KRONOS I: The $1{-}2.8μ$m JWST Transmission Spectrum of the 23 Myr V1298 Tau c

Abstract: While recent JWST observations of mature super-Earths and sub-Neptunes have frequently revealed featureless transmission spectra, their inflated progenitors offer a unique window into understanding their primordial compositions. As part of the KRONOS (Keys to Revealing the Origin and Nature Of sub-neptune Systems) JWST program, we present the NIRISS/SOSS transmission spectrum of V1298 Tau c, a $\sim$23 Myr super-Earth progenitor orbiting a young Solar analog. We detect H$<em>2$O in V1298 Tau c's atmosphere with a $\log</em>{10}$ volume mixing ratio of $-1.83<sup>{+0.68}_{-0.77}$, but no additional molecules from these data alone. We find consistent results for the planetary atmospheric properties in both retrievals with and without informed priors on stellar heterogeneities based on the observed stellar spectrum. We infer an atmospheric metallicity [O/H] of $14.8<sup>{+56.0}_{-12.28}\times$ the Solar value. This metallicity is similar to literature measurements for other young planets, including its massive outer companion V1298~Tau~b. In contrast, this measured metallicity is systematically lower than the metallicities of mature planets of similar mass and temperature. Altogether, these results provide tentative but growing evidence that the exoplanet mass--metallicity relation evolves with planetary age.

• The paper presents the first JWST NIRISS/SOSS transmission spectrum of 23 Myr V1298 Tau c, covering 0.85–2.83 μm and demonstrating reliable atmospheric characterization despite at least six starspot-crossing events.
• The analysis decisively detects H₂O with Δln𝒵 = 29.30, estimates an abundance of approximately 1.5% and an oxygen enrichment of about 15 times solar, while finding no significant constraints on other molecules or C/O ratio.
• The paper finds no compelling transit-light-source or limb-asymmetry signal and suggests that the planet’s relatively low metallicity may support an age-dependent mass–metallicity relation driven by atmospheric escape.

V1298 Tau c is a 23 Myr planet with a mass of $4.7 \pm 0.6\,M_\oplus$ and radius of $5.08 \pm 0.37\,R_\oplus$, yielding one of the lowest bulk densities ($0.20 \pm 0.05$ g/cm³) among known transiting exoplanets. Photoevaporation models predict it will contract to below $\sim$2.5 $R_\oplus$, making it a direct progenitor of the Galaxy's common super-Earth and sub-Neptune population. This paper presents the first JWST NIRISS/SOSS transmission spectrum of this planet (0.85–2.83 μm), obtained on UT 2025 September 6 as part of the KRONOS program (GO 5959), and demonstrates that robust atmospheric characterization is achievable despite severe stellar contamination from at least six starspot crossing events (SCEs) during the transit.

Observations and data reduction

The observation used SOSS in SUBSTRIP96 mode with NISRAPID readout, acquiring 2598 integrations over ~9.6 hr with a 4.78 hr transit duration. Because V1298 Tau is bright ($J = 8.687$), only the first spectral order was usable. The data were reduced with exoTEDRF v2.3.1, including group-level background subtraction, PCA removal of the thermal-control beating pattern, and clipping of a possible mirror-tilt event near integration 2390. Notably, the authors deliberately use a single reduction pipeline rather than multiple independent reductions, arguing that pipeline-to-pipeline differences are now demonstrably smaller than systematic effects from SCE modeling or limb darkening treatment—a methodological choice that shifts the burden of validation to intercomparison of light-curve fitting approaches.

Light curve fitting under heavy spot contamination

The broadband light curve shows a strongly curved out-of-transit baseline and numerous SCEs. Four independent treatments were implemented: (1) masking SCEs, (2) modeling them as Gaussian profiles, (3) physical modeling with fleck, and (4) a Gaussian process with an exponential-decay kernel via celerite. A cubic baseline polynomial was preferred by BIC across all tests, and a logarithmic limb darkening law with freely fit coefficients was adopted (fleck requires quadratic). The Gaussian model (Case 2) achieved the highest likelihood and lowest BIC among Cases 1–3 and was adopted as fiducial.

A key control test showed that a spot-free fit produces in-transit residuals elevated by a factor of 1.25 relative to out-of-transit (170 ppm vs. 137 ppm), whereas Cases 1–4 yield consistent in- and out-of-transit residuals—confirming that neglecting SCEs degrades the fit materially. The four methods produce transmission spectra whose shapes agree well within uncertainties; median depth offsets relative to Case 2 are only 21 ppm (masked), 30 ppm (fleck), and 120 ppm (GP, which appears to absorb some transit depth). Transit depth uncertainties reach ~40 ppm near 1.2 μm for Cases 1–3. The consistency across methods implies that the specific SCE treatment does not bias the inferred atmospheric properties.

Stellar surface heterogeneity

Fitting the flux-calibrated out-of-transit stellar spectrum with up to three components from the NewEra grid strongly favors a three-component model ($\Delta\log Z = 744$ over one component): a quiet photosphere at $T_\mathrm{phot} = 4893 \pm 10$ K, a cool spotted component at $3400 \pm 5$ K covering $22.4 \pm 0.6\%$ of the surface, and a hot component at $5.08 \pm 0.37\,R_\oplus$0 K covering $5.08 \pm 0.37\,R_\oplus$1. The spot coverage is twice that previously measured from Kepler photometry, though within the wide range seen on other young stars. The hot component is ~740 K above the photosphere—hotter than solar faculae—and may include plage contributions. The extinction posterior piled up at zero ($5.08 \pm 0.37\,R_\oplus$2 at 3σ); fixing $5.08 \pm 0.37\,R_\oplus$3 lowers the spot covering fraction to ~19%, so the exact coverage depends modestly on the assumed extinction.

Atmospheric retrieval

Retrievals with Aurora, using both informed priors on stellar heterogeneity (from the stellar spectrum fit) and wide uniform priors, detect H$5.08 \pm 0.37\,R_\oplus$4O decisively: $5.08 \pm 0.37\,R_\oplus$5 (informed) and 27.25 (uninformed), far exceeding conventional "strong evidence" thresholds. The water abundance is $5.08 \pm 0.37\,R_\oplus$6 (informed) versus $5.08 \pm 0.37\,R_\oplus$7 (uninformed)—consistent within 0.5σ. This corresponds to an oxygen metallicity of $5.08 \pm 0.37\,R_\oplus$8 solar. No other molecule (CH$5.08 \pm 0.37\,R_\oplus$9, NH$0.20 \pm 0.05$0, HCN, CO, CO$0.20 \pm 0.05$1) is constrained, so the C/O ratio remains undetermined pending the forthcoming G395H spectrum. Excess absorption at He I 1.083 μm and Paschen α may indicate an escaping H/He envelope, analyzed separately by Barat et al.

The retrievals prefer a haze scattering slope but no high-altitude cloud deck ($0.20 \pm 0.05$2 bar). Retrieved masses ($0.20 \pm 0.05$3 and $0.20 \pm 0.05$4) exceed the TTV mass of $0.20 \pm 0.05$5 at 1.3–1.8σ, illustrating the difficulty of mass determination around active stars; importantly, the chemical inference is robust to this mass uncertainty.

Ruling out the transit light source effect

Because stellar contamination can mimic planetary absorption, the authors performed dedicated tests. Retrievals including a TLSE component are preferred over no-TLSE models by only $0.20 \pm 0.05$6–0.7—negligible evidence. Forward-model grids show that reproducing even half the observed 1.4 μm feature via unocculted spots would require $0.20 \pm 0.05$7 K with $0.20 \pm 0.05$8, i.e., spots 1960–2450 K cooler than the photosphere—more extreme than any solar umbral contrast ($0.20 \pm 0.05$9–1872 K). Moreover, TLSE-induced water features have the wrong shape because starspot spectra at $\sim$0 K lack the molecular features of a cool planetary atmosphere. The conclusion that the H$\sim$1O is planetary rests partly on V1298 Tau being a warm (~4900 K) star where spots do not contain strong water bands—an advantage that does not extend to cooler M dwarfs, where such degeneracies remain acute.

Limb asymmetry

Tests using chromatic transit timing biases and catwoman two-limb fits reveal no significant evening–morning asymmetry, apart from noise-level hints near 1.4 μm that are inconsistent in magnitude with the analytic timing-bias prediction by a factor of two. This null result is notable because V1298 Tau c lies below the "asymmetry horizon" where comparable planets (WASP-39 b, WASP-94 Ab) do show asymmetry. The authors identify two plausible masking mechanisms: SCEs throughout the transit chord, and large TTVs (>2 hr) that force $\sim$2 to be measured from the same data used to probe asymmetry, introducing a self-masking bias. Which effect dominates is left unresolved.

Implications for the evolving mass–metallicity relation

Combining V1298 Tau c with its companion b and HIP 67522 b, all three young planets exhibit metallicities of only 1–10× solar, systematically below mature planets of similar mass. Fitting the mature-planet sample gives $\sim$3, while the two V1298 Tau planets give a nearly identical slope ($\sim$4) offset downward in metallicity. The interpretation offered is that preferential H/He loss through boil-off and photoevaporation enriches atmospheres over time, so the high metallicities of mature super-Earths and sub-Neptunes may be evolutionary products rather than primordial. This claim is explicitly tentative: it rests on three measurements from two young systems, drawn from heterogeneous methodologies and different elemental tracers, and TOI-421 b is an outlier whose data permit higher-metallicity solutions consistent with the trend. Magma-ocean volatile partitioning could further complicate early-time metallicity evolution, particularly in the first ~100 Myr.

Limitations and open questions

Several caveats bear directly on the results. The orbital solution ($\sim$5, inclination) varies by up to 2.7σ depending on the SCE treatment and correlates with spot parameters in the fleck fits, producing small vertical offsets between spectra—though not affecting spectral shape or retrieved chemistry. Elevated ingress/egress residuals suggest possible undetected limb spots, degenerate with limb darkening. The spot covering fraction depends on the assumed extinction. The limb-asymmetry null result cannot distinguish between genuine atmospheric homogeneity and systematic masking. And the age-evolution trend in metallicity awaits confirmation from additional young systems, including the planned KRONOS observations of V1298 Tau d, TOI-451 cd, and TOI-2076 bcd.

Conclusion

This work establishes that JWST transmission spectroscopy can recover planetary atmospheric signals from planets transiting heavily spotted young stars, provided multiple SCE treatments are cross-validated and the host star's spot temperatures are sufficiently distinct from planetary conditions. The decisive H$\sim$6O detection and super-solar metallicity for V1298 Tau c, combined with prior results for V1298 Tau b, provide the strongest current evidence that the exoplanet mass–metallicity relation evolves with age—but confirming this requires expanding the sample of characterized young planets across the first few hundred Myr, a question the broader KRONOS program is designed to answer.