Hot Jupiter Demographics

I led the TESS Grand Unified Hot Jupiter Survey, which has confirmed >100 new planets from TESS. Planets the size of Jupiter span more two orders of magnitude in mass, ranging from little more than a Neptune mass to super-Jupiters many times more massive than our Jupiter.

Hot Jupiters – giant planets that orbit their stars in less than ten days – were among the first exoplanets to be discovered. Today, we know of hundreds of them, and some individual planets have been studied in great detail, including atmospheric characterization by JWST (e.g. WASP-39b). The discovery of such large planets so close in to their stars came as a surprise to astronomers, and three decades later, we still do not have a good idea of how they formed.

Answering these questions requires studying hot Jupiters as a population. Most of the previously known hot Jupiters were discovered by various ground-based surveys, such as HAT and WASP. Each of these surveys had different selection criteria and biases, which has made combining their discoveries into a single sample difficult. Hot Jupiters are also relatively rare, occurring around only 1% or less of Sun-like stars. This meant that NASA’s Kepler mission discovered only 40 such planets. As a result, it has been challenging to study hot Jupiters as a population.

NASA’s Transiting Exoplanet Survey Satellite (TESS) mission is searching stars across the entire sky for planets. This means that it is “re-discovering” all of the previously known hot Jupiters, while also finding new ones that previous surveys missed. I intiated and led the TESS Grand Unified Hot Jupiter Survey, a coordinated effort to confirm and characterize TESS planet candidates in order to construct a magnitude-limited sample of 400 hot Jupiters orbiting Sun-like stars brighter than G < 12.5 mag. Working with members of the the TESS Follow-up Observing Program, I led the discovery of >100 new planets from TESS (Yee et al. 2022, 2023, 2025, 2026 in prep.). In combination with other members of the community, we have completed the largest ever statistical sample of hot Jupiters.

The first results confirmed the “three-day pile-up” of hot Jupiters and show that its position does not depend strongly on host star metallicity (Yee & Winn 2023). My collaborators have modeled the observed hot Jupiter period distribution to place constraints on tidal migration (Ma, Su, Yee et al. 2026). I am now investigating other trends in hot Jupiter occurrence, such as with stellar mass, where I have extended my survey one magnitude fainter around K stars. Stay tuned for more of these demographic results!

Transit Timing Variations and Orbital Architectures

While most hot Jupiters are found to be “lonely”, without any nearby planets, (e.g.Huang et al. 2016, Sha et al. 2026), an interesting subset of them have close-by planetary companions. One way to detect such companions is to see them transit their host stars, but non-transiting planets may be detected by their influence on the transiting hot Jupiter. Because of gravitational interactions between the hot Jupiter and its companion, the hot Jupiter’s transits may not arrive exactly like clockwork, but instead deviate in their timing (transit timing variations; TTVs) or shape (transit duration variations; TDVs).

I have been conducting a uniform search for TTVs and TDVs in the homogeneous hot Jupiter sample from the Grand Unified Survey. Initial results include the discovery of some of the largest TTVs and TDVs ever seen for hot Jupiters. One case is the TOI-3686 system (Yee et al. 2026b), which contains a hot Jupiter showing peak-to-peak TTVs of 40 minutes, and an inner sub-Neptune with TTVs of a whopping 10 hours, 16% of its orbital period. This is the first hot Jupiter system in a true mean-motion resonance with a nearby planet.

Transit timing variations of the sub-Neptune TOI-3686 b (P = 2.54 days) and the hot Jupiter TOI-3686 c (P = 5.08 days), which are locked in a 2:1 resonance.

Even more spectacular is the TOI-2535 system (Yee et al. 2026a), where not only does the hot Jupiter show peak-to-peak TTVs of 12 hours, but its transits gradually got shorter and shallower, until they vanished completely. This is the first planet that we have directly observed its transits to disappear, and is the result of nodal precession induced by a nearby, mutually inclined planet. Apart from TESS data, we were able to characterize this system in detail thanks to many collaborators across five continents who observed the transits from the ground.

The transits of TOI-2535 b gradually decreased in duration and depth over seven years of observations, the result of a changing sky-plane inclination. These orbital changes are due to the presence of a mutually inclined giant planet perturber, causing both planets' orbits to precess around the system invariable plane.

Other examples have been discovered by my search, including a planet whose transits have “appeared” for TESS, a nice complement to the disappearing transits of TOI-2535. I am also working on measuring the frequency of TTVs in hot Jupiter systems.

Sub-Saturns

Giant planets with masses comparable to Neptune may originally be giant planets that underwent significant mass loss during high-eccentricity migration. Such a scenario can help explain the period distribution of these sub-Saturns as well as solve challenges with the long tidal migration timescales for planets in this mass regime.

While hot Jupiters are mysterious, a subset of them – the sub-Saturns – are even more enigmatic. These are gas giants with masses less than Saturn, with some even as light as Neptune. These objects are thought to be below the mass threshold for runaway gas accretion, so how did they form?

One possibility is that these sub-Saturns are “hot Jupiters gone wrong”: they began as more massive giant planets but then lost a significant fraction of their mass. UCLA PhD student Grant Weldon’s simulations show that this is not only plausible, but can also solve other challenges with their formation, such as the long timescales required for them to migrate inward. We wrote a paper on the discovery of TOI-2195 A b, a planet only 1.5 times the mass of Neptune but with a radius close to Jupiter. Grant showed that this planet likely formed by losing up to 90% of its mass during high-eccentricity migration, and he is now turning this case study toward a broader population analysis.

Another line of evidence for a high-eccentricity formation history of sub-Saturns comes from the observation that many of them are on orbits highly misaligned with their host stars’ spin axes. This angle can be measured with the Rossiter-McLaughlin effect, an anomalous shift in the stellar radial-velocity as a planet occults parts of the stellar disk that are blue- or red-shifted by rotation. I have been leading a survey to measure sub-Saturn obliquities (Yee et al. 2025a, 2025c, Tamburo, Yee et al. 2025a, 2025b), with current results appearing to confirm a broad diversity in stellar obliquities. This work includes a close collaboration with Pat Tamburo and the Tierras team to measure stellar rotation periods for slowly-rotating stars and thereby infer true, three-dimensional stellar obliquities, making headway on the mysterious “preponderance of polar planets”.

The angle between a planet's orbit normal and the stellar spin axis -- the stellar obliquity -- can be measured through the Rossiter-McLaughlin effect, an anomalous RV shift while a planet is transiting its host star. The left plot shows the RM effect for well-aligned TOI-2364 b, while the right plot shows the misaligned sub-Saturn, TOI-2374 b.

Giant Planets in a Galactic Context

At low [Fe/H], Milky Way stars bifurcate into two populations with different abundances of elements produced by the alpha-process. In DiTomasso, Yee et al. (2026), we performed a homogeneous spectroscopic characterization of hot Jupiter host stars to identify new planets that orbit high-α stars that are part of the galactic thick disk.

One underexplored axis in exoplanet demographics is how planet formation and occurrence depend on galactic context. For example, our Milky Way contains three major populations of stars that differ in their kinematics and chemistry: the thin disk, thick disk, and halo. The different environments of stellar and planet formation, as well as the billions of years of evolution that the older thick disk and halo populations have experienced, likely affect the frequency and types of planets in each population.

I have been laying the groundwork to perform detailed demographics studies of giant planets in these different galactic contexts. Amongst the early results include the discovery of the first transiting brown dwarf around a thick disk star, led by Harvard PhD student Jéa Adams Redai. With Victoria DiTomasso, now an NPP Fellow at NASA Goddard, we conducted a homogeneous spectroscopic analysis of hot Jupiter host stars to identify new confirmed and candidate planets in the galactic thick disk, finding tentative differences in the properties of the two populations. We have since confirmed more such planets to be published soon (DiTomasso, Yee et al. in prep.).

Stay tuned for further updates in this area! I am actively working on understanding the relative occurrence rates of giant planets around thick disk and thin disk stars in the metal-poor regime.

Stellar Characterization

Performance of SpecMatch-Emp in an internal cross-validation test. Black points denote the library properties of a given star, red lines point to the derived properties. Especially along the main sequence, SpecMatch-Emp performs extremely well.

Our methods for detecting and characterizing exoplanets depend heavily on a detailed understanding of their host stars. For instance, the transit detection method measures the tiny dip in a star’s brightness as a planet passes in front of it, with the size of this dip depending on the ratio of the planet’s to star’s radius.

Working with Erik Petigura, I developed SpecMatch-Empirical, a new tool for characterizing the effective temperature, radius, and metallicity of a star from its spectrum. SpecMatch-Emp compares a target spectrum with a library of high resolution, high signal-to-noise observed spectra from stars with empirically-determined properties. Using empirical instead of synthetic spectra allows us to avoid model-dependent uncertainties particularly for late-type stars, and we achieve a precision of ~70 K in effective temperature and <10% in stellar radius.

SpecMatch-Emp has been applied to spectra from many high-resolution spectrographs around the world, including Keck/HIRES, Magellan/PFS, and HARPS. More information can be found in Yee, Petigura and Von Braun (2017), as well as at the documentation page. Please reach out to me if you run into any issues, bugs, or need any help applying it to your data!

SpecMatch-Emp identifies the five best-matching library spectra to the target, in this case the G-type star HD 190406, and synthesizes a linear combination to create an even better match to the target spectrum. The weights of the linear combination are used to derive a set of stellar parameters for the target.

Dynamical Histories of Planetary Systems

Early astronomers saw our solar system as a clockwork mechanism – planets orbiting the sun, moons orbiting the planets, forever and unchanging. But we have since learned that things are more complicated than that. In our Solar System, the orbits of the planets are thought to have migrated over long distances, while our own moon is thought to have formed as a result of a giant impact between a Mars-sized object and the Earth. And as the Sun ages and turns into a red giant, the inner planets (including possibly the Earth), will eventually be engulfed. Our Solar System is a dynamic and sometimes violent place. What can we learn about the history of other planetary systems?

One fascinating planet I studied is the hot Jupiter WASP-12b. Building on a decade of data, particularly those compiled by Patra et al. (2017), and adding new observations from NASA’s Spitzer Space Telescope, we found the decisive evidence showing that the orbit of this planet is decaying, and it will spiral into its star and be destroyed in just a few million years! See press coverage from Princeton and Popular Science for more details.

We obtained new transit and secondary eclipse times of the planet WASP-12b (black squares) and showed that the apparent variation in those times can be attributed to a slow shrinking of the planet's orbit (blue line) rather than apsidal precession of a slightly eccentric orbit (red).

Apart from hot Jupiters, another type of planetary system we have discovered that are quite different from our own contain many planets on relatively short, tightly-packed orbits. Typically, these planets are slightly bigger than the Earth (we call them super-Earths and sub-Neptunes), and a planetary system may have three or more such planets all orbiting closer to their star than the Earth is to the Sun. One hypothesis that had been put forward is that these systems may have been sculpted by constant gravitational interactions leading to chaos and instability. Working with Dan Tamayo and Sam Hadden, we showed that this cannot have been the only process at work, and some dissipiation must have taken place to reduce these planets’ eccentricities and inclinations (Yee et al. (2021a)).

We analyzed the planetary systems discovered by the Kepler mission that showed transit-timing variations (TTVs). The measured eccentricities in these systems are all significantly lower than the limits placed on them by stability, suggesting they must have been damped by some dissipative process.