HCN and C₂H₂ in the Atmosphere of a T8.5+T9 Brown Dwarf Binary

Brown dwarfs provide a laboratory for studying the physics and chemistry of cold atmospheres. Their atmospheres are similar to those of low-irradiation exoplanets, though with some differences due to their higher mass and surface gravity. Further, their atmospheres are relatively accessible with spectroscopic characterization, without some of the observational complications that arise for exoplanets (which are typically overwhelmed by the light of a much brighter host star). Brown dwarfs and exoplanets are both spectrally classified as late-M, L, T, and Y-type objects (M. C. Cushing et al. 2006, 2011). As they cool, more molecules are able to form in their atmospheres, leading to increasingly complex atmospheric chemistry and corresponding emission spectra, though at the very lowest temperatures chemistry is kinetically inhibited.

T-type objects have strong water and ammonia absorption features in the mid-IR (G. Suárez & S. Metchev 2022). In cool T dwarfs with high surface gravity, carbon is predicted to be primarily in the form of CH₄, with the CO atmospheric content correlating with the vertical mixing strength (e.g., K. J. Zahnle & M. S. Marley 2014). Mid- to late-T type brown dwarfs should have relatively clear atmospheres, since they are cold enough that silicate clouds in the atmosphere have sunk deep into the atmosphere, but warm enough that water- and ammonia-ice clouds are not expected.

K. J. Zahnle & M. S. Marley (2014) predicted that HCN might be present for high-gravity brown dwarfs with very strong vertical mixing (parameterized through an eddy diffusion coefficient (K_zz, with very strong vertical mixing corresponding to K_zz ⪆ 10¹⁰ cm² s⁻¹). HCN is only produced through chemical reactions at higher temperatures and pressures, but can become "quenched" (i.e., the mixing timescale becomes shorter than the chemical reaction timescale, leading to a vertically constant abundance; R. G. Prinn & S. S. Barshay 1977) and reach observable abundances higher in the atmosphere. However, S. Mukherjee et al. (2024) demonstrated that vertical mixing is relatively weak in field brown dwarfs. They studied the spectra of a sample of nine early- to late-T dwarfs observed with Spitzer and AKARI, and derived typical vertical mixing values in the atmosphere of K_zz ∼ 10¹–10⁴ cm² s⁻¹. A more recent study of a Y0 dwarf with JWST NIRSpec/PRISM and MIRI/LRS found a preference for models with ${\rm{log}}({K}_{{zz}}[{{\rm{cm}}}^{2}\,{{\rm{s}}}^{-1}])\geqslant 4$ (S. A. Beiler et al. 2023), but that work highlights that the strong vertical mixing is required to match the deep CO and NH₃ features. This strong mixing is inconsistent with the PH₃ nondetection in that object, possibly indicating the need for a more complex vertical mixing prescription (though see S. A. Beiler et al. 2024 for additional discussion of the PH₃ nondetection).

C₂H₂ was not predicted in the atmospheres of nonirradiated brown dwarfs, but both HCN and C₂H₂ should be formed by photodissociation in irradiated objects (e.g., A. G. Sharp et al. 2004; J. I. Moses et al. 2011), and in these cases they are a tracer of a carbon-rich atmosphere (O. Venot et al. 2015). In the case that C₂H₂ is produced by photodissociation, it should appear primarily at very high altitudes/low pressures (e.g., O. Venot et al. 2015 predicted C₂H₂ abundance to exceed a molar fraction of 10⁻¹² at above ∼10⁻³ bar, and peak at a molar fraction of ∼10⁻⁵–10⁻⁴ at ∼10⁻⁶ bar in a 500 K atmosphere with external radiation).

JWST provides the opportunity to characterize these T dwarfs to a level of detail not previously possible. The medium-resolution spectrograph (MRS; M. Wells et al. 2015; I. Argyriou et al. 2023) of the JWST/MIRI instrument (G. H. Rieke et al. 2015; G. S. Wright et al. 2015) provides high-sensitivity spectra at R ∼ 1500–3500 from 4.9 to 27.9 μm. For cold atmospheres, this unveils a rich forest of molecular absorption features. Throughout this work, we will refer to the 1–5 μm wavelength range as the near-IR and to λ > 5 μm as the mid-IR.

No MIRI/MRS spectra of T-type brown dwarfs have to date been published, though a few publications have studied L-type (B. E. Miles et al. 2023) and Y-type (D. Barrado et al. 2023; H. Kühnle et al. 2024) objects. In this work we present the first MIRI/MRS spectrum of a T dwarf: observations of the brown dwarf binary WISE J045853.90+643451.9 (hereafter WISE-0458), providing an opportunity to search for new chemistry and to constrain the atmospheric structure of a cold object. WISE-0458 was the first ultracool brown dwarf identified by the Wide-field Infrared Survey Explorer (WISE) mission (J. D. Kirkpatrick et al. 2011; A. Mainzer et al. 2011). It was originally classified as T9 or T8.5 by A. Mainzer et al. (2011) and M. C. Cushing et al. (2011) respectively, and has strong near-IR H₂O and CH₄ absorption bands, as well as tentative evidence for NH₃ absorption in the near-IR (A. J. Burgasser et al. 2012). The system is at a distance of $9.2{4}_{-0.15}^{+0.14}$ pc (S. K. Leggett et al. 2019), and has no identified markers of relative youth, suggesting it is a field-age object with high surface gravity.

WISE-0458 was soon resolved as a binary brown dwarf in Keck high-resolution images by C. R. Gelino et al. (2011), who found spectral types T8.5 and T9 and temperatures ∼600 K and ∼500 K respectively for the two components. The pair has a semimajor axis of $5.{0}_{-0.6}^{+0.3}$ au ($54{0}_{-70}^{+40}$ mas), and at the epoch of JWST observation were at a predicted mutual separation of 330 ± 40 mas, based on the orbit fit of S. K. Leggett et al. (2019). This corresponds to a separation of 1.3× the diffraction limit at 5 μm, and 0.45× the diffraction limit at 18 μm (D. R. Law et al. 2023), i.e., the pair is barely resolved at the shortest MRS wavelengths and unresolved at the longest. We therefore treat the pair as an unresolved binary throughout this work, and assume identical atmospheres for both objects; the similarity in spectral type and temperature between the two objects justifies this approach.

For any binary, in the extreme cases either (1) the flux of one object dominates, and models would constrain only the physical parameters of that object, or (2) the objects are identical, in which case models can also constrain all the physical parameters of both objects, but will return a radius inflated by $\sqrt{2}$ (corresponding to twice the emitting area of one object). In the intermediate case (both objects contribute to the flux, but have different atmospheres) retrievals may be more degenerate and measured parameters can be biased. For WISE-0458, the pair of brown dwarfs have similar fluxes and temperatures: S. K. Leggett et al. (2019) find the pair to be ∼1 mag different in brightness in the near-IR and derive temperatures of ∼600 K and ∼500 K respectively for the two atmospheres. If we infer that both formed via a star formation pathway and at the same time and location, they should also have similar atmospheric compositions. For this work, we therefore approximate the spectrum as a single atmosphere and assume a common effective temperature and composition, though we intend to revisit this assumption in a future work. As the library of late-T and Y dwarfs with JWST spectra grows, it will be important to understand the biases that are introduced when modeling the atmospheres of a binary brown dwarf in this way—this is likely particularly important in temperature ranges where the atmosphere is changing rapidly (e.g., the L–T transition and the T–Y transition).

The paper is structured as follows: we first present the observations and data reduction in Sections 2 and 3 respectively. In Section 4 we qualitatively describe the mid-IR spectrum of WISE-0458, and in Section 5 describe our retrieval analysis to constrain its atmospheric structure and composition. In Section 6 we discuss the molecular content of the atmosphere, and in particular the detections of HCN and surprisingly of C₂H₂, and the implications of these detections. Finally, we present some more general discussion and concluding remarks in Sections 7 and 8 respectively.

We observed WISE-0458 with the JWST/MIRI instrument on 2022 November 21, using the MRS. Observations were collected as part of the European MIRI GTO consortium. NIRSpec observations of WISE-0458 were also collected by another GTO team and will be presented in a future paper (B. W. P. Lew et al. 2025, in preparation). To reduce overheads, both the MRS and NIRSpec observations were collected through the same program (PID 1189).

The MRS has four channels (1 through 4) that cover consecutive wavelength regions and are observed simultaneously; together these four channels span 4.9–27.9 μm. These channels can be observed with any one of three grating settings (referred to as short, medium, and long or as A, B, and C), each covering one-third of each channel, for a total of 12 subchannels with unique grating/channel combinations. For WISE-0458, we integrated for 2997 s with each grating setting. At the longest wavelengths the brown dwarf signal becomes overwhelmed by thermal and background noise, and in this work we only consider channels 1–3 (4.9–18.0 μm). Observations were performed in the FASTR1 readout mode and in a two-point point-source optimized dither pattern to provide better sampling of the point-spread function; this dither also has the advantage of providing a contemporaneous observation that can be used for background subtraction. For each of the three grating settings, we collected one integration per dither position with 180 groups per integration.

Data reduction was performed using the public jwst pipeline²¹ (version 1.12.5; H. Bushouse et al. 2023) with the corresponding CRDS file version 11.17.10 and context files jwst_1183.pmap. The reduction consists of three distinct stages. The first stage of the pipeline calculates the rate of photoelectric charge accumulation from the ramp files. The second stage, after assigning a world coordinate system to the data, includes applying several calibrations such as correcting for the flat field, the characterized scattered light internal to the MIRI detectors, the fringes in the spectral direction, and the photometric flux calibration (for additional discussion of the flux calibration of MIRI/MRS, see D. R. Law et al. 2025). After this step, the background is removed by subtracting one dither from the other (i.e., a nod subtraction). The final stage uses the generated and corrected detector files to produce image cubes. The algorithm to project from the 2D to 3D cube in this case is the "drizzle" weighting algorithm (D. R. Law et al. 2023). Before generating the cubes, an outlier detection is performed to mask out cosmic rays and similar residual outliers.

From the cube, the one-dimensional spectrum is extracted by placing an aperture around the source. For WISE-0458, we center the aperture around the brightest point in the collapsed data cube for each channel, using the built-in ifu_autocen() function. We use a radius of twice the analytical FWHM around the source, which varies between ∼0$\mathop{.}\limits^{\unicode{x02033}}$3 and ∼1" over the MIRI wavelength range; this ensures flux from both brown dwarfs (separation ∼330 mas) is captured within the aperture. The spectrum is extracted with the extract_1d() function built into the jwst pipeline.

3.1. Residual Background Effects

The full MIRI/MRS spectrum of WISE-0458 is shown in Figure 1, with the key molecular absorptions highlighted. H₂O, CH₄, and NH₃ all have a forest of molecular lines in the mid-IR. In particular, H₂O dominates the shape of the spectrum between 5 and 7 μm and a very clear NH₃ feature is visible between 10 and 11 μm. CH₄ and NH₃ also contribute to the forest of molecular lines beyond ∼7 μm.

Zoom In Zoom Out Reset image size

Absorption features from HCN and C₂H₂ are visible at 13.75 μm and 14 μm respectively (Figure 2; these detections are both confirmed with a retrieval analysis (Sections 6.1, 6.2). Neither species has previously been observed in a brown dwarf atmosphere. While there are several mid-IR brown dwarf spectra measured by Spitzer (G. Suárez & S. Metchev 2022), we here present the first T dwarf spectrum with JWST/MIRI/MRS, and therefore the first T dwarf spectrum with this resolution and signal-to-noise ratio in the mid-IR, meaning that we are more sensitive than any previous observations to HCN and C₂H₂. Among the heavily irradiated exoplanets, detections of both HCN and C₂H₂ have been claimed in the atmosphere of HD 209458b (G. A. Hawker et al. 2018; P. Giacobbe et al. 2021), though the detections of both species are disputed based on JWST data (Q. Xue et al. 2024) and on high-resolution data at ∼3 μm (D. Blain et al. 2024). Finally, we note that both species are commonly seen as emission features in disks of young stars and brown dwarfs with Spitzer (I. Pascucci et al. 2013) and more recently JWST (B. Tabone et al. 2023; A. M. Arabhavi et al. 2024; T. Henning et al. 2024).

Zoom In Zoom Out Reset image size

We modeled the spectrum of WISE-0458 using a retrieval analysis. For simplicity, we assumed both brown dwarfs to have the same P–T structure and atmospheric composition; this is a valid approximation since the spectral types of the two components are very similar (T8.5+T9, S. K. Leggett et al. 2019). We used the petitRADTRANS²² retrieval package (hereafter pRT; P. Mollière et al. 2019; E. Nasedkin et al. 2024), which allows for emission spectra to be generated, and used a nested sampling technique (J. Skilling 2004) as implemented in pyMultiNest (F. Feroz & M. P. Hobson 2008; J. Buchner et al. 2014) to probe the parameter space and derive posterior distributions for atmospheric properties.

We binned the MRS data from a native resolution of between 1500 and 3500 to λ/Δλ = 1000 when running retrievals. This allows for an efficient opacity treatment with correlated-k opacities (rather than line-by-line opacities), where 1000 is pRT's finest wavelength spacing in correlated-k, and thereby minimizes computation time. We modeled the atmosphere with a free pressure–temperature (P–T) profile with 10 temperature nodes, spaced equidistantly between 10⁻⁶ and 10³ bar in log-space. The temperature at each "node" is a free parameter, and the full P–T profile is calculated by quadratic interpolation between these nodes. This led to a "wiggly" P–T profile, with several kinks in the posterior distribution, so we therefore also regularized the profile following M. R. Line et al. (2015; see also D. Barrado et al. 2023). Briefly, this approach penalizes rough profiles (defined based on their ${{d}}^{2}{\rm{log}}\,T/{d}{\rm{log}}\,{P}^{2}$) and includes a free parameter γ that dictates the degree of smoothing. We also inflated the pipeline-derived observational uncertainties, and retrieved an inflation term following M. R. Line et al. (2015), namely, we calculate inflated uncertainties for each data point following the formula $\sigma =\sqrt{{\sigma }_{{\rm{red}}}^{2}+1{0}^{b}}$, where b is a free parameter in the retrieval and σ_red is the observational uncertainty on each data point, derived from the jwst pipeline and binned to R = 1000 assuming uncorrelated errors. This allows a converged fit even in the presence of underestimated uncertainties in the data and/or missing physics in the model. A more physically accurate model is then expected to have a lower b parameter and smaller inflated uncertainties; this approach also accounts for the underestimated uncertainties in the jwst pipeline (see below).

Our basic model for the retrieval includes CH₄, CO₂, CO, H₂O, and NH₃, the main species that are expected in cold atmospheres. We assumed all molecules to have vertically constant abundances that were retrieved freely, which is a common assumption for brown dwarf modeling and is likely accurate across the pressures we probe with MRS spectra. We did not include clouds in our model. Line-lists for these species are from HITEMP for CO, CO₂ (L. S. Rothman et al. 2010), and CH₄ (R. J. Hargreaves et al. 2020), and from ExoMOL for H₂O (O. L. Polyansky et al. 2018) and NH₃ (P. A. Coles et al. 2019). Alongside these five abundance parameters, our model has 11 temperature/pressure free parameters (10 nodes and the γ parameter used in the P–T profile regularization), two additional physical parameters ($\mathrm{log}\,g$ and radius of the brown dwarf), and the b parameter to inflate uncertainties, for a total of 19 free parameters in the basic model. We also tested for the presence of other molecular species, in each case adding the molecule individually to the retrieval and repeating the analysis. Line-lists for additional species are from HITRAN2012 for H₂S and C₂H₂ (L. S. Rothman et al. 2013), HITRAN2020 for ¹⁵NH₃ and C₂H₆ (I. E. Gordon et al. 2022), ExoMol for C₂H₄ (B. P. Mant et al. 2018) and PH₃ (C. Sousa-Silva et al. 2015), and G. J. Harris et al. (2006) for HCN.

We found a well-fitting model that includes H₂O, CH₄, NH₃, HCN, and C₂H₂ with our retrieval approach. Neither CO nor CO₂ is detected; the nondetection of CO₂ is consistent with models while the nondetection of CO is expected only for relatively weak vertical mixing. This model is shown in Figure 1; no strong features remain unexplained with this model and there are no obvious wavelength-dependent systematics. There is some hint of CO₂ in the models without HCN, but this disappears when HCN is included and we infer it is a false-positive detection. Aside from the sharp 14 μm detection feature, HCN has some nonnegligible opacity between ∼12 and 16 μm (Figure 3), and correspondingly slightly changes the broadband spectral shape throughout this region; in a model without HCN, the retrieved CO₂ abundance is elevated to explain this broadband spectral shape. We also see some correlation between the retrieved HCN and C₂H₂ for the same reason: both have a similar impact on the broadband spectral shape between ∼12 and 16 μm, and in a model with only one of these species included, the abundance of that molecule is overestimated to better fit the broadband shape of the other molecule. A full table of posteriors and a corner plot of molecular abundances are included in Appendix A.

Zoom In Zoom Out Reset image size

The derived atmospheric C/O ratio for this retrieval is 0.35 ± 0.03 (driven mainly by the H₂O and CH₄ abundances), and the atmospheric metallicity is $1.3{5}_{-0.15}^{+0.19}\times$ solar, i.e., a slightly supersolar metallicity and slightly subsolar C/O ratio. The derived effective temperature is $56{6}_{-6}^{+7}$ K. For the retrieval without HCN and C₂H₂, we find consistent C/O values (0.33 ± 0.03) but a slightly higher metallicity and effective temperature ($1.7{1}_{-0.25}^{+0.31}\times$ solar; 580 ± 5 K). While these calculations formally include the CO and CO₂ abundances returned by the retrieval, the abundance upper limits on both species are sufficiently constraining that they do not impact the derived C/O and metallicity.

In the following text we discuss in detail the newly detected molecules HCN and C₂H₂, and then briefly also mention several molecules that are not detected in the current spectrum.

6.1. Hydrogen Cyanide Detection

6.2. Acetylene Detection

Aectylene (C₂H₂) is readily visible in the WISE-0458 spectrum, most notably with a clear absorption feature at 13.7 μm (Figure 2). In our retrieval analysis for WISE-0458, the best-fit model with C₂H₂ is favored with a Bayes factor of 18.4 over the model without (including HCN in both cases), corresponding to a C₂H₂ detection significance of 9.5σ. As with HCN, the only clear absorption features appear around 13–14 μm, but there is also some impact on the continuum flux level at other wavelengths. C₂H₂ also has a prominent 7.7 μm feature, as seen in some protoplanetary disk spectra (e.g B. Tabone et al. 2023), but in the WISE-0458 spectrum this feature is overwhelmed by the strong H₂O, CH₄, and NH₃ absorption.

C₂H₂ has not previously been seen in a brown dwarf atmosphere and is an unexpected discovery. At first glance, the presence of C₂H₂ is consistent with the detection of HCN: both species are indicators of disequilibrium chemistry and likely form together deeper in the brown dwarf atmosphere. However, several brown dwarf model atmospheres (M. S. Marley et al. 2021; S. Mukherjee et al. 2024) include C₂H₂ and suggest that the atmospheric abundance of C₂H₂ should be extremely low. Our observed abundance ($2.{7}_{-1.1}^{+1.4}\times 1{0}^{-7}$) of C₂H₂ in the WISE-0458 atmosphere cannot be explained within a "normal" disequilibrium chemistry model, which assumes that the species is in chemical equilibrium deep in the atmosphere and is quenched at higher altitudes due to vigorous vertical mixing.

C₂H₂ is observed in Jupiter (G. R. Gladstone 1999; T. Fouchet et al. 2000), where it is formed via photochemistry from CH₄ (S.-M. Tsai et al. 2021), but there is no star to drive photochemistry in WISE-0458, and the brown dwarfs (with mutual separation $5.{0}_{-0.6}^{+0.3}$ au, S. K. Leggett et al. 2019) are not emitting UV flux sufficient to drive this photochemistry. Note also that photodissociation driven by an external source should lead to C₂H₂ high in the upper atmosphere (above 10⁻³ bar), while our MIRI observations are primarily sensitive between 0.1 and 10 bar.

We modeled the WISE-0458 atmospheric chemistry with VULCAN to further explore the conditions under which C₂H₂ should be expected in a nonirradiated atmosphere with no thermal inversion. VULCAN is a chemical kinetics code that computes atmospheric composition considering mixing thermochemistry and optionally photochemistry, which has been validated against measurements of Jupiter as well as several exoplanets (S.-M. Tsai et al. 2017, 2021). The code assumes a one-dimensional model atmosphere, >300 chemical reactions (here we use the N–C–H–O chemical network), and uses eddy diffusion to mimic atmospheric dynamics. For this experiment, we used a P–T profile from E. F. Linder et al. (2019) with T_eff = 500 K and $\mathrm{log}\,g=5.0$. For our simple toy model we first assumed solar abundances, a vertically constant mixing profile with K_zz = 10¹¹ cm² s⁻¹, and no photochemistry. In this model, the expected C₂H₂ volume mixing ratio is ∼10⁻¹⁴—orders of magnitude below our measured value (Figure 4, left). We then tested how varying the vertical mixing (K_zz), the metallicity, or the C/O (via adding carbon atoms, that is, not conserving solar metallicity) impacts predicted C₂H₂ values. While increasing any of these parameters does promote C₂H₂ production, unphysically high values are required to reach close to the observed value of C₂H₂ (Figure 4, right).

Zoom In Zoom Out Reset image size

Perhaps, then, there is another unidentified process in the atmosphere. One possibility is that gravity waves are dissipating significant heat in the upper atmosphere, as modeled by C. Watkins & J. Y. K. Cho (2010). Alternatively, perhaps one or both brown dwarfs are magnetically active, driving aurorae and in turn generating UV photons and/or heating the upper atmosphere and driving a temperature inversion. For the brown dwarf WISE J193518.59-154620.3 (F. Marocco et al. 2019), an auroral process was inferred based on the observation of methane emission features as observed in JWST/NIRSpec G395H spectra (J. K. Faherty et al. 2024). However, NIRSpec data for WISE-0458 do not show the 3.325 μm emission feature that was used by J. K. Faherty et al. (2024) to identify aurorae. Galactic cosmic rays have also been theorized to be able to generate C₂H₂ in the upper atmosphere (P. B. Rimmer et al. 2014), though this would likely only produce smaller quantities of C₂H₂ than observed and at higher altitudes than our contribution function indicates. Another possibility is a more exotic scenario, such as a tidally heated exoplanet or exomoon (similar to Io in orbit around Jupiter) providing additional heat to the upper atmosphere. While occurrence rates of small bodies orbiting brown dwarfs are not known, in this case we would expect significant stochasticity in the occurrence and abundance of C₂H₂, based on the presence and exact properties of orbiting small bodies in different systems. Finally, substantial lightning discharges might ionize species in the atmosphere and drive the observed chemistry—a process that has previously been theorized to generate HCN in brown dwarf atmospheres (C. Helling & P. B. Rimmer 2019).

Even additional energy sources would only be expected to produce C₂H₂ in a carbon-rich (C/O > 1) environment, and not the carbon-poor atmosphere observed here. The detection of C₂H₂ remains a challenge for atmospheric models, suggesting a shortcoming in our understanding of vertical transport and disequilibrium processes, in our understanding of the energy sources of this atmosphere, or in our understanding of C₂H₂ chemistry (e.g., V. Saheb et al. 2018).

6.3. Other Hydrocarbons

6.4. Nondetections of ¹⁵NH₃, PH₃, and H₂S

The mid-IR spectrum of WISE-0458 is well fitted with a clear atmosphere model, and the retrieved temperature–pressure profile is broadly consistent with predictions from self-consistent forward models (see Figure 5), indicating that our P–T parameterization within the pRT setup is sufficient to describe this atmosphere. Our model shows evidence for five molecules: H₂O, CH₄, NH₃, HCN, and C₂H₂.

Zoom In Zoom Out Reset image size

We measure a radius of $0.80{7}_{-0.007}^{+0.008}{R}_{{\rm{Jup}}}$ for each brown dwarf (assuming the two binary components have equal radii and based on the effective radius of $1.14{1}_{-0.010}^{+0.011}{R}_{{\rm{Jup}}}$ determined for the pair in our retrievals), while the ATMO evolutionary models predict a radius of 0.84R_Jup for a single field-age 40M_Jup brown dwarf and 0.93R_Jup for a single field-age 20M_Jup brown dwarf. The measured uncertainty on the effective radius is very small (0.01R_Jup), while the difference in effective radius between our two models (Table 1) is 0.05R_Jup: this highlights the model dependence of the radius, which is also degenerate with the measured temperature and molecular abundances. Further, the distance is fixed to 9.24 pc in our calculations, and the distance uncertainty (±0.15 pc) has not been incorporated into the radius uncertainty. Taking this into account, the retrieved radius is broadly consistent with predictions of evolutionary models.

Table 1. Best-fit Parameters for our Retrievals

Parameter	Prior	Basic Model	Model with HCN, C2H2
Physical parameters
log(g[cm s−2])	Uniform [2.5, 6.0]	$5.5{6}_{-0.13}^{+0.13}$	$5.4{0}_{-0.12}^{+0.11}$
Measured radius [RJup]	Uniform [0.46, 2.74]	$1.09{4}_{-0.010}^{+0.010}$	$1.14{1}_{-0.010}^{+0.011}$
Pressure–temperature profile
Base temperature (Node0) [K]	Uniform [100, 9000]	$133{4}_{-41}^{+43}$	$133{5}_{-42}^{+45}$
Temp. ratio Node1/Node0	Uniform [0.2, 1.0]	$0.7{9}_{-0.01}^{+0.01}$	$0.7{9}_{-0.01}^{+0.01}$
Temp. ratio Node2/Node1	Uniform [0.2, 1.0]	$0.7{4}_{-0.01}^{+0.01}$	$0.7{3}_{-0.01}^{+0.01}$
Temp. ratio Node3/Node2	Uniform [0.2, 1.0]	$0.6{6}_{-0.01}^{+0.01}$	$0.6{6}_{-0.01}^{+0.01}$
Temp. ratio Node4/Node3	Uniform [0.2, 1.0]	$0.6{5}_{-0.01}^{+0.01}$	$0.6{6}_{-0.01}^{+0.02}$
Temp. ratio Node5/Node4	Uniform [0.2, 1.0]	$0.6{3}_{-0.03}^{+0.03}$	$0.6{6}_{-0.03}^{+0.04}$
Temp. ratio Node6/Node5	Uniform [0.2, 1.0]	$0.5{9}_{-0.06}^{+0.06}$	$0.6{3}_{-0.06}^{+0.06}$
Temp. ratio Node7/Node6	Uniform [0.2, 1.0]	$0.5{7}_{-0.09}^{+0.08}$	$0.6{0}_{-0.08}^{+0.08}$
Temp. ratio Node8/Node7	Uniform [0.2, 1.0]	$0.5{5}_{-0.09}^{+0.10}$	$0.5{7}_{-0.09}^{+0.10}$
Temp. ratio Node9/Node8	Uniform [0.2, 1.0]	$0.5{0}_{-0.17}^{+0.24}$	$0.4{7}_{-0.16}^{+0.23}$
γ	Gaussian, σ = 1, μ = 1	$1{5}_{-3}^{+4}$	$1{7}_{-3}^{+5}$
Molecular abundances [volume mixing ratio]
H2O	Uniform in log(mass fraction) between [−10, 0]	$2.{5}_{-0.4}^{+0.4}\times 1{0}^{-3}$	$1.{9}_{-0.2}^{+0.3}\times 1{0}^{-3}$
CH4	Uniform in log(mass fraction) between [−10, 0]	$8.{3}_{-1.4}^{+1.8}\times 1{0}^{-4}$	$6.{8}_{-1.0}^{+1.2}\times 1{0}^{-4}$
NH3	Uniform in log(mass fraction) between [−10, 0]	$3.{8}_{-0.6}^{+0.8}\times 1{0}^{-5}$	$3.{0}_{-0.4}^{+0.5}\times 1{0}^{-5}$
HCN	Uniform in log(mass fraction) between [−10, 0]	⋯	$1.{2}_{-0.3}^{+0.4}\times 1{0}^{-6}$
C2H2	Uniform in log(mass fraction) between [−10, 0]	⋯	$2.{9}_{-0.8}^{+1.0}\times 1{0}^{-7}$
Abundance upper limits [volume mixing ratio]
CO	Uniform in log(mass fraction) between [−10,0]	<4.1 × 10−6	<4.2 × 10−6
CO2	Uniform in log(mass fraction) between [−10,0]	<5.8 × 10−7	<2.0 × 10−7
Other parameters
b	Uniform [−14.259, −7.150]	$-8.1{5}_{-0.02}^{+0.02}$	$-8.2{1}_{-0.02}^{+0.02}$
Derived parameters
Teff [K]	⋯	$580.{3}_{-5.1}^{+5.0}$	$565.{7}_{-5.5}^{+6.5}$
C/O [atomic ratio]	⋯	$0.3{3}_{-0.03}^{+0.03}$	$0.3{5}_{-0.03}^{+0.03}$
Fe/H [× solar]	⋯	$1.7{1}_{-0.25}^{+0.31}$	$1.3{5}_{-0.15}^{+0.19}$
Radius/√2 [RJup]	⋯	$0.77{4}_{-0.007}^{+0.007}$	$0.80{7}_{-0.007}^{+0.008}$
Mass [MJup]	⋯	$8{8}_{-23}^{+31}$	$6{6}_{-15}^{+19}$
Fixed parameters
Distance [pc]	⋯	9.24	9.24

Note. We show the values for both our "basic model" (with five molecules: H₂O, CH₄, NH₃, CO, and CO₂) as well as our preferred model, which is the same as the basic model except for the addition of HCN and C₂H₂. Listed uncertainties are 1σ.

Download table as:  ASCIITypeset image

We derive a mass of $6{6}_{-15}^{+19}{M}_{{\rm{Jup}}}$ for each brown dwarf based on the retrieved values of $\mathrm{log}\,g$ and radius values, corresponding to a total mass of $13{2}_{-28}^{+38}{M}_{{\rm{Jup}}}$ for the pair. This is higher than expected from evolutionary models, based on the low temperature of the pair. The retrieved mass is also somewhat higher than the dynamical mass estimated for the pair of brown dwarfs ($7{0}_{-24}^{+15}{M}_{{\rm{Jup}}}$, S. K. Leggett et al. 2019), though we note that using a single $\mathrm{log}g$ value in our retrievals for the binary is an oversimplification. Future works should use the resolved brown dwarf astrometry from MIRI and/or NIRSpec to improve this orbital fit, and make a more robust comparison between the spectroscopically derived and dynamically derived masses.

Analysis of the NIRSpec observations of this target (B. W. P. Lew et al. 2025, in preparation) will likely improve our CO and CO₂ constraints and help determine a more precise C/O ratio for the atmosphere. This may shed further light on the carbon chemical network of WISE-0458, which will be relevant for explaining the presence of C₂H₂. We do not expect this to significantly alter the constraints, since H₂O and CH₄ are the main carriers of oxygen and carbon respectively in a cool, high-$\mathrm{log}g$ atmosphere (e.g., K. J. Zahnle & M. S. Marley 2014), and our CO and CO₂ upper limits are much lower than the measured CH₄ abundance, which suggests that CH₄ is a sufficient proxy for atmospheric carbon content. Note also that sufficiently high CO₂ abundance would lead to an absorption feature at 15 μm, which would be readily detectable with MIRI/MRS data but is not seen in the WISE-0458 spectrum. If there is any sequestration of oxygen in cloud species, then this would lead to an even lower C/O ratio than our measured value (and not the strongly supersolar C/O regime that might be needed to promote C₂H₂ formation in current models).

7.1. Uncertainty Inflation

7.2. Optimal Wavelengths for HCN and C₂H₂ Detection

7.3. Future Prospects for HCN and C₂H₂ Detection

In this work, we have presented a mid-IR spectrum of WISE-0458 at resolution >1000 between 4.9 and 18.0 μm and treated the system as an unresolved binary. This is the first JWST/MIRI/MRS spectrum of any T-type object and is well modeled by a clear, molecule-rich atmosphere in our retrieval analysis. Our key findings are the following.

• 1.  
  HCN is detected at 13.4σ in the WISE-0458 atmosphere, suggesting very strong vertical mixing (K_zz ∼ 10¹¹ cm² s⁻¹), in tension with previous studies of T dwarf vertical mixing.
• 2.  
  C₂H₂ is detected at 9.5σ in the WISE-0458 atmosphere; this discovery is surprising and cannot be explained with a standard prescription of quenching, i.e., a deep-atmosphere equilibrium species brought to higher altitudes (lower pressure levels) with vertical mixing using a single K_zz value.
• 3.  
  PH₃ and ¹⁵NH₃ are not detected. The PH₃ nondetection is consistent with other JWST spectra of late-type brown dwarfs. The ¹⁵NH₃ nondetection is consistent with the observed ¹⁵N enrichment of other cold brown dwarfs, and we highlight that ¹⁵NH₃ will be most readily detectable in cold brown dwarfs (T_eff ≲ 500 K).
• 4.  
  The atmosphere of WISE-0458 is well fitted by our retrievals with a cloud-free model. The required uncertainty inflation is consistent with the known underestimation of MRS uncertainties in the jwst pipeline.

This spectrum demonstrates the power of MIRI MRS to characterize cold brown dwarfs. Future works should study the HCN and C₂H₂ in more detail and determine whether these species are present in other cold brown dwarfs of similar temperature to WISE-0458. Our data challenge existing prescriptions for atmospheric mixing and/or C₂H₂ chemistry and hint toward a new era in brown dwarf chemistry with an increasing number of carbon species detectable in these cold atmospheres.

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 #1189. All the JWST data used in this paper can be found in MAST under the DOI 10.17909/0tbz-1635.

We used pyMultiNest (J. Buchner et al. 2014), which in turn builds on MultiNest (F. Feroz & M. P. Hobson 2008; F. Feroz et al. 2009, 2019) for the nested sampling analysis. We used various functions from astropy (Astropy Collaboration et al. 2013, 2018, The Astropy Collaboration et al. 2022), pandas (W. McKinney 2010; pandas development team 2021), scipy (P. Virtanen et al. 2020), matplotlib (J. D. Hunter 2007), and seaborn (M. L. Waskom 2021) to carry out analysis and create figures.

P.P. thanks the Swiss National Science Foundation (SNSF) for financial support under grant No. 200020_200399. N.W. acknowledges support from NSF awards #2238468 and #1909776, and NASA Award #80NSSC22K0142 P.-O.L acknowledges funding support by CNES. S.-M.T. acknowledge support from NASA through Exobiology grant No. 80NSSC20K1437. B.V., O.A, I.A, and P.R. thank the European Space Agency (ESA) and the Belgian Federal Science Policy Office (BELSPO) for their support in the framework of the PRODEX Programme. O.A. is a Senior Research Associate of the Fonds de la Recherche Scientifique – FNRS. D.B. is supported by Spanish MCIN/AEI/10.13039/501100011033 grants PID2019-107061GB-C61 and PID2023-150468NB-I00. G.O. acknowledge support from the Swedish National Space Board and the Knut and Alice Wallenberg Foundation. J.P.P. acknowledges financial support from the UK Science and Technology Facilities Council and the UK Space Agency. For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) licence to the Author Accepted Manuscript version arising from this submission. E.v.D. acknowledges support from A-ERC grant 101019751 MOLDISK. T.P.R. acknowledges support from the ERC grant 743029 EASY. G.Ö. acknowledges support from SNSA.

Here we present the full posteriors of the retrieval analysis, as described in Section 5. In Table 1 we provide posterior values (median and 1σ confidence intervals) for all parameters, for our "basic" retrieval (molecules CH₄, CO₂, CO, H₂O, and NH₃) and our retrieval with HCN and C₂H₂ (and otherwise identical to the "basic" retrieval).

In Figure 6 we provide a corner plot for the retrieved volume mixing ratios for all molecules in the atmospheres, for our "basic" model and for models with HCN added, and with C₂H₂ added, and with both HCN and C₂H₂ added. The abundances of HCN and C₂H₂ are somewhat correlated, since both have a similar effect on the continuum between 12 and 16 μm. The basic model and the model with C₂H₂ added show a hint of a CO₂ detection, in the form of a peak in the posterior at a volume mixing ratio of ∼10⁻⁷ but a long tail toward lower abundances. However, this posterior peak disappears when including HCN; we infer that the CO₂ detection was a false positive to explain the effect of HCN on the broadband spectral shape.

Zoom In Zoom Out Reset image size

¹⁵NH₃ is not observed in the WISE-0458 atmosphere, even though it has been seen in other MRS observations of cold brown dwarfs. For the cold brown dwarf WISE J1828 (380 K) ¹⁵NH₃ was detected (D. Barrado et al. 2023), and for the coldest known brown dwarf WISE J0855 (∼285 K), the ¹⁵NH₃/¹⁴NH₃ ratio is very well constrained (H. Kühnle et al. 2024). We infer this is a temperature-dependent effect (similar to that described for isotopologues of C and H by P. Mollière & I. A. G. Snellen 2019).

In this appendix we provide a comparison of the ¹⁵NH₃ and ¹⁴NH₃ opacities, at a reference pressure of 0.1 bar and three reference temperatures of 300 K, 500 K, and 700 K, and generated with pRT (Figure 7). For this comparison we assume an abundance ratio of 250, similar to the ISM value. At higher temperatures, the ¹⁴NH₃ opacity is much higher than the ¹⁵NH₃ opacity at all wavelengths, and ¹⁵NH₃ is not detectable. At lower temperatures, selected ¹⁵NH₃ opacity features begin to protrude over the ¹⁴NH₃ opacity, meaning that a sufficiently high-quality mid-IR spectrum would detect both isotopes. For sufficiently cold planets (e.g., the 300 K example in Figure 7), many ¹⁵NH₃ features clearly emerge over the ¹⁴NH₃ opacity. These plots highlight the strong temperature dependence of the detectability of ¹⁵NH₃, which will likely only be detectable in planets and brown dwarfs colder than ∼500 K.

Zoom In Zoom Out Reset image size

After dither subtraction, the background remains poorly corrected: there is significant spatial structure, primarily in the form of "striping," i.e., linear regions across the detector that are significantly nonzero. Even more concerning, this poorly corrected background varies as a function of wavelength, meaning that wavelength-dependent systematics introduce spurious "features" into the spectrum or change the shape of true molecular features. Given the 2D nature of this structure, an annulus subtraction is insufficient to correct the background contribution to extracted photometry. This in turn adds a systematic to the 1D spectrum: aperture photometry on each slice of the 3D cube is biased since the local background contribution is not zero within the aperture, and instead contributes a wavelength-dependent positive or negative flux to the photometry procedure. When the MRS cube is built in the ifualign orientation (i.e., in the instrument plane, as opposed to the skyalign, where the cube is built with north up and east left), these background variations appear as horizontal and vertical stripes in each slice of the 3D cube. This suggests that they are associated with large-scale detector systematics, with pixels that are physically close in the detector space showing correlated values.

Figure 8 illustrates this issue: here we show an example integral field unit (IFU) cube (*s3d.fits), extracted in the ifualign orientation and collapsed across wavelength and with the source masked. Significant striping in the background is clear, especially in the horizontal direction. The right and lower panels show the mean and variation of the pixels in each row/column, with colored traces demonstrating how this background evolves with wavelength through a single IFU cube.

Zoom In Zoom Out Reset image size

To correct for this effect, we performed a data-driven postprocessing on the 3D cubes. The code for this correction is available at https://github.com/ecmatthews/mirimrs_destripe. We found that a relatively simple treatment was effective: in the ifualign orientation, we subtracted the mean value of each row and each column from that row/column, in order to "flatten" the background and remove large-scale structures. We then used the standard Extract1dStep from the jwst pipeline to extract a spectrum from these "flattened" *s3d.fits files. While this does not account for systematics that change within a row/column, we found that these are rarely present in the background of our frames. In Figure 9 we highlight the impact of this correction on the extracted spectrum. The correction can impact the overall slope of the spectrum (e.g., around 7.6 μm) as well as introduce sharper features and change the shapes of absorption features (e.g., around 8 μm).

Zoom In Zoom Out Reset image size

These changes in turn bias the outcome of retrieval analyses on the measured spectrum. We illustrate this in Figure 10 with a retrieval on data without this background correction, where extended wavelength-dependent systematics are visible, with the same physical model as for Figure 1. There are clear, wavelength-dependent systematics remaining in the residuals (unlike in Figure 1 above) that cannot be explained with an atmosphere model. Further, the shapes of key molecular features are distorted, and the retrieved abundances of molecules in the atmosphere are correspondingly biased.

Zoom In Zoom Out Reset image size

Future works should explore more sophisticated corrections to this residual background term. In particular, a data-driven postprocessing in the calibrated detector images produced by the stage 2 pipeline (*cal.fits) might allow for large-scale structures in the residual background term to be modeled and removed. We further recommend that several different dither positions be used when collecting spectra (as opposed to the two dither positions used in the present work) to provide a larger data set with which to model and remove the background emission.