HAT-P-12 b: Sub-Saturn Exoplanet Analysis

• HAT-P-12 b is a transiting exoplanet with a sub-Saturn mass and near-Jovian radius, making it a prime candidate for testing gas-giant structure models.
• High-precision transit photometry has refined its nearly edge-on orbital geometry and compact orbit, solidifying our understanding of its stellar and planetary parameters.
• Atmospheric studies reveal muted water features and haze-dominated spectra, while transit-timing variations suggest potential dynamical influences from additional companions.

HAT-P-12 b is a transiting exoplanet orbiting the K-dwarf star HAT-P-12 A, also known as GSC 3033-706 and 2MASS J13573347+4329367. It is classified as a sub-Saturn-mass, gas-giant planet: its mass is significantly below Saturn’s but its radius is close to that of Jupiter, so it is a very low-density, H/He-dominated planet. Orbiting a metal-poor K4 dwarf on a short-period orbit of about 3.21306 d, HAT-P-12 b has been used to test models of irradiated gas-giant structure, atmospheric aerosols, spin-orbit geometry, and transit-timing behavior (Lee et al., 2012).

1. System identification and observational setting

HAT-P-12 b was originally discovered by Hartman et al. (2009), and subsequent work consistently describes the host as a relatively cool, late-K dwarf with subsolar metallicity. Parameters adopted or remeasured in later studies include $T_{\rm eff}=4650\pm 60$ K and $[\mathrm{Fe/H}]=-0.29\pm 0.05$, while HARPS-N spectroscopy gave $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$ and $[\mathrm{Fe/H}] = -0.20 \pm 0.09$. The star is faint in the optical, with $V \approx 12.8$, and the system geometry is nearly edge-on, enabling repeated high-quality transit measurements (Mancini et al., 2018).

The planet has been described as a “warm Saturn,” a “hot Saturn,” and a “sub-Saturn type” planet in different analyses. Those labels refer to the same basic physical regime: a low-mass gas giant with a near-Jovian radius, low surface gravity, and equilibrium temperature close to $10^3$ K. This combination makes the system simultaneously useful for interior-structure studies and for transmission spectroscopy, because a clear H/He atmosphere would be expected to have a relatively large atmospheric scale height (Line et al., 2013).

The orbital architecture is compact. Early physical analyses placed the semi-major axis at $a = 0.03829 \pm 0.00046\,\mathrm{AU}$, while later homogeneous modeling gave $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$. The orbit is generally treated as circular in photometric and Rossiter-McLaughlin analyses, following the conclusion that the eccentricity is consistent with zero in the literature, although later transit-timing work explored a small nonzero eccentricity in an apsidal-precession framework (Lee et al., 2012).

2. Transit photometry and ephemeris refinement

A major early reanalysis was based on three new transit light curves obtained in 2011 with the 1.0-m telescope at Mt. Lemmon Optical Astronomy Observatory using a Cousins $R_{\rm C}$ filter. The telescope was defocused to improve photometric precision; exposure times were 40–70 s, and the resulting transit series contained several hundred points per event with an rms scatter of about 2.1 mmag. These data were modeled with JKTEBOP using a quadratic limb-darkening law, with the linear coefficient fitted and the nonlinear coefficient fixed and perturbed in the error analysis (Lee et al., 2012).

The transit fit used the standard fractional radii $r_{\rm A}=R_{\rm A}/a$ and $[\mathrm{Fe/H}]=-0.29\pm 0.05$0, but the fitted combinations were the sum and ratio in order to reduce parameter correlations: $[\mathrm{Fe/H}]=-0.29\pm 0.05$1 The inclination was found to be almost perfectly edge-on, $[\mathrm{Fe/H}]=-0.29\pm 0.05$2, with fitted or derived values $[\mathrm{Fe/H}]=-0.29\pm 0.05$3, $[\mathrm{Fe/H}]=-0.29\pm 0.05$4, $[\mathrm{Fe/H}]=-0.29\pm 0.05$5, and $[\mathrm{Fe/H}]=-0.29\pm 0.05$6. Uncertainties were estimated from 10,000 Monte Carlo simulations and residual-permutation analysis, with the larger error adopted (Lee et al., 2012).

Using 18 mid-transit times spanning about 4.2 yr, an improved linear ephemeris was derived: $[\mathrm{Fe/H}]=-0.29\pm 0.05$7 In explicit form, this corresponds to $[\mathrm{Fe/H}]=-0.29\pm 0.05$8 BJD and $[\mathrm{Fe/H}]=-0.29\pm 0.05$9 d, with all times expressed in BJD(TDB). The timing set combined the three 2011 LOAO transits, four mid-transit times re-derived from Hartman et al. (2009), and eleven ETD timings transformed from HJD(UTC) to BJD(TDB). At that stage, the $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$0 diagram showed short-term oscillations, but no statistically significant periodic TTV signal was found; systematic errors and starspot-induced asymmetries were preferred over a clear dynamical interpretation (Lee et al., 2012).

Later photometric work within the GAPS programme added 17 new light curves obtained with INT, Calar Alto, and Cassini facilities, again using JKTEBOP with quadratic limb darkening and Monte Carlo plus prayer-bead error analysis. The resulting weighted photometric solution gave $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$1, $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$2, $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$3, $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$4, and $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$5, fully compatible with earlier determinations while tightening the uncertainties (Mancini et al., 2018).

3. Stellar and planetary properties

Combining transit geometry, spectroscopic inputs, empirical calibrations from eclipsing binaries, and stellar evolutionary models, one detailed analysis concluded that the BCAH models with $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$6 gave the most physically consistent solution for this metal-poor, low-mass star. That preference was explicitly motivated by the fact that empirical calibrations are relatively poorly populated at subsolar metallicity and $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$7 (Lee et al., 2012).

The adopted BCAH stellar solution gave $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$8, $T_\mathrm{eff}=4665 \pm 50\ \mathrm{K}$9, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$0, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$1, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$2, and an age of $[\mathrm{Fe/H}] = -0.20 \pm 0.09$3 Gyr. On that scale, HAT-P-12 A is a cool, low-mass K dwarf, about 30% less massive and smaller than the Sun, but nearly three times denser and with only about 20% of the Sun’s luminosity (Lee et al., 2012).

The corresponding planetary parameters were $[\mathrm{Fe/H}] = -0.20 \pm 0.09$4, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$5, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$6, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$7, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$8, $[\mathrm{Fe/H}] = -0.20 \pm 0.09$9, and $V \approx 12.8$0. The mass is therefore below Saturn’s, whereas the radius is almost Jupiter’s, implying a puffy, low-density H/He envelope rather than an ice-giant-like bulk structure (Lee et al., 2012).

A later homogeneous analysis based on HARPS-N spectroscopy and new photometry returned $V \approx 12.8$1, $V \approx 12.8$2, $V \approx 12.8$3, $V \approx 12.8$4, $V \approx 12.8$5, and $V \approx 12.8$6. Those values were explicitly described as fully compatible with the existing literature, so the physical picture of HAT-P-12 b as a cool, low-density short-period gas giant has remained stable across different analyses (Mancini et al., 2018).

Interior-structure comparisons with Fortney et al. (2007) irradiated gas-giant models initially placed HAT-P-12 b among Saturn-mass planets with negligible inferred core masses, $V \approx 12.8$7, around subsolar-metallicity hosts (Hartman et al., 2010). A later interpolation using a solar-equivalent semi-major axis of $V \approx 12.8$8 AU and an age of 3.2 Gyr found that the measured $V \approx 12.8$9 lies closest to a model with $10^3$0, and that the system agrees well with standard models of irradiated gas giants (Lee et al., 2012). This suggests a transition in interpretation from “negligible or very small core” to “moderately massive heavy-element core,” rather than a qualitative change in the planet’s classification.

4. Atmosphere, transmission spectrum, and dayside emission

HAT-P-12 b became an important target for transmission spectroscopy because its low mass, near-Jovian radius, and moderate temperature imply a relatively large scale height for any clear H/He atmosphere. A first near-infrared transmission spectrum with HST/WFC3 G141 covered 1.037–1.721 µm and used one full transit observed in 2011 in staring mode, with 111 exposures over four HST orbits. After explicit modeling of WFC3 staring-mode systematics, including the ramp or hook, the measured transmission spectrum appeared essentially flat across the 1.1–1.7 µm range and lacked the expected 1.4 µm water band of a clear, solar-composition atmosphere. A frequentist hypothesis test ruled out a hydrogen-dominated cloud-free atmosphere at 4.9$10^3$1, while a solar-composition atmosphere with a high, gray, opaque cloud deck at 1 mbar was statistically consistent with the data (Line et al., 2013).

Warm Spitzer/IRAC secondary-eclipse photometry at 3.6 and 4.5 µm initially yielded only $10^3$2 upper limits, $10^3$3 and $10^3$4, with corresponding brightness temperature limits $10^3$5 and $10^3$6. The non-detection was interpreted as more likely reflecting weak planetary emission than a large eclipse-timing offset caused by eccentricity, and inverted Burrows models were described as poor fits to those limits. Non-inverted, moderately redistributed atmospheres were closer to the allowed range, but the cloud-free model grids still tended to overpredict the observed fluxes (Todorov et al., 2013).

A broader 0.3–5.0 $10^3$7m analysis later combined HST/STIS, HST/WFC3, and Spitzer transit data and obtained a more differentiated atmospheric picture. That study detected a muted 1.4 $10^3$8m water vapor absorption feature attenuated by clouds, as well as a Rayleigh scattering slope in the optical indicative of small particles. The same work reanalyzed secondary eclipses and measured depths of $10^3$9 at 3.6 $a = 0.03829 \pm 0.00046\,\mathrm{AU}$0m and $a = 0.03829 \pm 0.00046\,\mathrm{AU}$1 at 4.5 $a = 0.03829 \pm 0.00046\,\mathrm{AU}$2m, consistent with a blackbody temperature of $a = 0.03829 \pm 0.00046\,\mathrm{AU}$3 K and with efficient day-night heat recirculation (Wong et al., 2020).

Atmospheric interpretation in that joint study used the retrieval framework SCARLET and the aerosol microphysics model CARMA. SCARLET indicated that the atmosphere is consistent with a broad range of metallicities between several tens to a few hundred times solar, a roughly solar C/O ratio, and moderately efficient vertical mixing. The preferred aerosol particles were sub-micron in size, with retrieval constraints corresponding to $a = 0.03829 \pm 0.00046\,\mathrm{AU}$4, and a cloud-top pressure that remained uncertain but clearly placed opacity in the upper atmosphere. CARMA showed that condensate clouds do not readily generate the sub-micron particles required to reproduce the observed Rayleigh scattering slope, whereas photochemical hazes composed of soot or tholins are able to match the full transmission spectrum (Wong et al., 2020).

The atmospheric literature therefore evolved from an early WFC3-only result favoring a flat spectrum and high-altitude clouds to a broader, multi-instrument picture in which water is present but muted and the optical continuum is haze-dominated. A plausible implication is that the earlier “water-free” appearance was a consequence of limited wavelength coverage and precision rather than a genuinely molecule-free atmosphere.

5. Rossiter-McLaughlin effect, obliquity, and stellar activity

The first measurement of the Rossiter-McLaughlin signal in the system was obtained with HARPS-N from in-transit radial velocities taken on 2015-03-13 and 2015-04-24, with 900 s exposures. Modeling of the anomaly yielded a sky-projected obliquity

$a = 0.03829 \pm 0.00046\,\mathrm{AU}$5

and a projected stellar rotation velocity from the RM fit of $a = 0.03829 \pm 0.00046\,\mathrm{AU}$6, consistent with the spectroscopic value $a = 0.03829 \pm 0.00046\,\mathrm{AU}$7. The RM detection was statistically significant, with $a = 0.03829 \pm 0.00046\,\mathrm{AU}$8 in favor of including the RM effect (Mancini et al., 2018).

The sign and magnitude of $a = 0.03829 \pm 0.00046\,\mathrm{AU}$9 imply a substantially misaligned but still prograde orbit in projection. HAT-P-12 b was therefore identified as an exception to the simplest empirical trend in which cool stars with $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$0 K tend to host aligned planets. Because the host rotates very slowly in projection and the RM amplitude is correspondingly small, the exact projected angle remains only moderately constrained, but alignment is disfavored (Mancini et al., 2018).

The stellar-activity diagnostics support the picture of an old, slowly rotating K dwarf. HARPS-N Ca II H&K measurements yielded $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$1 and $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$2, with $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$3 and $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$4. Activity-based calibrations implied rotation periods of about 43–45 d, and dedicated photometric monitoring in later atmospheric work showed no detectable variability at $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$5, consistent with a quiet star (Mancini et al., 2018).

No true three-dimensional obliquity $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$6 has been derived, because there is no secure photometric rotation period and the measured $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$7 is very small. Thus, for HAT-P-12 b the available spin-orbit information remains the projected angle $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$8, not the true obliquity.

6. Transit-timing variations, dynamical hypotheses, and population context

The timing history of HAT-P-12 b illustrates a genuine chronological shift in interpretation. The 18-timing analysis spanning about 4.2 yr found no statistically significant periodic TTV signal and favored systematic errors, red noise, or starspot-induced asymmetries as the explanation for the residual structure (Lee et al., 2012). A much longer baseline study, however, analyzed 46 light curves from 2009–2022, including 7 new ground-based transits, three TESS sectors, and 23 previously published light curves, and concluded that a linear ephemeris is inadequate (Parthasarathy et al., 13 Mar 2025).

In that later work, the linear model gave $a = 0.03767 \pm 0.00057 \pm 0.00027\ \mathrm{au}$9. An orbital-decay model improved the fit to $R_{\rm C}$0, but the implied stellar tidal quality factor,

$R_{\rm C}$1

was described as far below the theoretical expectations of $R_{\rm C}$2–$R_{\rm C}$3, making orbital decay an unlikely explanation. An apsidal-precession model performed better, with $R_{\rm C}$4, $R_{\rm C}$5, and $R_{\rm C}$6, implying a slight orbital eccentricity and a measurable precession rate (Parthasarathy et al., 13 Mar 2025).

The preferred description in that analysis was a sinusoidal TTV model. A Generalized Lomb-Scargle periodogram identified a significant frequency at $R_{\rm C}$7 cycles/day with false alarm probability $R_{\rm C}$8. The best sinusoidal solution had amplitude $R_{\rm C}$9 minutes and $r_{\rm A}=R_{\rm A}/a$0, corresponding to an observed TTV amplitude of 156 s. Under the assumption of a 2:1 mean-motion resonance and using the Lithwick-Xie-Wu analytic scaling, the perturbing body was estimated to have a mass of approximately $r_{\rm A}=R_{\rm A}/a$1 and an orbital period of about 6.24 d. The same study argued that the Applegate mechanism is too small by nearly three orders of magnitude to explain the signal (Parthasarathy et al., 13 Mar 2025).

This later dynamical interpretation remains distinct from the earlier null result, and the difference is attributable to the much longer timing baseline and the inclusion of TESS plus homogeneous reanalysis. A plausible implication is that HAT-P-12 b may not be dynamically solitary, although the proposed companion remains an inference from timing rather than a direct RV detection.

In the broader sub-Saturn population, HAT-P-12 b has repeatedly served as a comparison object. It was grouped early with WASP-21b in an emerging class of low-density Saturn-mass planets orbiting subsolar-metallicity stars, apparently consistent with a correlation between planetary core mass and host-star metallicity. The discovery of HAT-P-18b and HAT-P-19b, both low-density Saturn-mass planets around super-solar-metallicity K stars, was then used to argue that this simple correlation is not universal (Hartman et al., 2010). HAT-P-12 b therefore remains important less as a unique outlier than as a well-observed case at the intersection of low density, metal-poor host composition, cloudy or hazy atmosphere, and potentially nontrivial orbital dynamics.