betaFeedback

WorksPhysics and Astronomy

The deuterium-rich atmosphere of a mature super-Jupiter

Abstract

The deuterium-to-hydrogen ratio (D/H) has provided key insights into the origin and evolution of the Solar System. Using the James Webb Space Telescope, we detected deuterated methane at 48σ48σ in a mature giant exoplanet, εε Ind Ab. We measure the planet's D/H ratio to be (4.4±0.7)×10−5(4.4 \pm 0.7) \times 10^{-5}, which is similar to those of Uranus and Neptune but about twice those of Jupiter, Saturn, and the protosolar nebula. εε Ind Ab's heavy elements (C, O, N, and S) are uniformly enriched to a level similar to Jupiter's. These measurements require εε Ind Ab to have accreted solids with D/H≳1%\mathrm{D/H} \gtrsim 1\%, at least 30 times more deuterium-rich than typical Solar System comets. Such D/H values are found in the interstellar comet 3I/ATLAS and in organic molecules in star-forming regions, indicating that εε Ind Ab accreted pristine solids formed in a very cold environment.

Deuterium is a heavy isotope of hydrogen with one proton and one neutron in its nucleus. It was formed in the Big Bang with a primordial abundance relative to hydrogen of (2.527±0.030)×10−5(2.527 \pm0.030) \times10^{-5} [1]. Since then, deuterium has been gradually destroyed by burning in stellar interiors, and the local interstellar medium (ISM) today has D/H ≥(2.0±0.1)×10−5\geq(2.0 \pm0.1) \times10^{-5} [2]. In cold star-forming regions, chemical fractionation redistributes deuterium among molecules and can raise the D/H of certain ices and organics by orders of magnitude above the bulk abundance [3]. In the Solar System, D/H has been used to study the formation conditions of comets [4, 5], the origin of Earth’s ocean water [6, 7], and the loss of water from Venus and Mars due to atmospheric escape [8, 9].

Jupiter and Saturn have D/H close to the protosolar value of ∼2×10−5\sim2 \times10^{-5} [10, 11], whereas Uranus and Neptune are enriched in deuterium by a factor of two, usually attributed to the accretion of icy planetesimals, whose remnants survive as comets [12, 13]. For giant planets, D/H could therefore constrain the nature of the solids they accreted during formation.

ϵ\epsilon Ind Ab is the first mature (≈4\approx4 Gyr) cold Jupiter imaged around another star [14]. It has a dynamical mass of 6 – 8 Jupiter masses (MJupM_{\mathrm{Jup}}) [15, 16] and orbits with a semi-major axis of 16±216 \pm2 astronomical units (AU) around a nearby K dwarf (3.6 parsec). Given its cold temperature of Teff≈275T_{\mathrm{eff}} \approx275 K, ϵ\epsilon Ind Ab opens new frontiers in exoplanet atmospheric studies. For example, deuterated methane is predicted to be most readily detectable in cold exoplanets and brown dwarfs with Teff≲600T_{\mathrm{eff}} \lesssim600 K [17], and was recently detected in the field brown dwarf WISE 0855 which has an effective temperature similar to that of ϵ\epsilon Ind Ab [18].

We present moderate-resolution spectroscopy of ϵ\epsilon Ind Ab observed with the integral field unit (IFU; R≈2700R \approx2700, 2.85−5.30 μm2.85 - 5.30\,\mu\mathrm{m}) of JWST’s near-infrared spectrograph (NIRSpec). The observations were obtained on 25 October 2025 UT as part of the cycle 4 program 8714 (PI J. Xuan) and had a total integration time of 4.4 hours. The observations were equally split into two sequences at different observatory roll angles, and each roll includes data on two detectors, NRS1 and NRS2. The planet was placed at the center of the IFU’s field-of-view (FoV; 3″-by-3″). At the time of observations, ϵ\epsilon Ind Ab’s projected separation from the host star was 3.55″, so the core of the stellar point-spread function (PSF) is outside the FoV. Nevertheless, because the star is nearly 1 million times brighter than the planet at 4.1 μm4.1\,\mu\mathrm{m} [15], diffraction features from the stellar PSF dominate the raw data cube (Figure 1). Due to the large projected separation, however, the local stellar halo is only about 2 times brighter than the planet in NRS2 (4.1−5.3 μm4.1 - 5.3\,\mu\mathrm{m}).

Figure 1. Detection of ϵ\epsilon Ind Ab with moderate-resolution mode of JWST/NIRSpec IFU in the 3−53 - 5 μ\mum spectral range. a, The median spectral cube for one of the two NIRSpec rolls from the standard JWST pipeline before starlight subtraction. Here, the stellar diffraction pattern dominates the image. The host star ϵ\epsilon Ind A is just outside the field of view of the IFU, and the red arrow marks its direction. b, S/N map for ϵ\epsilon Ind Ab combining the two NIRSpec rolls and both NIRSpec detectors. Summed over all wavelength channels, the planet is detected with a S/N of approximately 600.

To detect ϵ\epsilon Ind Ab and extract its spectrum, we take advantage of the spectral difference between the Teff≈275T_{\mathrm{eff}} \approx275 K planet and Teff≈4700T_{\mathrm{eff}} \approx4700 K star, which show vastly different absorption lines. Specifically, we use the breads package [19] which is frequently used in similar high-contrast observations from JWST/NIRSpec [20, 21, 22, 23, 24, 25]. To detect ϵ\epsilon Ind Ab, we simultaneously model the planet and the starlight directly in the NIRSpec detector frame to avoid systematic extraction issues from the spatial undersampling of the NIRSpec IFU (see Methods). To extract its spectrum, we first subtract the starlight row-by-row on the detector using a spline-based model, and then fit a model PSF to the combined dither-sequence at the planet’s position in the field of view for each wavelength [21]. This provides a high-pass filtered spectrum of ϵ\epsilon Ind Ab, in which the spectral features of the star have been removed. The detection map of ϵ\epsilon Ind Ab is shown in Figure 1, while the spectrum is shown in Figure 2 and Figure S1.

Figure 2. Detection of CH₃D in ε Ind Ab. a, A segment of the NIRSpec spectrum (black). The best-fit full model including all species is in blue, and the best-fit model from a separate retrieval that excludes CH₃D opacity is in orange. The difference between the blue and orange models is due to CH₃D; we mark locations of CH₃D lines with red dashed lines and arrows. b, The residuals after subtracting the best-fit model without CH₃D from the data (black), plotted against the CH₃D-only model (red). A cross-correlation analysis between the residuals and CH₃D model indicates a detection S/N ≈48\approx48 for CH₃D in ε Ind Ab’s atmosphere. Equivalent plots to panel b for other species are shown in figure S2.

Figure S1. JWST/NIRSpec spectrum and archival photometry of ϵ\epsilon Ind Ab used in the atmospheric retrievals. a,b,c, NIRSpec spectrum from 2.8 − 5.3 μ\mum in black and best-fit model in orange. In each panel, we highlight the key absorbers for specific wavelength ranges. The bottom insets of each panel show the data residuals in gray, and inflated data uncertainties as orange contours. d, Circles show the JWST NIRCam and MIRI photometry covering 4 − 25 μ\mum [14, 15, 16]. The gray curve is the best-fit model (R=300R = 300) made from correlated-k opacities, and open squares denote the synthetic photometry. The bottom inset shows data residuals in σ\sigma units. All points are fit by the model well and agree at the <2σ< 2\sigma level.

To measure the elemental and isotopic abundances of ϵ\epsilon Ind Ab, we carried out Bayesian retrieval analyses using the radiative transfer code petitRADTRANS [26, 27]. In addition to the NIRSpec spectrum, we also jointly model ϵ\epsilon Ind Ab’s 44–25 μm25\,\mu\mathrm{m} photometry [14, 15, 16], which anchors the broadband continuum shape (see Figure S1). The two NIRSpec rolls are fitted separately, which allows us to compare the consistency between two independent datasets. The parameters of the atmospheric model include the planet’s mass, radius, radial velocity (RV) shift, and those describing the thermal structure, chemical abundances, and clouds in the atmosphere (see Methods). We considered both free retrievals, where the mass-mixing ratios of each molecule are fitted independently and constant with pressure, as well as chemically-consistent retrievals [21, 28], where the individual molecular abundances are determined by four elemental abundance parameters [C/H][\mathrm{C}/\mathrm{H}], [O/H][\mathrm{O}/\mathrm{H}], [S/H][\mathrm{S}/\mathrm{H}], and [N/H][\mathrm{N}/\mathrm{H}] and an equilibrium chemistry network. In the chemically-consistent retrievals, disequilibrium chemistry is accounted for by fitting the vertical eddy diffusion coefficient, KzzK_{\mathrm{zz}}, using the quench approximation [29] (see Methods). To account for possible underestimation of the data uncertainties, we fit multiplicative error inflation factors [30]. NRS1 requires no inflation whereas the NRS2 uncertainties need to be inflated by a factor of 2 to yield residuals with a reduced χ2\chi^2 of unity (Figure 2).

We detect twelve different molecular isotopologues – 12CH4{}^{12}\mathrm{CH}_{4}, 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D}, 13CH4{}^{13}\mathrm{CH}_{4}, H216O\mathrm{H}_{2}^{16}\mathrm{O}, H218O\mathrm{H}_{2}^{18}\mathrm{O}, H217O\mathrm{H}_{2}^{17}\mathrm{O}, 12CO{}^{12}\mathrm{CO}, 13CO{}^{13}\mathrm{CO}, C18O\mathrm{C}^{18}\mathrm{O}, CO2\mathrm{CO}_{2}, H2S\mathrm{H}_{2}\mathrm{S}, and NH3\mathrm{NH}_{3} – in the atmosphere of ϵ\epsilon Ind Ab (Figure 2 and Figure S2). The molecular detections were validated using leave-one-out tests and cross-correlation analyses [22, 31, 32]. The elemental abundances of [C/H][\mathrm{C}/\mathrm{H}], [O/H][\mathrm{O}/\mathrm{H}], [S/H][\mathrm{S}/\mathrm{H}], and [N/H][\mathrm{N}/\mathrm{H}], as measured by the chemically-consistent retrievals, are consistent between the two NIRSpec rolls within 0.1 dex. The statistical error on the abundances can be as small as 0.02 dex. Therefore, for all planet parameters reported here, we quote the mean of the posterior medians from the two rolls, and add the scatter between rolls in quadrature as a systematic error. From the chemically-consistent retrievals, we measure C/H=5.7−0.9+1.0\mathrm{C}/\mathrm{H}=5.7^{+1.0}_{-0.9}, O/H=5.2−0.9+1.1\mathrm{O}/\mathrm{H}=5.2^{+1.1}_{-0.9}, S/H=5.3−1.8+2.7\mathrm{S}/\mathrm{H}=5.3^{+2.7}_{-1.8}, and N/H=4.5−1.1+1.4\mathrm{N}/\mathrm{H}=4.5^{+1.4}_{-1.1} times the stellar values [33]. The free retrievals yield consistent elemental abundances [32]. Therefore, ϵ\epsilon Ind Ab exhibits a nearly uniform heavy element enrichment pattern across volatile species (C, O, N) and refractory species (S). The same pattern is seen in Jupiter [34] and systems of young super-Jupiters [21, 35].

Figure S2. Detection of NH3\mathrm{NH_3}, H2S\mathrm{H_2S}, H218O\mathrm{H_2^{18}O}, 13CO{}^{13}\mathrm{CO}, H217O\mathrm{H_2^{17}O}, 13CH4{}^{13}\mathrm{CH_4}, and C18O\mathrm{C^{18}O} in ϵ\epsilon Ind Ab. In each row, the left panel shows the data residuals after subtracting the corresponding leave-one-out retrieval (black). The colored dashed lines show the molecular template. On the right panels, the cross-correlation function between the data residuals and molecular templates are plotted in color. This provides an estimate of the detection S/N for each species. The dashed gray line shows the auto-correlation function of the molecular template.

From the retrieved abundances of 12CH4{}^{12}\mathrm{CH}_{4} and 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D}, we obtain (D/H)CH4=(5.4±0.2)×10−5(\mathrm{D}/\mathrm{H})_{\mathrm{CH}_{4}} = (5.4 \pm0.2) \times10^{-5}.

Since the atmosphere of a gas giant like ϵ\epsilon Ind Ab is mainly composed of H2\mathrm{H}_{2}, the D/H from H2\mathrm{H}_{2} and HD is a better tracer of the overall D/H. However, H2\mathrm{H}_{2} and HD are not directly detectable in our NIRSpec spectrum. Thus, a correction factor f=(D/H)CH4/(D/H)H2f = (\mathrm{D}/\mathrm{H})_{\mathrm{CH}_{4}}/(\mathrm{D}/\mathrm{H})_{\mathrm{H}_{2}} is required to account for the difference between the D/H measured from CH4\mathrm{CH}_{4} and D/H in hydrogen [36]. From both analytic and numerical calculations [32], we estimate that f=1.10f = 1.10–1.351.35 for ϵ\epsilon Ind Ab, which yields a corrected (D/H)H2=(4.4±0.7)×10−5(\mathrm{D}/\mathrm{H})_{\mathrm{H}_{2}} = (4.4 \pm0.7) \times10^{-5} for the planet’s atmosphere. We note that our inferred ff values are similar to those found for Jupiter and Saturn, which are f≈1.25f \approx1.25 and f≈1.38f \approx1.38, respectively [36].

The D/H of ϵ\epsilon Ind Ab is comparable to those of Uranus and Neptune and is approximately twice the protosolar value of 2×10−52 \times10^{-5} (Fig. 3). The elevated D/H of the ice giants can be explained by accretion of ice with D/H at or below cometary values [12, 13], and their atmospheric C/H of ∼70\sim70–100×100\times solar [34] is consistent with this model. The similarity in D/H is therefore striking because ϵ\epsilon Ind Ab is a gas giant with a mass of 66–8 MJup8\,M_{\mathrm{Jup}}, and C/H and O/H only ≈5×\approx5\times stellar. Indeed, Jupiter and Saturn have D/H ratios that agree with the protosolar value within 1σ1\sigma error bars [10, 11]. The nearby brown dwarf WISE 0855 also has a D/H close to the protosolar value [18]; as validation of our model, we reproduce the published D/H of WISE 0855 [32]. Below, we discuss how ϵ\epsilon Ind Ab may have acquired a high D/H ratio.

First, we consider if the ϵ\epsilon Ind system may have formed in a molecular cloud with higher bulk D/H than the ones which formed the Solar System or WISE 0855. As the host star and binary brown dwarf [37] in the ϵ\epsilon Ind system have long burned their deuterium, we cannot directly test this hypothesis. In the local diffuse ISM, the gas-phase D/H varies from ≈(0.5\approx(0.5–2.2)×10−52.2) \times10^{-5} along different sightlines [2, 38], with the lower values generally attributed to the depletion of deuterium onto dust grains. After correction, the bulk, undepleted D/H of the local ISM is estimated to be D/H≥(2.0±0.1)×10−5\mathrm{D}/\mathrm{H} \ge(2.0 \pm0.1) \times10^{-5} [2]. Most importantly, since deuterium is gradually destroyed by nuclear burning after the Big Bang, no natal cloud should have a bulk D/H above the primordial value of (2.527±0.030)×10−5(2.527 \pm0.030) \times10^{-5} [1]. Thus, a high D/H from a molecular cloud cannot explain ϵ\epsilon Ind Ab’s deuterium enrichment.

Next, we consider accretion of deuterium-rich solids, the same scenario invoked for Uranus and Neptune. In the Solar System, comets exhibit a range of D/HD/H between ≈(1.4−5.3)×10−4\approx(1.4 - 5.3) \times10^{-4} measured from water [39]. The interstellar comet 3I/ATLAS shows a much higher D/H≈10−2D/H \approx10^{-2} from water [40, 41], and D/H≈3×10−2D/H \approx3 \times10^{-2} from methane [42], which supports its formation from a colder and less thermally processed environment than that experienced by the Solar System comets. Around the outbursting protostar V883 Ori, a D/H≈1×10−3D/H \approx1 \times10^{-3} was measured from H2_2O ice that sublimated into gas phase in the outer disk at ∼100\sim100 AU distances [43]. This value is a factor of a few higher than the D/HD/H of Solar System comets. Subsequently, a high D2_2O/H2_2O ratio comparable to values measured in protostellar envelopes was found for V883 Ori, supporting the inheritance of the water in its disk from molecular clouds [44]. Such a conclusion has also been reached for water in the Solar System comets [7, 45]. In this context, ϵ\epsilon Ind Ab could have acquired its high D/HD/H ratio through accretion of deuterium-rich solids from the cold outer regions of its protoplanetary disk, where pristine material from the molecular cloud phase can be preserved.

To test this hypothesis, we use the composition model accomp [46] to fit the elemental abundances (C/H, O/H, S/H, and N/H) of ϵ\epsilon Ind Ab along with its D/HD/H ratio [32]. The uniform enrichment in C, O, N, and S is most consistent with formation outside the N2_2 or CO snowlines (∼56\sim56 AU and ∼36\sim36 AU, respectively). Both these snowlines lie beyond the current semi-major axis at 16±216 \pm2 AU [15], which implies ϵ\epsilon Ind Ab could have formed farther out and migrated inward. Alternatively, icy pebbles originating beyond these snowlines could have drifted inward to enrich the planet at its current location if they consisted of amorphous ice with trapped volatiles such as CO and N2_2, a mechanism also proposed for Jupiter [47, 48]. Accretion of solids from the cold outer disk also helps explain ϵ\epsilon Ind Ab’s high D/HD/H, since such material is more likely to have been inherited from the protostellar stage without subsequent thermal processing. Assuming that ϵ\epsilon Ind Ab formed beyond the N2_2 snowline, the measured elemental abundances imply that it accreted 89−15+18 M⊕89^{+18}_{-15}\ M_{\oplus} of metals through solids. To simultaneously match the metal enrichment and the D/HD/H, the accreted solids must contain either water ice with D/H∼1.5×10−2D/H \sim1.5 \times10^{-2} or refractory organics with D/H∼5×10−2D/H \sim5 \times10^{-2} (Figure S6). These D/HD/H ratios substantially exceed those of Solar System comets (Figure 3) but are comparable to the D/HD/H of 3I/ATLAS [40, 41, 42] and the D/HD/H of organic molecules observed in protostellar envelopes [49, 50]. The required D/HD/H can be a factor of two lower if accreted solids contained both D-rich water ice and organics (Figure S7).

Figure 3. D/H measurements in various astrophysical environments, including planets, comets, circumstellar disks, and protostars. The black markers indicate estimates of bulk D/H, either measured directly from atomic or molecular hydrogen or inferred from other molecules with appropriate corrections (CH3_3D/CH4_4 for ϵ\epsilon Ind Ab and WISE 0855). The other colors denote D/H in specific molecular tracers, and do not represent bulk D/H: H2_2O (blue), CH4_4 (teal), HCN (yellow), gas-phase deuterated methanol, CH2_2DOH (magenta), and refractory organics (brown). Data sources: primordial value [1]; WISE 0855 [18]; local ISM, Solar System planets, most Solar System comets, and protostars, from a published compilation [39]; 3I/ATLAS [41, 42]; 67P/Churyumov–Gerasimenko, using updated values [54, 55]; V883 Ori [43]; HCN in other protoplanetary disks [56].

Figure S6. Accretion of highly deuterium-rich solids is required to explain ϵ\epsilon Ind Ab’s D/H. In the two panels, we plot the planetary bulk D/H as a function of the O and C enrichment. The colored lines correspond to different adopted D/H for water ice (a) and refractory organics (b). The D/H values adopted are chosen to represent objects in Figure 3 (see references there). For water, 3×10−43 \times10^{-4} represent a typical Solar System cometary value, 10−310^{-3} is a typical value for protostars, and 10−210^{-2} is the water D/H of 3I/ATLAS. For organics, 1.5×10−31.5 \times10^{-3} is from comet 67P/C-G, 10−210^{-2} corresponds to CH2_2DOH in protostars, and 5×10−25 \times10^{-2} represents HCN in protoplanetary disks. If accreting material with Solar System comet-like D/H, matching the observed D/H requires very high enrichment in C/H and O/H of ∼200×\sim200\times stellar, which is ruled out for ϵ\epsilon Ind Ab.

Figure S7. The D/H required in water ice or refractory organics to match ϵ Ind Ab’s composition is consistent with measurements of the interstellar comet 3I/ATLAS. The color map shows log⁡10(χ2)\log_{10}(\chi^2) for combinations of water ice and refractory organic D/H that reproduce the planet’s D/H, C/H, and O/H. Lower χ2\chi^2 indicates better agreement. The black diamond shows the D/H measurements for 67P/C–G from water and refractory organics [54, 55], while the pink diamond shows measurements for 3I/ATLAS from H2_2O and CH4_4 [41, 42]. If considered independently, water ice or refractory organics would need to have D/H of ∼1.5×10−2\sim1.5 \times10^{-2} or ∼5×10−2\sim5 \times10^{-2}, respectively.

The calculation above assumes that ϵ\epsilon Ind Ab is well mixed and that the atmospheric D/HD/H reflects the bulk D/H\mathrm{D}/\mathrm{H}. Another possibility we consider is late stage accretion of solids combined with inefficient mixing between the upper atmosphere and interior, which could preserve the imprints of any accreted solids in the observed atmosphere. Such a scenario has been proposed to explain Jupiter’s atmospheric enrichment [51, 52]. This scenario could reduce the total mass of accreted solids by confining them to a thin atmospheric shell. However, the relative abundance ratios between C, O, S, N, D, and H (Figure 4) still require the planet to accrete highly deuterium-rich material. Regardless of whether ϵ\epsilon Ind Ab is well-mixed or not, the planet’s high D/H\mathrm{D}/\mathrm{H} cannot be explained by accreting Solar System-like comets as doing so would increase the planet’s O/H\mathrm{O}/\mathrm{H} and C/H\mathrm{C}/\mathrm{H} to much higher values of ∼200×\sim200\times stellar, inconsistent with the observed value of ≈5×\approx5\times stellar (Figure S6). Finally, we also consider if disk gas accreted by the planet can become enriched in D and metals during disk dispersal but find that it is implausible [32].

Figure 4. ϵ\epsilon Ind Ab shows similar levels of C, O, S and N enrichment as Jupiter, but a D/HD/H ratio comparable to those of the ice giants. The abundances of C, O, S, N (circles), and D (squares) relative to H in ϵ\epsilon Ind Ab and the Solar System gas and ice giants. All values are normalized to the stellar values: for ϵ\epsilon Ind A, we use the stellar abundances [33], and for the Solar System planets we normalize to the Sun’s photospheric abundances [57]. The C, O, S, and N abundances of the Solar System planets are from a published compilation [34], and their D/HD/H values are from previous measurements [10, 11, 13].

As a cold and mature gas giant in the solar neighborhood, ϵ\epsilon Ind Ab provides a rare point of comparison with the giant planets of the Solar System. The high D/H\mathrm{D}/\mathrm{H} of ϵ\epsilon Ind Ab suggests that its planetary system inherited pristine ice and refractory organics from the molecular cloud stage, and that these minimally processed solids were accreted by the planet. In the Solar System, deuterium has revealed the inheritance of interstellar ices and organics in meteorites, comets, and Earth’s oceans [5, 7, 53]. Around other stars, protoplanetary disks appear to inherit much of their water ice from the molecular cloud [43, 44], and the interstellar comet 3I/ATLAS shows that such pristine material can be preserved as planet building blocks [40, 41]. ϵ\epsilon Ind Ab traces this inheritance to a mature super-Jupiter, demonstrating that interstellar material can be incorporated into planets and leave a sizable imprint even for a gas giant. Future JWST measurements of D/H\mathrm{D}/\mathrm{H} across a larger sample of cold planets are poised to reveal how planetary systems inherit and process material from molecular clouds, placing the formation of the Solar System in a broader context.

Materials and methods are available as supplementary material.

Acknowledgments

J.W.X. thanks Sam de Regt for advice on using fastchem in retrievals, Melanie Rowland for sharing the WISE 0855 spectrum, and Aneesh Baburaj for advice on stellar abundances. We thank Jeehyun Yang, Teresa Paneque-Carreño, Ngoc Tuan Truong, Edward Young, Peng Ni, Jonathan Fortney, Eliza Kempton, and Mark Giovinazzi for helpful discussions. J.W.X. would also like to thank Jingwen Zhang for her support and constructive suggestions. This work is based on observations made with the NASA/ESA/CSA James Webb Space Telescope. The data were obtained from the Mikulski Archive for Space Telescopes at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-03127 for JWST. These observations are associated with program 8714. This work benefited from the 2026 Exoplanet Summer Program in the Other Worlds Laboratory (OWL) at the University of California, Santa Cruz, a program partially supported by funding from NASA.

Funding: J.W.X. acknowledges support from the Heising-Simons Foundation 51 Pegasi b Fellowship (grant #2025-5887). Y.Z. acknowledges support from the Heising-Simons Foundation 51 Pegasi b Fellowship (grant #2023-4298).

Author contributions: J.W.X. led the data reduction and atmospheric retrievals, coordinated the analyses, and wrote the manuscript. Y.C. led the composition modeling of the planet and wrote the corresponding section. J-B.R. provided guidance with the data reduction and contributed text. K.O. contributed to the planet formation interpretation and helped with defining the chemical framework used in the retrievals. S-M.T. led the chemical kinetics models with VULCAN. J.M. helped with defining the P-T profile priors. J.W.X., J-B.R., A.S., and Y.Z. designed the GO 8714 observing program. B.B., A.B., G.A.B., M.M., E.M., C.V.M., R.M-C., M.P., J.J.W., B.P.B., H.A.K. provided advice on various aspects of the project. Other authors are part of the GO 8714 team or commented on the manuscript.

Competing interests: There are no competing interests to declare.

Data and materials availability: The JWST/NIRSpec data used in this paper are from JWST GO program 8714 (PI J. Xuan) and are publicly available from the Mikulski Archive for Space Telescopes (MAST) [58]. The NIRSpec data reduction was performed using the breads package available at https://breads.readthedocs.io/ [59]. The atmospheric models were generated using the petitRADTRANS radiative transfer tool [26] available at https://petitradtrans.readthedocs.io/.

Supplementary materials

Materials and Methods

Figs. S1 to S7

Tables S1 to S3

References (60-133)

Supplementary Materials for The deuterium-rich atmosphere of a mature super-Jupiter

Jerry W. Xuan; Yayaati Chachan; Jean-Baptiste Ruffio; Kazumasa Ohno; Shang-Min Tsai; James J. Mang; Yapeng Zhang; Aniket Sanghi; Björn Benneke; Alexis Bidot; Geoffrey A. Blake; Brendan P. Bowler; Mathilde Malín; Elisabeth Matthews; Caroline V. Morley; Ruth Murray-Clay; Marshall Perrin; Jason J. Wang; Carrie He; Katelyn Horstman; Heather A. Knutson; Alexander Madurowicz; Christian Marois; Dimitri Mawet; Rebecca Oppenheimer; William Thompson

This PDF file includes:

Materials and Methods

Figures S1 to S7

Tables S1 to S3

Materials and Methods

Observations

In this study, we analyze JWST/NIRSpec IFU observations of ϵ\epsilon Ind Ab from 2.85−5.30 μm2.85-5.30\ \mu\mathrm{m} in the medium-resolution spectroscopy mode (R∼2,700R \sim2{,}700; filter: F290LP; grating: G395H) from JWST Cycle 4 GO program 8714 (PI: J. Xuan). The observations took place on 25 October 2025 (UT). The observations were equally split over two observatory roll angles spanning ∼10\sim10 degrees. Each roll includes 2 detectors (NRS1 and NRS2), 33 groups, 1 integration, and 16 dithers with NIRSpec’s small cycling pattern totaling ∼132\sim132 min per roll. The 16 small dithers are chosen to improve the spatial sampling of the NIRSpec instrument. To minimize saturation effects from the bright host star (K≈2.2K \approx2.2), the planet is placed near the center of the IFU, and the PSF core of the star is outside the FoV.

NIRSpec data reduction

Starting with the raw uncalibrated detector images, we first performed reductions using stage 1 and stage 2 of the JWST Science Calibration Pipeline v2.0.1 [60]. We used the Calibration Reference Data System (CRDS) software version 13.1.16 (CRDS_VER), and the CRDS context jwst_1535.pmap (CRDS_CTX). From stage 2, the resulting flux-calibrated detector images (*_cal.fits) are further processed using the open-source breads package [19] to subtract the starlight and obtain the planet detection map (Figure 1) and spectrum (Figure 2). A detailed description of the breads reduction process for JWST can be found in previous work [19, 21], and we briefly summarize the main steps and highlight a few updates.

In raw NIRSpec frames, the planet is a few times fainter than the stellar PSF diffraction features. To detect the planet, we fit a joint forward model of the planet and diffracted starlight directly in the detector images. For the star, we first derive a continuum-normalized stellar spectrum from the science data after masking the planet location. This continuum-normalized stellar spectrum is later combined with a flexible 1D spline model following detector rows to model the starlight in the data. The planet is modeled using a model PSF combined with a Sonora Flame Skimmer model [61] with Teff=275T_{\mathrm{eff}} = 275 K, log⁡g=4.0\log g = 4.0, log⁡Kzz=4.0\log K_{\mathrm{zz}} = 4.0, [M/H]=0.5=0.5 dex, and C/O=0.458=0.458.

In previous studies using the breads framework, the model PSF was computed from a numerical model of JWST/NIRSpec with STPSF [62]. However, given the high S/NS/N of ϵ\epsilon Ind Ab (detected at ≈600 σ\approx600\,\sigma, Figure 1), inaccuracies in the STPSF model lead to systematics in the extracted flux of the planet. In this work, we use an empirical PSF model derived from observations acquired by the JWST PSF calibration program GO 3399 (PI: M. Perrin) as follows. The program GO 3399 observed two stars using a sequence of 16 small cycling dithers. One of the stars is faint enough to not saturate (2MASS J17571324+6703409; K = 11.15). It is used to probe the core of the PSF. The second star is brighter so its central core saturates (HD 163466; K = 6.33); however, it is used to obtain higher S/NS/N data for the fainter tails of the PSF. The 16 dithers of each star were combined to create a super-sampled effective PSF (ePSF) [63] by fitting a 3D (xx, yy, λ\lambda) spline model to the point cloud. The sampling of the ePSF is 20 mas compared to the original 100 mas of the NIRSpec spatial pixels. In this context, the point cloud refers to the ensemble of illuminated pixels on the NIRSpec detectors, which irregularly samples the NIRSpec PSF. The best-fit 3D spline model of both stars were then combined to create the final high-S/NS/N empirical model of the NIRSpec PSF in the F290LP/G395H observing mode. The resulting model PSF provides a more accurate spectral extraction compared to STPSF.

By fitting the forward model of the star and planet across the IFU in each spatial location, we estimate the planet’s detection S/NS/N at each pixel. We normalize the S/NS/N maps to achieve a standard deviation of unity in the absence of planet signal. Combining all wavelengths, we obtain a planet detection S/NS/N of ≈600\approx600, shown in Figure 1.

To extract the planet spectrum, we apply the spline-based high-pass filter to fit and subtract the starlight from everywhere in the detector. Note that this step effectively removes the continuum of the planet while preserving its high-resolution spectral lines. We then fit the empirical PSF model at each wavelength over a grid of right ascension (RA) and declination (Dec) to construct an initial planet flux cube. Each element of the cube stores the fitted flux for a planet placed at that position and wavelength. This initial cube uses sky coordinates derived from the World Coordinate System (WCS), without the correction for the wavelength-dependent centroid [19]. We use this initial cube to locate the planet and measure this wavelength-dependent drift before repeating the cube extraction with corrected coordinates.

Unlike previous NIRSpec datasets, where the IFU is centered on the star [21, 35], in these observations the star is outside the FoV. In this case, we found that determining the wavelengthdependent centroid shift of the PSF [19] using the stellar diffraction features is challenging. As the planet is detected with a high S/NS/N, we instead solve for the chromatic centroid shift using the planet signal itself. In short, we cross-correlate the same Sonora Flame Skimmer model mentioned above with the high-pass-filtered planet spectrum at each spatial location in the spectral cube, in contiguous wavelength intervals of 0.1 μm0.1\,\mu\mathrm{m}. We determine the planet position from the peak of the cross-correlation function in each interval and fit first-order polynomials to the measured ΔRA\Delta\mathrm{RA} and ΔDec\Delta\mathrm{Dec} offsets as a function of wavelength. This procedure recovers a ≈15\approx15 mas centroid shift, predominantly in the Dec direction. The size of the centroid shift is found to be consistent between the two NIRSpec rolls. We apply these wavelength-dependent polynomial corrections to the WCS-derived coordinates to build the corrected planet flux cube. We then cross-correlate the corrected cube with the Sonora Flame Skimmer template to determine the final planet position. The final 1D planet spectrum is extracted at the peak of the resulting cross-correlation map. We estimate the flux uncertainties and covariance matrix of the planet spectrum from starlight residuals within an annulus centered on the host star at the same projected separation as the planet, after masking out a 0.3′′0.3^{\prime\prime} radius region around the planet’s position.

Atmospheric retrieval setup

Here we describe the process of inferring atmospheric parameters from the observed planet spectra. We use the radiative transfer code petitRADTRANS [26, 27] to calculate synthetic planet templates, which are compared to the observed spectra in a Bayesian retrieval analysis using nested sampling. The atmospheric retrieval framework used in this work closely follows previous work [35], which we refer to for details.

For the NIRSpec spectra, we use the line-by-line opacity sampling mode with R=100,000R = 100{,}000 to construct a model from 2.85−5.30 μm2.85 - 5.30\,\mu\mathrm{m}. For photometry, we use the correlated-kk sampling mode with R=300R = 300 to generate a model covering the remaining wavelengths. A separate model at R=300R = 300 is used for the photometry to reduce the computational cost. The parameters we fit for include the planet mass and radius, radial velocity, temperature profile, chemical abundances, and cloud structure, as well as nuisance parameters for the spectral resolution and error inflation terms. All fitted parameters and their priors are summarized in Extended Data Table S1. We sample the posterior distributions with the nested sampling package pymultinest [64, 65].

ParameterPriorParameterPrior
Common parameters
Mass (MJupM_{\mathrm{Jup}})N(6.5,0.7)\mathcal{N}(6.5, 0.7)NRS1 error multipleU(1,10)\mathcal{U}(1, 10)
Radius (RJupR_{\mathrm{Jup}})N(1.04,0.04)\mathcal{N}(1.04, 0.04)NRS2 error multipleU(1,10)\mathcal{U}(1, 10)
RV (km s−1^{-1})U(−60,−20)\mathcal{U}(-60, -20)rbluer_{\mathrm{blue}}U(2400,3600)\mathcal{U}(2400, 3600)
log(gray opacity)U(−6,0)\mathcal{U}(-6, 0)rredr_{\mathrm{red}}U(3600,4600)\mathcal{U}(3600, 4600)
P–T profile parameters
TrefT_{\mathrm{ref}} [log(PP/bar) = −1-1] (K)U(75,475)\mathcal{U}(75, 475)∇7\nabla_{7} [0.0]N(0.30,0.15)\mathcal{N}(0.30, 0.15)
∇1\nabla_{1} [2.7]N(0.26,0.06)\mathcal{N}(0.26, 0.06)∇8\nabla_{8} [-1.0]N(0.13,0.10)\mathcal{N}(0.13, 0.10)
∇2\nabla_{2} [2.0]N(0.28,0.06)\mathcal{N}(0.28, 0.06)∇9\nabla_{9} [-2.0]N(0.10,0.10)\mathcal{N}(0.10, 0.10)
∇3\nabla_{3} [1.7]N(0.29,0.06)\mathcal{N}(0.29, 0.06)∇10\nabla_{10} [-3.0]N(0.07,0.10)\mathcal{N}(0.07, 0.10)
∇4\nabla_{4} [1.3]N(0.29,0.06)\mathcal{N}(0.29, 0.06)∇11\nabla_{11} [-4.0]U(−0.40,0.50)\mathcal{U}(-0.40, 0.50)
∇5\nabla_{5} [1.0]N(0.30,0.10)\mathcal{N}(0.30, 0.10)∇12\nabla_{12} [-5.0]U(−0.40,0.50)\mathcal{U}(-0.40, 0.50)
∇6\nabla_{6} [0.7]N(0.30,0.10)\mathcal{N}(0.30, 0.10)∇13\nabla_{13} [-6.0]U(−0.40,0.50)\mathcal{U}(-0.40, 0.50)
∇7\nabla_{7} [0.0]N(0.30,0.15)\mathcal{N}(0.30, 0.15)
Free retrieval
log mass-mixing ratio, log XiX_iU(−12,−1)\mathcal{U}(-12, -1)
Chemically-consistent retrieval
[C/H]U(−0.5,1.0)\mathcal{U}(-0.5, 1.0)log(KzzK_{\mathrm{zz}}/cm2^{2} s−1^{-1})U(2,9)\mathcal{U}(2, 9)
[O/H]U(−0.5,1.0)\mathcal{U}(-0.5, 1.0)log10 isotopologue ratioU(1,10)\mathcal{U}(1, 10)
[N/H]U(−0.5,1.0)\mathcal{U}(-0.5, 1.0)
[S/H]U(−0.5,1.0)\mathcal{U}(-0.5, 1.0)

Table S1. Fitted parameters and priors for the atmospheric retrievals.

We use 1000 live points and stop sampling when the estimated contribution of the remaining prior volume to the total evidence is <1%< 1\%. The Bayesian evidence from each fit allows us to calculate the Bayes factor BB, which quantifies the relative probability of different models.

For each synthetic spectrum from petitRADTRANS, we apply the following steps before comparing the model with the observed NIRSpec spectra. First, we apply a radial velocity shift to the model to align the model spectra with the data. Second, we apply instrumental broadening to the model and resample the model to the data wavelengths. We fit the NIRSpec resolving power as a linear function of wavelength [35], parameterized here by its values at the blue and red endpoints, rbluer_{\mathrm{blue}} and rredr_{\mathrm{red}}. Third, we apply the same spline-based high-pass filtering on the model that was applied on the data. This means we remove the best-fit spline from the model using the same node positions in the wavelength direction. After these steps, we compute the log likelihood as follows:

ln⁡L=−12(RTC−1R+nln⁡(2π)+ln⁡det⁡(C))(1)\ln\mathcal{L} = -\frac{1}{2}\left(R^{\mathsf{T}} C^{-1} R + n \ln(2\pi) + \ln\det(C)\right) \tag*{(1)}

where RR is the residual array, CC is the covariance matrix, nn is the number of data points, and det⁡(C)\det(C) is the determinant of the covariance matrix. Here, CC has been multiplied by a factor of emulte_{\mathrm{mult}}, which is the error inflation term that we fit. Due to the significantly different flux levels of the planet in NRS1 and NRS2 wavelengths, we fit two separate error inflation terms, one for NRS1 and one for NRS2.

To constrain the continuum flux of the planet, we jointly model the literature photometry with the NIRSpec spectra (Extended Data Fig. S1). The photometry data span 4 to 25 μ\mum, and are obtained by JWST/NIRCam and JWST/MIRI [14, 15, 16]. In the Bayesian atmospheric retrieval, the log likelihoods from the photometry and NIRSpec components are added.

To parametrize the thermal structure, or pressure-temperature (P-T) profile, of the planet’s atmosphere, we use the gradient-based profile [28], which fits for (dln⁡T/dln⁡P)(d \ln T/d \ln P) values at fixed pressure nodes and a reference temperature, TrefT_{\mathrm{ref}}, which we set at 0.1 bars. We define 13 pressure nodes spanning 10−610^{-6}–102.710^{2.7} bar to resolve the pressure range relevant to the planet’s emission contribution function. The exact node locations are log⁡10(P/bar)=[−6,−5,−4,−3,−2,−1,0,0.7,1.0,1.3,1.7,2.0,2.7]\log_{10}(P/\mathrm{bar}) = [-6, -5, -4, -3, -2, -1, 0, 0.7, 1.0, 1.3, 1.7, 2.0, 2.7]. The node spacing is finer near the photosphere and in the deep atmosphere. The denser grid in the deep atmosphere prevents interpolation across wide pressure ranges from imposing an artificially rigid P–T profile.

We consider three different cloud models: clear, gray clouds, and the EddySed model [66] with water clouds implemented in petitRADTRANS [67]. Given the cold temperature of ϵ\epsilon Ind Ab, water clouds are expected to be the most relevant cloud species at near-infrared wavelengths [14, 68]. Previous work modeling ϵ\epsilon Ind Ab’s photometry data has also hinted at the presence of water clouds in its atmosphere [15, 16].

We consider two different chemical setups for the retrievals. The first setup is known as “free retrievals”, where the mass-mixing ratio of each species is fitted as a separate parameter and assumed to be constant with pressure. The advantage of this approach is mainly flexibility, as it assumes less prior knowledge about the chemistry. We also tested an alternative free retrieval where we allow the gas abundance of H2_2O (and H218_2^{18}O, H217_2^{17}O) to decrease as a power law towards the top of the atmosphere [69]. Such water depletion is motivated by the expected formation of water clouds, and this alternative free retrieval is used to test whether the NIRSpec data find evidence of water-depletion.

In the second setup, which is known as chemically-consistent retrievals, the individual mass-mixing ratios are determined by elemental abundance parameters and an equilibrium chemistry network [22, 70]. Compared to free retrievals, this approach is more physically-motivated. In this work, we use the equilibrium chemistry code FastChem [71], specifically version 4.0.3 (https://newstrangeworlds.github.io/FastChem/). For each iteration of the retrieval, we run a new FastChem calculation given the P-T profile, and four fitted parameters: [C/H], [O/H], [N/H], and [S/H]. We include rainout condensation [71]. For computational efficiency, we use a simplified FastChem network [70] with 46 gas-phase species and 25 condensates (nearly an order of magnitude less than the full FastChem network). We verified that our reduced network reproduces the major gas-phase abundances relevant for ϵ\epsilon Ind Ab-like P-T profiles, and runs 20×20\times faster. We adopt the solar elemental abundances [57] for the reported [C/H], [O/H], [N/H], and [S/H].

To account for disequilibrium chemistry in the second setup, we fit the vertical eddy diffusion coefficient, KzzK_{\mathrm{zz}}, and apply the quench approximation [29]. We calculate the quench pressures by equating the relevant chemical timescale to the vertical mixing timescale

τmix=L2Kzz.(2)\tau_{\mathrm{mix}} = \frac{L^2}{K_{\mathrm{zz}}}. \tag*{(2)}

Here, LL is characteristic mixing length and is typically assumed to be the pressure scale height

H=kBTμmgH = \frac{k_B T}{\mu m g}

(kBk_B: Boltzmann constant; μ\mu: mean molecular weight in amu; mm: atomic mass unit; gg: surface gravity). We discuss the choice of L=HL = H further in the ‘Vertical eddy diffusion coefficient’ section. There are three relevant quench pressures for ϵ\epsilon Ind Ab: one for CO, CH4_4, and H2_2O, one for NH3_3 and N2_2, and one for CO2_2. The abundances of CO, CH4_4, H2_2O, CO2_2, and NH3_3 at P<PquenchP < P_{\mathrm{quench}} are set equal to their abundances at the respective quench points. The chemical reaction timescales are taken from previous work [29, 72]. For CO2_2, we use the disequilibrium H2_2O and CO profiles to calculate a pseudo-equilibrium CO2_2 profile, to which we then apply the CO2_2 quench approximation [73].

Since ϵ\epsilon Ind Ab is cold enough for volatile condensation to occur, we additionally limit the post-quench gas abundances such that the partial pressure of each condensable species does not exceed its saturation vapor pressure. We apply saturation caps for all volatile species (H2_2O, CH4_4, CO, NH3_3, H2_2S, and CO2_2) using vapor pressure relations [74]. We note that the saturation caps are applied only in the chemically-consistent retrieval.

We detect several minor isotopologues in the NIRSpec spectrum of ϵ\epsilon Ind Ab. In the retrievals, we fit the mass fraction of the major isotopologue relative to the minor isotopologue as a log mass fraction ratio. For example, the parameter for CH3_3D is

log⁡10(XCH4XCH3D),(3)\log_{10}\left(\frac{X_{\mathrm{CH}_4}}{X_{\mathrm{CH}_3\mathrm{D}}}\right), \tag*{(3)}

where XiX_i is the mass fraction of species ii. The corresponding number ratio is

nCH4nCH3D=XCH4XCH3DμCH3DμCH4,(4)\frac{n_{\mathrm{CH}_4}}{n_{\mathrm{CH}_3\mathrm{D}}} = \frac{X_{\mathrm{CH}_4}}{X_{\mathrm{CH}_3\mathrm{D}}} \frac{\mu_{\mathrm{CH}_3\mathrm{D}}}{\mu_{\mathrm{CH}_4}}, \tag*{(4)}

where μi\mu_i is the molecular mass. For the other isotopologues besides CH3_3D, the number ratios from the equation above yields the isotope ratios of 12^{12}C/13^{13}C from CO and CH4_4, 16^{16}O/18^{18}O from CO and H2_2O, and 16^{16}O/17^{17}O from H2_2O. To obtain the D/H ratio from CH4_4, the molecular number ratio needs to be converted to an atomic ratio as CH4_4 contains four H atoms and CH3_3D contains one D atom and three H atoms:

(D/H)CH4=nCH3D4nCH4+3nCH3D.(5)(\mathrm{D}/\mathrm{H})_{\mathrm{CH}_4} = \frac{n_{\mathrm{CH}_3\mathrm{D}}}{4n_{\mathrm{CH}_4}+3n_{\mathrm{CH}_3\mathrm{D}}}. \tag*{(5)}

The line opacities included in the retrievals are summarized in Extended Data Table S2. For continuum opacities, we include the collision-induced absorption (CIA) from H2_2-H2_2 and H2_2-He. Given the large number of isotopologues and molecules detected in ϵ\epsilon Ind Ab’s spectrum, we calculate new opacity tables for all line species using line-racer [75]. In this process, we noted the default opacity tables in petitRADTRANS versions 3 or below use a 1 dex pressure spacing, which created noticeable interpolation errors. Therefore, for the new calculations, we compute opacities on a grid of log⁡10(P/bar)\log_{10}(P/\mathrm{bar}) from −7-7 to 3 in steps of 1 dex, with finer steps of 1/31/3 dex between log⁡10(P/bar)=−2\log_{10}(P/\mathrm{bar})=-2 and 1, near the peak of the emission contribution function.

Atmospheric retrieval results

In the NIRSpec spectra of ϵ\epsilon Ind Ab, we detect 12 molecular isotopologues spanning six parent molecules: 12CH4{}^{12}\mathrm{CH}_{4}, 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D}, 13CH4{}^{13}\mathrm{CH}_{4}, H216O\mathrm{H}_{2}^{16}\mathrm{O}, H218O\mathrm{H}_{2}^{18}\mathrm{O}, H217O\mathrm{H}_{2}^{17}\mathrm{O}, 12CO{}^{12}\mathrm{CO}, 13CO{}^{13}\mathrm{CO}, C18O\mathrm{C}^{18}\mathrm{O}, CO2\mathrm{CO}_{2}, H2S\mathrm{H}_{2}\mathrm{S}, and NH3\mathrm{NH}_{3}.

To validate these detections, we run leave-one-out retrievals where, each time, one isotopologue is left out from the line opacities. Next, we compare the residuals of the leave-one-out models with single-species templates [22]. For example, the residuals of the leave-one-out model without 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D} would contain 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D} lines if the data shows evidence of this isotopologue (as shown in panel b of Figure 2). From cross-correlating the data residuals of the model without 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D}, and a 12CH3D{}^{12}\mathrm{CH}_{3}\mathrm{D}-only template, we calculate a cross-correlation function (CCF) S/N [76]. From the CCF analysis, CH3_{3}D is detected at 48σ48\sigma, and the five carbon and oxygen isotopologues are detected at significances ranging from 8σ8\sigma to 44σ44\sigma, as shown in Extended Data Fig. S2. The CCF S/N values for NH3_{3} and H2_{2}S are ≈70\approx70. We do not perform this CCF exercise for species such as 12CO{}^{12}\mathrm{CO} and H216O\mathrm{H}_{2}^{16}\mathrm{O} whose lines are clearly evident in the data.

The isotopic ratios of 12C/13C{}^{12}\mathrm{C}/{}^{13}\mathrm{C}, 16O/18O{}^{16}\mathrm{O}/{}^{18}\mathrm{O}, 16O/17O{}^{16}\mathrm{O}/{}^{17}\mathrm{O} from CO, CH4_{4}, and H2_{2}O are shown in Extended Data Fig. S3. Unlike the case of D/H, the planet’s carbon and isotopic isotopic ratios measured from different molecules are consistent with the corresponding solar ratios [77] at the <1σ<1\sigma level. In addition, the isotopic ratios measured from different molecules (i.e. 13CH4{}^{13}\mathrm{CH}_{4} and 13CO{}^{13}\mathrm{CO}; H218O\mathrm{H}_{2}^{18}\mathrm{O} and C18O\mathrm{C}^{18}\mathrm{O}) are also consistent with each other.

We also searched for additional species in the NIRSpec spectrum. We tested but found no significant evidence for HDO, C17{}^{17}O, 13{}^{13}CO2_{2}, PH3_{3}, C2_{2}H2_{2}, HCN, SiH4_{4}, 13{}^{13}C18{}^{18}O, GeH4_{4}, and AsH3_{3}.

We note that the non-detection of HDO implies a 3σ3\sigma upper limit of (D/H)H2O<6×10−5_{\mathrm{H}_{2}\mathrm{O}}<6\times10^{-5}, which is consistent with the measured (D/H)CH4_{\mathrm{CH}_{4}}} value,

To place the planet’s heavy element enrichment into context, we need to know the host star’s abundances. Two recent studies report C, O, and S abundances for the star ϵ\epsilon Ind A [33, 78], finding slightly sub-solar abundances. The values from these studies agree at the 1σ1\sigma level, so we adopt the newer values [33], which are [C/H]⋆_\star = −0.22±0.07-0.22 \pm0.07, [O/H]⋆_\star = −0.14±0.08-0.14 \pm0.08, and [S/H]⋆_\star = −0.15±0.18-0.15 \pm0.18. There is no literature measurement for [N/H] in ϵ\epsilon Ind A, so we assume [N/H]⋆_\star = [M/H]⋆_\star = −0.12±0.08-0.12 \pm0.08 [33].

In the chemically-consistent retrievals, we directly retrieve [C/H], [O/H], [N/H], and [S/H], and find that ϵ\epsilon Ind Ab is enriched by ≈5×\approx5\times stellar in C, O, N, and S (Figure 4). From the free retrievals, we can add up the retrieved gas-phase molecular abundances to estimate [C/H], [O/H], and [S/H]. This yields C/H = 4.9±1.04.9 \pm1.0, O/H = 4.1−0.7+0.94.1^{+0.9}_{-0.7}, and S/H = 4.5−1.5+2.3×4.5^{+2.3}_{-1.5}\times stellar, which are slightly lower by ∼20%\sim20\% but consistent within 1σ1\sigma of the results from chemically-consistent retrievals. We do not quote N/H from free retrievals as they do not account for N2_2, which constitutes a substantial part of the nitrogen budget even at Teff≈275T_{\mathrm{eff}} \approx275 K. In addition, the free retrieval neglects elements sequestered in condensates. Therefore, the resulting elemental abundances from the two retrieval frameworks are not directly comparable. Nevertheless, the broadly consistent [C/H], [O/H], and [S/H] values suggest that gas-phase CH4_4, H2_2O, CO, CO2_2, and H2_2S are the dominant carriers of C, O, and S in ϵ\epsilon Ind Ab’s atmosphere, as expected from self-consistent atmospheric models [61].

We measure a planetary radial velocity of −45.6±0.9-45.6 \pm0.9 km s−1^{-1} at Oct 25 (UT) 2025. This value is close to the host star RV of −40.43±0.13-40.43 \pm0.13 km s−1^{-1} reported by the Gaia DR3 data release [79], which has a reference epoch of 2016.0. Since the change in the host star’s RV over two decades due to a combination of orbital motion and perspective acceleration is <0.1< 0.1 km s−1^{-1} [15], our measurement would imply a relative RV (defined as planet RV - stellar RV) of −5.2±0.9-5.2 \pm0.9 km s−1^{-1}, implying the planet is currently moving towards Earth in its orbit. This is consistent with the prediction from the latest orbit fits of the planet [15].

The retrieved PP–TT profiles from the chemically-consistent and free retrievals are shown in Extended Data Fig. S4, and broadly match self-consistent PP–TT profiles from Sonora Flame Skimmer [61]. The NIRSpec spectra are mainly sensitive to regions between 10110^{1} to 10−210^{-2} bars in the atmosphere, as indicated by the weighted emission contribution function. At pressures of ≈10\approx10–100100 bar, near the inferred CO2_2, CO/CH4_4/H2_2O, and NH3_3 quench points, the chemically-consistent retrieval yields a narrower range of P–T profiles compared to the free retrieval. This is because in the chemically-consistent retrieval, the photosphere and deeper atmosphere are linked through quenching and the chemical equilibrium abundances. In other words, when quenching happens below the photosphere, the molecular abundances in the photospheric region are controlled by both the equilibrium abundances below and the KzzK_{zz}.

To estimate the planet’s bolometric luminosity, we integrate the model spectra from 0.3–50 μ\mum.

To account for the remaining flux at longer wavelengths, we additionally integrate a blackbody spectrum normalized to match the model flux at 50 μ\mum and integrate it to an arbitrarily large wavelength of 10610^{6} μ\mum. We find the planet’s bolometric luminosity to be log⁡(Lbol/L⊙)=−7.23±0.01\log(L_{\mathrm{bol}}/L_{\odot})=-7.23\pm0.01. Given the retrieved radius of 1.02±0.011.02\pm0.01 RJupR_{\mathrm{Jup}}, this implies Teff=277±2T_{\mathrm{eff}}=277\pm2 K. These values are very consistent with those derived previously by modeling only ϵ\epsilon Ind Ab’s 4–25 μ\mum photometry data with self-consistent atmospheric models [15].

By comparing the three clouds models – clear, gray cloud, and EddySed with water clouds – we find that the gray cloud model is strongly preferred over the other two with ln⁡B≈20\ln B\approx20. The clear model PT profile shows an isothermal region around 10 bars and is highly discrepant with the self-consistent PT shapes, suggesting the clear model adjusts the PT profile to compensate for the lack of an gray opacity source from clouds. The clear and EddySed models also find a planet radius that is too small (≈0.9\approx0.9 RJupR_{\mathrm{Jup}}). While water clouds could suppress and alter the 4–5 μ\mum flux of cold planets like ϵ\epsilon Ind Ab [16, 68, 80], the effect of clouds is generally much easier to discern in the spectral continuum. Additional work on improving PSF subtraction techniques for NIRSpec [19] could allow the recovery of the planet’s 3–5 μ\mum continuum spectra and further test the presence of water clouds.

As noted in the ‘Chemistry’ section, we accounted for volatile condensation by applying a saturation cap in the chemically-consistent retrievals. However, in the free retrieval where we allow the gas-phase water abundance to decrease as a power law, we do not find evidence that such water depletion is preferred by the data. This is not surprising since for WISE 0855, only the MIRI/MRS spectra >15>15 μ\mum were sensitive to the effects of water depletion [69]. A future study analyzing the JWST MIRI/MRS spectroscopy of ϵ\epsilon Ind Ab from GO 8714 will explore the existence of water clouds in its atmosphere in more detail.

We infer log⁡10(Kzz/cm2s−1)=5.6±0.4\log_{10}(K_{zz}/\mathrm{cm}^{2}\mathrm{s}^{-1})=5.6\pm0.4 by combining the chemically-consistent retrievals of the two NIRSpec rolls. Note that this log⁡10(Kzz)\log_{10}(K_{zz}) value assumes L=HL=H, which is the commonly adopted convention for brown dwarf quenching calculations [72, 81]. If we instead adopt L=0.14HL=0.14H, as inferred for Jupiter [82], the same mixing timescale would correspond to log⁡10(Kzz/cm2s−1)=\log_{10}(K_{zz}/\mathrm{cm}^{2}\mathrm{s}^{-1})= 3.9 ±\pm 0.4. The corresponding quench pressures are 53 ±\pm 11, 27 ±\pm 4, and 90 ±\pm 22 bars for CO–CH4_4, CO2_2, and N2_2–NH3_3, respectively.

Methane–hydrogen fractionation correction for D/H

From the retrievals, we measure the D/H in methane for ϵ\epsilon Ind Ab. Because the atmosphere is dominated by H2_2, the D/H in molecular hydrogen more closely traces the bulk atmospheric D/H. We therefore correct for chemical fractionation between CH4_4 and H2_2 using two complementary methods, described below.

We first perform an analytic calculation following an established method for Solar System giant planets [36]. The methane–hydrogen enrichment factor is defined as

fCH4−H2=(D/H)CH4(D/H)H2.(6)f_{\mathrm{CH_4-H_2}} = \frac{(\mathrm{D/H})_{\mathrm{CH_4}}}{(\mathrm{D/H})_{\mathrm{H_2}}}. \tag*{(6)}

Deuterium is exchanged between methane and molecular hydrogen through the net reaction

CH4+HD⇌CH3D+H2.(7)\mathrm{CH_4 + HD \rightleftharpoons CH_3D + H_2}. \tag*{(7)}

In thermochemical equilibrium, the partitioning of D between methane and hydrogen can be described by the temperature-dependent fractionation factor α(T)\alpha(T) [36]. At temperatures exceeding 1200 K, α(T)\alpha(T) approaches unity so if the isotopic exchange reaction quenches in this regime fCH4−H2f_{\mathrm{CH_4-H_2}} would also be close to unity [36]. As gas is mixed upward in the atmosphere, the isotopic exchange reaction becomes slower until its chemical timescale τchem\tau_{\mathrm{chem}} exceeds the vertical mixing timescale τmix\tau_{\mathrm{mix}}. The CH3_3D/CH4_4 ratio is then approximately fixed at its equilibrium value at the quench point.

To estimate ff, we need to estimate the quench point of the isotopic exchange reaction [83]. Under the quench approximation, the enrichment in the observed atmosphere is

fCH4−H2≃α(Tquench).(8)f_{\mathrm{CH_4-H_2}} \simeq\alpha(T_{\mathrm{quench}}). \tag*{(8)}

To find TquenchT_{\mathrm{quench}}, we compare the isotope exchange timescale

τchem=[k−(T)nH2]−1(9)\tau_{\mathrm{chem}} = \left[k^{-}(T)n_{\mathrm{H_2}}\right]^{-1} \tag*{(9)}

with the vertical mixing timescale given by Equation (2). In Equation (9), k−(T)k^{-}(T) is the isotopic exchange rate constant [36, 84] and nH2n_{\mathrm{H_2}} is number density of H2_2, which we estimate using the H2_2 volume-mixing ratio from FastChem calculations in the retrieval.

From independent retrievals of the two NIRSpec rolls, we obtain (D/H)CH4=(5.4±0.2)×10−5(\mathrm{D}/\mathrm{H})_{\mathrm{CH}_4} = (5.4 \pm0.2) \times10^{-5}.

Using the inferred KzzK_{zz} from the retrievals, we find that the isotopic exchange reaction quenches at P=9±2P = 9 \pm2 bar and T=620±50T = 620 \pm50 K, which gives fCH4−H2≈1.35f_{\mathrm{CH}_4-\mathrm{H}_2} \approx1.35. For comparison, a similar calculation for Jupiter gives fCH4−H2=1.25±0.12f_{\mathrm{CH}_4-\mathrm{H}_2} = 1.25 \pm0.12 [36].

The analytic calculations above used the isotopic exchange rate constant, k−(T)k^{-}(T), inferred from experimental data with initial CD4_4–H2_2 mixtures [84]. To verify the isotopic fractionation of methane within a full chemical system, we also ran a kinetics model incorporating the relevant C–H–O elementary reactions. We used the open-source photochemical kinetics model VULCAN [85, 86], expanding its existing chemical network1 to include the following deuterated species: D, HD, D2_2, OD, D2_2O, HDO, CH3_3D, and CH2_2D (see Extended Data Table S3). To ensure consistency, the VULCAN models were run using the median retrieved elemental abundances and median PP–TT profile from the retrieval. For the relevant conditions in ε Ind Ab’s atmosphere, the kinetics model shows that the dominant pathway for isotope exchange between molecular hydrogen and methane is

CH4+H→CH3+H2CH3+HD→CH3D+H,(10)\begin{aligned} \mathrm{CH}_4 + \mathrm{H} &\rightarrow\mathrm{CH}_3 + \mathrm{H}_2 \\ \mathrm{CH}_3 + \mathrm{HD} &\rightarrow\mathrm{CH}_3\mathrm{D} + \mathrm{H}, \tag*{(10)} \end{aligned}

where the reaction between the methyl radical and HD controls the isotope exchange efficiency.

The kinetic results from VULCAN for two KzzK_{zz} values are shown in Extended Data Fig. S5, where we overplot the retrieved volume-mixing ratios for various species. The VULCAN model with log⁡10(Kzz/cm2 s−1)=5\log_{10}(K_{zz}/\mathrm{cm}^2\,\mathrm{s}^{-1}) = 5 matches the retrieved abundances best. This KzzK_{zz} value is consistent with the inferred KzzK_{zz} from the retrievals. The equilibrium CH3_3D/CH4_4 ratio increases with decreasing temperature but varies only gradually with pressure near the inferred quench region. As a result, the fractionation factor fCH4−H2f_{\mathrm{CH}_4-\mathrm{H}_2} is only weakly sensitive to KzzK_{zz} here. For log⁡10(Kzz/cm2 s−1)=5\log_{10}(K_{zz}/\mathrm{cm}^2\,\mathrm{s}^{-1}) = 5, VULCAN places the isotope exchange quench point at approximately 18 bar and 809 K, giving fCH4−H2=1.07f_{\mathrm{CH}_4-\mathrm{H}_2} = 1.07; for log⁡10(Kzz/cm2 s−1)=4\log_{10}(K_{zz}/\mathrm{cm}^2\,\mathrm{s}^{-1}) = 4, the corresponding values are 14 bar and 752 K, giving fCH4−H2=1.10f_{\mathrm{CH}_4-\mathrm{H}_2} = 1.10.

In summary, the full kinetics calculation finds f≈1.10f \approx1.10, whereas the analytic calculation yields f≈1.35f \approx1.35. This demonstrates that for ϵ\epsilon Ind Ab, isotopic fractionation only causes a modest 10–35% difference between (D/H)CH4_{\mathrm{CH}_4} and (D/H)H2_{\mathrm{H}_2}. Neither value brings (D/H)H2_{\mathrm{H}_2} close to the protosolar value. The difference between the ff values from the two calculations may arise from differences in the underlying chemical kinetics data. To conservatively account for model uncertainties, we choose to adopt a range of fCH4−H2=1.10f_{\mathrm{CH}_4-\mathrm{H}_2} = 1.10–1.351.35. After applying the fractionation correction, we obtain (D/H)H2=(4.4±0.7)×10−5_{\mathrm{H}_2} = (4.4 \pm0.7) \times10^{-5}, which represents the bulk atmospheric D/H of ϵ\epsilon Ind Ab.

Modeling the atmospheric composition of ϵ\epsilon Ind Ab

We use the open-source composition framework accomp [46, 87] to relate the composition of ϵ\epsilon Ind Ab to its building blocks. The code is available at https://github.com/y-chachan/accomp. In brief, this framework leverages measurements of volatile and refractory elements in a giant planet’s atmosphere to deduce how the planet acquired its metals and the quantity of metals accreted via different sources such as solids, metals in local gas, or metal-enhanced disk gas or solids through processes such as radial drift of pebbles. The disk thermal structure is adopted from previous work [88] with specific choices for stellar mass dependence of disk temperature [89]. In the outer part of the disk where irradiation sets the disk temperature, our adopted temperature profile is Tdisk≈43(r/16au)−3/7 KT_{\mathrm{disk}} \approx43\left(r/16\mathrm{au}\right)^{-3/7}\ \mathrm{K}.

For ϵ\epsilon Ind Ab, sulfur is a tracer of refractory solids as the planet is far out enough (16 ±\pm 2 AU) that all of it must have been in solid form at the planet’s formation location [90]. Thus, S/H informs the solid-to-gas accretion rate during envelope accretion and indicates that the planet acquired its metals through solid accretion. However, the planet’s S abundance once normalized by the relatively imprecise stellar S abundance does not allow us to discriminate between different models. Instead, we rely on the more precisely measured C, O, and N abundances to put constraints on the planet’s formation history. We vary the fraction of volatile species C, O and N that were present in disk solids at different locations to infer in which regions accretion would produce a composition that is compatible with the observed composition of ϵ\epsilon Ind Ab.

The near uniformity of C, O, N, and S enrichment suggests that the planet most likely accreted solids beyond the N2N_2 (∼25\sim25 K, 56 au) or CO snowline (∼30\sim30 K, 36 au). Accretion of material between the CO2_2 (∼70\sim70 K, 5 au) and CO snowlines would also match the observed composition. We proceed with the simplest model of accretion beyond the N2N_2 snowline, where all the metals are condensed into solids and their accretion produces uniform enrichment of C, O, N, and S. We note that the potential effects of additional processes, such as radial dust transport [91, 92, 93], and local disk cooling due to self-shadowing [94], are left for future studies. As discussed below, the simplest model of accretion beyond the N2\mathrm{N}_2 snowline is also compatible with the accretion of cold and unprocessed D-rich material inherited from the parent molecular cloud.

The planet’s C, O, S, and N atmospheric abundances constrain how much metals it accreted during envelope accretion and provide a lower limit on its bulk metallicity and the metal mass in the planet [22, 46]. For super-Jupiters such as ϵ\epsilon Ind Ab, most of the metals are expected to be in the envelope and the core’s contribution is small in comparison [95]. Assuming that the enrichment of Fe, Mg, and Si (i.e. other refractory species) follow that of S, we find that the planet’s bulk metallicity is 0.043−0.006+0.0070.043^{+0.007}_{-0.006}. This translates to 89−15+18 M⊕89^{+18}_{-15}\,M_{\oplus} of metals in the planet assuming a planet mass of 6.5±0.7 MJ6.5 \pm0.7\,M_{\mathrm{J}}. The planet’s bulk metallicity is ∼4.5×\sim4.5 \times the stellar Z⋆=0.0096Z_{\star} = 0.0096, which is commensurate with the super-stellar bulk metallicity found for the transiting giant planet population as well as several directly imaged super-Jupiters [35, 95].

The measured elemental abundances allow us to infer the D/H of accreted material needed to match the planet’s D/H. We consider two sources of D-rich material: water ice and organics. In our composition model, a fraction fO,H2O=0.45f_{\mathrm{O,H_2O}} = 0.45 of the disk’s O inventory is in the form of water and all of the planet’s O comes from the accretion of solids. So, the amount fO,H2O×xOf_{\mathrm{O,H_2O}} \times x_{\mathrm{O}} comes from the accretion of water ice, where xOx_{\mathrm{O}} is the measured enrichment of O. Each O atom brings 2 H atoms along with it, some of which are substituted with D. Similarly, a fraction fC,organics=0.5f_{\mathrm{C,organics}} = 0.5 of C is assumed to be in organics in the disk and they contribute fC,organics×xCf_{\mathrm{C,organics}} \times x_{\mathrm{C}} of the planet’s observed enrichment assuming all of C comes from solid accretion. The amount of H each C atom brings along in organics is determined by (H/C)organics(\mathrm{H}/\mathrm{C})_{\mathrm{organics}}, and depends on how hydrogenated the organics are. The ratio (H/C)organics(\mathrm{H}/\mathrm{C})_{\mathrm{organics}} can vary from ∼0.1−2\sim0.1 - 2 for various organics based on observations of Solar System comets [96, 97]. We adopt an intermediate (H/C)organics=1(\mathrm{H}/\mathrm{C})_{\mathrm{organics}} = 1 as our fiducial value because it represents the value of the (H/C) in complex organic matter, which constitutes most of the refractory carbon.

We calculate the planet’s D/H from the D/H of its building blocks as

(D/H)p=∑inH+D,i(D/H)i/[1+(D/H)i]∑inH+D,i/[1+(D/H)i],(11)(\mathrm{D}/\mathrm{H})_{\mathrm{p}} = \frac{\sum_i n_{\mathrm{H+D},i}(\mathrm{D}/\mathrm{H})_i/[1 + (\mathrm{D}/\mathrm{H})_i]}{\sum_i n_{\mathrm{H+D},i}/[1 + (\mathrm{D}/\mathrm{H})_i]}, \tag*{(11)}

where nH+D,in_{\mathrm{H+D},i} and (D/H)i(\mathrm{D}/\mathrm{H})_i are the total number of H and D atoms and the D/H, respectively, of species ii. The nH+D,in_{\mathrm{H+D},i} terms are normalized such that ∑inH+D,i=1\sum_i n_{\mathrm{H+D},i}=1. The contributions of water and organics to H and D atoms are calculated as 2fO,H2OxO(O/H)⋆2f_{\mathrm{O,H_2O}}x_{\mathrm{O}}(\mathrm{O}/\mathrm{H})_{\star} and (H/C)organicsfC,organicsxC(C/H)⋆(\mathrm{H}/\mathrm{C})_{\mathrm{organics}}f_{\mathrm{C,organics}}x_{\mathrm{C}}(\mathrm{C}/\mathrm{H})_{\star}, respectively. Stellar values for (O/H)⋆=108.55/1012=10−3.45(\mathrm{O}/\mathrm{H})_{\star}=10^{8.55}/10^{12}=10^{-3.45} and (C/H)⋆=108.21/1012=10−3.79(\mathrm{C}/\mathrm{H})_{\star}=10^{8.21}/10^{12}=10^{-3.79} [33] follow the same convention used for solar composition.

Considering water ice and organics in isolation at first, we find that water ice needs to have a D/H =1.5−0.5+0.6×10−2=1.5^{+0.6}_{-0.5}\times10^{-2} and organics need to have a D/H =5.4−1.7+2.1×10−2=5.4^{+2.1}_{-1.7}\times10^{-2} for the planet’s enrichment in C, O, and D to be mutually compatible. This can also be seen using an order-of-magnitude treatment (neglecting the [1+(D/H)i][1+(\mathrm{D}/\mathrm{H})_i] terms in Equation (11)), which yields (D/H)water∼(4.4×10−5−2×10−5)/(2×0.45×5.2×10−3.45)∼1.5×10−2(\mathrm{D}/\mathrm{H})_{\mathrm{water}}\sim(4.4\times10^{-5}-2\times10^{-5})/(2\times0.45\times5.2\times10^{-3.45})\sim1.5\times10^{-2} and (D/H)organics∼(4.4×10−5−2×10−5)/(1×0.5×5.8×10−3.79)∼5×10−2(\mathrm{D}/\mathrm{H})_{\mathrm{organics}}\sim(4.4\times10^{-5}-2\times10^{-5})/(1\times0.5\times5.8\times10^{-3.79})\sim5\times10^{-2}. For refractory organics, C/Si could be as high as 5.85, which corresponds to 70% of C being in organics [98] (as opposed to our assumed 50%), and H/C ∼1.5\sim1.5 for pristine organics [55]. With these assumptions, our order-of-magnitude calculation yields a lower required value of (D/H)organics∼2.5×10−2(\mathrm{D}/\mathrm{H})_{\mathrm{organics}}\sim2.5\times10^{-2}. Extended Data Fig. S6 shows the D/H obtained for the planet as a function of O/H and C/H for a range of D/H of water ice and refractory organics (H/C =1=1), respectively. Accretion of material with Solar System-like D/H would require a much higher (∼200×\sim200\times stellar) C/H or O/H to match the observed D/H of ϵ\epsilon Ind Ab.

For a more comprehensive analysis, we create a grid of values for the D/H of water and organics (with H/C =1=1) in the range 10−4−10−110^{-4}-10^{-1}. We calculate the planet’s expected D/H over this grid and compare it to the measured D/H of the planet for the full set of retrieved abundance chains by calculating a χ2\chi^2 value. Extended Data Fig. S7 shows the χ2\chi^2 map obtained for the D/H of water and organics to fit the planet’s measured D/H. It agrees with the order-of-magnitude estimate above and suggests that, when considered independently, the D/H values of water and organics need to be ∼1.5×10−2\sim1.5\times10^{-2} and ∼5×10−2\sim5\times10^{-2}, respectively, to match the planet’s D/H given its O and C enrichment.

Feasibility for accretion of deuterium-rich gas

We also investigate whether the disk gas can become enriched in deuterium and heavy elements during disk dispersal, and whether this alternative scenario could explain the observed enrichment of D and metals in ϵ\epsilon Ind Ab. The vertical gravitational force in the disk is so weak that the critical molecular/atomic mass of a species above which fractionation would occur is much larger the mass of most relevant molecules (HD\mathrm{HD}, CO\mathrm{CO}, H2O\mathrm{H_2O}, N2\mathrm{N_2}), i.e., the outflow itself is expected to be hydrodynamic and not lead to any fractionation [99, 100]. Enrichment of disk gas in D and metals would only occur if these species are vertically settled relative to the background disk gas H2\mathrm{H_2} such that these species are depleted relative to H2\mathrm{H_2} at the wind-launching height above the midplane. This is equivalent to the concept of the homopause in planetary atmospheres above which the scale heights of different species depend on their molecular or atomic weight [99]. We can calculate the conditions under which such vertical separation would occur by comparing molecular diffusivity DmD_m with turbulent diffusivity DtD_t. For heavier elements to settle relative to H2\mathrm{H_2}, we need Dm>DtD_m > D_t.

The molecular diffusivity of a binary mixture of H2\mathrm{H_2} and HD\mathrm{HD} is given by Dm=bH2,HD/nD_m = b_{\mathrm{H_2,HD}}/n, where n is the total number density of the gas and bH2,HD≈5×1017T0.75b_{\mathrm{H_2,HD}} \approx5 \times10^{17} T^{0.75} molecules cm−1 s−1\mathrm{cm^{-1}\,s^{-1}} is the binary diffusion coefficient [101, 102] for HD\mathrm{HD} and H2\mathrm{H_2} at temperature TT. For turbulent diffusivity, we use Dt=αtcsHD_t = \alpha_t c_s H, where αt\alpha_t is the Shakura-Sunyaev parameter, csc_s is the sound speed, and HH is the disk scale height. The number density nn of the gas can be written as n=Σ/(2πHμ)n = \Sigma/(\sqrt{2\pi}H\mu) for gas surface density Σ\Sigma, and mean particle mass μ=2.3×\mu= 2.3 \times the mass of a proton, assuming a vertically isothermal and hydrostatic disk. For αt=10−4\alpha_t = 10^{-4} and disk thermal structure adopted above (Tdisk≈43 (r/16au)−3/7 KT_{\mathrm{disk}} \approx43\,(r/16\mathrm{au})^{-3/7}\ \mathrm{K}), we find that vertical settling comes into play when

Σ≤bH2,HD2πμαtcs≈1.85×10−5(r16 au)−3/28(αt10−4)−1 g cm−2.(12)\Sigma\le\frac{b_{\mathrm{H_2,HD}}\sqrt{2\pi}\mu}{\alpha_t c_s} \approx1.85 \times10^{-5}\left(\frac{r}{16\ \mathrm{au}}\right)^{-3/28}\left(\frac{\alpha_t}{10^{-4}}\right)^{-1}\ \mathrm{g\ cm^{-2}}. \tag*{(12)}

The required surface density is very weakly dependent on orbital distance and given that Σdisk\Sigma_{\mathrm{disk}} varies as r−1r^{-1} to r−3/2r^{-3/2}, it is easier to meet this criterion further out in the disk. However, the pre-factor of ∼2×10−5\sim2 \times10^{-5} implies a highly depleted nebula, e.g., a depletion of ∼5×10−7\sim5 \times10^{-7} relative to the MMSN of Chiang & Youdin [103] at 16 au. By the time a disk reaches such a low surface density, a giant planet is expected to have accreted almost the entirety of its mass. We verify this by calculating ϵ\epsilon Ind Ab’s mass doubling timescale using the maximum accretion rate from numerical simulations [104] and find that tdouble=Mp/M˙≈2 Gyrt_{\mathrm{double}} = M_p/\dot{M} \approx2\ \mathrm{Gyr}, far longer than the disk lifetime. The planet would barely accrete any mass after the disk surface density has become low enough for vertical settling of deuterium and metals to matter. Therefore, accretion of deuterium-rich nebular gas is an implausible explanation for ϵ\epsilon Ind Ab’s elevated D/H.

Benchmarking the retrieval framework with WISE 0855

To benchmark our atmospheric retrieval framework, we ran a retrieval on the public NIRSpec/G395M spectrum of WISE 0855 from GO 2327 (PI A. Skemer). For this dataset, a value of (D/H)CH4=1.27−0.19+0.22×10−5(\mathrm{D}/\mathrm{H})_{\mathrm{CH}_4} = 1.27^{+0.22}_{-0.19} \times10^{-5} was measured from CHIMERA retrievals [18, 105]. Using our free retrieval framework with petitRADTRANS, we fit the same WISE 0855 spectrum and obtain (D/H)CH4=1.27−0.14+0.15×10−5(\mathrm{D}/\mathrm{H})_{\mathrm{CH}_4} = 1.27^{+0.15}_{-0.14} \times10^{-5}, fully consistent with the literature value [18].

Figure S3. Carbon and oxygen isotope ratios measured from different isotopologues in ϵ\epsilon Ind Ab. CO is blue, H2_2O is green, and CH4_4 is orange. In each panel, the circle marker denotes the solar ratios [77]. 12^{12}C/13^{13}C is measured from both 13^{13}CO and 13^{13}CH4_4, whereas 16^{16}O/18^{18}O is measured from both C18^{18}O and H218_2^{18}O. The isotopic ratios are consistent with the solar value at the <1σ< 1\sigma level.

Figure S4. Retrieved P-T profiles for ϵ\epsilon Ind Ab. Profiles from the chemically-consistent retrieval (left) and free retrieval (right). The orange line is the median retrieved P-T while the orange shade denotes the 2σ2\sigma interval on the retrieved profiles. In both panels, we overplot P-T profiles from the Sonora Flame Skimmer model [61] with Teff=275T_{\rm eff}=275 K, C/O=0.45, [M/H]=0.5 dex, log⁡Kzz=7.0\log K_{zz}=7.0, and two values of surface gravity: log⁡(g)=4.25\log(g)=4.25 (dashed line) and log⁡(g)=4.0\log(g)=4.0 (dotted line). We show the weighted emission contribution function probed by NIRSpec using the dashdot line on the pressure axis. From P≈10−100P\approx10-100 bars, the profiles are more constrained in the chemically-consistent retrieval since the deeper atmosphere and photosphere are connected via quenching of the chemical equilibrium abundances (see ‘P−TP-T profile and bulk properties’).

Figure S5. Volume-mixing ratio profiles for a subset of species detected in ϵ\epsilon Ind Ab. The shaded contours show the 2σ2\sigma range from the chemically-consistent retrievals, whereas the black dashdot and solid lines show model abundance profiles from the VULCAN chemical kinetics code for two different KzzK_{zz} values. Overall, the log⁡10(Kzz/cm2 s−1)=5\log_{10}(K_{zz}/\mathrm{cm}^{2}\,\mathrm{s}^{-1}) = 5 VULCAN model matches the retrieved abundances well. Depending on the assumption of mixing length, the retrieval finds log⁡10(Kzz)=3.9−5.6\log_{10}(K_{zz}) = 3.9 - 5.6, which is consistent with the VULCAN value. The retrieval H2O\mathrm{H}_{2}\mathrm{O} and H2S\mathrm{H}_{2}\mathrm{S} abundance profiles decrease at low pressures due to condensation, which we implement using the saturation vapor pressure relations (74).

U\mathcal{U} denotes a uniform prior, and N\mathcal{N} denotes a Gaussian prior. The P–T parameters ∇i\nabla_i are the fitted dln⁡T/dln⁡Pd\ln T/d\ln P values, with the associated pressure nodes given in square brackets in log(bar). All logarithms are base 10. In the free retrieval, each molecular species has an independent constant mass-mixing ratio. In the chemically-consistent retrieval which uses FastChem, the molecular abundances are instead determined from the elemental abundances and KzzK_{\mathrm{zz}}, while isotopologue ratios are fitted as log10 mass-fraction ratios relative to their parent isotopologues.

SpeciesReference
12C16O{}^{12}\mathrm{C}{}^{16}\mathrm{O}, 13C16O{}^{13}\mathrm{C}{}^{16}\mathrm{O}, 12C18O{}^{12}\mathrm{C}{}^{18}\mathrm{O}[106]
H216O\mathrm{H}_{2}^{16}\mathrm{O}[107]
H218O\mathrm{H}_{2}^{18}\mathrm{O}, H217O\mathrm{H}_{2}^{17}\mathrm{O}[108]
12CH4{}^{12}\mathrm{C}\mathrm{H}_{4}[109]
13CH4{}^{13}\mathrm{C}\mathrm{H}_{4}, CH3D\mathrm{CH}_{3}\mathrm{D}[110]
14NH3{}^{14}\mathrm{N}\mathrm{H}_{3}[111]
12C16O2{}^{12}\mathrm{C}{}^{16}\mathrm{O}_{2}[112]
H2S\mathrm{H}_{2}\mathrm{S}[113]

Table S2. Line opacities used in the atmospheric retrievals.

No.Reactiona^{a}AAnnEaE_a (K)Ref.
Rate expression: k=ATnexp⁡(−Ea/T)k = A T^n \exp(-E_a/T)
Bimolecular (AA in cm3 s−1\mathrm{cm^3\,s^{-1}})
R1H2+D→H+HD\mathrm{H_2 + D \to H + HD}3.98×10−183.98 \times10^{-18}2.2902630.0[114]
R3D2+O→OD+D\mathrm{D_2 + O \to OD + D}1.99×10−161.99 \times10^{-16}1.7204880.0[115]
R5D2+OD→D2O+D\mathrm{D_2 + OD \to D_2O + D}1.60×10−171.60 \times10^{-17}1.8701724.0[116]
R7D2+H→HD+D\mathrm{D_2 + H \to HD + D}1.69×10−171.69 \times10^{-17}2.1003530.0[117]
R9D2+OH→HDO+D\mathrm{D_2 + OH \to HDO + D}3.30×10−173.30 \times10^{-17}1.7701830.0[116]
R11O(1D)+D2→OD+D\mathrm{O(^{1}D) + D_2 \to OD + D}1.10×10−101.10 \times10^{-10}0.0000.0[118]
R13OH+HD→HDO+H\mathrm{OH + HD \to HDO + H}7.87×10−127.87 \times10^{-12}0.0002810.0[119]
R15OH+HD→H2O+D\mathrm{OH + HD \to H_2O + D}4.70×10−124.70 \times10^{-12}0.0002260.0[119]
R17OH+D→OD+H\mathrm{OH + D \to OD + H}1.90×10−91.90 \times10^{-9}−0.630-0.6300.0[120]
R19H2+OD→H+HDO\mathrm{H_2 + OD \to H + HDO}2.80×10−172.80 \times10^{-17}1.8301419.0[116]
R21CH4+OD→CH3+HDO\mathrm{CH_4 + OD \to CH_3 + HDO}3.90×10−193.90 \times10^{-19}2.3901270.0[121]
R23CH3D+OH→CH2D+H2O\mathrm{CH_3D + OH \to CH_2D + H_2O}5.19×10−185.19 \times10^{-18}2.0001332.0[122]
R25CH3D+OH→CH3+HDO\mathrm{CH_3D + OH \to CH_3 + HDO}6.70×10−196.70 \times10^{-19}2.3201430.0[121]
R27CH3+D→CH2D+H\mathrm{CH_3 + D \to CH_2D + H}2.31×10−102.31 \times10^{-10}0.0000.0[123]
R29CH3+D→CH3D\mathrm{CH_3 + D \to CH_3D}2.74×10−102.74 \times10^{-10}0.000106.0[124]
R31CH3+HD→CH3D+H\mathrm{CH_3 + HD \to CH_3D + H}2.23×10−202.23 \times10^{-20}2.5144295.0[125]
R33CH4+D→CH3+HD\mathrm{CH_4 + D \to CH_3 + HD}7.47×10−117.47 \times10^{-11}0.0005590.0[126]
R35H2O+OD→HDO+OH\mathrm{H_2O + OD \to HDO + OH}7.23×10−167.23 \times10^{-16}1.980−527.0-527.0[127]
R37H+CH3D→CH2D+H2\mathrm{H + CH_3D \to CH_2D + H_2}1.65×10−201.65 \times10^{-20}3.0004040.0assumed 34×H+CH4\frac{3}{4} \times\mathrm{H + CH_4}
R39CH3+D2→CH3D+D\mathrm{CH_3 + D_2 \to CH_3D + D}2.36×10−112.36 \times10^{-11}0.0007810.0[128]
Termolecular (first row: low-pressure limit k0k_0 in cm6 s−1\mathrm{cm^6\,s^{-1}}; second row: high-pressure limit k∞k_\infty in cm3 s−1\mathrm{cm^3\,s^{-1}})
R41D+D+M→D2+M\mathrm{D + D + M \to D_2 + M}6.09×10−336.09 \times10^{-33}<br>5.00×10−165.00 \times10^{-16}0.000<br>0.0000.0<br>0.0[129, 130]<br>[130]
R43D+H+M→HD+M\mathrm{D + H + M \to HD + M}2.70×10−312.70 \times10^{-31}<br>3.31×10−63.31 \times10^{-6}−0.600-0.600<br>−1.000-1.0000.0<br>0.0assumed same as H+H→H2\mathrm{H + H \to H_2}
R45CH2D+H+M→CH3+D+M\mathrm{CH_2D + H + M \to CH_3 + D + M}1.99×10−311.99 \times10^{-31}<br>5.15×10−225.15 \times10^{-22}−1.090-1.090<br>2.3001100.0<br>207.0[131]
R47CH3D+M→CH2D+H+M\mathrm{CH_3D + M \to CH_2D + H + M}1.16×10121.16 \times10^{12}<br>5.57×10165.57 \times10^{16}−4.480-4.480<br>−0.320-0.32053 800.0<br>52 300.0[131]
R49CH3D+M→CH3+D+M\mathrm{CH_3D + M \to CH_3 + D + M}9.69×1089.69 \times10^{8}<br>1.49×10181.49 \times10^{18}−3.850-3.850<br>−0.890-0.89054 700.0<br>53 800.0[131]
Photodissociation
R51HDO→H+OD\mathrm{HDO \to H + OD}cross-sections from [132, 133]; branching ratios assumed the same as H2O\mathrm{H_2O}
R53HDO→D+OH\mathrm{HDO \to D + OH}
R55CH3D→CH2D+H\mathrm{CH_3D \to CH_2D + H}assumed same as CH4\mathrm{CH_4}
R57CH3D→CH3+D\mathrm{CH_3D \to CH_3 + D}
R59CH3D→CH2+HD\mathrm{CH_3D \to CH_2 + HD}

Table S3. Deuterium reactions included in VULCAN.

References

  1. [1]R. J. Cooke, M. Pettini, C. C. Steidel, One Percent Determination of the Primordial Deuterium Abundance. Astrophys. J. 855 (2), 102 (2018), doi:10.3847/1538-4357/aaab53.DOI
  2. [2]T. Prodanović, G. Steigman, B. D. Fields, The deuterium abundance in the local interstellar medium. Mon. Not. R. Astron. Soc. 406 (2), 1108–1115 (2010), doi:10.1111/j.1365-2966.2010.16734.x.DOI
  3. [3]C. Ceccarelli, et al., Deuterium Fractionation: The Ariadne’s Thread from the Precollapse Phase to Meteorites and Comets Today, in Protostars and Planets VI, H. Beuther, R. S. Klessen, C. P. Dullemond, T. Henning, Eds. (2014), pp. 859–882, doi:10.2458/azu_uapress_9780816531240-ch037.DOI
  4. [4]D. Bockelée-Morvan, et al., Cometary Isotopic Measurements. Space Sci. Rev. 197 (1-4), 47–83 (2015), doi:10.1007/s11214-015-0156-9.
  5. [5]D. R. Müller, et al., High D/H ratios in water and alkanes in comet 67P/Churyumov-Gerasimenko measured with Rosetta/ROSINA DFMS. Astron. Astrophys. 662, A69 (2022), doi:10.1051/0004-6361/202142922.DOI
  6. [6]C. M. O. Alexander, et al., The Provenances of Asteroids, and Their Contributions to the Volatile Inventories of the Terrestrial Planets. Science 337 (6095), 721–723 (2012), doi:10.1126/science.1223474.DOI
  7. [7]L. I. Cleeves, et al., The ancient heritage of water ice in the solar system. Science 345 (6204), 1590–1593 (2014), doi:10.1126/science.1258055.DOI
  8. [8]T. M. Donahue, J. H. Hoffman, R. R. Hodges, A. J. Watson, Venus Was Wet: A Measurement of the Ratio of Deuterium to Hydrogen. Science 216 (4546), 630–633 (1982), doi:10.1126/science.216.4546.630.DOI
  9. [9]G. L. Villanueva, et al., Strong water isotopic anomalies in the martian atmosphere: Probing current and ancient reservoirs. Science 348 (6231), 218–221 (2015), doi:10.1126/science.aaa3630.DOI
  10. [10]E. Lellouch, et al., The deuterium abundance in Jupiter and Saturn from ISO-SWS observations. Astron. Astrophys. 370, 610–622 (2001), doi:10.1051/0004-6361:20010259.DOI
  11. [11]J. D. R. Pierel, et al., D/H Ratios on Saturn and Jupiter from Cassini CIRS. Astron. J. 154 (5), 178 (2017), doi:10.3847/1538-3881/aa899d.DOI
  12. [12]W. Hubbard, J. MacFarlane, Theoretical predictions of deuterium abundances in the Jovian planets. Icarus 44 (3), 676–682 (1980), doi:https://doi.org/10.1016/0019-1035(80)90136-0.
  13. [13]H. Feuchtgruber, et al., The D/H ratio in the atmospheres of Uranus and Neptune from Herschel-PACS observations. Astron. Astrophys. 551, A126 (2013), doi:10.1051/0004-6361/201220857.DOI
  14. [14]E. C. Matthews, et al., A temperate super-Jupiter imaged with JWST in the mid-infrared. Nature 633 (8031), 789–792 (2024), doi:10.1038/s41586-024-07837-8.DOI
  15. [15]A. Sanghi, et al., Worlds Next Door. IV. Mapping the Late Stages of Giant Planet Evolution with a Precise Dynamical Mass and Luminosity for Eps Ind Ab. Astron. J. 172 (2), 115 (2026), doi:10.3847/1538-3881/ae74c8.DOI
  16. [16]E. C. Matthews, et al., A Second Visit to Eps Ind Ab with JWST: New Photometry Confirms Ammonia and Suggests Thick Clouds in the Exoplanet Atmosphere of the Closest Super-Jupiter. Astrophys. J. Lett. 1002 (1), L5 (2026), doi:10.3847/2041-8213/ae5823.DOI
  17. [17]C. V. Morley, et al., Measuring the D/H Ratios of Exoplanets and Brown Dwarfs. Astrophys. J. Lett. 882 (2), L29 (2019), doi:10.3847/2041-8213/ab3c65.DOI
  18. [18]M. J. Rowland, et al., Protosolar D-to-H Abundance and One Part per Billion PH₃ in the Coldest Brown Dwarf. Astrophys. J. Lett. 977 (2), L49 (2024), doi:10.3847/2041-8213/ad9744.DOI
  19. [19]J.-B. Ruffio, et al., JWST-TST High Contrast: Achieving Direct Spectroscopy of Faint Sub-stellar Companions Next to Bright Stars with the NIRSpec Integral Field Unit. Astron. J. 168 (2), 73 (2024), doi:10.3847/1538-3881/ad5281.DOI
  20. [20]A. Madurowicz, et al., Direct Spectroscopy of 51 Eridani b with JWST NIRSpec. Astron. J. 170 (6), 326 (2025), doi:10.3847/1538-3881/ae1028.
  21. [21]J.-B. Ruffio, et al., Jupiter-like uniform metal enrichment in a system of multiple giant exoplanets. Nature Astronomy 10, 511–521 (2026), doi:10.1038/s41550-026-02783-z.DOI
  22. [22]J. W. Xuan, et al., The Compositions of the HR 8799 Planets Reflect Accretion of Both Solids and Metal-enriched Gas. Astrophys. J. 1000 (1), 27 (2026), doi:10.3847/1538-4357/ae448f.DOI
  23. [23]A. Baburaj, et al., JWST-TST High Contrast: First Direct Spectroscopy of GJ 504 b Reveals Clouds and Possible Metal Enrichment. Astron. J. 172 (1), 28 (2026), doi:10.3847/1538-3881/ae6919.DOI
  24. [24]K. K. W. Hoch, et al., The Dyn-Atmo Survey: High-contrast Imaging Spectroscopy of the Substellar Companion HD 13724 B with the JWST NIRSpec IFU. arXiv e-prints arXiv:2609.11827 (2026), doi:10.48550/arXiv.2609.11827.DOI
  25. [25]A. Madurowicz, et al., The Dyn-Atmo Survey: JWST/NIRSpec spectroscopy of dynamical benchmark GJ 758 B. arXiv e-prints arXiv:2609.11828 (2026), doi:10.48550/arXiv.2609.11828.DOI
  26. [26]P. Mollière, et al., petitRADTRANS. A Python Radiative Transfer Package for Exoplanet Characterization and Retrieval. Astron. Astrophys. 627, A67 (2019), doi:10.1051/0004-6361/201935470.DOI
  27. [27]E. Nasedkin, P. Mollière, D. Blain, Atmospheric Retrievals with petitRADTRANS. The Journal of Open Source Software 9 (96), 5875 (2024), doi:10.21105/joss.05875.DOI
  28. [28]Z. Zhang, et al., Elemental abundances of Planets and brown dwarfs Imaged around Stars (ELPIS). I. Potential Metal Enrichment of the Exoplanet AF Lep b and a Novel Retrieval Approach for Cloudy Self-luminous Atmospheres. Astron. J. 166 (5), 198 (2023), doi:10.3847/1538-3881/acf768.DOI
  29. [29]K. J. Zahnle, M. S. Marley, Methane, Carbon Monoxide, and Ammonia in Brown Dwarfs and Self-Luminous Giant Planets. Astrophys. J. 797, 41 (2014), doi:10.1088/0004-637X/797/1/41.DOI
  30. [30]J. W. Xuan, et al., A Clear View of a Cloudy Brown Dwarf Companion from High-resolution Spectroscopy. Astrophys. J. 937, 54 (2022), doi:10.3847/1538-4357/ac8673.DOI
  31. [31]Y. Zhang, et al., The ¹³CO-rich atmosphere of a young accreting super-Jupiter. Nature 595 (7867), 370–372 (2021), doi:10.1038/s41586-021-03616-x.DOI
  32. [32]Materials and methods are available as supplementary material.
  33. [33]A. Baburaj, et al., A High-resolution Spectroscopic Survey of Directly Imaged Companion Hosts. III. Characterization of the Cold Imaged Planet Hosts AF Lep A and Eps Indi A. Astron. J. 172 (4), 227 (2026), doi:10.3847/1538-3881/ae9afe.DOI
  34. [34]T. Guillot, et al., Giant Planets from the Inside-Out, in Protostars and Planets VII, S. Inutsuka, Y. Aikawa, T. Muto, K. Tomida, M. Tamura, Eds., vol. 534 of Astronomical Society of the Pacific Conference Series (2023), p. 947, doi:10.48550/arXiv.2205.04100.arxiv.org/abs/2205.04100
  35. [35]J. W. Xuan, et al., A Sulfur-rich Atmosphere for the Young Jupiter Analog AF Lep b Reveals Significant Solid Accretion. Astron. J. 172 (4), 233 (2026), doi:10.3847/1538-3881/ae9d73.DOI
  36. [36]C. Lecluse, F. Robert, D. Gautier, M. Guiraud, Deuterium enrichment in giant planets. Planet. Space Sci. 44 (12), 1579–1592 (1996), doi:10.1016/S0032-0633(96)00070-0.DOI
  37. [37]M. J. McCaughrean, et al., ε Indi Ba,Bb: The nearest binary brown dwarf. Astron. Astrophys. 413, 1029–1036 (2004), doi:10.1051/0004-6361:20034292.DOI
  38. [38]J. L. Linsky, et al., What Is the Total Deuterium Abundance in the Local Galactic Disk? Astrophys. J. 647 (2), 1106–1124 (2006), doi:10.1086/505556.DOI
  39. [39]H. Nomura, et al., The Isotopic Links from Planet Forming Regions to the Solar System, in Protostars and Planets VII, S. Inutsuka, Y. Aikawa, T. Muto, K. Tomida, M. Tamura, Eds., vol. 534 of Astronomical Society of the Pacific Conference Series (2023), p. 1075.arxiv.org/abs/2203.10863
  40. [40]L. E. Salazar Manzano, et al., Water D/H in 3I/ATLAS as a probe of formation conditions in another planetary system. Nature Astronomy 10, 988–1000 (2026), doi:10.1038/s41550-026-02850-5.DOI
  41. [41]M. Cordiner, et al., Isotopic evidence for a cold and distant origin of 3I/ATLAS. Nature 655 (8124), 870–874 (2026), doi:10.1038/s41586-026-10771-6.DOI
  42. [42]N. X. Roth, et al., An Enriched Methane D/H Ratio in the Interstellar Object 3I/ATLAS. arXiv e-prints arXiv:2603.20445 (2026), doi:10.48550/arXiv.2603.20445.DOI
  43. [43]J. J. Tobin, et al., Deuterium-enriched water ties planet-forming disks to comets and protostars. Nature 615 (7951), 227–230 (2023), doi:10.1038/s41586-022-05676-z.DOI
  44. [44]M. Leemker, et al., Pristine ices in a planet-forming disk revealed by heavy water. Nature Astronomy 9, 1486–1494 (2025), doi:10.1038/s41550-025-02663-y.DOI
  45. [45]K. Altwegg, et al., D2O and HDS in the coma of 67P/Churyumov-Gerasimenko. Philosophical Transactions of the Royal Society of London Series A 375 (2097), 20160253 (2017), doi:10.1098/rsta.2016.0253.DOI
  46. [46]Y. Chachan, R. Murray-Clay, J. J. Fortney, Quantitative Links between Formation and Atmospheric Composition for Giant Planets. arXiv e-prints arXiv:2608.23682 (2026), doi:10.48550/arXiv.2608.23682.DOI
  47. [47]N. Monga, S. Desch, External Photoevaporation of the Solar Nebula: Jupiter’s Noble Gas Enrichments. Astrophys. J. 798 (1), 9 (2015), doi:10.1088/0004-637X/798/1/9.DOI
  48. [48]O. Mousis, T. Ronnet, J. I. Lunine, Jupiter’s Formation in the Vicinity of the Amorphous Ice Snowline. Astrophys. J 875 (1), 9 (2019), doi:10.3847/1538-4357/ab0a72.DOI
  49. [49]B. Parise, et al., Testing grain surface chemistry: a survey of deuterated formaldehyde and methanol in low-mass class 0 protostars. Astron. Astrophys. 453 (3), 949–958 (2006), doi:10.1051/0004-6361:20054476.DOI
  50. [50]V. Taquet, et al., Interferometric observations of warm deuterated methanol in the inner regions of low-mass protostars. Astron. Astrophys. 632, A19 (2019), doi:10.1051/0004-6361/201936044.DOI
  51. [51]S. Howard, et al., Exploring the Hypothesis of an Inverted Z Gradient inside Jupiter. Astron. Astrophys. 680, L2 (2023), doi:10.1051/0004-6361/202348129.DOI
  52. [52]S. Müller, R. Helled, Can Jupiter’s Atmospheric Metallicity Be Different from the Deep Interior? Astrophys. J. 967, 7 (2024), doi:10.3847/1538-4357/ad3738.DOI
  53. [53]H. Busemann, et al., Interstellar Chemistry Recorded in Organic Matter from Primitive Meteorites. Science 312 (5774), 727–730 (2006), doi:10.1126/science.1123878.DOI
  54. [54]K. E. Mandt, et al., A nearly terrestrial D/H for comet 67P/Churyumov-Gerasimenko. Science Advances 10 (46), eadp2191 (2024), doi:10.1126/sciadv.adp2191.DOI
  55. [55]J. A. Paquette, et al., D/H in the refractory organics of comet 67P/Churyumov-Gerasimenko measured by Rosetta/COSIMA. Mon. Not. R. Astron. Soc. 504 (4), 4940–4951 (2021), doi:10.1093/mnras/stab1028.DOI
  56. [56]J. Huang, et al., An ALMA Survey of DCN/H¹³CN and DCO⁺/H¹³CO⁺ in Protoplanetary Disks. Astrophys. J. 835 (2), 231 (2017), doi:10.3847/1538-4357/835/2/231.DOI
  57. [57]M. Asplund, N. Grevesse, A. J. Sauval, P. Scott, The Chemical Composition of the Sun. Annu. Rev. Astron. Astrophys. 47 (1), 481–522 (2009), doi:10.1146/annurev.astro.46.060407.145222.DOI
  58. [58]J. W. Xuan, JWST GO 8714: NIRSpec IFU observations of ε Indi Ab, Mikulski Archive for Space Telescopes (MAST) (2026), doi:10.17909/anqj-a832.
  59. [59]S. Agrawal, J.-B. Ruffio, M. Perrin, A. Madurowicz, B. Sappey, jruffio/breads: Accepted HD19467B paper (2024), doi:10.5281/zenodo.11391503.DOI
  60. [60]H. Bushouse, et al., JWST Calibration Pipeline (2026), doi:10.5281/zenodo.19558572.
  61. [61]J. Mang, et al., The Sonora Substellar Atmosphere Models. VII. Flame Skimmer: Cloud-free Atmospheric and Evolutionary Models for the Coldest Substellar Objects. arXiv e-prints arXiv:2608.06454 (2026), doi:10.48550/arXiv.2608.06454.arxiv.org/abs/2608.06454
  62. [62]M. D. Perrin, et al., Updated point spread function simulations for JWST with WebbPSF, in Space Telescopes and Instrumentation 2014: Optical, Infrared, and Millimeter Wave, J. Oschmann, Jacobus M., M. M. Clampin, G. G. Fazio, H. A. MacEwen, Eds., vol. 9143 of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series (2014), p. 91433X, doi:10.1117/12.2056689.
  63. [63]J. Anderson, I. R. King, Toward High-Precision Astrometry with WFPC2. I. Deriving an Accurate Point-Spread Function. PASP 112 (776), 1360–1382 (2000), doi:10.1086/316632.DOI
  64. [64]J. Buchner, et al., X-ray spectral modelling of the AGN obscuring region in the CDFS: Bayesian model selection and catalogue. Astron. Astrophys. 564, A125 (2014), doi:10.1051/0004-6361/201322971.DOI
  65. [65]F. Feroz, M. P. Hobson, E. Cameron, A. N. Pettitt, Importance Nested Sampling and the MultiNest Algorithm. The Open Journal of Astrophysics 2 (1), 10 (2019), doi:10.21105/astro.1306.2144.DOI
  66. [66]A. S. Ackerman, M. S. Marley, Precipitating Condensation Clouds in Substellar Atmospheres. Astrophys. J. 556, 872–884 (2001), doi:10.1086/321540.DOI
  67. [67]P. Mollière, et al., Retrieving Scattering Clouds and Disequilibrium Chemistry in the Atmosphere of HR 8799e. Astron. Astrophys. 640, A131 (2020), doi:10.1051/0004-6361/202038325.DOI
  68. [68]B. Lacy, A. Burrows, Self-consistent Models of Y Dwarf Atmospheres with Water Clouds and Disequilibrium Chemistry. Astrophys. J. 950 (1), 8 (2023), doi:10.3847/1538-4357/acc8cb.DOI
  69. [69]H. Kühnle, et al., Water depletion and ¹⁵NH₃ in the atmosphere of the coldest brown dwarf observed with JWST/MIRI. Astron. Astrophys. 695, A224 (2025), doi:10.1051/0004-6361/202452547.DOI
  70. [70]S. de Regt, et al., The ESO SupJup Survey: IX. Isotopic evidence of a recent formation for Luhman 16AB. Astron. Astrophys. 707, A210 (2026), doi:10.1051/0004-6361/202558089.DOI
  71. [71]D. Kitzmann, J. W. Stock, A. B. C. Patzer, FASTCHEM COND: equilibrium chemistry with condensation and rainout for cool planetary and stellar environments. Mon. Not. R. Astron. Soc. 527 (3), 7263–7283 (2024), doi:10.1093/mnras/stad3515.DOI
  72. [72]S. Mukherjee, et al., Probing the Extent of Vertical Mixing in Brown Dwarf Atmospheres with Disequilibrium Chemistry. Astrophys. J. 938 (2), 107 (2022), doi:10.3847/1538-4357/ac8dfb.DOI
  73. [73]N. F. Wogan, et al., The Sonora Substellar Atmosphere Models. V. A Correction to the Disequilibrium Abundance of CO₂ for Sonora Elf Owl. Research Notes of the American Astronomical Society 9 (5), 108 (2025), doi:10.3847/2515-5172/add407.DOI
  74. [74]N. Fray, B. Schmitt, Sublimation of ices of astrophysical interest: A bibliographic review. Planet. Space Sci. 57 (14-15), 2053–2080 (2009), doi:10.1016/j.pss.2009.09.011.DOI
  75. [75]D. B. Hägele, P. Mollière, M. Min, Iine racer: Rapid Calculation of Exoplanetary Radiative Opacities (2026), doi:10.5281/zenodo.20367386.
  76. [76]J.-B. Ruffio, et al., Improving and Assessing Planet Sensitivity of the GPI Exoplanet Survey with a Forward Model Matched Filter. Astrophys. J. 842 (1), 14 (2017), doi: 10.3847/1538-4357/aa72dd.DOI
  77. [77]J. R. Lyons, E. Gharib-Nezhad, T. R. Ayres, A light carbon isotope composition for the Sun. Nature Communications 9, 908 (2018), doi:10.1038/s41467-018-03093-3.DOI
  78. [78]J. R. Kolecki, J. Wang, Measuring Elemental Abundances of JWST Target Stars for Exoplanet Characterization. I. FGK Stars. Astron. J. 164 (3), 87 (2022), doi:10.3847/1538-3881/ac7de3.DOI
  79. [79]Gaia Collaboration, et al., Gaia Data Release 3. Summary of the content and survey properties. Astron. Astrophys. 674, A1 (2023), doi:10.1051/0004-6361/202243940.
  80. [80]C. V. Morley, et al., Water Clouds in Y Dwarfs and Exoplanets. Astrophys. J. 787 (1), 78 (2014), doi:10.1088/0004-637X/787/1/78.DOI
  81. [81]C. Visscher, J. I. Moses, Quenching of Carbon Monoxide and Methane in the Atmospheres of Cool Brown Dwarfs and Hot Jupiters. Astrophys. J. 738 (1), 72 (2011), doi:10.1088/0004-637X/738/1/72.DOI
  82. [82]M. D. Smith, Estimation of a Length Scale to Use with the Quench Level Approximation for Obtaining Chemical Abundances. Icarus 132 (1), 176–184 (1998), doi:10.1006/icar.1997.5886.DOI
  83. [83]B. Fegley, R. G. Prinn, The Predicted Abundances of Deuterium-bearing Gases in the Atmospheres of Jupiter and Saturn. Astrophys. J. 326, 490 (1988), doi:10.1086/166111.DOI
  84. [84]C. Lécluse, F. Robert, Hydrogen isotope exchange reaction rates: Origin of water in the inner solar system. Geochim. Cosmochim. Acta 58 (13), 2927–2939 (1994), doi:10.1016/0016-7037(94)90126-0.DOI
  85. [85]S.-M. Tsai, et al., VULCAN: an Open-Source, Validated Chemical Kinetics Python Code for Exoplanetary Atmospheres. Astrophys. J. Suppl. Ser. 228 (2), 1–26 (2017), doi:10.3847/1538-4365/228/2/20.DOI
  86. [86]S.-M. Tsai, et al., A Comparative Study of Atmospheric Chemistry with VULCAN. The Astrophysical Journal 923 (2), 264 (2021), doi:10.3847/1538-4357/ac29bc.DOI
  87. [87]Y. Chachan, H. A. Knutson, J. Lothringer, G. A. Blake, Breaking Degeneracies in Formation Histories by Measuring Refractory Content in Gas Giants. Astrophys. J. 943 (2), 112 (2023), doi:10.3847/1538-4357/aca614.DOI
  88. [88]S. Ida, T. Guillot, A. Morbidelli, The radial dependence of pebble accretion rates: A source of diversity in planetary systems. I. Analytical formulation. Astron. Astrophys. 591, A72 (2016), doi:10.1051/0004-6361/201628099.DOI
  89. [89]Y. Chachan, E. J. Lee, Small Planets around Cool Dwarfs: Enhanced Formation Efficiency of Super-Earths around M Dwarfs. Astrophys. J. Lett. 952 (1), L20 (2023), doi:10.3847/2041-8213/ace257.DOI
  90. [90]M. Kama, et al., Abundant Refractory Sulfur in Protoplanetary Disks. Astrophys. J. 885 (2), 114 (2019), doi:10.3847/1538-4357/ab45f8.DOI
  91. [91]R. A. Booth, C. J. Clarke, N. Madhusudhan, J. D. Ilee, Chemical enrichment of giant planets and discs due to pebble drift. Mon. Not. R. Astron. Soc. 469 (4), 3994–4011 (2017), doi:10.1093/mnras/stx1103.DOI
  92. [92]A. D. Schneider, B. Bitsch, How drifting and evaporating pebbles shape giant planets. I. Heavy element content and atmospheric C/O. Astron. Astrophys. 654, A71 (2021), doi:10.1051/0004-6361/202039640.DOI
  93. [93]K. Ohno, M. Ikoma, S. Okuzumi, T. Kimura, A dichotomy of the mass–metallicity relation of exoplanetary atmospheres demarcated by their birthplace. Publ. Astron. Soc. Japan 78 (2), 493–523 (2026), doi:10.1093/pasj/psaf157.DOI
  94. [94]K. Ohno, T. Ueda, Jupiter’s “cold” formation in the protosolar disk shadow. An explanation for the planet’s uniformly enriched atmosphere. Astron. Astrophys. 651, L2 (2021), doi: 10.1051/0004-6361/202141169.DOI
  95. [95]Y. Chachan, J. J. Fortney, K. Ohno, D. Thorngren, R. Murray-Clay, Revising the Giant Planet Mass-Metallicity Relation: Deciphering the Formation Sequence of Giant Planets. Astrophys. J. 994 (1), 43 (2025), doi:10.3847/1538-4357/ae0cbf.DOI
  96. [96]M. Fomenkova, S. Chang, L. Mukhin, Carbonaceous Components in the Comet Halley Dust. Geochimica et Cosmochimica Acta 58 (20), 4503–4512 (1994), doi:10.1016/0016-7037(94) 90351-4.pubmed.ncbi.nlm.nih.gov/11539150
  97. [97]H.-P. Gail, M. Trieloff, Spatial Distribution of Carbon Dust in the Early Solar Nebula and the Carbon Content of Planetesimals. Astronomy & Astrophysics 606, A16 (2017), doi: 10.1051/0004-6361/201730480.DOI
  98. [98]E. A. Bergin, G. A. Blake, F. Ciesla, M. M. Hirschmann, J. Li, Tracing the ingredients for a habitable earth from interstellar space through planet formation. Proceedings of the National Academy of Sciences 112 (29), 8965–8970 (2015), doi:10.1073/pnas.1500954112.DOI
  99. [99]D. M. Hunten, The Escape of Light Gases from Planetary Atmospheres. Journal of the Atmospheric Sciences 30 (8), 1481–1494 (1973), doi:10.1175/1520-0469(1973)030⟨1481:TEOLGF⟩2.0.CO;2.doi.org/10.1175/1520-0469(1973)030<1481:teolgf>2.0.co;2
  100. [100]T. Guillot, R. Hueso, The composition of Jupiter: sign of a (relatively) late formation in a chemically evolved protosolar disc. Mon. Not. R. Astron. Soc. 367 (1), L47–L51 (2006), doi:10.1111/j.1745-3933.2006.00137.x.DOI
  101. [101]E. A. Mason, T. R. Marrero, The diffusion of atoms and molecules. Advances in Atomic and Molecular Physics 6, 155–232 (1970), doi:10.1016/S0065-2199(08)60205-5.DOI
  102. [102]H. Genda, M. Ikoma, Origin of the ocean on the Earth: Early evolution of water D/H in a hydrogen-rich atmosphere. Icarus 194 (1), 42–52 (2008), doi:10.1016/j.icarus.2007.09.007.DOI
  103. [103]E. Chiang, A. Youdin, Forming Planetesimals in Solar and Extrasolar Nebulae. Annual Review of Earth and Planetary Sciences 38 (1), 493–522 (2010), doi:10.1146/annurev-earth-040809-152513.DOI
  104. [104]N. Choksi, E. Chiang, J. Fung, Z. Zhu, The maximum accretion rate of a protoplanet: how fast can runaway be? Monthly Notices of the Royal Astronomical Society 525 (2), 2806–2819 (2023), doi:10.1093/mnras/stad2269.DOI
  105. [105]M. R. Line, J. Teske, B. Burningham, J. J. Fortney, M. S. Marley, Uniform Atmospheric Retrieval Analysis of Ultracool Dwarfs. I. Characterizing Benchmarks, GJ 570D and HD 3651B. Astrophys. J. 807 (2), 183 (2015), doi:10.1088/0004-637X/807/2/183.DOI
  106. [106]G. Li, et al., Rovibrational Line Lists for Nine Isotopologues of the CO Molecule in the X ¹Σ⁺ Ground Electronic State. Astrophys. J. Suppl. Ser. 216 (1), 15 (2015), doi:10.1088/0067-0049/216/1/15.
  107. [107]O. L. Polyansky, et al., ExoMol molecular line lists XXX: a complete high-accuracy line list for water. Mon. Not. R. Astron. Soc. 480, 2597–2608 (2018), doi:10.1093/mnras/sty1877.DOI
  108. [108]O. L. Polyansky, et al., ExoMol molecular line lists XIX: high-accuracy computed hot line lists for H218O and H217O. Mon. Not. R. Astron. Soc. 466, 1363–1371 (2017), doi:10.1093/mnras/stw3125.DOI
  109. [109]S. N. Yurchenko, A. Owens, K. Kefala, J. Tennyson, ExoMol line lists - LVII. High accuracy ro-vibrational line list for methane (CH₄). Mon. Not. R. Astron. Soc. 528 (2), 3719–3729 (2024), doi:10.1093/mnras/stae148.DOI
  110. [110]R. J. Hargreaves, et al., An Accurate, Extensive, and Practical Line List of Methane for the HITEMP Database. Astrophys. J. Suppl. Ser. 247 (2), 55 (2020), doi:10.3847/1538-4365/ab7a1a.DOI
  111. [111]P. A. Coles, S. N. Yurchenko, J. Tennyson, ExoMol molecular line lists - XXXV. A rotation-vibration line list for hot ammonia. MNRAS 490, 4481–4488 (2019), doi:10.1093/mnras/stz2778.DOI
  112. [112]S. N. Yurchenko, et al., ExoMol line lists – LXIII. ExoMol line lists for 12 isotopologues of CO₂. Mon. Not. R. Astron. Soc. 545 (3), staf2135 (2026), doi:10.1093/mnras/staf2135.DOI
  113. [113]A. A. Azzam, J. Tennyson, S. N. Yurchenko, O. V. Naumenko, ExoMol molecular line lists - XVI. The rotation-vibration spectrum of hot H2S. Mon. Not. R. Astron. Soc. 460, 4063–4074 (2016), doi:10.1093/mnras/stw1133.DOI
  114. [114]J. V. Michael, J. R. Fisher, Rate Constants for the Reaction D + H₂ → HD + H, over the Temperature Range 655–1979 K, by the Flash Photolysis–Shock Tube Technique. J. Phys. Chem. 94, 3318–3323 (1990), doi:10.1021/j100371a022.
  115. [115]H.-X. Yang, K. S. Shin, W. Gardiner, Rate Coefficients for O + H₂ → OH + H and O + D₂ → OD + D by Kinetic Laser Absorption Spectroscopy in Shock Waves. Chem. Phys. Lett. 207, 69–74 (1993), doi:10.1016/0009-2614(93)85013-E.DOI
  116. [116]T. L. Nguyen, J. F. Stanton, J. R. Barker, Ab Initio Reaction Rate Constants Computed Using Semiclassical Transition-State Theory: HO + H₂ → H₂O + H and Isotopologues. J. Phys. Chem. A 115, 5118–5126 (2011), doi:10.1021/jp2022743.DOI
  117. [117]J. V. Michael, Rate Constants for the Reaction H + D₂ → HD + D, over the Temperature Range 724–2061 K, by the Flash Photolysis–Shock Tube Technique. J. Chem. Phys. 92, 3394–3402 (1990), doi:10.1063/1.457850.DOI
  118. [118]R. K. Talukdar, A. R. Ravishankara, Rate Coefficients for O(¹D) + H₂, D₂, HD Reactions and H Atom Yield in O(¹D) + HD Reaction. Chem. Phys. Lett. 253, 177–183 (1996), doi:10.1016/0009-2614(96)00203-5.
  119. [119]D. H. Zhang, J. Z. H. Zhang, Y. Zhang, D. Wang, Q. Zhang, Quantum Dynamics Study of the Reaction HD + OH → H + DOH, D + HOH. J. Chem. Phys. 102, 7400–7408 (1995), doi:10.1063/1.469052.DOI
  120. [120]M. J. Howard, I. W. M. Smith, Direct Rate Measurements on the Reaction D + OH → OD + H from 300 to 515 K. J. Chem. Soc., Faraday Trans. 2 78, 1403–1412 (1982), doi: 10.1039/F29827801403.DOI
  121. [121]L. Masgrau, À. González-Lafont, J. M. Lluch, The Reactions CHₙD₄−ₙ + OH → P and CH₄ + OD → CH₃ + HOD as a Test of Current Direct Dynamics Multicoefficient Methods to Determine Variational Transition State Rate Constants. II. J. Chem. Phys. 115, 4515–4526 (2001), doi:10.1063/1.1389848.DOI
  122. [122]R. Atkinson, Kinetics of the Gas-Phase Reactions of OH Radicals with Alkanes and Cycloalkanes. Atmos. Chem. Phys. 3, 2233–2307 (2003), doi:10.5194/acp-3-2233-2003.DOI
  123. [123]P. W. Seakins, et al., Temperature and Isotope Dependence of the Reaction of Methyl Radicals with Deuterium Atoms. J. Phys. Chem. A 101, 9974–9987 (1997), doi:10.1021/jp9720348.DOI
  124. [124]W. L. Hase, S. L. Mondro, R. J. Duchovic, D. M. Hirst, Thermal Rate Constant for H + CH₃ → CH₄ Recombination. 3. Comparison of Experiment and Canonical Variational Transition State Theory. J. Am. Chem. Soc. 109, 2916–2922 (1987), doi:10.1021/ja00244a010.
  125. [125]A. Fernández-Ramos, E. Martínez-Núñez, Z. Smedarchina, S. A. Vázquez, Rate constants and isotope effects for the CH₃+ H₂ → CH₄+ H reaction by an approximate semiclassical initial-value representation method. Chemical Physics Letters 341 (3), 351–357 (2001), doi: 10.1016/S0009-2614(01)00489-4.
  126. [126]M. J. Kurylo, G. A. Hollinden, R. B. Timmons, ESR Study of the Kinetic Isotope Effect in the Reaction of H and D Atoms with CH₄. J. Chem. Phys. 52, 1773–1781 (1970), doi: 10.1063/1.1673216.DOI
  127. [127]L. Masgrau, À. González-Lafont, J. M. Lluch, Mechanism of the Gas-Phase HO + H₂O → H₂O + OH Reaction and Several Associated Isotope Exchange Reactions: A Canonical Variational Transition State Theory Plus Multidimensional Tunneling Calculation. J. Phys. Chem. A 103, 1044–1053 (1999), doi:10.1021/jp983749t.DOI
  128. [128]A. E. Rodriguez, P. D. Pacey, Kinetics of the Reaction of Methyl Radicals with Molecular Deuterium at 829 and 930 K. J. Phys. Chem. 90, 6298–6300 (1986), doi:10.1021/j100281a046.DOI
  129. [129]D. W. Trainor, D. O. Ham, F. Kaufman, Gas Phase Recombination of Hydrogen and Deuterium Atoms. J. Chem. Phys. 58, 4599–4609 (1973), doi:10.1063/1.1679024.DOI
  130. [130]S. Takahashi, S. Miyazaki, The Reactions of the Nitrogen Atom with Ethylene and Ethylene-d₄. Bull. Chem. Soc. Jpn. 50, 1627–1628 (1977), doi:10.1246/bcsj.50.1627.DOI
  131. [131]P. H. Stewart, G. P. Smith, D. M. Golden, The pressure and temperature dependence of methane decomposition. Int. J. Chem. Kinet. 21 (10), 923–945 (1989).DOI
  132. [132]C.-Y. Chung, E. P. Chew, B.-M. Cheng, M. Bahou, Y.-P. Lee, Temperature dependence of absorption cross-section of H₂O, HOD, and D₂O in the spectral region 140–193 nm. Nucl. Instrum. Methods Phys. Res. A 467–468, 1572–1576 (2001), doi:10.1016/S0168-9002(01)00762-8.DOI
  133. [133]B.-M. Cheng, et al., Quantitative spectroscopic and theoretical study of the optical absorption spectra of H₂O, HOD, and D₂O in the 125-145 nm region. J. Chem. Phys. 120 (1), 224–229 (2004).pubmed.ncbi.nlm.nih.gov/15267281

Paper details

vlnc:W_h4pfqf6n
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents