Kepler-90 g: A Super-Puff Exoplanet

• Kepler-90 g is a gas-rich super-puff exoplanet in the eight-planet Kepler-90 system, characterized by a low density and strong transit-timing variations.
• Its orbital parameters, including a ~211-day period and semi-major axis near 0.7 AU, are precisely constrained using combined TTV and RV analyses.
• Dynamical studies show that Kepler-90 g plays a critical role in the system's near-resonant architecture and long-term stability.

Searching arXiv for the cited Kepler-90 and Kepler-90 g papers to ground the article in the relevant literature. Kepler-90 g is the second-outermost known planet in the Kepler-90 system, an eight-planet transiting architecture around a roughly solar-type star. In the observational literature it is the outer gas-rich planet interior to Kepler-90 h, with a semi-major axis near $0.7$ AU, a period of about $211$ days, and some of the largest transit-timing variations measured in a Kepler multi-planet system. Subsequent dynamical analyses constrained its mass to $15.0^{+0.9}_{-0.8}\,M_\oplus$ and later to $15.0 \pm 1.3\,M_\oplus$, while its radius of $8.1 \pm 0.8\,R_\oplus$ implies an apparent bulk density of $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$, placing it in the “super-puff” regime (Liang et al., 2020, Shaw et al., 18 Jul 2025).

1. Discovery, designation, and observational history

Kepler-90 g entered the literature as KIC 11442793 g / KOI-351.02 in the seven-planet discovery and validation study of the Kepler-90 system. In that work, planets d, g, and h had been previously reported, while b, c, e, and f were newly presented; the system was described as a compact analogue to the Solar System, with small planets inside and gas giants outside (Cabrera et al., 2013). Within that seven-planet architecture, Kepler-90 g was already identified as one of the two outer gas giants and as the object showing especially strong dynamical interactions with Kepler-90 h (Cabrera et al., 2013).

A later nomenclature issue is sometimes a source of confusion. The deep-learning discovery paper that brought Kepler-90 to eight known transiting planets did not concern Kepler-90 g itself; it added a distinct inner planet, designated Kepler-90 i, with $P = 14.44912 \pm 0.00020$ days and $R_p = 1.32 \pm 0.21\,R_\oplus$ (Shallue et al., 2017). That eighth-planet discovery matters for Kepler-90 g chiefly because it changed the system from the original seven-planet architecture used in earlier dynamical and polytropic analyses to the eight-planet architecture used in later work.

The observational baseline for Kepler-90 g has remained unusually sparse for such a dynamically informative object. Over the four years of the Kepler mission, six transits of g were recorded, and later work extended that record with one Swift transit, one Spitzer transit, and a 2024 ground-based transit recovered after prediction from joint transit-timing and radial-velocity modeling (Liang et al., 2020, Shaw et al., 18 Jul 2025). This long but sparsely sampled baseline is central to both the power and the limitations of the planet’s dynamical characterization.

2. Orbital and transit properties

In the original seven-planet transit analysis, Kepler-90 g was fitted with a mean orbital period of $P_g = 210.60697 \pm 0.00043$ days, a transit duration of $D_g = 12.593 \pm 0.045$ hours, a scaled semi-major axis $211$0, a semi-major axis $211$1 AU, an impact parameter $211$2, and an inclination $211$3 (Cabrera et al., 2013). These values place the planet on a nearly edge-on transiting orbit at a moderate orbital distance, interior to Kepler-90 h but exterior to the inner compact chain.

The later full dynamical TTV solution for the outer pair g and h reported $211$4 days, $211$5, $211$6 deg, $211$7 deg, and $211$8 deg, with the longitude of ascending node of g fixed to $211$9 as a reference direction (Liang et al., 2020). That analysis emphasized that the eccentricity is modest but significantly non-zero and larger than what simple long-term stability arguments had previously assumed for the system (Liang et al., 2020).

The most precise updated linear ephemeris came from the later joint RV + all-transits analysis. That fit gave $15.0^{+0.9}_{-0.8}\,M_\oplus$0 days and $15.0^{+0.9}_{-0.8}\,M_\oplus$1, or equivalently

$15.0^{+0.9}_{-0.8}\,M_\oplus$2

The same analysis derived $15.0^{+0.9}_{-0.8}\,M_\oplus$3 and $15.0^{+0.9}_{-0.8}\,M_\oplus$4 deg from the fitted variables $15.0^{+0.9}_{-0.8}\,M_\oplus$5 and $15.0^{+0.9}_{-0.8}\,M_\oplus$6 (Shaw et al., 18 Jul 2025). The literature therefore contains multiple period estimates for g, each tied to a specific data span and ephemeris model rather than to a single immutable orbital parameterization.

3. Transit-timing variations and dynamical determination

Kepler-90 g is notable for the scale and structure of its transit-timing variations. The seven-planet discovery paper reported that the orbit of planet g is perturbed such that its orbital period changes by $15.0^{+0.9}_{-0.8}\,M_\oplus$7 hours between two consecutive transits during the observing baseline, described there as the largest such perturbation found so far (Cabrera et al., 2013). The later dedicated TTV study similarly described TTVs “up to 25 hours,” with the sixth observed transit showing a $15.0^{+0.9}_{-0.8}\,M_\oplus$8-hour anomaly relative to a linear ephemeris (Liang et al., 2020).

The 2020 TTV analysis treated the outer system as a three-body problem consisting of the star, Kepler-90 g, and Kepler-90 h. Transit times and durations were extracted from Kepler light curves with the Fourier Gaussian process method, using PDCSAP flux, long-cadence data, and short-cadence data when available, with non-Gaussian noise modeled through a mixture model. The dynamical interpretation used TTVFast as a non-relativistic $15.0^{+0.9}_{-0.8}\,M_\oplus$9-body integrator over $15.0 \pm 1.3\,M_\oplus$0 days with a $15.0 \pm 1.3\,M_\oplus$1-day step size, and a 13-parameter fit optimized by basin-hopping with L-BFGS-B and then sampled with emcee using 128 walkers and more than $15.0 \pm 1.3\,M_\oplus$2 million samples (Liang et al., 2020).

That fit yielded

$15.0 \pm 1.3\,M_\oplus$3

for Kepler-90 g and

$15.0 \pm 1.3\,M_\oplus$4

for Kepler-90 h, while reproducing the large TTV excursion of g with residuals of order $15.0 \pm 1.3\,M_\oplus$5 minute and achieving a total $15.0 \pm 1.3\,M_\oplus$6 for 18 data points (Liang et al., 2020). The resonance angles

$15.0 \pm 1.3\,M_\oplus$7

were found to circulate over $15.0 \pm 1.3\,M_\oplus$8–$15.0 \pm 1.3\,M_\oplus$9, implying that the pair lies close to but is not locked in a 2:3 resonance, despite a period ratio of about $8.1 \pm 0.8\,R_\oplus$0 (Liang et al., 2020).

The 2025 update combined 34 HIRES/Keck radial velocities spanning 2011–2022 with Kepler transit times and later post-Kepler transit detections. The key result for g was a mass of $8.1 \pm 0.8\,R_\oplus$1, exactly consistent in central value with the earlier TTV-only determination, together with an ephemeris precise enough to predict transit midpoints through 2029 with uncertainties of order $8.1 \pm 0.8\,R_\oplus$2 minutes (Shaw et al., 18 Jul 2025). A notable empirical result was that the 2024 transit occurred about $8.1 \pm 0.8\,R_\oplus$3 days later than the initial joint RV + Kepler TTV ephemeris had predicted, demonstrating that the Kepler-era baseline alone did not adequately constrain the long-term average period (Shaw et al., 18 Jul 2025).

4. Radius, density, and super-puff interpretation

The radius adopted for Kepler-90 g from independent transit-depth and RV work is

$8.1 \pm 0.8\,R_\oplus$4

a value already present in the original seven-planet study and carried forward into later physical interpretation (Cabrera et al., 2013, Liang et al., 2020). Combined with the TTV mass, this yields an apparent bulk density

$8.1 \pm 0.8\,R_\oplus$5

using $8.1 \pm 0.8\,R_\oplus$6 (Liang et al., 2020). In the terminology of the TTV paper, a super-puff is any planet with apparent density $8.1 \pm 0.8\,R_\oplus$7, so Kepler-90 g is not merely within that class but lies well below the threshold (Liang et al., 2020).

The later RV + TTV study summarized the same result more qualitatively by describing Kepler-90 g as Saturn-sized and Neptune-mass, and therefore consistent with the super-puff class (Shaw et al., 18 Jul 2025). In comparative terms supplied in the TTV analysis, its density is far below that of Neptune and even below that of Saturn (Liang et al., 2020).

Several physical interpretations have been discussed, but none is established as definitive. The TTV study considered a highly extended low-mean-molecular-weight atmosphere, dusty atmospheres or outflows that inflate the apparent transit radius, formation at larger distances with subsequent migration, and large optically thick rings. The authors reported no convincing evidence for a tilted ring in the Kepler photometry, while emphasizing that the data were not sensitive enough to rule out most ring configurations; they therefore remained agnostic about whether dust, rings, or a genuinely high-entropy envelope dominates the explanation (Liang et al., 2020). A plausible implication is that the term “super-puff” for Kepler-90 g is robust as a phenomenological density label, even though the underlying atmospheric or structural mechanism remains unresolved.

5. Position in the Kepler-90 architecture and dynamical context

Kepler-90 is one of the highest-multiplicity exoplanet systems known and has been repeatedly compared with the Solar System because it contains small inner planets and larger outer planets, but compressed so that all eight known planets lie within 1 AU (Shallue et al., 2017, Shaw et al., 18 Jul 2025). Within that architecture, Kepler-90 g sits between the inner compact chain and the outer giant h, at a moderate distance near $8.1 \pm 0.8\,R_\oplus$8 AU (Liang et al., 2020).

The original seven-planet dynamical study emphasized that g is dynamically linked both inward and outward. It described g and h as close to an 8:5 mean-motion resonance, g and f as close to 5:3, and the d-e-f triplet as near a 2:3:4 chain, leading to a network of coupled interactions and a system close to instability under some assumed masses and eccentricities (Cabrera et al., 2013). By contrast, the later TTV analysis of the outer pair found the observed g-h configuration to be stable over billions of years, Hill stable under the quoted two-planet criterion, and apsidally aligned with $8.1 \pm 0.8\,R_\oplus$9 librating within $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$0, while not presently in a 2:3 resonant lock (Liang et al., 2020).

A further dynamical case study examined Kepler-90 as a system with tightly packed inner planets in the presence of a hypothetical outer Jupiter-like perturber. In that paper, Kepler-90 g was assigned an assumed mass of $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$1, semi-major axis $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$2 au, initial eccentricity $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$3, and inclination $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$4, with the authors explicitly noting that the nominal masses of Kepler-90 planets were unknown (Contreras et al., 2018). Under those adopted parameters the study identified a near-3:2 g-h mean-motion resonance with a librating resonant angle, and found that an outer perturber can reorganize the system’s secular structure without destabilizing g itself (Contreras et al., 2018). This suggests an important methodological distinction: measured TTV-based masses for g are near $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$5, whereas some secular experiments used substantially larger assumed masses as part of exploratory stability analyses (Liang et al., 2020, Contreras et al., 2018).

6. Alternative orbital-spacing models and principal caveats

Kepler-90 g has also been modeled in the “global polytropic model,” which treats a planetary system as if it were in hydrostatic equilibrium and solves the Lane-Emden equation in the complex plane to define polytropic shells (Geroyannis, 2014, Geroyannis, 2020). In the 2014 seven-planet application with fixed stellar radius, the optimum Kepler-90 model used $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$6 and placed g in shell 9 at the maximum-density orbit $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$7 AU, versus an observed $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$8 AU, for a percentage error of about $0.15 \pm 0.05\,{\rm g\,cm^{-3}}$9 (Geroyannis, 2014).

In the 2020 two-dimensional version, which treated the host-star radius and polytropic index as free scanning variables and included the eight-planet system, the optimum fit had $P = 14.44912 \pm 0.00020$0, $P = 14.44912 \pm 0.00020$1, and placed Kepler-90 g in shell 10 at the maximum-density orbit $P = 14.44912 \pm 0.00020$2 AU, compared with the adopted $P = 14.44912 \pm 0.00020$3 AU and $P = 14.44912 \pm 0.00020$4 (Geroyannis, 2020). In that framework, g is one of the better-fitted planets in the system, and no additional planets are explicitly predicted in its shell even though the formalism allows up to three special orbits per shell (Geroyannis, 2020).

These shell assignments should be read as model-dependent constructs rather than as dynamical measurements. The authors themselves characterize the shell structure as arising from the complex-plane Lane-Emden solution and note that the hydrostatic-equilibrium assumption at the scale of an entire planetary system is non-standard; the method aims to match orbital radii, not planetary masses, stability, or formation pathways (Geroyannis, 2020). More broadly, several recurring caveats apply to Kepler-90 g across the literature: the number of observed transits remains small; RVs alone do not detect g; some dynamical papers used assumed rather than measured masses; and even where the mass is now well constrained, the physical origin of the planet’s extreme apparent low density remains uncertain (Shaw et al., 18 Jul 2025, Contreras et al., 2018).

Kepler-90 g therefore occupies a distinctive position in exoplanet research: it is simultaneously a benchmark TTV system, a well-measured low-density outer transiting planet, a test case for near-resonant long-term stability, and an object whose internal structure remains open to competing interpretations.