betaFeedback

WorksPhysics and Astronomy

The Same But Different: SiO2_2 Clouds in Two L-Dwarfs with JWST

Abstract

The precision of brown dwarf spectra from JWST enables unprecedented insight into the atmospheric and cloud properties of sub-stellar atmospheres. Here, we compare both self-consistent radiative-convective equilibrium (RCE) and retrieval atmosphere models to high-SNR 1–18~μμm spectra of two field L-dwarfs from JWST, J0624 (L5) and J2148 (L6–L6.5). Our observations include spectral features from H2_2O, CO, CO2_2, CH4_4, Na, and K and cover approximately 98% of the outgoing thermal emission from the two objects. We first compare the data to three independent grids of 1D-RCE models: BT-Settl, exoREM, and Sonora Diamondback. While exoREM provides the best fit of the grid models, no 1D-RCE model provides a satisfactory fit to the data, highlighting the complexity of fitting cloudy L-dwarf spectra. Nonetheless, the temperature and cloud properties can be constrained with the 1D-RCE models, while metallicity, C/O ratio, and surface gravity are not well-measured. Next, we use petitRADTRANS to retrieve atmosphere and cloud properties, finding strong evidence for the presence of SiO2_2 clouds in both objects. The SiO2_2 is found to be amorphous in structure for J0624 but crystalline in J2148, highlighting important differences in objects of similar spectral type. We also used a non-Bayesian optimization framework to explore a wider variety of cloud scenarios, finding evidence for the inclusion of iron-rich magnesium silicates (Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3, MgFeSiO4_4) in addition to SiO2_2 in both objects. These results show both the complexity and diversity of brown dwarf cloud properties, while also highlighting the constraining power of broad-wavelength JWST observations of cloudy L-dwarfs.

Introduction

Condensate clouds are expected to dominate the thermal emission spectra of L-type brown dwarfs. Atmosphere models have long predicted the formation of solid high-temperature minerals at the infrared photospheres of substellar objects with effective temperatures between approximately 1300 and 1600 K (e.g., Allard et al. [2]; Tsuji [102]; Ackerman & Marley [1]; Visscher et al. [104]). Such particles tend to have significant opacity throughout the infrared, suppressing flux emerging from deeper atmospheric layers and yielding characteristically red near-infrared colors (Cushing et al. [25, 26]).

As an L-type brown dwarf continues to cool over time, it will transition to a T-type dwarf, marked by a sudden shift towards bluer colors (Tsuji & Nakajima [103]). This has long been interpreted as a gradual “clearing” of the clouds as they become patchy and begin to sink below the photosphere (Saumon & Marley [92]; Marley et al. [65]). This hypothesis is supported by the population of highly-variable objects along the L-T transition thought to be showing cloud “patchiness” (e.g., Apai et al. [5]; Vos et al. [105]; Biller et al. [15]; Oliveros-Gomez et al. [80]; Nasedkin et al. [78]) and by the depth of the 10 μm10\,\mu\mathrm{m} silicate feature, maximized for late L-dwarfs and gradually disappearing in T-dwarfs (Cushing et al. [26]; Suárez & Metchev [98]; Barber et al. [12]).

Nonetheless, other effects may also contribute to the L-T transition, including fingering convection mimicking clouds in the outgoing spectrum by making the temperature structure more isothermal (Tremblin et al. 2015), though this has also been contested (Leconte [58]). Observations of both the cloud features themselves in the MIR and their effect on the emergent spectrum across the IR can chart the formation and evolution of these clouds through parameter space.

The composition, crystallinity, particle size distribution, vertical extent, and patchiness of clouds in brown dwarfs and exoplanets can all be measured with high-SNR infrared spectra (Cushing et al. [26]; Luna & Morley [63]; Moran et al. [73]). The shape of the spectral energy distribution, combined with multiple molecular spectral features, can help constrain these properties, but the most information-rich measurements come from mid-IR spectra of the 10 μm10\,\mu\mathrm{m} silicate feature. Differences in composition (e.g., SiO2\mathrm{SiO_2} vs. MgSiO3\mathrm{MgSiO_3} vs. Mg2SiO4\mathrm{Mg_2SiO_4}), crystallinity (amorphous versus crystalline), and particle size all influence the level of solid-state absorption at mid-IR wavelengths (Cushing et al. [26]). For example, observations of the cloudy L-dwarf J2224 with IRTF/SpeX, Gemini/GNIRS, and Spitzer were fit with atmosphere retrievals, identifying sub-micron particles of SiO2\mathrm{SiO_2} and MgSiO3\mathrm{MgSiO_3} above a layer of cloud deck of Fe clouds (Burningham et al. [18]). Such detailed constraints can then inform microphysical models of cloud formation.

In this work, we fit JWST observations from 1–28 μm\mu\mathrm{m} with atmosphere model grids and retrievals to measure the atmospheric and cloud properties in two L-dwarfs, J21481633+4003594 (hereafter J2148) (Looper et al. [61]) and J06244595-4521548 (hereafter J0624). These two objects were chosen due to their relative proximity to Earth (∼10pc\sim10\mathrm{pc}) and their classification as somewhat dissimilar cloudy objects. While J2148 appears to have a very strong 10 μm10\,\mu\mathrm{m} silicate absorption feature (Looper et al. [61]) and very red colors (J−K=2.38J-K=2.38), J0624 has a somewhat more muted MIR absorption (Suárez & Metchev [98]) and bluer colors (J−K=1.9J-K=1.9), indicating a potential difference in cloud properties. Table 1 lists fundamental properties of our two targets from past literature.

NameJ0624RefJ2148Ref
Right Ascension (HH MM SS.sss)06 24 45.9449285271 ±\pm 0.3122 mas121 48 16.2986583891 ±\pm 0.2342 mas1
Declination-45 21 54.758017706 ±\pm 0.3951 mas1+40 03 59.331855665 ±\pm 0.2807 mas1
Parallax (mas)82.03 ±\pm 0.361123.67 ±\pm 0.361
Distance (pc)12.191 ±\pm 0.05418.086 ±\pm 0.0231
Opt SpTL52L63
IR SpTL54L6.5p5
J2MASS_{2MASS}14.480 ±\pm 0.029614.147 ±\pm 0.0296
H2MASS_{2MASS}13.335 ±\pm 0.028612.78 ±\pm 0.036
K2MASS_{2MASS}12.595 ±\pm 0.026611.7655 ±\pm 0.0236
log(L/Lsun)-4.33 ±\pm 0.057-4.394 ±\pm 0.0047
Age (Gyr)0.500–10.0070.500–10.007
log(g)5.14 ±\pm 0.2775.13 ±\pm 0.287
Radius (RJ_J)0.99 ±\pm 0.1070.99 ±\pm 0.107
Teff_{\mathrm{eff}} (K)1501 ±\pm 8571446 ±\pm 727
Mass (MJ_J)56.04 ±\pm 18.52755.03 ±\pm 18.557
Inclination (°)Not Available–88−2+888^{+8}_{-2}8

Table 1. Previous Literature Target Properties

§ 2 describes our observations, followed by a comparison to self-consistent model grids in § 3. § 4 describes our retrieval analysis methods and § 5 details the results for each object. A discussion of these results, comparing the two objects to themselves and the broader substellar population, is presented in § 6. We conclude with a brief summary in § 7.

Observations

JWST observed J0624 on 2023-Apr-04 and J2148 on 2023-Sep-20 in program 2288 (Lothringer et al. [62]). We obtained NIRSpec (Jakobsen et al. [52]) spectra with all three R∼2700R\sim2700 gratings: G140H/F100LP (0.97–1.82 μm\mu\mathrm{m}), G235H (1.66–3.05 μm\mu\mathrm{m}), G395H (2.87–5.14 μm\mu\mathrm{m}). For each NIRSpec grating, we obtained spectra through two NIRSpec slits: S200A1 and S200A2. For each NIRSpec slit, we obtained spectra at two dither locations. Thus, for each target we obtained four dithered spectra with each NIRSpec grating and 12 spectra in total. For all the NIRSpec spectra, we used the NRSIRS2RAPID readout pattern, which reads the entire detector and mitigates 1/f noise to some extent.

We also obtained R∼3000R\sim3000 spectra with the MIRI Medium Resolution Spectrometer (MRS; Wells et al. [108]) using three settings (short, medium, long). Each setting produces four separate spectra (channels) that cover disjoint wavelength ranges. Taken together, the three settings cover 4.9–28 μm\mu\mathrm{m} in 12 segments, but we ignore wavelengths longer than 18 μm\mu\mathrm{m} where SNR is below 20. For each setting, we obtained spectra at four dither positions. Thus, for each target, we obtained a total of 48 MIRI spectra. We also obtained simultaneous images of an adjacent field with the F1800W, F2100W, and F2550W filters, but do not report on those images here.

We sequenced exposures in ascending (J0624) or descending (J2148) order, so that consecutive exposures had some spectral overlap. The order was dictated by observatory constraints. Flux in the spectral overlap regions agrees to within the expected accuracy of flux calibration, including the order overlap between the long-wavelength end of NIRSpec and the short wavelength end of MIRI.

We used version 1.16.0 of the JWST pipeline [20] with calibration reference files in CRDS context jwst_1296.pmap to produce extracted and calibrated 1D spectra. Prior to running stage 3 of the pipeline, we masked out pixel (set DQ=1) with zero flux and the last 20 pixels of NRS2 spectra, which did not seem to be well calibrated.

After running the JWST science calibration pipeline, we combined the three pipeline-produced NIRSpec spectra (one per grating) and nine pipeline-produced MIRI spectra (one per segment, ignoring three segments beyond 18 μ\mum) for each target into a single spectrum with a log-uniform wavelength scale Δλ/λ=3000\Delta\lambda/\lambda= 3000. For each input spectrum, we found a cubic b-spline that exactly reproduced the input spectrum after integrating the b-spline analytically over the wavelength domain of each input wavelength bin. We then integrated that same b-spline over the wavelength domain of each output wavelength bin. To estimate uncertainty in each output bin, we calculated the standard deviation of trial fluxes in each output bin from 100 Monte Carlo trials with input fluxes perturbed according to the input uncertainty. Figure 1 and 2 show the spectra.

Figure 1. Best-fit 1D-RCE and retrieval models for J0624 compared to our observations. All spectra are shown at R∼1,000R\sim1{,}000, though the retrieval (orange) was fit at R∼100R\sim100 for computational efficiency. All fitting routines were done in FλF_{\lambda}-space, but the flux shown here is scaled as FνF_{\nu} to better visualize the dynamic range across the entire wavelength range.

Radiative-Convective Equilibrium Models

We first compare the observed spectra of J0624 and J2148 to three different grids of self-consistent 1-D models: BT-Settl (Allard et al. 2012; Allard et al. 2012), Sonora-Diamondback (Version 2 Morley et al. 2024; Morley et al. 2024), and 2021-era Exo-REM [14, 22]. Each grid uses a different set of opacity sources, model resolution, radiative transfer algorithm, cloud parameterization, and grid dimensions. We show the range, dimensions, and sampling of the grids in Table 2.

GridTeffT_{\mathrm{eff}} (K)ΔTeff\Delta T_{\mathrm{eff}} (K)log⁡(gcgs)\log(g_{\mathrm{cgs}})Δlog⁡g\Delta\log g[M/H]Δ\Delta[M/H]fsedf_{\mathrm{sed}}C/OΔ\DeltaC/O
BT-Settl400–26001003.5–5.50.50.00.0a——0.55 (Solara)—
Sonora-DB900–24001003.5–5.50.5−0.5 − 0.5b0.51,2,3,4,80.458 (Solarb)—
Exo-REM1000–2000503.5–5.00.5−0.5 − 1.0c0.5—0.1–0.80.05

Table 2. Parameters of the Forward Model Grids

We compare each model to the observations by first convolving the model to the approximate instrumental resolution and then summing the flux at the wavelengths associated with each pixel or wavelength bin. We compare to the absolute flux from the observations, in which we scale the observed fluxes to the distance of the objects and assume a radius of 0.95 RJupR_{\mathrm{Jup}} for each object. We also compare the grids to the normalized fluxes at 6.3 μ\mum to account for uncertainties in the object radii and distances. After either treatment, a χ2\chi^2 is calculated.

1D-RCE Grid Comparison Results

Figure 1 and Figure 2 show the best-fit models from each grid to the observations of J0624 and J2148, respectively. The best-fit parameters from each grid are listed in Table 3 and 4, along with the χ2\chi^2 goodness-of-fit.

Figure 2. Same as Figure 1, but for J2148.

PropertyBT SettlBT SettlSonora DiamondbackSonora DiamondbackExo-REMExo-REM
Normalized 6.3μ\mumAbsolute fluxNormalized 6.3μ\mumAbsolute fluxNormalized 6.3μ\mumAbsolute flux
TeffT_{\mathrm{eff}} (K)1500.01500.01400.01500.01500.01450.0
log⁡(g)\log(g)4.54.54.54.05.04.5
[M/H]0.00.00.50.53.161.0
C/O--0.50.50.80.65
fsedf_{\mathrm{sed}}--3.03.0--
χν2\chi_{\nu}^{2}a1094.311311.68881.36623.06567.02554.65

Table 3. Best-fit grid values for J0624

In general, the 1D-RCE model grids do not replicate the observed fluxes well. This, combined with the high precision of the data, result in χ2\chi^2 values in the hundreds to thousands. The best-fitting model grid is Exo-REM with χ2\chi^2 values in the 500s for J0624 and between 900 and 2000 for J2148. The Sonora Diamondback models fit the second-best while BT-Settl had the highest χ2\chi^2 values. These results were consistent between J0624 and J2148, as well as whether the spectral comparison was between absolute or normalized fluxes. In general, the grids with the more sampling of parameter space (i.e., closer space grid points) faired best. For example, Exo-REM’s flexibility with composition via both [M/H] and C/O free parameters, likely helped fit the observations better.

In each fit to the observations, the grids often fit a portion of the spectrum well (e.g., Petrus et al. [83]), but miss large features of the observed spectra with no model adequately fitting the whole wavelength range. The shape of the NIR continuum regions (J-, H-, K-bands) are particularly poorly fit. Historically, the 4 μ\mum continuum region has been especially hard to fit [97], but these objects seem to give poor fits in the other continuum ranges (e.g., 1.7, 2.2 μ\mum) to the grids, perhaps because our targets were chosen for their unique cloudiness.

Furthermore, no 1D-RCE model from any grid fit the MIRI absorption seen at around 8–10 μ\mum. A spectral feature of this strength is likely caused by abundant small-particle silicates, which forward models general have difficulty producing self-consistently [26, 63]. For example, the Sonora Diamondback models calculate the particle size from the Ackerman & Marley [1] model with a given fixed fsedf_{\mathrm{sed}} and a fixed log-normal particle size distribution of 2 (Morley et al. 2024). BT-Settl and Exo-REM both use different simplified versions of the Rossow [90] microphysical cloud model to self-consistently model the particle sizes [23] (Allard et al. 2012). In all

References—(1) Gaia Collaboration et al. [38]; (2) Reid et al. [89]; (3) Looper et al. [61]; (4) Manjavacas et al. [64]; (5) Kirkpatrick et al. [53]; (6) Cutri et al. [27]; (7) Filippazzo et al. [37]; (8) Vos et al. [105].

PropertyBT Settl Normalized 6.3µmBT Settl Absolute fluxSonora Diamondback Normalized 6.3µmSonora Diamondback Absolute fluxExo-REM Normalized 6.3µmExo-REM Absolute flux
TeffT_{\mathrm{eff}} (K)1500.01500.01200.01500.01350.01550.0
log⁡(g)\log(g)4.03.53.55.54.05.0
[M/H]0.00.00.00.010.01.0
C/O--0.50.50.30.8
fsedf_{\mathrm{sed}}--1.01.0--
χ2\chi^2 a4910.564440.461550.84082.93912.971907.14

Table 4. Best-fit grid values for 2148

aN=8727a_N = 8727.

cases, these forward models either predict too large of particles or wide size distributions, which may wash out the silicate feature, supporting the use of more complex or flexible particle size parameterizations [87].

Despite the relatively poor fits to the data, the best-fitting models in each case often agree across grids. For both objects, when comparing to the absolute fluxes, each model grid gives a best-fit TeffT_{\mathrm{eff}} within 50 K of 1500 K, which agrees well with the L5 and L6-L6.5 spectral type of J0624 and J2148, respectively [97]. This is not too surprising, since the observations cover the vast majority of the outgoing flux from each object, and thus the mean flux level will be highly sensitive to TeffT_{\mathrm{eff}} given an assumed radius, which the grids ought to agree on.

For BT-Settl and Exo-REM, the best-fit TeffT_{\mathrm{eff}} agreed to within 50 K between normalized and absolute flux fits for J0624 and to within 200 K for J2148. The Sonora Diamondback grid consistently preferred a lower TeffT_{\mathrm{eff}} in normalized flux fits relative to absolute flux fits. We hypothesize that this is an attempt by the grid to produce a redder spectrum, more in agreement with these cloudy L-dwarfs.

Maps of how χ2\chi^2 varies between log⁡(g)\log(g) and TeffT_{\mathrm{eff}} in the normalized flux fits are shown in Figure 3, indicating the temperature can be constrained, while log⁡(g)\log(g) less so. Interestingly, the same is not clearly true for [M/H] and C/O. Exo-REM was the only grid to directly vary these two parameters and the χ2\chi^2 map shown in Figure 4 does not show a clear range that is well constrained by the grid, despite this being the best-fitting grid for each object. This suggests equilibrium [M/H] and C/O compositions are inadequately describing the chemistry in these two objects, potentially implying disequilibrium chemistry, non-solar elemental abundances, or both.

Figure 3. χ2\chi^2 maps for the TeffT_{\mathrm{eff}} and log⁡10(gcgs)\log_{10}(g_{\mathrm{cgs}}) for each grid in our comparison to the J0624 (left) and J2148 (right) 1-18µm, R∼3,000R\sim3{,}000 data. Bar charts above and to the right of the χ2\chi^2 maps show the minimum χ2\chi^2 found in that that row or column.

Figure 4. χ2\chi^2 map comparing [M/H] and C/O from the exoREM grid for fits to J0624 (left) and J2148 (right). Bar charts above and to the right of the χ2\chi^2 maps show the minimum χ2\chi^2 found in that that row or column. Note the lack of a clearly constrained region of parameter space.

Figure 5 shows the χ2\chi^2 map for TeffT_{\mathrm{eff}} and fsedf_{\mathrm{sed}}. For both J0624 and J2148, the Sonora Diamondback grid prefers cloudy models to cloud-free models (fsed=0f_{\mathrm{sed}} = 0) and the range of preferred fsedf_{\mathrm{sed}} for J2148 is markedly lower than for J0624, indicating a more vertically-extended cloud layer. This is also seen in the goodness-of-fit: in both absolute and normalized flux comparisons, a lower fsedf_{\mathrm{sed}} is fit for J2148 than for J0624. This interpretation is consistent with a qualitative understanding of the spectra, i.e., J2148 appears redder with a larger MIR spectral feature likely caused by a more extended cloud of smaller grains, as well as with the retrieval result as is described below in Section 4.

Figure 5. TeffT_{\mathrm{eff}} versus fsedf_{\mathrm{sed}} χ2\chi^2 map for J0624 (left) and J2148 (right) from Sonora DB. Bar charts above and to the right of the χ2\chi^2 maps show the minimum χ2\chi^2 found in that that row or column. Fits for J0624 cluster around fsed=3.0f_{\mathrm{sed}} = 3.0, while the fits clearly prefer fsed=1.0f_{\mathrm{sed}} = 1.0 for J2148. This implies more lofted clouds in J2148 compared to J0624.

Low-Resolution Retrieval Methods

To find a better fit to our observations than those provided by the self-consistent modeling, we analyzed the data in a retrieval framework, which includes flexible parameterizations for the temperature-pressure structure, the chemical abundances, and the clouds. We performed a suite of retrievals with petitRADTRANS [70, 76, 78] on the dataset from 0.98 to 18µm at R∼\sim100. We chose the low-resolution version of the data to perform the full-wavelength retrievals for computational efficiency in order to test many retrieval set ups and evaluate the inclusion of different cloud species. Our main goal here is to constrain the cloud species responsible for the MIR absorption seen in each object, as well as the cloud crystallinity, particle sizes, and vertical distribution. Our setup is described below with all priors listed in Table 5.

ParameterPrior
log⁡g\log g (log⁡10[cm s−2]\log_{10}[\mathrm{cm}\,\mathrm{s}^{-2}])U(3.0, 6.0)U(3.0,\ 6.0)
RplR_{\mathrm{pl}} (RJR_{\mathrm{J}})U(0.733, 1.222)U(0.733,\ 1.222)
TintT_{\mathrm{int}} (K)U(750, 3000)U(750,\ 3000)
T3T_{3}U(0, 1)U(0,\ 1)
T2T_{2}U(0, 1)U(0,\ 1)
T1T_{1}U(0, 1)U(0,\ 1)
α\alphaU(1, 2)U(1,\ 2)
log⁡δ\log\deltaU(0, 1)U(0,\ 1)
log⁡Pquench\log P_{\mathrm{quench}} (log⁡10[bar]\log_{10}[\mathrm{bar}])U(−6, 3)U(-6,\ 3)
[M/H] (dex)U(−1, 2)U(-1,\ 2)
C/OU(0.1, 1.5)U(0.1,\ 1.5)
σlnorma\sigma_{\mathrm{lnorm}}^{a}U(1.05, 3.0)U(1.05,\ 3.0)
bhansbb_{\mathrm{hans}}^{b}U(0.001, 0.5)U(0.001,\ 0.5)
log⁡Kzz\log K_{zz} (log⁡10[cm2 s−1]\log_{10}[\mathrm{cm}^{2}\,\mathrm{s}^{-1}])U(5, 13)U(5,\ 13)
fsedf_{\mathrm{sed}}U(0.01, 10.01)U(0.01,\ 10.01)
log⁡Xcbc\log X_{\mathrm{cb}}^{c}U(−6, 0)U(-6,\ 0)
log⁡Pbase\log P_{\mathrm{base}} (log⁡10[bar]\log_{10}[\mathrm{bar}])c^{c}U(−5, 3)U(-5,\ 3)

Table 5. Uniform prior constraints adopted in the pRT retrieval framework.

Each retrieval uses the three-zone temperature profile parameterization designed for self-luminous objects in Mollière et al. [71]. The deep atmosphere is parameterized by an adiabat with an internal temperature, TintT_{\mathrm{int}}. The middle atmosphere is then parameterized by

an Eddington approximation where the optical depth, τ\tau, is parameterized as:

τ=δPα,(1)\tau= \delta P^{\alpha}, \tag*{(1)}

where δ\delta and α\alpha are free parameters and PP is the pressure. The middle atmosphere extends from τ=0.1\tau= 0.1 down to the radiative convective boundary (determined from an interpolated grid in petitRADTRANS’s chemistry table). Finally, the upper atmosphere (τ<0.1\tau< 0.1) is parameterized by a three-point spline from three temperatures, T1T_{1}, T2T_{2}, T3T_{3}, spread equidistant to the top of the model.

For the chemistry, we assume chemical equilibrium abundances from an interpolated grid of [M/H] and C/O assuming solar abundances according to Asplund et al. [7], with an additional parameter, PquenchP_{\mathrm{quench}}, which quenches the abundances of H2_{2}O, CO, CO2_{2}, and CH4_{4} below this pressure to their values at PquenchP_{\mathrm{quench}}. This is different than a timescale-based quenching using KzzK_{zz} that gives each molecule a different quench pressure based on their chemical conversion timescale (e.g., Prinn & Barshay [88]).

From these abundances, opacities were calculated for H2_{2}O (Polyansky et al. [86]), CO (Li et al. (2015)), CH4_{4} (Yurchenko & Tennyson [112]), CO2_{2} (Yurchenko et al. [111]), NH3_{3} (Coles et al. [24]), HCN (Barber et al. [11]), H2_{2}S (Azzam et al. [9]), FeH (Wende et al. [109]), PH3_{3} (Sousa-Silva et al. [96]), Na (Piskunov et al. [84]), K (Piskunov et al. [84]), TiO (McKemmish et al. [66]), VO (McKemmish et al. [67]), and SiO (Barton et al. [13]). Additionally, H2_{2}-H2_{2} (Borysow [16]) and H2_{2}-He (Borysow & Frommhold [17]) collision-induced opacity was added, as well as H2_{2} and He Rayleigh scattering.

As mentioned above, various cloud species and set-ups were tested, summarized in Table 6. We included different combinations of opacity from Fe(a) (Henning & Stognienko [47]), SiO(a) (Palik [81]; Kitzmann et al. [55]), SiO2_{2}(a,c) (Palik [81]; Henning & Mutschke [46]; Zeidler et al. [114]; Kitzmann et al. [55]), MgSiO3_{3}(a) (Jäger et al. [51]), Mg2_{2}SiO4_{4}(a) (Jäger et al. [51]), and Al2_{2}O3_{3}(c) (Koike et al. [56]), with a or c denoting amorphous or crystalline structures, respectively. We also tested SiO2_{2} polymorphs cristobalite and tridymite, which are chemically identical to quartz and amorphous SiO2_{2} but exhibit a different crystalline structure (Moran et al. [73]). In addition to these clouds, we also test MgSiO3_{3}(c) (Scott & Duley [93]; Jaeger et al. [49]), Mg2_{2}SiO4_{4}(c) (Servoin & Piriou [94]), Mg0.5_{0.5}Fe0.5_{0.5}SiO3_{3}(a) (Jaeger et al. [50]), and MgFeSiO4_{4}(a) (Dorschner et al. [29]; Kitzmann & Heng [54]) in our best-fit optimization framework, described below in § 4.1. These cloud species were selected based on their expected dominance of the cloud mass at the temperatures found in L-dwarf atmospheres (Wakeford et al. [106]).

Cloud SpeciesRetrievalCMA-ESOptical Constants
Fe(a)✓✓Pollack et al. [85]
SiO(a)✓✓Palik [81]; Kitzmann et al. [55]
SiO2_2(a)✓✓Palik [81]; Henning & Mutschke [46]; Zeidler et al. [114]; Kitzmann et al. [55]
SiO2_2(c) (quartz)✓✓Palik [81]; Henning & Mutschke [46]; Zeidler et al. [114]; Kitzmann et al. [55]
SiO2_2(c) (cristobalite)✓✓Palik [81]; Koike et al. [57]; Moran et al. [73]
SiO2_2(c) (tridymite)✓✓Palik [81]; Sitarz et al. [95]; Moran et al. [73]
MgSiO3_3(a)✓✓Jäger et al. [51]
MgSiO3_3(c)—✓Scott & Duley [93]; Jaeger et al. [49]
Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3(a)—✓Jaeger et al. [50]
Mg2_2SiO4_4(a)✓✓Jäger et al. [51]
Mg2_2SiO4_4(c)—✓Servoin & Piriou [94]
MgFeSiO4_4(a)—✓Dorschner et al. [29]; Kitzmann & Heng [54]
Al2_2O3_3(c)✓✓Koike et al. [56]

Table 6. Cloud species explored in the retrieval and CMA-ES optimization frameworks.

All cloud opacity for the retrievals were calculated using the Distributed Hollow Sphere (DHS) approximation (Min et al. [69]) as opposed to simple solid sphere Mie scattering.6 We use the 1D condensate (Ackerman & Marley [1]) model in pRT, where vertical mixing parameterized by an eddy diffusion coefficient, KzzK_{zz}, and gravitational settling parameterized by a sedimentation efficiency, fsedf_{\mathrm{sed}} compete to determine the vertical extent and particle sizes of the clouds. Our fiducial setup used a single fsedf_{\mathrm{sed}} to describe different cloud species, but we also tested species-specific fsedf_{\mathrm{sed}} values. The fiducial retrieval setup also assumes crystalline/quartz SiO2_{2} and amorphous Fe, MgSiO3_{3}, Mg2_{2}SiO4_{4}.

The fiducial setup also assumed a log-normal particle size distribution with a free parameter for its width given by σlnorm\sigma_{\mathrm{lnorm}}. Additionally, we tested a Hansen particle distribution ([41]), which has been shown to potentially better describe the particle size distribution in substellar atmospheres ([18]), as it more heavily weights small, sub-micron particles with a longer tail of less frequent larger particles. For the Hansen distribution we parameterize the width of particle sizes with the bhansb_{\mathrm{hans}} parameter, with the particle sizes and vertical distribution then calculated with KzzK_{zz} and fsedf_{\mathrm{sed}} according to Appendix D of Nasedkin et al. [77]. Lastly, the radius and log (gg) of the objects are freely fit parameters. The distance is set to the Gaia distances listed in Table 1.

The retrieval is run in a nested sampling framework ([34, 35]) with 4000 live points for J0624 and 5000 live points for J2148.7 An error inflation term was also included in the fits that scaled the observed uncertainties to achieve χν2≈1.0\chi_\nu^2 \approx1.0. This effectively increased the width of the retrieved posteriors to better match the disagreement between the retrieved model and the observations.

CMA-ES Cloud Scenario Optimization

With our fiducial full-wavelength nested sampling retrieval framework, individual runs can take up to about a week on a cluster of a few dozen cores. While we run several different cloud scenarios with the full setup, this computational inefficiency makes it difficult to evaluate all possible cloud species permutations. We therefore built a custom parameter optimization framework to more quickly quantify the best fitting cloud scenarios, allowing us to search a wider array of cloud species, crystallinities, particle size distributions, and opacity computation method.

We use the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) using the pycma package ([42]). CMA-ES is an optimization algorithm designed for non-linear, noisy, multimodal, and high-dimensional parameter spaces with broad application to optimization problems ([43]). We choose CMA-ES as a more robust algorithm for our potentially complex parameter space compared to simpler local optimization algorithms. For each cloud scenario, we repeated the fit five times, each with different starting conditions, taking the best-fitting result. We verified CMA-ES’s behavior in the low-dimension limit by fitting [M/H] and C/O with petitRADTRANS to the observed spectrum of J0624 with both CMA-ES and Nelder-Mead algorithms, obtaining the same solution to better than 1%.

To further balance computational efficiency, we fit only the 5-18μ\mum portion of the observed spectra, keeping all non-cloud-related parameters fixed to the best-fit solution found from the fiducial full-wavelength nested sampling retrieval. However, we compared the spectra at a higher resolution (R∼1000R\sim1000) compared to the lower-resolution, full-wavelength retrieval (R∼100R\sim100), leveraging any additional information on the cloud species from the higher resolution enabled by MRI/MRS.

Compared to the retrieval setup, we also test 1) both amorphous and crystalline MgSiO3_3 and Mg2_2SiO4_4 (rather than just amorphous), 2) both DHS and Mie scattering opacity computation, and 3) every possible combination SiO2_2, one of MgSiO3_3 and Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3, and one of Mg2_2SiO4_4 and MgFeSiO4_4 on top of a Fe deep cloud base. Cloud species data sources are listed in Table 6.

Low-Resolution Retrieval Results

Figures 1 and 2 show the best-fit retrieved model to the observations of J0624 and J2148 using the fiducial setup (see above), alongside the best fits from the self-consistent grids. The fits obtained in the retrieval analysis are much better than the self-consistent grids, thanks to the flexibility in the temperature structure, gas chemistry, and cloud properties. All major spectral features are fit, including the shape of the J, H, and K-bands, which the self-consistent grids struggled to fit. Importantly, the full-wavelength retrieval is able to fit the MIR silicate feature between 8-10 μ\mum, albeit not perfectly. At the R∼100R\sim100 resolution used for the full-wavelength retrievals, the best-fit J0624 retrieval achieves a χ2/Ndata\chi^2/N_{\mathrm{data}} of 1648.7 without error inflation, compared to 12,913.2 from the best-fitting grid model from Exo-REM. Meanwhile, the higher-resolution MIR-specific CMA-ES fits were able to fit the cloud absorption even better, with a minimum χ2/Ndata\chi^2/N_{\mathrm{data}} of 50.248. A similar story holds for J2148 where the retrieval obtains χ2/Ndata=1506.4\chi^2/N_{\mathrm{data}} = 1506.4 compared to 48,560.0 from Exo-REM. The corresponding best-fitting CMA-ES scenario resulted in χ2/Ndata=96.836\chi^2/N_{\mathrm{data}} = 96.836.

In summary, the best-fitting models for J0624 contained small, amorphous SiO2_2 grains. The CMA-ES fits also clearly preferred the inclusion of iron and magnesium-rich silicate grains as well. Additional Al2_2O3_3 grains were suggested by the full-wavelength retrieval, **NOTE—**The fiducial retrieval framework uses Fe(c), SiO(a), quartz SiO2_2(c), MgSiO3_3(a), and Mg2_2SiO4_4(a). Additional retrieval tests include amorphous SiO2_2 and the SiO2_2 polymorphs cristobalite and tridymite. The CMA-ES optimization additionally explores crystalline MgSiO3_3 and Mg2_2SiO4_4, Fe-bearing silicates (Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3 and MgFeSiO4_4) in lieu of Mg-only silicates, and all allowed combinations of silicate cloud species atop a deep Fe cloud.

while Fe clouds matched the data well in the CMA-ES fits. Likewise for J2148, small SiO2_2 grains best explain the MIR spectrum, but quartz rather than amorphous SiO2_2 grains are favored by both the full-wavelength retrieval and the CMA-ES fits. We describe these results in more detail below.

J0624 Results

The median-fit parameters and 1σ\sigma uncertainties from each retrieval of the J0624 observations are listed in Table 7 and Table 8. The former table compares retrievals with different cloud properties including species-specific fsedf_{\mathrm{sed}} values, the absence of SiO2_2, Mgx_xSiOy_y (i.e., magnesium silicates), and Fe, and the presence of Al2_2O3_3. Table 8 compares additional retrievals computed in order to identify the specific properties of the SiO2_2 clouds that appear to fit the observations best. Despite the qualitative quality of the fit to the spectra, all retrievals fit for an error inflation term of about 30, implying that the deviation of our residuals of the fits were about 30 times worse than photon noise alone, likely representing systematic inadequacies of our relatively simple 1D modeling approach.

ParameterFiducialMulti-fsedf_{\mathrm{sed}}No SiO2_2No Mgx_xSiOy_yNo FeWith Al2_2O3_3
Live Points400040004000400040004000
ln⁡Za\ln Z^{a}10642.510632.9±\pm0.910643.310699.510673.110663.3±\pm0.0
χν2\chi_\nu^21648.71691.11713.01526.01541.51687.4
TintT_{\mathrm{int}}2202±\pm132182±\pm152206±\pm192252±\pm772170±\pm162186±\pm35
log⁡g\log g4.85±\pm0.054.85±\pm0.074.90±\pm0.054.48±\pm0.044.78±\pm0.064.31±\pm0.05
RJupR_{\mathrm{Jup}}0.897±\pm0.0060.897±\pm0.0030.896±\pm0.0050.924±\pm0.0050.924±\pm0.0070.937±\pm0.006
MJupbM_{\mathrm{Jup}}^{b}23±323 \pm323±323 \pm326±326 \pm310±310 \pm321±321 \pm37±37 \pm3
log⁡pquench\log p_{\mathrm{quench}}−3.29±0.92-3.29\pm0.92−3.77±0.79-3.77\pm0.79−2.57±0.70-2.57\pm0.70−0.137±0.035-0.137\pm0.035−0.172±0.071-0.172\pm0.071−0.358±0.030-0.358\pm0.030
M/H0.58±\pm0.030.63±\pm0.030.65±\pm0.020.70±\pm0.040.70±\pm0.040.55±\pm0.03
C/O0.655±\pm0.0020.657±\pm0.0020.658±\pm0.0020.75±\pm0.020.685±\pm0.0070.71±\pm0.01
σlnorm\sigma_{\mathrm{lnorm}}2.38±\pm0.132.02±\pm0.162.13±\pm0.192.74±\pm0.192.68±\pm0.212.76±\pm0.16
log⁡kzz\log k_{zz}7.77±\pm0.177.08±\pm0.139.72±\pm0.577.50±\pm0.187.09±\pm0.136.86±\pm0.13
fsedf_{\mathrm{sed}}1.34±\pm0.29-1.17±\pm0.271.19±\pm0.172.04±\pm0.234.98±\pm0.60
fsed,SiO2f_{\mathrm{sed,SiO_2}}-1.15±\pm0.24----
fsed,MgSiO3f_{\mathrm{sed,MgSiO_3}}-4.13±\pm0.96----
fsed,Mg2SiO4f_{\mathrm{sed,Mg_2SiO_4}}-1.25±\pm0.64----
fsed,Fef_{\mathrm{sed,Fe}}-4.42±\pm0.59----
log⁡Xcb,SiO2\log X_{\mathrm{cb,SiO_2}}-3.07±\pm0.11-3.60±\pm0.14--3.84±\pm0.11-3.64±\pm0.11-3.00±\pm0.14
log⁡Pbase,SiO2\log P_{\mathrm{base,SiO_2}}-1.370±\pm0.085-1.06±\pm0.16--0.594±\pm0.033-0.823±\pm0.031-1.009±\pm0.015
log⁡Xcb,MgSiO3\log X_{\mathrm{cb,MgSiO_3}}-3.99±\pm0.87-3.79±\pm0.69-0.35±\pm0.46--3.01±\pm0.12-4.89±\pm0.67
log⁡Pbase,MgSiO3\log P_{\mathrm{base,MgSiO_3}}-3.45±\pm0.78-3.69±\pm0.84-3.78±\pm0.46--1.025±\pm0.023-3.2±\pm1.2
log⁡Xcb,Mg2SiO4\log X_{\mathrm{cb,Mg_2SiO_4}}-4.90±\pm0.61-4.44±\pm0.49-4.11±\pm0.92--4.54±\pm0.85-2.6±\pm1.5
log⁡Pbase,Mg2SiO4\log P_{\mathrm{base,Mg_2SiO_4}}-1.85±\pm0.71-3.21±\pm0.891.13±\pm0.98--3.41±\pm0.982.06±\pm0.52
log⁡Xcb,Fe\log X_{\mathrm{cb,Fe}}-3.34±\pm0.13-3.23±\pm0.20-4.52±\pm0.46-4.28±\pm0.91--3.67±\pm0.82
log⁡Pbase,Fe\log P_{\mathrm{base,Fe}}-2.059±\pm0.096-2.10±\pm0.150.9±\pm2.01.63±\pm0.71-0.76±\pm0.17
log⁡Xcb,Al2O3\log X_{\mathrm{cb,Al_2O_3}}------4.85±\pm0.71
log⁡Pbase,Al2O3\log P_{\mathrm{base,Al_2O_3}}------3.3±\pm1.1

Table 7. Retrieved properties of J0624 from retrievals with alternative cloud properties.

Fiducial (Quartz)Amorphous*TridymiteCristobaliteSiO(a)Am. SiO2_2 (Hansen)
Live Points400040004000400040004000
ln⁡Z\ln Z10642.510697.310678.2 ±\pm 0.010690.0 ±\pm 0.010594.110643.1 ±\pm 0.0
χν2\chi_\nu^21648.71427.41522.61684.725751717.2
TintT_{\mathrm{int}}2202 ±\pm 132179 ±\pm 182175 ±\pm 342177 ±\pm 151588 ±\pm 182154 ±\pm 16
log⁡g\log g4.854 ±\pm 0.0484.676 ±\pm 0.0304.384 ±\pm 0.0474.591 ±\pm 0.0424.410 ±\pm 0.0734.705 ±\pm 0.053
RplR_{\mathrm{pl}}0.897 ±\pm 0.0060.915 ±\pm 0.0040.924 ±\pm 0.0050.919 ±\pm 0.0050.880 ±\pm 0.0050.880 ±\pm 0.003
MJupbM_{\mathrm{Jup}}^{b}23 ±\pm 316 ±\pm 38 ±\pm 313 ±\pm 38 ±\pm 316 ±\pm 3
log⁡pquench\log p_{\mathrm{quench}}−3.29±0.92-3.29 \pm0.92−0.020±0.032-0.020 \pm0.032−0.250±0.031-0.250 \pm0.031−0.134±0.031-0.134 \pm0.0310.96 ±\pm 0.21−4.34±0.87-4.34 \pm0.87
[M/H]0.575 ±\pm 0.0250.770 ±\pm 0.0260.599 ±\pm 0.0300.703 ±\pm 0.0300.724 ±\pm 0.0410.603 ±\pm 0.028
C/O0.6551 ±\pm 0.00150.7505 ±\pm 0.00860.728 ±\pm 0.0150.728 ±\pm 0.0130.8275 ±\pm 0.00430.6563 ±\pm 0.0017
σlnorm\sigma_{\mathrm{lnorm}}2.38 ±\pm 0.132.85 ±\pm 0.112.859 ±\pm 0.0942.84 ±\pm 0.101.40 ±\pm 0.15-
bhansb_{\mathrm{hans}}-----0.442 ±\pm 0.037
log⁡kzz\log k_{zz}7.77 ±\pm 0.176.86 ±\pm 0.127.08 ±\pm 0.146.709 ±\pm 0.08010.406 ±\pm 0.0917.32 ±\pm 0.10
fsedf_{\mathrm{sed}}1.34 ±\pm 0.294.55 ±\pm 0.634.03 ±\pm 0.464.50 ±\pm 0.477.7 ±\pm 1.62.22 ±\pm 0.42
fsed,SiO2f_{\mathrm{sed,SiO_2}}------
fsed,MgSiO3f_{\mathrm{sed,MgSiO_3}}------
fsed,Mg2SiO4f_{\mathrm{sed,Mg_2SiO_4}}------
fsed,Fef_{\mathrm{sed,Fe}}------
log⁡Xcb,SiO\log X_{\mathrm{cb,SiO}}----−3.1±1.3-3.1 \pm1.3-
log⁡Pbase,SiO\log P_{\mathrm{base,SiO}}----−0.9±2.7-0.9 \pm2.7-
log⁡Xcb,SiO2\log X_{\mathrm{cb,SiO_2}}−3.07±0.11-3.07 \pm0.11−2.88±0.18-2.88 \pm0.18−3.31±0.16-3.31 \pm0.16−2.83±0.13-2.83 \pm0.13-−3.23±0.12-3.23 \pm0.12
log⁡Pbase,SiO2\log P_{\mathrm{base,SiO_2}}−1.370±0.085-1.370 \pm0.085−0.912±0.025-0.912 \pm0.025−0.685±0.024-0.685 \pm0.024−1.015±0.017-1.015 \pm0.017-−0.98±0.10-0.98 \pm0.10
log⁡Xcb,MgSiO3\log X_{\mathrm{cb,MgSiO_3}}−3.99±0.87-3.99 \pm0.87−3.8±1.7-3.8 \pm1.7−4.2±1.1-4.2 \pm1.1−4.66±0.82-4.66 \pm0.82−3.0±1.3-3.0 \pm1.3−3.82±0.88-3.82 \pm0.88
log⁡Pbase,MgSiO3\log P_{\mathrm{base,MgSiO_3}}−3.45±0.78-3.45 \pm0.781.86 ±\pm 0.66−3.29±0.93-3.29 \pm0.93−3.4±1.0-3.4 \pm1.0−0.9±2.7-0.9 \pm2.7−3.4±1.0-3.4 \pm1.0
log⁡Xcb,Mg2SiO4\log X_{\mathrm{cb,Mg_2SiO_4}}−4.90±0.61-4.90 \pm0.61−3.6±1.7-3.6 \pm1.7−4.64±0.83-4.64 \pm0.83−4.68±0.81-4.68 \pm0.81−3.1±1.3-3.1 \pm1.3−4.61±0.70-4.61 \pm0.70
log⁡Pbase,Mg2SiO4\log P_{\mathrm{base,Mg_2SiO_4}}−1.85±0.71-1.85 \pm0.711.87 ±\pm 0.64−3.3±1.0-3.3 \pm1.0−3.47±0.95-3.47 \pm0.95−1.0±2.7-1.0 \pm2.7−3.24±0.96-3.24 \pm0.96
log⁡Xcb,Fe\log X_{\mathrm{cb,Fe}}−3.34±0.13-3.34 \pm0.13−4.84±0.75-4.84 \pm0.75−2.6±1.3-2.6 \pm1.3−4.93±0.69-4.93 \pm0.69−3.0±1.3-3.0 \pm1.3−4.86±0.60-4.86 \pm0.60
log⁡Pbase,Fe\log P_{\mathrm{base,Fe}}−2.059±0.096-2.059 \pm0.096−3.1±1.3-3.1 \pm1.31.21 ±\pm 0.34−3.53±0.88-3.53 \pm0.88−1.2±2.6-1.2 \pm2.6−2.68±0.88-2.68 \pm0.88

Table 8. Retrieved properties of J0624 from retrievals with different SiO2_2 polymorphs.

J0624 Bulk Properties

All retrievals tend to find an elevated [M/H], ranging from 0.548 to 0.770. Similarly, a somewhat supersolar C/O ratio is consistently retrieved, ranging from 0.655 to 0.828, not unlike the T-dwarf retrievals of Zalesky et al. [113]. Retrieved radii ranged from 0.880 to 0.924 RJupR_{\mathrm{Jup}}, with log⁡(g)cgs\log(g)_{\mathrm{cgs}} varying between 4.384 to 4.928. This corresponds to a range of implied masses between 8 and 26 MJupM_{\mathrm{Jup}}. While the retrieval fits are not outright unphysical, we do not expect a mass lower than the deuterium burning limit and metallicities significantly above solar. The retrieved surface gravity in particular appears to be consistent with the range of best-fit self-consistent models, which ranged from 4.0 to 5.0, with the Sonora models further agreeing with the somewhat elevated metallicity, pushing up against the edge of the grid boundary (see Table 3). We discuss possible biases in J0624’s bulk retrieved properties in our comparison to J2148 below (see § 5.2.2).

J0624 Cloud Properties

For our main goal of identifying the cloud species and properties present in J0624, we compare our various cloud scenarios in terms of both the Bayesian Evidence (ln Z) as well as the least squares of the residuals in the best-fit spectrum (χν2\chi^2_\nu). By comparing to our fiducial retrieval, which includes crystalline SiO2_2, MgSiO3_3, Mg2_2SiO4_4, and Fe clouds with a single fsedf_{\mathrm{sed}}, we can quantify whether certain cloud species or treatments are statistically justified. For example, the treatment of clouds

with species-specific fsedf_{\mathrm{sed}} parameters resulted in a lower Bayesian evidence (ln⁡B=ln⁡Z1−ln⁡Z2=9.6\ln B = \ln Z_1 - \ln Z_2 = 9.6), implying the additional parameters to control species-specific vertical profiles and size distributions were not statistically justified. Interestingly, this is confirmed with the fact that the best-fit retrieved spectrum was somewhat worse when fitting species-specific fsedf_{\mathrm{sed}} parameters, perhaps implying that the higher dimensional retrieval had trouble finding an equivalent best-fit compared to the single-fsedf_{\mathrm{sed}} case.

The removal of Fe and magnesium-silicate clouds both resulted in higher Bayesian evidences by >30> 30, implying that their inclusion in the retrievals of J0624 was not justified. However, we removed both MgSiO3_3 and Mg2_2SiO4_4 simultaneously; as shown below, the inclusion of magnesium and iron silicates is suggested by fits with the CMA-ES optimization framework (see Section 5.1.2.0). The lack of clear evidence for Fe or the combined magnesium-silicate clouds is somewhat surprising as both silicate and iron clouds were necessary to explain the spectrum of J2224 ([18]) and PSO J316 ([72]). Meanwhile, the inclusion of Al2_2O3_3 resulted in a higher evidence by >20> 20, implying that its presence was justified by the data. We suspect that the retrieval is using Al2_2O3_3 as an additional gray opacity source, preferable over Fe perhaps due to a different implied vertical location, extent, and/or particle size distribution.

The best-fit spectra from these scenarios compared to the observations are shown in Figure 6. While the

Figure 6. Top: Comparison of the best-fit full-wavelength retrievals of J0624 with different cloud species compared to the R∼100R\sim100 observations (black). Bottom: Residuals between the retrieved best-fit spectra and the observations.

combination of species explored here does not perfectly match the silicate feature, especially the deepest part of the absorption at 8.5 μ\mum, it is clear that the inclusion of SiO2_2 greatly improves the fit. In particular, the red edge of the feature is much better matched with SiO2_2 compared to the spectra without SiO2_2. Because the inclusion of SiO2_2 in the retrievals is highly preferred by the retrieval and somewhat unexpected, we ran additional retrievals to identify additional details of the cloud properties, including different crystal structures and particle size distributions. We also replaced SiO2_2 with amorphous SiO as recently seen in PSO J316 [72]. We run these additional SiO2_2 tests with the fiducial setup, including Fe, MgSiO3_3 and Mg2_2SiO4_4 in the event that these species become an important component of the fit. These results are listed in Table 8.

Both tridymite and cristobalite are preferred over quartz for J0624’s spectrum, with Bayes factors of ln⁡B=35.7\ln B = 35.7 and 47.5, respectively, aligning with the expected microphysical stability of the crystal structures at J0624’s relevant atmospheric pressure and temperatures ([73]). However, we find a strong preference for amorphous SiO2_2 over all other cloud types or crystalline structures, with a Bayes factor of ln⁡B=54.7\ln B = 54.7 over the fiducial quartz retrievals and a reduced χ2\chi^2 almost 100 less than the second best-fitting polymorph, tridymite. Compared to the other crystal structures, the amorphous SiO2_2 provides a better fit between 9 and 9.5 μ\mum and longward of 11μ\mum, but regions of disagreement still exist, particularly around 8.5 and 10.5 μ\mum.

We next test for a preference between a log-normal and Hansen particle size distribution by running our preferred amorphous SiO2_2 scenario with a Hansen particle size distribution. We find a significantly worse fit to the data and a significantly lower Bayesian evidence (ln⁡B=54.2\ln B = 54.2) for the Hansen size distribution scenario. Figure 6 shows that the best-fit Hansen distribution model absorbs too heavily at both the short and long wavelength edges of the silicate absorption feature, despite matching the 8.5μ\mum better than other models.

Figure 9 shows the retrieved particle size distributions and vertical profiles for the fiducial retrieval, the preferred amorphous SiO2_2 retrieval, the retrieval using species-specific fsedf_{\mathrm{sed}} values, and the Hansen distribution retrieval. In general, particles between 0.1 and 1.0 μm\mu\mathrm{m} are retrieved. The Hansen distribution allows for larger particles are higher pressures, while preserving the low particle sizes near the IR photosphere.

Figure 9. Cloud mass fraction (left) and particle size(right) as a function of pressure for J0624 (top) and J2148 (bottom) from the best-fit retrieval model. We show results from both the fiducial retrieval, where one fsedf_{\mathrm{sed}} is shared between cloud species (solid), as well as from the retrieval where an individual fsedf_{\mathrm{sed}} was retrieved for each cloud species (see Table 7 and Table 9).

The SiO2_2 cloud deck is retrieved between 10 and 100 millibar with the exact location of the other cloud decks dependent on the retrieval setup. Interestingly, the non-SiO2_2 clouds are placed below the SiO2_2 clouds in the amorphous SiO2_2 case, while the opposite is true for the fiducial crystalline SiO2_2 case.

The large oscillation in the particle size in the amorphous SiO2_2 retrieval appears to be from a subtle increase in the temperature in the upper atmosphere compared to the fiducial retrieval. The atmosphere at pressures below 10 millibar is poorly constrained by the observations as evidenced by the contribution function (see Figure 8), so we do not consider the temperature or particle sizes in this regime to be physical.

Figure 8. Retrieved (yellow) versus best-fit 1D-RCE (black) pressure-temperature profiles of J0624 (left) and J2148 (right). The best-fit 1D-RCE models are from exoREM (see § 3 and § 4). We show the best-fit TP profiles after both normalizing the data at 6.3 µm (solid) and using the absolute fluxes, assuming 1 Jupiter radius for the brown dwarf radius. We also show contribution functions (CF) for the retrieved model at probing CO (blue), the continuum (green), and the silicate cloud feature (purple).

J0624 CMA-ES Best-Fits—To search more broadly for the best-fitting cloud scenario, we performed a sweep of cloud composition, crystallinity, particle size distribution, and opacity computation method with the CMA-ES algorithm framework. Figure 7 shows the best fitting model scenarios compared to the data, as well as compared to the preferred scenario from the full-wavelength retrievals, in brightness temperature space. We compare the brightness temperatures as this removes the general blackbody shape, helping to isolate the shape of the absorption features. Appendix Table 11 ranks the various combinations that were tried based on their respective χ2/Ndata\chi^2/N_{\mathrm{data}} goodness-of-fit.

Figure 7. Top: Fits to the R∼1000R\sim1000 mid-IR spectrum of J0624 (black) using CMA-ES. Our best-fit retrieval (computed at R∼100R\sim100) is shown in red, which included amorphous Fe, SiO2_2, MgSiO3_3, and Mg2_2SiO4_4 using DHS for the opacity computation. We show the best fitting scenarios without SiO2_2 and without magnesium-iron silicates for comparison. Bottom: Residuals between the observations and each model fit.

The best-fitting CMA-ES scenarios contain both amorphous SiO2_2 and iron-rich pyroxene or olivine. In agreement with the retrieval, the CMA-ES fits generally indicate the presence of amorphous SiO2_2 cloud particles over crystalline particles. Additionally, the difference in goodness-of-fit between the Hansen and log-normal particle size distributions was insignificant. The best-fitting model without SiO2_2 clouds has a χ2/Ndata\chi^2/N_{\mathrm{data}} about 4.5 worse than the best-fitting model with SiO2_2 clouds (55.25 versus 50.74, respectively), particularly driven by a poor match to the blue edge of the silicate absorption (≈7.5−8.5 μm\approx7.5 - 8.5\,\mu\mathrm{m}). However, in addition to SiO2_2, Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3 or MgFeSiO4_4 is also suggested by the fit; the best-fit is improved from χ2/Ndata=59.24\chi^2/N_{\mathrm{data}} = 59.24 to 50.74 with the inclusion of these silicates. While iron-rich silicates are slightly preferred, iron-poor silicates provide comparable fits with χ2/Ndata=53.60\chi^2/N_{\mathrm{data}} = 53.60. The inclusion of these other silicates improves the fit to the red edge of the silicate feature, particularly between 8.5 and 9.5 μm\mu\mathrm{m}. The worst fitting 29 spectra all contained no SiO2_2 and the three worst-of-the-worst contained only the Fe base and Mg2_2SiO4_4.

J0624 TP Profile

In Figure 8, we compare the model temperature profiles. Compared to the best-fit 1D-RCE temperature profile, the retrieved TP profile is significantly steeper, i.e., dT/dPdT/dP is less for the retrieved TP profile. The TP profile parameterization of Mollière et al. [71] qualitatively recreates the general shape of the 1D-RCE profiles with a deep adiabat and a radiative photosphere around 1 bar. However, at all observed pressures, the retrieved profile remains steeper. In general, the pressures probed by our observations vary between 10 mbar and 1 bar according to the contribution function. Therefore, the TP profile above and below those pressures is constrained mainly through the parameterization rather than empirical observation.

J2148 Results

As a direct comparison to J0624, we run many of the same fits for J2148. As mentioned above, J2148 has a very similar optical and IR spectral type as J0624 – just 1 and 1.5 spectral types later, respectively. However, their full IR spectra are quite different, with J2148 being significantly redder. This suggests that J2148 is likely cloudier than J0624, perhaps due to the equator-on inclination at which we observe J2148 ([105, 99]). Similar to J0624, an error inflation term of about 30 was required for the χ2\chi^2 of the fits to be about 1.0.

J2148 Cloud Properties

The median-fit parameters and 1σ1\sigma uncertainties from each J2148 retrieval are listed in Table 9 and Table 10 with the best-fit spectra from the preferred scenario shown in Figure 10. Overall, a similar picture to J0624 is suggested, with SiO2_2 clouds strongly preferred over magnesium-silicates. The scenario without SiO2_2 had a significantly worse fit and ln⁡B=37.0\ln B = 37.0. Not including magnesium silicates is moderately preferred at ln⁡B=4.0\ln B = 4.0 (though again see results from the CMA fits below in Section 5.2.1.0). Fe clouds are moderately justified with ln⁡B=3.5\ln B = 3.5. Al2_2O3_3 is only marginally suggested with ln⁡B=0.4\ln B = 0.4. Fitting for species-specific fsedf_{\mathrm{sed}} parameters is justified by the data with ln⁡B=9.0\ln B = 9.0.

ParameterFiducialMulti-fsedf_{\mathrm{sed}}No SiO2_2No Mgx_xSiOy_yNo FeWith Al2_2O3_3
Live Points500050005000500050005000
ln⁡Za\ln Z^{a}10491.110500.110454.110495.110487.610491.5
χν2\chi^2_\nu1506.41610.81781.01686.91688.11484.5
TintT_{\mathrm{int}}2353±162615±572523±302292±192375±242304±18
log⁡g\log g4.389±0.0534.011±0.0614.214±0.0534.352±0.0604.241±0.0744.554±0.050
RJupR_{\mathrm{Jup}}1.0480±0.00431.0593±0.00351.1119±0.00651.0365±0.00321.0525±0.00521.0851±0.0046
MJupbM_{\mathrm{Jup}}^{b}11 ± 35 ± 38 ± 310 ± 38 ± 317 ± 3
log⁡pquench\log p_{\mathrm{quench}}-5.32±0.37-5.23±0.52-3.81±0.85-5.15±0.49-4.72±0.81-4.98±0.68
M/H0.424±0.0240.354±0.0280.439±0.0270.397±0.0300.404±0.0340.440±0.023
C/O0.6444±0.00510.6560±0.00220.6639±0.00180.6477±0.00440.6471±0.00460.6242±0.0051
σlnorm\sigma_{\mathrm{lnorm}}2.937±0.0472.959±0.0322.870±0.0862.945±0.0442.952±0.0422.84±0.10
log⁡kzz\log k_{zz}7.51±0.157.438±0.0888.473±0.0717.61±0.107.054±0.0617.567±0.092
fsedf_{\mathrm{sed}}0.61±0.18-0.74±0.110.75±0.130.90±0.170.460±0.091
fsed,SiO2f_{\mathrm{sed,SiO_2}}-0.418±0.094----
fsed,MgSiO3f_{\mathrm{sed,MgSiO_3}}-4.1±1.1----
fsed,Mg2SiO4f_{\mathrm{sed,Mg_2SiO_4}}-1.9±1.1----
fsed,Fef_{\mathrm{sed,Fe}}-4.29±0.51----
log⁡Xcb,SiO2\log X_{\mathrm{cb,SiO_2}}-2.924±0.055-3.392±0.074--2.696±0.081-3.14±0.14-2.968±0.062
log⁡Pbase,SiO2\log P_{\mathrm{base,SiO_2}}-1.495±0.030-1.386±0.030--1.727±0.033-1.296±0.070-1.377±0.028
log⁡Xcb,MgSiO3\log X_{\mathrm{cb,MgSiO_3}}-3.49±0.87-3.0±1.3-1.674±0.060--3.84±0.81-3.83±0.74
log⁡Pbase,MgSiO3\log P_{\mathrm{base,MgSiO_3}}-3.72±0.772.05±0.56-2.756±0.026--3.79±0.85-3.75±0.69
log⁡Xcb,Mg2SiO4\log X_{\mathrm{cb,Mg_2SiO_4}}-1.72±0.52-1.36±0.41-4.09±0.33--2.63±0.68-1.84±0.41
log⁡Pbase,Mg2SiO4\log P_{\mathrm{base,Mg_2SiO_4}}-4.17±0.52-4.41±0.411.61±0.41--3.20±0.79-3.88±0.43
log⁡Xcb,Fe\log X_{\mathrm{cb,Fe}}-4.76±0.37-1.14±0.77-4.036±0.096-3.99±0.42--4.73±0.17
log⁡Pbase,Fe\log P_{\mathrm{base,Fe}}-1.95±0.481.05±0.15-1.25±0.12-2.79±0.41--2.03±0.25
log⁡Xcb,Al2O3\log X_{\mathrm{cb,Al_2O_3}}------4.820±0.093
log⁡Pbase,Al2O3\log P_{\mathrm{base,Al_2O_3}}------0.260±0.099

Table 9. Retrieved properties of J2148 from retrievals with alternative cloud properties.

Fiducial (Quartz)AmorphousTridymiteCristobaliteSiO₂ (Hansen)*SiO(a)
Live Points500050005000500050005000
ln⁡Z\ln Z10491.110479.410456.4 ± 0.010475.8 ± 0.010505.5 ± 0.010381.4
χν2\chi^2_\nu1506.41679.71782.21628.71469.32566.2
TintT_{\mathrm{int}}2353 ± 162602 ± 572708 ± 982456 ± 722544 ± 571727 ± 21
log⁡g\log g4.389 ± 0.0534.015 ± 0.0584.075 ± 0.0624.062 ± 0.0664.266 ± 0.0614.376 ± 0.070
RJupR_{\mathrm{Jup}}1.05 ± 0.0041.10 ± 0.0051.16 ± 0.0061.10 ± 0.0041.11 ± 0.0041.15 ± 0.01
MJupM_{\mathrm{Jup}}11 ± 35 ± 36 ± 36 ± 39 ± 313 ± 3
log⁡pquench\log p_{\mathrm{quench}}−5.32 ± 0.37−1.97 ± 0.25−2.82 ± 0.85−1.95 ± 0.22−3.0 ± 1.2−4.1 ± 1.2
[M/H]0.424 ± 0.0240.272 ± 0.0290.402 ± 0.0310.306 ± 0.0280.411 ± 0.0270.533 ± 0.038
C/O0.6444 ± 0.00510.6552 ± 0.00200.6583 ± 0.00190.6604 ± 0.00180.6496 ± 0.00370.6143 ± 0.0058
bhansb_{\mathrm{hans}}----0.145 ± 0.045-
σlnorm\sigma_{\mathrm{lnorm}}2.937 ± 0.0472.922 ± 0.0551.415 ± 0.0922.934 ± 0.046-2.21 ± 0.20
log⁡kzz\log k_{zz}7.51 ± 0.157.794 ± 0.0738.281 ± 0.0937.211 ± 0.0607.787 ± 0.08110.58 ± 0.13
fsedf_{\mathrm{sed}}0.61 ± 0.180.271 ± 0.0460.170 ± 0.0361.50 ± 0.160.183 ± 0.0420.451 ± 0.076
fsed,SiO2f_{\mathrm{sed,SiO_2}}------
fsed,MgSiO3f_{\mathrm{sed,MgSiO_3}}------
fsed,Mg2SiO4f_{\mathrm{sed,Mg_2SiO_4}}------
fsed,Fef_{\mathrm{sed,Fe}}------
log⁡Xcb,SiO\log X_{\mathrm{cb,SiO}}-----−3.05 ± 0.61
log⁡Pbase,SiO\log P_{\mathrm{base,SiO}}-----−3.54 ± 0.63
log⁡Xcb,SiO2\log X_{\mathrm{cb,SiO_2}}−2.924 ± 0.055−3.995 ± 0.089−4.101 ± 0.089−3.193 ± 0.089−3.209 ± 0.071-
log⁡Pbase,SiO2\log P_{\mathrm{base,SiO_2}}−1.495 ± 0.030−1.377 ± 0.068−1.952 ± 0.034−1.961 ± 0.028−1.385 ± 0.032-
log⁡Xcb,MgSiO3\log X_{\mathrm{cb,MgSiO_3}}−3.49 ± 0.87−3.24 ± 0.53−2.821 ± 0.070−4.26 ± 0.48−3.54 ± 0.97−0.032 ± 0.028
log⁡Pbase,MgSiO3\log P_{\mathrm{base,MgSiO_3}}−3.72 ± 0.77−4.03 ± 0.52−2.076 ± 0.030−4.08 ± 0.52−3.78 ± 0.78−4.60 ± 0.11
log⁡Xcb,Mg2SiO4\log X_{\mathrm{cb,Mg_2SiO_4}}−1.72 ± 0.52−3.781 ± 0.080−4.922 ± 0.063−3.217 ± 0.074−2.970 ± 0.066−3.92 ± 0.70
log⁡Pbase,Mg2SiO4\log P_{\mathrm{base,Mg_2SiO_4}}−4.17 ± 0.52−1.292 ± 0.0712.57 ± 0.31−1.502 ± 0.028−1.846 ± 0.032−2.8 ± 1.2
log⁡Xcb,Fe\log X_{\mathrm{cb,Fe}}−4.76 ± 0.37−3.91 ± 0.31−4.841 ± 0.081−3.35 ± 0.25−4.79 ± 0.14−4.5 ± 0.35
log⁡Pbase,Fe\log P_{\mathrm{base,Fe}}−1.95 ± 0.48−3.35 ± 0.290.81 ± 0.29−3.54 ± 0.212.39 ± 0.29−2.1 ± 1.1

Table 10. Retrieved properties of J2148 from retrievals with different SiO₂ polymorphs.

Figure 10. Same as Figure 6, but for J2148.

Of all SiO2_2 polymorphs, the fiducial crystalline quartz particles appear to be the most preferred by the retrieval. Amorphous SiO2_2 has the second-best Bayesian evidence but with ln⁡B=11.7\ln B=11.7, while the cristobalite scenario has a slightly worse Bayesian evidence and a slightly better fitting χ2\chi^2. SiO cloud particles are not favored with ln⁡B≈110\ln B\approx110.

Unlike J0624, a Hansen particle size distribution is preferred over a log-normal size distribution with ln⁡B=14.4\ln B=14.4. In fact, the Hansen distribution scenario has a higher Bayesian evidence than the species-specific fsedf_{\mathrm{sed}} scenario. In particular, the Hansen distribution scenario can better fit the heart of the solid-state cloud feature and the red edge, compared to the fiducial log-normal particle size distribution. Figure 9 shows that while the Hansen distribution scenario has a similar vertical profile for the clouds, the effective particle sizes are significantly larger than the fiducial log-normal scenario. We hypothesize that the asymmetric nature of the Hansen distribution allows more small particles to fit the cloud feature while retaining some larger particles. The species-specific fsedf_{\mathrm{sed}} allows for larger magnesium-silicate and Fe particles, while keeping the SiO2_2 particles small. The retrieval is likely using the small SiO2_2 particles to fit the MIR silicate feature, while using the other cloud species to adjust the rest of the spectral energy distribution.

J2148 CMA-ES Best-Fits—As with J0624, we ran CMA-ES fits to explore a wider range of possible cloud scenarios. Figure 11 shows the fits to the observations from various best-fit CMA-ES scenarios. Similar to J0624, both SiO2_2 and a magnesium-iron silicate are required to adequately fit the silicate absorption. In J2148’s case, the pyroxene Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3 is preferred slightly over olivine MgFeSiO4_4. The behavior of the fit is similar to J0624, if not even more clear, due to the deeper and sharper silicate feature: without SiO2_2, the blue edge of the silicate feature cannot be fit, while a magnesium-iron silicate is necessary to fit the red edge of the feature.

Figure 11. Top: Fits to the R∼1000R\sim1000 mid-IR spectrum of J2148 (black) using CMA-ES. Our best-fit retrieval (computed at R∼100R\sim100) is shown in red, which included amorphous Fe, SiO2_2, MgSiO3_3, and Mg2_2SiO4_4 using DHS for the opacity computation. We show the best fitting scenarios with and without SiO2_2 and MgSiO3_3. Bottom: Residuals between the observations and each model fit.

Appendix Table 12 ranks the various combinations that were trialed based on their respective χ2/Ndata\chi^{2}/N_{\mathrm{data}} goodness-of-fit. The best-fitting model without SiO2_2 had a χ2/Ndata\chi^{2}/N_{\mathrm{data}} value of 115.17, far worse than the best-fitting model with SiO2_2 at 87.58. Including magnesium silicates was even more important with the best-fitting SiO2_2 + Fe scenario only reaching χ2/Ndata=179.84\chi^{2}/N_{\mathrm{data}} = 179.84. Like with J0624, the worst-fitting scenarios were isolated MgSiO4_4, plus the Fe base. The bottom 13 worst-fitting models all did not contain SiO2_2, highlighting its importance to the fit.

χ2\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
62.215Crys./Mie⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsLog-normal
67.559Crys./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
70.995Crys./Mie⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsHansen
78.660Amorph./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
82.605Crys./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
83.615Amorph./Mie⋯\cdots⋯\cdots⋯\cdotsAmorph./MieHansen
84.379Crys./Mie⋯\cdots⋯\cdots⋯\cdotsAmorph./MieLog-normal
85.891Crist./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
86.028Amorph./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsLog-normal
86.507Trid./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
87.584Crys./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
88.029Crys./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
90.317Crys./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsLog-normal
91.451Amorph./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
92.469Crys./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSHansen
92.790Crist./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
93.209Crys./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
94.413Crist./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
94.784Amorph./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
95.184Crys./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
95.349Trid./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
97.830Crist./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
98.803Amorph./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
99.114Trid./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
99.392Crist./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
100.072Crys./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsHansen
102.674Amorph./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
104.126Amorph./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
104.470Crist./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
104.619Amorph./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
105.102Crys./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
105.809Crys./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
107.536Amorph./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
108.739Crys./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
109.184Crys./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
109.347Amorph./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
111.137Crys./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
111.437Crys./Mie⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsLog-normal
113.354Amorph./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
114.121Amorph./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
115.166⋯\cdots⋯\cdotsCrys./MieAmorph./Mie⋯\cdotsLog-normal
χ2\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
116.485Amorph./Mie⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsLog-normal
117.363Crys./Mie⋯\cdots⋯\cdots⋯\cdotsAmorph./MieHansen
118.091Amorph./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsHansen
119.325⋯\cdots⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsLog-normal
120.224Crist./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
122.587⋯\cdots⋯\cdots⋯\cdotsAmorph./MieAmorph./MieHansen
123.240Trid./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
124.302Crys./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
124.627Trid./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
126.050Crist./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
127.819Crys./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
129.106Amorph./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
129.905Crist./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSHansen
131.024Crys./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
131.130Trid./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
133.840Amorph./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
134.296Amorph./Mie⋯\cdots⋯\cdots⋯\cdotsAmorph./MieLog-normal
134.365⋯\cdots⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsHansen
134.586Trid./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
134.856Trid./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
135.347⋯\cdots⋯\cdotsAmorph./MieAmorph./Mie⋯\cdotsLog-normal
138.579Trid./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
140.463⋯\cdots⋯\cdots⋯\cdotsAmorph./MieAmorph./MieLog-normal
140.774Amorph./Mie⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsHansen
142.144Amorph./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSHansen
147.685Trid./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
147.702⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsAmorph./MieHansen
150.534⋯\cdotsAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
152.248⋯\cdotsAmorph./DHSAmorph./DHS⋯\cdots⋯\cdotsHansen
152.373Trid./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsLog-normal
152.957⋯\cdotsAmorph./DHSCrys./DHS⋯\cdots⋯\cdotsLog-normal
153.325⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsAmorph./MieLog-normal
153.338Crist./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
154.696⋯\cdotsAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
154.872⋯\cdotsAmorph./MieAmorph./Mie⋯\cdots⋯\cdotsLog-normal
156.664⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsAmorph./DHSLog-normal
156.862⋯\cdotsAmorph./DHSAmorph./DHS⋯\cdots⋯\cdotsLog-normal
157.982⋯\cdots⋯\cdotsCrys./MieAmorph./Mie⋯\cdotsHansen
160.093Amorph./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
160.776⋯\cdotsAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
161.500Crys./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
χ2\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
162.181...Amorph./MieAmorph./Mie......Hansen
169.229......Amorph./MieAmorph./Mie...Hansen
169.459Amorph./MieAmorph./Mie.........Hansen
171.833...Amorph./Mie.........Hansen
172.523Trid./DHS......Amorph./DHS...Hansen
176.491Trid./DHS.........Amorph./DHSHansen
176.665Crys./Mie............Log-normal
179.174...Amorph./DHS......Amorph./DHSHansen
179.551Crys./Mie...Crys./Mie......Hansen
179.837Crys./DHS............Log-normal
183.048Amorph./Mie...Crys./Mie......Log-normal
183.222...Amorph./MieCrys./Mie......Hansen
184.128Crys./Mie............Hansen
184.355...Amorph./DHSCrys./DHS......Hansen
188.645...Amorph./MieCrys./Mie......Log-normal
192.797Crys./DHS............Hansen
195.641Crist./DHS............Hansen
206.376Crist./DHS...Amorph./DHS......Log-normal
212.921Amorph./Mie...Crys./Mie......Hansen
214.778Amorph./Mie............Log-normal
215.660Amorph./DHS............Log-normal
216.787Crist./DHSAmorph./DHS.........Hansen
219.783.........Amorph./DHSAmorph./DHSLog-normal
222.494.........Amorph./DHS...Hansen
225.831......Amorph./DHSAmorph./DHS...Log-normal
225.937Amorph./Mie............Hansen
229.182Amorph./DHS............Hansen
232.635Trid./DHS............Hansen
236.479.........Amorph./DHSAmorph./DHSHansen
238.503Crist./DHS............Log-normal
244.593......Crys./DHSAmorph./DHS...Log-normal
250.764.........Amorph./DHS...Log-normal
263.212......Amorph./DHSAmorph./DHS...Hansen
271.914......Crys./DHSAmorph./DHS...Hansen
283.048...Crys./MieCrys./Mie......Log-normal
284.089...Crys./Mie.........Log-normal
284.522...Crys./Mie......Amorph./MieLog-normal
284.613...Crys./MieCrys./Mie......Hansen
286.274...Crys./MieAmorph./Mie......Log-normal
289.366...Crys./MieAmorph./Mie......Hansen
293.927...Crys./Mie.........Hansen
χ2\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
299.543…Crys./Mie……Amorph./MieHansen
320.940…Crys./DHSAmorph./DHS……Log-normal
323.772…Crys./DHS……Amorph./DHSLog-normal
324.471…Crys./DHSCrys./DHS……Log-normal
325.552…Crys./DHS………Log-normal
332.576…Crys./DHS………Hansen
335.634…Crys./DHSCrys./DHS……Hansen
342.593…Crys./DHSAmorph./DHS……Hansen
348.238……Amorph./Mie……Hansen
350.193……Amorph./Mie……Log-normal
371.081…Crys./DHS……Amorph./DHSHansen
445.358……Amorph./DHS……Log-normal
452.032……Amorph./DHS……Hansen
457.398…………Amorph./MieLog-normal
457.770…………Amorph./MieHansen
506.010…………Amorph./DHSLog-normal
506.337…………Amorph./DHSHansen
513.695……Crys./Mie……Log-normal
533.954……Crys./Mie……Hansen
557.006……Crys./DHS……Log-normal
559.020……Crys./DHS……Hansen

Table 12. J2148 Best-Fit Cloud Scenarios with CMA-ES

As with the retrieval analysis, the best-fit scenario for J2148 included quartz SiO2_2. Amorphous SiO2_2 fit significantly worse with χ2/Ndata=78.66\chi^{2}/N_{\mathrm{data}} = 78.66 versus the best-fit at 62.22. Interestingly, cristobalite and tridymite is preferred thermodynamically [73], yet each fit worse than quartz at χ2/Ndata=85.89\chi^{2}/N_{\mathrm{data}} = 85.89 and 86.507, respectively.

J2148 Bulk Properties

We also note that J2148 showed a similar metallicity and C/O ratio enhancement as J0624. Specifically, the metallicity ranges from [M/H] = 0.272 to 0.533 and the C/O ratio varies between 0.6143 and 0.6604 across all scenarios tested. While the C/O ratio is close enough to solar to be explained by primordial abundance variations, the enhanced metallicity in both objects is somewhat is surprising. We discuss possible reasons for this elevated metallicity in § 6.2.

In a similar sense, the retrieved surface gravity appear to be somewhat low compared to what we would expect for a field brown dwarf. With log⁡g\log g ranging from 4.01 to 4.55 and radii between 1.04 and 1.15 RJupR_{\mathrm{Jup}}, our retrievals imply a mass for J2148 between 5 and 17 MJupM_{\mathrm{Jup}}. Like J0624, this is compared to the inference from Filippazzo et al. [37] of log⁡g≈5.13\log g \approx5.13 and ≈50MJup\approx50M_{\mathrm{Jup}}. It is not clear what in the observed spectrum is driving the retrieval to prefer larger radii, lower surface gravity, and lower masses, but these biases have commonly appeared for retrievals of self-luminous objects, which we discuss in § 6.2. We further interpret these bulk properties in comparison to brown dwarf evolution tracks in § 6.2.3.

J2148 TP Profile

Lastly, we find the same steep TP profile for J2148 compared to the best-fit self-consistent 1D-RCE models as we did for J0624 as shown in Figure 8. This suggests a common reason (or bias) in the retrieval parameterization for a steep profile in cloudy L-dwarfs. While J2148 more closely matches the 1D-RCE models at pressures above 1 bar, the retrieved profile becomes significantly steeper at pressures below 1 bar. Further investigation of the TP profile with additional retrieval parameterizations is likely necessary to robustly assess the disagreement between retrieval and self-consistent models.

Discussion

Cloud Properties

Our observing program (JWST-GO-2288) originally selected J0624 and J2148 due to their close proximity (∼10\sim10 pc), their similar spectral type (L5 and L6, respectively), and their seemingly different colors: J2148 appears to be a redder, slightly later-type object compared to J0624. Our hypothesis was that J2148 should be cloudier and that by comparing their entire infrared emission spectra, we could determine if this “cloudiness” corresponded with changes in particle composition, size, crystallinity, or vertical distribution.

Figure 12 compares the posterior distribution for our fiducial retrievals of each object. Our retrieval results broadly support the picture that J2148 is qualitatively cloudier than J0624 as illustrated in Table 7-Table 10 and Figure 9. While constraining the total cloud mass is difficult, J2148 appears to have more lofted clouds, made up of smaller particles. The KzzK_{zz} measurements for J2148 are generally somewhat higher than J0624, but the most consequential difference is the lower fsedf_{\mathrm{sed}} value for J2148. Values for fsedf_{\mathrm{sed}} are almost uniformly below 1.0 for J2148, yet are above 1.0 for J0624, reaching 4.55±0.634.55 \pm0.63 for the best-fit amorphous SiO2\mathrm{SiO}_{2} retrieval. This picture of a lower fsedf_{\mathrm{sed}} for J2148 compared to J0624 is further supported by the Sonora Diamondback grid comparisons shown in Figure 5. This difference in mixing and sedimentation may have tangible effects on the crystallinity of the particles with the amorphous particles in J0624 falling out of the atmosphere before they can anneal into

Figure 12. Posterior distributions from our fiducial retrievals of J0624 (top labels, blue) and J2148 (bottom labels, orange).

a crystalline structure ([31, 100, 57, 73]).

Future observations of other L-dwarfs may shed on the physical processes causing these differences. For example, J2148 is known to be observed equator-on ([105, 99]), while J0624 is unconstrained and most likely observed at a mid-latitude. Thus, latitudinal differences in atmospheric circulation may be consequential for cloud formation and microphysics. Other JWST programs (e.g., JWST-GO-5474) are observing more cloudy L-dwarfs with differing inclinations and surface gravities, potentially revealing population trends.

As for cloud composition, a picture is already emerging that more simple silicate minerals may be present in substellar atmospheres than previously thought. Early kinetic and chemical equilibrium theoretical studies indicated that iron- and magnesium silicates were most likely to form in the atmospheres of brown dwarfs and exoplanets (e.g., [104]), though some studies had noted SiO2_2 as an important constituent ([45]). Yet, early observations with JWST have revealed a different picture. MIRI/LRS transmission spectroscopy measurements of hot Jupiter WASP-17b were best fit by a SiO2_2 cloud ([40]). As mentioned above, planetary mass PSO-513 appears to be best-fit by a SiO cloud.

Our observations agree with this picture; both our retrievals and CMA-ES fits of J0624 and J2148 found strong evidence for SiO2_2 clouds. In both cases, the presence of magnesium silicates was not justified in the re-

trievals according to the Bayesian evidence, but iron- and magnesium-rich pyroxene were strongly favored in the CMA-ES optimization framework. Meanwhile, the retrieval found evidence for Fe clouds in J2148 and Al₂O₃ clouds in J0624 (though this is likely just a preference for a deep, gray cloud deck). Together, this implies multiple cloud species and layers are likely present, in agreement with previous analyses ([18]). Future modeling and observation can help explore whether the preference for simple silicates reflects a elemental abundance variation ([21]) or is microphysical in origin.

The suggested presence of iron-rich silicates in both objects from the CMA-ES optimization is intriguing because such species have previously been unexpected due to the initial expected condensation of Fe into metal and the low oxygen fugacity of substellar atmospheres ([30, 104]). While such a detection would be novel and would challenge current models of cloud formation, we note that the precise form of iron and magnesium silicate is not unambiguously constrained despite our high-SNR observations, with small differences between different cloud scenarios. Perhaps non-solar Mg, Si, and Fe abundances ([21]) or microphysical details may result in a wider set of of observed cloud species than expected. We suggest future model comparisons of cloudy L-dwarf spectra fit a wide variety of cloud scenarios and species.

Another consideration is whether either brown dwarf we have observed is variable in time. Here, we present a single snapshot of each object, but other cloudy objects have shown significant variability through complex interactions of the atmospheric temperature, chemistry, and clouds [15, 78, 80]. Future time-series observations of these two objects would reveal whether the cloud properties we measure here are stable with time or whether these worlds are more dynamic. A measurement of J0624’s inclination via its rotation rate and vsin⁡iv \sin i would also be helpful for interpreting the difference in clouds between J0624 and J2148.

Bulk Properties

Our analysis highlights the fact that even with unprecedented wavelength coverage and signal-to-noise, the inference of a substellar object’s bulk properties, like radius, mass, structure, and composition, remains a difficult problem. Here, we chose to focus on the identification of the cloud properties and were unable to explore the full range of possible bulk property parameterizations. Most glaringly, our desire to calculate many retrieval scenarios across the full wavelength regime in an efficient manner dictated a reduction in resolving power to R∼100R\sim100, while our observations were taken at much higher spectral resolution (R∼3,000R\sim3{,}000). In future work, we hope to expand on the analysis of L-dwarf bulk properties now that statistically-justified cloud property scenarios have been identified.

Metallicity

Our measurement of super-solar metallicity in both J0624 and J2148 is surprising. While the intrinsic scatter of stellar metallicities does reach up to [M/H]≈0.5[M/H]\approx0.5 for young objects [36], it seems unlikely that two separate field dwarfs would exhibit metallicities around this upper limit by chance. We note that retrieval analyses of cloudy L-dwarf J2224 also indicated a super-solar metallicity of [M/H]=0.38−0.06+0.07[M/H]=0.38^{+0.07}_{-0.06} and C/O =0.83−0.07+0.06=0.83^{+0.06}_{-0.07} [18]. Retrievals of other substellar companions also show elevated metallicities approaching and sometimes exceeding exceeding 10×10\times solar, including HR 8799 c, d, and e (Nasedkin et al. 2024b, but see Xuan et al. 2026), HD 72946 B [10], and AF Lep b (Balmer et al. 2025).

Free retrievals of less-cloudy T-dwarfs do not appear to show any significant metallicity enhancement [113], suggesting either a bias in our retrievals of cloudy brown dwarfs or a bias in retrieval with chemical equilibrium due to disequilibrium processes remaining relevant to L-dwarfs. While we do include a quench pressure parameterization to account for disequilibrium chemistry, only the J0624 retrievals found a PquenchP_{\rm quench} that was not above the IR photosphere. This was likely used by the retrieval to help fit the CH4_4 qqbranch absorption at 3.3 μm\mu\mathrm{m} seen in J0624. More physically motivated disequilibrium chemistry treatments (i.e., using a time-scale based setup) may be an avenue to more robustly constraining the bulk composition of these objects.

An additional source of bias in the composition constraints may be the choice of temperature profile parameterization: mock L-dwarf retrievals of cloudy L-dwarfs indicated that physically motivated temperature profile parameterizations can bias metallicity measurements by over 0.1 dex compared to more flexible layer-by-layer parameterizations [107, 91]. The retrieved TP profiles for both J0624 and J2148 are much steeper than the 1D-RCE models (see § 5.1.3 and § 5.2.3), so it is possible that the retrieval increases the molecular abundances to produce features of the right strength to compensate. Future sensitivity analyses may shed light on this issue. One last possible bias affecting the metallicity measurement may be with the cloud parameterization ([10] (Balmer et al. 2025)). The lack of flexibility in setting the particle size and its distribution may be requiring the retrieval to adjust the atmospheric opacity by increasing [M/H] to match the data. This may be especially true in our data, where the retrieval may have have to choose between fitting the MIR silicate feature with small particles and having the correct opacity at the photosphere through large particles.

Gravity, Radius, and Mass

Lastly, we consider our retrieved surface gravity, radius, and mass. Retrievals of self-luminous objects have historically run into issues with fitting radius; the retrieval will often use the radius as a dilution factor to adjust the effective temperature up or down. This can easily lead to unphysical retrieval results.

For example, retrievals of cloudy L-dwarf J2224 ([18]) found a radius of 0.75±0.020.75 \pm0.02, which is significantly lower than the 0.99±0.080.99 \pm0.08 inferred from the planets distance and photometry ([37]). Our retrievals of J0624 find a radius about 11–2σ2\sigma smaller than in [37], while the radii retrieved for J2148 generally agree within 1σ1\sigma.

A bigger discrepancy is found in the retrieved surface gravity, which also results in unexpectedly low masses. For J0624, the surface gravity is about 0.30.3–0.50.5 dex lower than expected from a normal field dwarf or from [37]. For J2148, the discrepancy is worse, at about 0.70.7–0.80.8 dex in surface gravity. For both objects, this results in masses between 8 and 23 MJupM_{\rm Jup}, depending on the retrieval scenario. Below, we evaluate these results compared to evolution models and our empirically-measured flux.

Evolution Analysis of J0624 and J2148

With our retrieved radii for J0624 and J2148, we can use the spectrum to calculate a semi-empirical effective temperature. Using these two quantities, we can then compare to evolutionary models to assess the reasonableness of our retrieved surface gravity and mass constraints, as well as to estimate an approximate age. We first integrate the measured flux from our observations between 0.98 and 17.6 μ\mum. This flux is then scaled by the distance modulus, taking the radius of the preferred retrieval scenario for each object (RJup=0.915±0.004R_{\rm Jup} = 0.915 \pm0.004 and 1.05±0.0041.05 \pm0.004 for J0624 and J2148, respectively) and the Gaia distance (12.191 ±0.054\pm0.054 and 8.086 ±0.023\pm0.023 pc, respectively). From the Stefan-Boltzman law, we obtain a preliminary effective temperature of 1525.5 and 1438.2 K, respectively. We then use this preliminary temperature to calculate a bolometric correction to find the amount of flux estimated to be outside our observed bandpass by integrating a blackbody at our preliminary temperatures between 0.98 and 17.6 μ\mum. We estimate that we measure 98.1% and 98.4% of the flux from each object in our bandpass. We use this factor to estimate a final bolometric flux from which we measure an empirical effective temperature of 1532.8 ±\pm 7.3 K and 1443.9 ±\pm 5.7 K for J0624 and J2148, respectively, where the measurement uncertainty is taken to be the difference between the preliminary and bolometrically-corrected temperature.

With a measured radius and temperature, we can now compare to brown dwarf evolution tracks to constrain possible ages and masses for the objects. Crucially, this will not depend on the potentially biased surface gravity estimated from the retrieval, though it still depends on the inferred radius. We use the Red Diamondback Sonora substellar atmosphere grid ([28]) to find a best matching isochrone in radius versus effective temperature space, shown in Figure 13. We perform this calculation for both the solar metallicity models and, to match our retrieval results, the [M/H]=0.5[M/H] = 0.5 elevated metallicity models.

Figure 13. Brown dwarf evolution tracks from Davis et al. [28] compared to the radii of J0624 and J2148 from the best-fitting retrieval with effective temperature then measured from the flux from each object. Solid lines show tracks for objects with solar metallicity, while dashed lines show tracks for elevated metallicity objects.

J0624 lies between two solar metallicity isochrones at 0.06MSun=62.9MJup0.06M_{\rm Sun} = 62.9M_{\rm Jup} and 0.065MSun=68.1MJup0.065M_{\rm Sun} = 68.1M_{\rm Jup}, indicating an age between 1.2 and 1.9 Gyr and log⁡g\log g between 5.26 and 5.32 – near the values found in [37]. No elevated metallicity track is near J0624, as the 0.07MSun=73.3MJup0.07M_{\rm Sun} = 73.3M_{\rm Jup} track ends at a temperature of 1700 K, but J0624’s location above the 0.065MSun0.065M_{\rm Sun} would imply and age greater than 2.2 Gyr and log⁡g>5.28\log g > 5.28.

J2148 lies within 1σ1\sigma of the 0.04MSun=41.9MJup0.04M_{\rm Sun} = 41.9M_{\rm Jup} solar metallicity track, corresponding to an age of 0.42 Gyr and log⁡g=4.98\log g = 4.98. For the elevated metallicity isochrones, J2148 lies between the 0.045MSun=47.1MJup0.045M_{\rm Sun} = 47.1M_{\rm Jup} and 0.05MSun=52.4MJup0.05M_{\rm Sun} = 52.4M_{\rm Jup} tracks, implying an age between 0.57 and 0.77 Gyr and log⁡g\log g between 5.01 and 5.09. The inferred surface gravity from the isochrones and retrieved radius imply a much higher surface gravity than obtained in the retrievals by more than half a dex. If the surface gravity of J0624 and J2148 were truly as low as the retrievals imply, we would expect them to be very young and much larger if at the same effective temperature.

Conclusion

We have presented a modeling analysis of 1-28μ\mum observations of two cloudy L-dwarfs, J0624 and J2148. We first compared our observations to grids of 1-D radiative-convective equilibrium models from BT-Settl, Sonora Diamondback, and Exo-REM. While these self-consistent models did not have the necessary flexibility to fit our high-signal to noise observations, reasonable effective temperatures and fsedf_{\mathrm{sed}} parameters were found. Compositional information in the form of [M/H] and C/O, along with surface gravity, were poorly constrained by the self-consistent models.

We next presented a variety of 1-D atmospheric retrievals and CMA-ES optimization analyses, focusing on testing different cloud property scenarios. Our findings indicate that both J0624 and J2148 have SiO2_2 clouds. For J2148, these clouds appear to be lofted, sub-μ\mum crystalline SiO2_2 particles. Both the retrieval and CMA-ES optimization prefer quartz particles compared to tridymite or cristobalite polymorphs. For J0624, the observations were better fit by slightly larger, more vertically-compact amorphous SiO2_2 particles. The CMA-ES also showed strong evidence for amorphous iron-rich magnesium silicates, with both Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3 and MgFeSiO4_4 fitting well for both objects. A comparison to more L-dwarfs and other sub-stellar objects may shed light on these compositional and microphysical differences.

Lastly, we discuss other aspects of the retrieval, including our measurement of elevated metallicity and low surface-gravity in both objects. We interpret these measurements in comparison to other brown dwarf measurements, and suggest that remaining retrieval biases may need to be mitigated for more robust bulk property inferences. We also compare our retrieved bulk properties to evolution models, finding reasonable constraints on age and mass if we use the empirically measured flux and our retrieved radii but ignore the retrieved surface gravity, which appear to biased low in both objects.

In future works, we plan to connect J0624 and J2148 to other substellar objects through empirical comparisons. While this work focused specifically on the cloud properties of the objects, we also plan to also understand the bulk properties of J0624 and J2148, like metallicity and the TP structure, compared to the broader population of brown dwarfs through a more comprehensive set of retrievals, specifically testing many of the parameterizations used in this work. We also plan to leverage the high-resolution of the JWST data to measure the isotope ratios (e.g., [39]).

As mentioned above, further observation of L-dwarfs will help establish trends with temperature, observed latitude, and mass, while also defining the expected intrinsic diversity we may expect from field brown dwarfs. Already, JWST has shown a surprising variety among brown dwarfs in the rich spectra already observed (e.g., [68, 32, 15, 33, 72, 78, 80]). Our observations here highlight the remaining open-questions in our understanding of cloudy L-dwarfs, while simultaneously demonstrating the explanatory power of wide-wavelength JWST observations.

Acknowledgements

We thank the referee for a constructive and helpful review that improved the manuscript. We thank Everett Nasedkin for sharing quartz and amorphous SiO2_2 cloud opacities. We thank Jacqueline Faherty for continual encouragement towards publication of these results. This work is based on observations made with the NASA/ESA/CSA James Webb Space Telescope. All of the data presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute. The specific observations analyzed can be accessed via https://doi.org/10.17909/xa61-yf45. STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5–26555. Support to MAST for these data is provided by the NASA Office of Space Science via grant NAG5–7584 and by other grants and contracts. These observations are associated with program JWST-GO-2288. Support for program JWST-GO-2288 was provided by NASA through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-03127. This research has made use of the VizieR catalog access tool, CDS, Strasbourg, France. The original description of the VizieR service was published in Ochsenbein et al. [79]. This research has made use of NASA’s Astrophysics Data System Bibliographic Services.

Facilities: JWST(NIRSpec, MIRI)

Software: astropy (Astropy Collaboration et al. [8]), matplotlib (Hunter [48]), NumPy (Harris et al. [44]), iPython (Perez & Granger [82]), petitRADTRANS (Mollière et al. [70]; Nasedkin et al. [76])

Cloud Scenario Best Fits with CMA-ES

χ2\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
50.737Amorph./Mie………Amorph./MieHansen
51.387Amorph./Mie……Amorph./Mie…Hansen
52.540Amorph./DHS……Amorph./DHS…Log-normal
52.885Amorph./Mie……Amorph./Mie…Log-normal
53.604Crys./DHSCrys./DHS………Log-normal
53.756Amorph./DHS………Amorph./DHSHansen
54.562Amorph./DHS………Amorph./DHSLog-normal
55.229Amorph./DHSCrys./DHS………Hansen
55.250………Amorph./MieAmorph./MieLog-normal
55.311Crys./DHS………Amorph./DHSLog-normal
55.424Amorph./Mie…Amorph./Mie……Log-normal
55.659Crys./DHSCrys./DHS………Hansen
55.765Crys./DHS………Amorph./DHSHansen
56.011Crys./DHS……Amorph./DHS…Hansen
56.031Amorph./DHS…Crys./DHS……Log-normal
56.641Crys./DHS……Amorph./DHS…Log-normal
56.685Crist./DHS………Amorph./DHSHansen
56.728Amorph./Mie………Amorph./MieLog-normal
56.844Crys./Mie……Amorph./Mie…Hansen
57.022Amorph./DHS…Amorph./DHS……Log-normal
57.070………Amorph./Mie…Hansen
57.448……Crys./MieAmorph./Mie…Log-normal
57.462………Amorph./Mie…Log-normal
57.612Amorph./Mie…Crys./Mie……Hansen
57.752Crys./Mie………Amorph./MieHansen
57.945……Crys./MieAmorph./Mie…Hansen
58.180Amorph./Mie…………Log-normal
58.255Crys./MieCrys./Mie………Hansen
58.730Amorph./DHS…Crys./DHS……Hansen
χ2/N\chi^2/NSiO2\mathrm{SiO}_2MgSiO3\mathrm{MgSiO}_3Mg2SiO4\mathrm{Mg}_2\mathrm{SiO}_4Mg0.5Fe0.5SiO3\mathrm{Mg}_{0.5}\mathrm{Fe}_{0.5}\mathrm{SiO}_3MgFeSiO4\mathrm{MgFeSiO}_4Size Dist.
59.210Amorph./Mie⋯\cdots⋯\cdots⋯\cdots⋯\cdotsHansen
59.232Amorph./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsHansen
59.300Crys./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
59.537Crys./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
59.855⋯\cdots⋯\cdots⋯\cdotsAmorph./MieAmorph./MieHansen
60.300Trid./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
60.350⋯\cdots⋯\cdotsAmorph./MieAmorph./Mie⋯\cdotsHansen
60.546Amorph./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
60.687Amorph./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
60.738Crys./Mie⋯\cdots⋯\cdots⋯\cdotsAmorph./MieLog-normal
60.743Crist./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
60.763Crys./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
60.806Trid./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
60.841Amorph./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
61.822Trid./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
61.894Crist./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
61.964Amorph./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsLog-normal
62.016Amorph./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsHansen
62.057Amorph./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
62.133Trid./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
62.457Crys./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsHansen
62.715Amorph./Mie⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsLog-normal
62.739Trid./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
62.810Crys./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
62.847Trid./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
62.936Trid./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
64.123Crist./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
64.468Trid./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsLog-normal
64.650Amorph./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
64.894Crist./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsLog-normal
65.238Crist./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsHansen
65.327Crys./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsHansen
65.663Trid./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
65.702Trid./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
65.950Amorph./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
65.973Crist./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
66.095Crys./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsLog-normal
66.258Crist./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
66.260Crist./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
66.269Trid./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
66.488Crist./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
χ2\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
67.019⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsAmorph./MieLog-normal
67.370Crys./Mie⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsHansen
67.677⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsAmorph./MieHansen
68.040Trid./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSHansen
68.377⋯\cdotsAmorph./MieCrys./Mie⋯\cdots⋯\cdotsLog-normal
68.482Trid./DHS⋯\cdots⋯\cdots⋯\cdots⋯\cdotsHansen
68.584Crist./DHSCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
68.970Crys./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
68.977⋯\cdotsAmorph./DHSCrys./DHS⋯\cdots⋯\cdotsLog-normal
69.097⋯\cdotsAmorph./MieAmorph./Mie⋯\cdots⋯\cdotsLog-normal
69.173⋯\cdotsAmorph./MieCrys./Mie⋯\cdots⋯\cdotsHansen
69.275Crys./Mie⋯\cdots⋯\cdots⋯\cdots⋯\cdotsLog-normal
69.622Crys./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
70.148⋯\cdotsAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
70.174Amorph./MieAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
70.468Crist./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
70.622⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsAmorph./DHSLog-normal
70.626⋯\cdotsAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
70.886⋯\cdotsAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
71.199⋯\cdotsAmorph./DHSAmorph./DHS⋯\cdots⋯\cdotsHansen
72.365Crist./DHS⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
72.545Crys./Mie⋯\cdots⋯\cdots⋯\cdots⋯\cdotsHansen
72.860⋯\cdotsAmorph./MieAmorph./Mie⋯\cdots⋯\cdotsHansen
72.898Amorph./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
72.936⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSAmorph./DHSHansen
73.077Crys./Mie⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsLog-normal
73.195⋯\cdots⋯\cdotsCrys./DHSAmorph./DHS⋯\cdotsLog-normal
73.500⋯\cdots⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsHansen
73.642⋯\cdots⋯\cdotsAmorph./DHSAmorph./DHS⋯\cdotsHansen
73.683⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsAmorph./DHSHansen
73.981Crys./Mie⋯\cdots⋯\cdotsAmorph./Mie⋯\cdotsLog-normal
74.421Crys./MieCrys./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
74.721⋯\cdotsAmorph./DHSAmorph./DHS⋯\cdots⋯\cdotsLog-normal
75.023Crys./Mie⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsLog-normal
75.484⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSAmorph./DHSLog-normal
75.514⋯\cdots⋯\cdotsCrys./DHSAmorph./DHS⋯\cdotsHansen
75.910⋯\cdotsAmorph./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
77.093⋯\cdots⋯\cdots⋯\cdotsAmorph./DHS⋯\cdotsLog-normal
77.235Crist./DHS⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
79.859Trid./DHS⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
84.127⋯\cdots⋯\cdotsAmorph./MieAmorph./Mie⋯\cdotsLog-normal
χ2/N\chi^2/NSiO2_2MgSiO3_3Mg2_2SiO4_4Mg0.5_{0.5}Fe0.5_{0.5}SiO3_3MgFeSiO4_4Size Dist.
85.124⋯\cdots⋯\cdotsAmorph./DHSAmorph./DHS⋯\cdotsLog-normal
95.541Crys./DHSAmorph./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
95.790Amorph./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
97.610Crys./DHS⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
102.093⋯\cdotsCrys./MieCrys./Mie⋯\cdots⋯\cdotsLog-normal
102.274⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsAmorph./MieHansen
103.003⋯\cdotsCrys./Mie⋯\cdots⋯\cdots⋯\cdotsLog-normal
104.257⋯\cdots⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsLog-normal
104.867⋯\cdots⋯\cdotsAmorph./Mie⋯\cdots⋯\cdotsHansen
104.951⋯\cdotsCrys./MieAmorph./Mie⋯\cdots⋯\cdotsLog-normal
105.702⋯\cdotsCrys./MieAmorph./Mie⋯\cdots⋯\cdotsHansen
106.602⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsAmorph./MieLog-normal
108.333⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsAmorph./DHSLog-normal
109.525⋯\cdotsCrys./DHS⋯\cdots⋯\cdots⋯\cdotsLog-normal
110.293⋯\cdotsCrys./Mie⋯\cdots⋯\cdots⋯\cdotsHansen
110.617⋯\cdots⋯\cdots⋯\cdots⋯\cdotsAmorph./MieHansen
111.228⋯\cdotsCrys./DHSCrys./DHS⋯\cdots⋯\cdotsHansen
111.843⋯\cdots⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSHansen
112.110⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsAmorph./DHSHansen
113.496⋯\cdots⋯\cdots⋯\cdots⋯\cdotsAmorph./DHSLog-normal
115.284⋯\cdots⋯\cdots⋯\cdots⋯\cdotsAmorph./MieLog-normal
115.706⋯\cdotsCrys./DHSCrys./DHS⋯\cdots⋯\cdotsLog-normal
116.000⋯\cdotsCrys./DHSAmorph./DHS⋯\cdots⋯\cdotsLog-normal
118.428⋯\cdotsAmorph./DHSCrys./DHS⋯\cdots⋯\cdotsHansen
122.498⋯\cdotsCrys./DHS⋯\cdots⋯\cdots⋯\cdotsHansen
123.261⋯\cdotsCrys./MieCrys./Mie⋯\cdots⋯\cdotsHansen
125.609⋯\cdots⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsLog-normal
132.254⋯\cdots⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsLog-normal
135.958⋯\cdots⋯\cdotsAmorph./DHS⋯\cdots⋯\cdotsHansen
137.260⋯\cdotsCrys./DHSAmorph./DHS⋯\cdots⋯\cdotsHansen
140.005⋯\cdots⋯\cdotsCrys./DHS⋯\cdots⋯\cdotsHansen
155.167⋯\cdots⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsLog-normal
161.814⋯\cdots⋯\cdotsCrys./Mie⋯\cdots⋯\cdotsHansen

Table 11. Table 11 continued

**Note—**The size distribution column applies to all cloud species in that model. Each dust species column gives crystallinity/opacity treatment. SiO2_2 crystallinity: Qtz.=Quartz, Trid.=Tridymite, Crist.=Cristobalite, Amorph.=Amorphous. Other species: Crys.=Crystalline, Amorph.=Amorphous. Opacity: Pure Mie or DHS (Distribution of Hollow Spheres) [69].

References

  1. [1]Ackerman, A. S., & Marley, M. S. 2001, ApJ, 556, 872, doi: 10.1086/321540DOI
  2. [2]Allard, F., Hauschildt, P. H., Alexander, D. R., Tamanai, A., & Schweitzer, A. 2001, ApJ, 556, 357, doi: 10.1086/321547DOI
  3. [5]Apai, D., Radigan, J., Buenzli, E., et al. 2013, ApJ, 768, 121, doi: 10.1088/0004-637X/768/2/121DOI
  4. [6]Asplund, M., Amarsi, A. M., & Grevesse, N. 2021, A&A, 653, A141, doi: 10.1051/0004-6361/202140445DOI
  5. [7]Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P. 2009, ARA&A, 47, 481, doi: 10.1146/annurev.astro.46.060407.145222DOI
  6. [8]Astropy Collaboration, Price-Whelan, A. M., Sipőcz, B. M., et al. 2018, AJ, 156, 123, doi: 10.3847/1538-3881/aabc4f
  7. [9]Azzam, A. A. A., Tennyson, J., Yurchenko, S. N., & Naumenko, O. V. 2016, MNRAS, 460, 4063, doi: 10.1093/mnras/stw1133DOI
  8. [10]Balmer, W. O., Pueyo, L., Stolker, T., et al. 2023, ApJ, 956, 99, doi: 10.3847/1538-4357/acf761 Balmer, W. O., Franson, K., Chomez, A., et al. 2025, AJ, 169, 30, doi: 10.3847/1538-3881/ad9265DOI
  9. [11]Barber, R. J., Strange, J. K., Hill, C., et al. 2014, MNRAS, 437, 1828, doi: 10.1093/mnras/stt2011DOI
  10. [12]Barber, S., Metchev, S., Suárez, G., et al. 2026, MNRAS, 550, stag1387, doi: 10.1093/mnras/stag1387DOI
  11. [13]Barton, E. J., Yurchenko, S. N., & Tennyson, J. 2013, MNRAS, 434, 1469, doi: 10.1093/mnras/stt1105DOI
  12. [14]Baudino, J.-L., Bézard, B., Charnay, B., & Dorianin, B. 2024, Exo-REM: 1D self-consistent radiative-equilibrium model for exoplanetary atmospheres, Astrophysics Source Code Library, record ascl:2410.012
  13. [15]Biller, B. A., Vos, J. M., Zhou, Y., et al. 2024, MNRAS, 532, 2207, doi: 10.1093/mnras/stael602
  14. [16]Borysow, A. 2002, A&A, 390, 779, doi: 10.1051/0004-6361:20020555DOI
  15. [17]Borysow, A., & Frommhold, L. 1986, ApJ, 304, 849, doi: 10.1086/164221DOI
  16. [18]Burningham, B., Faherty, J. K., Gonzales, E. C., et al. 2021, MNRAS, doi: 10.1093/mnras/stab1361DOI
  17. [20]Bushouse, H., Eisenhamer, J., Dencheva, N., et al. 2024, JWST Calibration Pipeline, 1.16.1, Zenodo, doi: 10.5281/zenodo.14153298
  18. [21]Calamari, E., Faherty, J. K., Visscher, C., et al. 2024, ApJ, 963, 67, doi: 10.3847/1538-4357/adf16d
  19. [22]Charnay, B. 2021, Exo-REM YGP Model Grids, LESIA, Observatoire de Paris. https://lesia.obspm.fr/exorem/YGP_grids/old_grids_2021/
  20. [23]Charnay, B., Bézard, B., Baudino, J.-L., et al. 2018, ApJ, 854, 172, doi: 10.3847/1538-4357/aaaс7d
  21. [24]Coles, P. A., Yurchenko, S. N., & Tennyson, J. 2019, MNRAS, 490, 4638, doi: 10.1093/mnras/stz2778DOI
  22. [25]Cushing, M. C., Rayner, J. T., & Vacca, W. D. 2005, ApJ, 623, 1115, doi: 10.1086/428040DOI
  23. [26]Cushing, M. C., Roellig, T. L., Marley, M. S., et al. 2006, ApJ, 648, 614, doi: 10.1086/505637DOI
  24. [27]Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, 2MASS All Sky Catalog of point sources.
  25. [28]Davis, C. E., Fortney, J. J., Iyer, A., et al. 2025, ApJ, 994, 198, doi: 10.3847/1538-4357/ae1015DOI
  26. [29]Dorschner, J., Begemann, B., Henning, T., Jaeger, C., & Mutschke, H. 1995, A&A, 300, 503
  27. [30]Ebel, D. S. 2006, in Meteorites and the Early Solar System II, ed. D. S. Lauretta & H. Y. McSween, 253
  28. [31]Fabian, D., Jäger, C., Henning, T., Dorschner, J., & Mutschke, H. 2000, A&A, 364, 282
  29. [32]Faherty, J. K., Burningham, B., Gagné, J., et al. 2024, Nature, 628, 511, doi: 10.1038/s41586-024-07190-wDOI
  30. [33]Faherty, J. K., Meisner, A. M., Burningham, B., et al. 2025, Nature, 645, 62, doi: 10.1038/s41586-025-09369-1DOI
  31. [34]Feroz, F., & Hobson, M. P. 2008, MNRAS, 384, 449, doi: 10.1111/j.1365-2966.2007.12353.xDOI
  32. [35]Feroz, F., Hobson, M. P., & Bridges, M. 2009, MNRAS, 398, 1601, doi: 10.1111/j.1365-2966.2009.14548.xDOI
  33. [36]Feuillet, D. K., Frankel, N., Lind, K., et al. 2019, MNRAS, 489, 1742, doi: 10.1093/mnras/stz2221DOI
  34. [37]Filippazzo, J. C., Rice, E. L., Faherty, J., et al. 2015, ApJ, 810, 158, doi: 10.1088/0004-637X/810/2/158DOI
  35. [38]Gaia Collaboration, Vallenari, A., Brown, A. G. A., et al. 2023, A&A, 674, A1, doi: 10.1051/0004-6361/202243940
  36. [39]Gandhi, S., de Regt, S., Snellen, I., et al. 2023, ApJL, 957, L36, doi: 10.3847/2041-8213/ad07e2DOI
  37. [40]Grant, D., Lewis, N. K., Wakeford, H. R., et al. 2023, ApJL, 956, L29, doi: 10.3847/2041-8213/acfc3bDOI
  38. [41]Hansen, J. E. 1971, Journal of the Atmospheric Sciences, 28, 1400, doi: 10.1175/1520-0469(1971)028<1400:MSOPLI>2.0.CO;2
  39. [42]Hansen, N., Akimoto, Y., & Baudis, P. 2019, CMA-ES/pycma on Github, Zenodo, DOI:10.5281/zenodo.2559634, doi: 10.5281/zenodo.2559634
  40. [43]Hansen, N., & Ostermeier, A. 2001, Evolutionary Computation, 9, 159, doi: 10.1162/106365601750190398
  41. [44]Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357–362, doi: 10.1038/s41586-020-2649-2DOI
  42. [45]Helling, C., & Woitke, P. 2006, A&A, 455, 325, doi: 10.1051/0004-6361:20054598DOI
  43. [46]Henning, T., & Mutschke, H. 1997, A&A, 327, 743
  44. [47]Henning, T., & Stognienko, R. 1996, A&A, 311, 291
  45. [48]Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, doi: 10.1109/MCSE.2007.55
  46. [49]Jaeger, C., Molster, F. J., Dorschner, J., et al. 1998, A&A, 339, 904
  47. [50]Jaeger, C., Mutschke, H., Begemann, B., Dorschner, J., & Henning, T. 1994, A&A, 292, 641
  48. [51]Jäger, C., Il’in, V. B., Henning, T., et al. 2003, JQSRT, 79-80, 765, doi: 10.1016/S0022-4073(02)00301-1DOI
  49. [52]Jakobsen, P., Ferruit, P., Alves de Oliveira, C., et al. 2022, A&A, 661, A80, doi: 10.1051/0004-6361/202124663
  50. [53]Kirkpatrick, J. D., Looper, D. L., Burgasser, A. J., et al. 2010, ApJS, 190, 100, doi: 10.1088/0067-0049/190/1/100DOI
  51. [54]Kitzmann, D., & Heng, K. 2018, MNRAS, 475, 94, doi: 10.1093/mnras/stx3141DOI
  52. [55]Kitzmann, D., Heng, K., Rimmer, P. B., et al. 2018, ApJ, 863, 183, doi: 10.3847/1538-4357/aace5aDOI
  53. [56]Koike, C., Kaito, C., Yamamoto, T., et al. 1995, Icarus, 114, 203, doi: 10.1006/icar.1995.1055DOI
  54. [57]Koike, C., Noguchi, R., Chihara, H., et al. 2013, ApJ, 778, 60, doi: 10.1088/0004-637X/778/1/60DOI
  55. [58]Leconte, J. 2018, ApJL, 853, L30, doi: 10.3847/2041-8213/aaaa61DOI
  56. [60]Lodders, K. 2010, in Astrophysics and Space Science Proceedings, Vol. 16, Principles and Perspectives in Cosmochemistry, ed. A. Goswami & B. E. Reddy, 379, doi: 10.1007/978-3-642-10352-0_8
  57. [61]Looper, D. L., Gelino, C. R., Burgasser, A. J., & Kirkpatrick, J. D. 2008, ApJ, 685, 1183, doi: 10.1086/590382DOI
  58. [62]Lothringer, J. D., Valenti, J. A., Gao, P., et al. 2021, Formation and impact of silicate clouds on L dwarfs, JWST Proposal. Cycle 1, ID. #2288
  59. [63]Luna, J. L., & Morley, C. V. 2021, ApJ, 920, 146, doi: 10.3847/1538-4357/ac1865DOI
  60. [64]Manjavacas, E., Apai, D., Zhou, Y., et al. 2019, AJ, 157, 101, doi: 10.3847/1538-3881/aaf88fDOI
  61. [65]Marley, M. S., Saumon, D., & Goldblatt, C. 2010, ApJL, 723, L117, doi: 10.1088/2041-8205/723/1/L117DOI
  62. [66]McKemmish, L. K., Masseron, T., Hoeijmakers, H. J., et al. 2019, MNRAS, 488, 2836, doi: 10.1093/mnras/stz1818DOI
  63. [67]McKemmish, L. K., Yurchenko, S. N., & Tennyson, J. 2016, MNRAS, 463, 771, doi: 10.1093/mnras/stw1969DOI
  64. [68]Miles, B. E., Biller, B. A., Patapis, P., et al. 2023, ApJL, 946, L6, doi: 10.3847/2041-8213/acb04aDOI
  65. [69]Min, M., Hovenier, J. W., de Koter, A., Waters, L. B. F. M., & Dominik, C. 2005, Icarus, 179, 158, doi: 10.1016/j.icarus.2005.05.015DOI
  66. [70]Mollière, P., Wardenier, J. P., van Boekel, R., et al. 2019, A&A, 627, A67, doi: 10.1051/0004-6361/201935470DOI
  67. [71]Mollière, P., Stolker, T., Lacour, S., et al. 2020, A&A, 640, A131, doi: 10.1051/0004-6361/202038325DOI
  68. [72]Mollière, P., Kühne, H., Matthews, E. C., et al. 2025, arXiv e-prints, arXiv:2507.18691, doi: 10.48550/arXiv.2507.18691DOI
  69. [73]Moran, S. E., Marley, M. S., & Crossley, S. D. 2024, ApJL, 973, L3, doi: 10.3847/2041-8213/ad72e7DOI
  70. [76]Nasedkin, E., Mollière, P., & Blain, D. 2024a, The Journal of Open Source Software, 9, 5875, doi: 10.21105/joss.05875
  71. [77]Nasedkin, E., Mollière, P., Lacour, S., et al. 2024b, A&A, 687, A298, doi: 10.1051/0004-6361/202449328DOI
  72. [78]Nasedkin, E., Schrader, M., Vos, J. M., et al. 2025, A&A, 702, A1, doi: 10.1051/0004-6361/202555370DOI
  73. [79]Ochsenbein, F., Bauer, P., & Marcout, J. 2000, A&AS, 143, 23, doi: 10.1051/aas:2000169DOI
  74. [80]Oliveros-Gomez, N., Manjavacas, E., Karalidi, T., et al. 2025, arXiv e-prints, arXiv:2511.15667, doi: 10.48550/arXiv.2511.15667
  75. [81]Palik, E. D., ed. 1985, Handbook of Optical Constants of Solids, Vol. 1 (Orlando, FL: Academic Press)
  76. [82]Perez, F., & Granger, B. E. 2007, Computing in Science & Engineering, 9, 21, doi: 10.1109/MCSE.2007.53
  77. [83]Petrus, S., Whiteford, N., Patapis, P., et al. 2024, ApJL, 966, L11, doi: 10.3847/2041-8213/ad3e7cDOI
  78. [84]Piskunov, N. E., Kupka, F., Ryabchikova, T. A., Weiss, W. W., & Jeffery, C. S. 1995, A&AS, 112, 525
  79. [85]Pollack, J. B., Hollenbach, D., Beckwith, S., et al. 1994, ApJ, 421, 615, doi: 10.1086/173677DOI
  80. [86]Polyansky, O. L., Kyuberis, A. A., Zobov, N. F., et al. 2018, MNRAS, 480, 2597, doi: 10.1093/mnras/sty1877DOI
  81. [87]Powell, D., Zhang, X., Gao, P., & Parmentier, V. 2018, ApJ, 860, 18, doi: 10.3847/1538-4357/aac215DOI
  82. [88]Prinn, R. G., & Barshay, S. S. 1977, Science, 198, 1031, doi: 10.1126/science.198.4312.1031
  83. [89]Reid, I. N., Cruz, K. L., Kirkpatrick, J. D., et al. 2008, AJ, 136, 1290, doi: 10.1088/0004-6256/136/3/1290DOI
  84. [90]Rossow, W. B. 1978, Icarus, 36, 1, doi: 10.1016/0019-1035(78)90072-6DOI
  85. [91]Rowland, M. J., Morley, C. V., & Line, M. R. 2023, ApJ, 947, 6, doi: 10.3847/1538-4357/acbb07DOI
  86. [92]Saumon, D., & Marley, M. S. 2008, ApJ, 689, 1327, doi: 10.1086/592734DOI
  87. [93]Scott, A., & Duley, W. W. 1996, ApJS, 105, 401, doi: 10.1086/192321
  88. [94]Servoin, J. L., & Piriou, B. 1973, Physica Status Solidi B Basic Research, 55, 677, doi: 10.1002/pssb.2220550224
  89. [95]Sitarz, M., Handke, M., & Mozgawa, W. 2000, Spectrochimica Acta Part A: Molecular Spectroscopy, 56, 1819, doi: 10.1016/S1386-1425(00)00241-9
  90. [96]Sousa-Silva, C., Al-Refaie, A. F., Tennyson, J., & Yurchenko, S. N. 2015, MNRAS, 446, 2337, doi: 10.1093/mnras/stu2246
  91. [97]Stephens, D. C., Leggett, S. K., Cushing, M. C., et al. 2009, ApJ, 702, 154, doi: 10.1088/0004-637X/702/1/154DOI
  92. [98]Suárez, G., & Metchev, S. 2022, MNRAS, 513, 5701, doi: 10.1093/mnras/stac1205DOI
  93. [99]Suárez, G., Vos, J. M., Metchev, S., Faherty, J. K., & Cruz, K. 2023, ApJL, 954, L6, doi: 10.3847/2041-8213/acec4bDOI
  94. [100]Toppani, A., Libourel, G., Robert, F., & Ghanbaja, J. 2006, GeoCoA, 70, 5035, doi: 10.1016/j.gca.2006.05.020DOI
  95. [102]Tsuji, T. 2002, ApJ, 575, 264, doi: 10.1086/341262DOI
  96. [103]Tsuji, T., & Nakajima, T. 2003, ApJL, 585, L151, doi: 10.1086/374388DOI
  97. [104]Visscher, C., Lodders, K., & Fegley, Bruce, J. 2010, ApJ, 716, 1060, doi: 10.1088/0004-637X/716/2/1060DOI
  98. [105]Vos, J. M., Allers, K. N., & Biller, B. A. 2017, ApJ, 842, 78, doi: 10.3847/1538-4357/aa73cfDOI
  99. [106]Wakeford, H. R., Visscher, C., Lewis, N. K., et al. 2017, MNRAS, 464, 4247, doi: 10.1093/mnras/stw2639DOI
  100. [107]Wang, J., Kolecki, J. R., Ruffio, J.-B., et al. 2022, AJ, 163, 189, doi: 10.3847/1538-3881/ac56e2DOI
  101. [108]Wells, M., Pel, J.-W., Glasse, A., et al. 2015, PASP, 127, 646, doi: 10.1086/682281DOI
  102. [109]Wende, S., Reiners, A., Seifahrt, A., & Bernath, P. F. 2010, A&A, 523, A58, doi: 10.1051/0004-6361/201015220DOI
  103. [111]Yurchenko, S. N., Mellor, T. M., Freedman, R. S., & Tennyson, J. 2020, MNRAS, 496, 5282, doi: 10.1093/mnras/staa1874DOI
  104. [112]Yurchenko, S. N., & Tennyson, J. 2014, MNRAS, 440, 1649, doi: 10.1093/mnras/stu326DOI
  105. [113]Zalesky, J. A., Saboi, K., Line, M. R., et al. 2022, ApJ, 936, 44, doi: 10.3847/1538-4357/ac786cDOI
  106. [114]Zeidler, S., Posch, T., & Mutschke, H. 2013, A&A, 553, A81, doi: 10.1051/0004-6361/201220459DOI

Paper details

vlnc:W_g3wfyvgy
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 1

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