or Table 2, the TESS catalogue, or our own choice. Run node scripts/blog-figures/toi-2535b-vanishing-transits.mjs and it prints every number quoted in this article and redraws the charts.

How we test the maths

The same functions are checked against published results every time fibub’s code changes (tests/sci.mjs, about 18,600 checks in all). For the tools used here:

  • The planet mass maths must give Jupiter’s pull on the Sun as 12.47 m/s, and the classic hot Jupiter 51 Pegasi b (a wobble of 55.94 m/s over 4.23 days) as 0.477 Jupiter masses at 0.053 au, matching the published values.
  • The transit model must give HD 209458 b, the first transiting planet found, a transit length of 3.07 hours from its published orbit.
  • The fitter must find the middle of a simulated transit within three of its own error bars.

If any of these fail, the change is not published.

Check it yourself

You can repeat Step 1 in a minute on the exoplanet mass calculator:

  1. Open the calculator and choose Mass from the wobble.
  2. Type 0.864 in Star mass (Suns), 8.236 in Orbital period (days) and 0 in Eccentricity.
  3. Type 100 in Velocity semi-amplitude K (m/s).
  4. The planet mass reads about 0.90 Jupiters (at least; type 90 in Inclination since it transits).
  5. To get the planet’s size, type 0.877 in Star radius (Suns) and 1.06 in Transit depth (%) (the depth of a planet 0.103 times the star’s width). The radius reads about 0.88 Jupiter radii.

To try the fitter on a real-looking transit, open the transit fitter, press Load the example, choose Near infrared or TESS band under Limb darkening, and compare the fitted transit length with the length you get after changing the impact parameter in your own data.

What researchers can use it for

  • Sanity-check a planet table: type a paper’s K, period and star mass into the exoplanet mass calculator and see whether the quoted planet mass follows.
  • Plan follow-up: the transit model shows how fast a grazing transit will fade, which tells you how soon to observe.
  • Teach transit shapes: the fitter’s example data and the impact parameter make a ready class exercise on why grazing transits look V-shaped.
  • Fit your own photometry: paste time and brightness from a small telescope into the transit fitter; it runs in the browser, so the data never leave your computer.

The fine print

  • The star’s mass and radius, the wobble and the seven impact parameters are the paper’s values; the planet radius is the TESS catalogue value; limb darkening is our choice. Small changes in the planet radius move our end date by a month or two.
  • Reading the 2019 crossings as being on the far side of the centre is our interpretation of the paper’s numbers, not something the paper states. It fits a straight line far better, but the paper’s full model, with more data, gives 0.16 a year.
  • Our light curves in Step 4 are simulated, not the TESS data. They show that the method works at this signal size, not that the real data say so.
  • The companion is inferred from its pull; it has not been seen directly. Its mass is a lower limit.
  • The swing model in Figure 7 treats the change in b as a simple to-and-fro. Real precession around two planets’ combined orbit is more complicated.
  • This is one system. It does not tell us on its own how often other transiting planets come and go.

Sources

  • Yee, S. W., Petigura, E. A., Brahm, R. et al. (2026), A TESS Disappearing Act: The Vanishing Transits of TOI-2535 b, The Astrophysical Journal Letters 1011, L60; arXiv:2611.13343 (posted 9 November 2026). Licence CC BY 4.0, checked on the arXiv page, 2026-11-12. Values used here (read 2026-11-16): orbital period 8.236 days and rate 0.16 per year (abstract, Sections I and II.1); star mass 0.864 (+0.036/-0.028) and radius 0.877 (+0.028/-0.027) solar, Teff 5256 ± 80 K (Table 1, EXOFASTv2); star wobble K ≈ 100 m/s (Section II.2); the seven TESS impact parameters in Table 2 (sectors 10, 37 and 63).
  • Planet radius 9.855 Earth radii: TESS Object of Interest catalogue entry TOI-2535.01 on ExoFOP (exofop.ipac.caltech.edu), read 2026-11-16. The paper's own fitted radius ratio is in its Appendix C, which we could not read.
  • Transits gone in early 2026, distance about 1,000 light-years and the 30 to 50 year wait: press coverage of the paper (UCLA newsroom, Sci.News), 2026-11-12 and 2026-11-16.
  • Transit length formula: Winn, J. N. (2010), Transits and Occultations, in Exoplanets (University of Arizona Press), equations 14 and 15; stellar density from a transit: Seager, S. and Mallén-Ornelas, G. (2003), ApJ 585, 1038.

If you are an author or publisher and want a credit or a passage changed or removed, tell us and we will act quickly.

Tools in this article

Read next