{"id":"6c85f217-5560-4651-8557-8582146a06f3","arxiv_id":"2508.01281","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"High-resolution REACH spectra of the brown dwarf HR 7672 B reveal water and iron hydride absorption, and a preference for an optically thick cloud layer near 10^1.16 bar.","lead":"Astronomers took very sharp near-infrared spectra of the brown dwarf HR 7672 B with the new REACH instrument on Subaru, and found water and iron hydride in its atmosphere. The data also point to an optically thick cloud layer with a temperature around 2100 K, consistent with clouds of titanium oxide, aluminum oxide, or iron.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The fixed-TP telluric model (Section 3.2) admits residuals up to 0.1 in normalized flux, yet is fit simultaneously with the brown dwarf spectrum; this can bias the H2O abundance and cloud-top pressure that anchor the central claims.","rationale":"The reader's weakest assumption and mine coincide: the telluric model's fixed T-P is the most load-bearing simplification. The paper's own validation (Section 3.2.1) documents residuals up to 0.1 in normalized flux, which is large compared to the per-pixel noise and is asserted to be negligible without a supporting test. Since telluric absorption is modeled simultaneously with the brown dwarf spectrum, any systematic error in the telluric template propagates directly into the retrieved molecular abundances and, through the continuum shape, into the cloud-top pressure and the cloudy-vs-clear model comparison. The FeH detection (CCF S/N = 3.6, Appendix B) is also marginal, but it is a secondary issue; if the telluric concern is resolved, the FeH detection can be reassessed with rotationally broadened templates. A re-run with a flexible telluric model (free T,P) is a decisive, low-cost check: it either confirms the reported values or reveals a bias larger than the quoted 1-sigma intervals. Therefore I agree with the conditional verdict and recommend no change to it.","tokens_in":31312,"tokens_out":11434,"duration_ms":152277,"concrete_test":"Re-run the fiducial mass-constrained cloudy retrieval with the telluric transmission recomputed at T=280 K and P=0.8 bar (or with temperature and pressure included as free parameters), leaving all other settings unchanged. If the posterior medians of log H2O, log FeH, or log Ptop shift by more than the combined 1-sigma uncertainties listed in Table 2, the fixed-TP telluric model is a load-bearing systematic and the conditional verdict should require a new retrieval before quantitative adoption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The telluric transmission model in Equation (2) fixes pressure and temperature at 0.6005 bar and 273 K. Section 3.2.1 reports residuals up to 0.1 in normalized flux around several telluric lines, which is more than 100 times the mean per-pixel uncertainty (7.8-8.1e-4) and is dismissed without quantitative justification. Because the telluric model is fitted jointly with the brown dwarf spectrum, a systematic shape error (for example, wrong temperature-dependent H2O line strengths or pressure-broadened wings) cannot be absorbed by the single scaling coefficients beta_i; it can shift the retrieved molecular abundances and the pseudo-continuum level. The H band retrieval is especially vulnerable because telluric H2O lines overlap the rotationally broadened (v sin i ~ 41 km/s) brown dwarf H2O lines in the 1.48-1.54 um range. The jitter term (sigma = 0.06) inflates the noise but does not correct for a coherent shape error. The cloud-vs-clear conclusion (Delta BIC ~ 22) rests on the wavelength dependence of the continuum, so a telluric-induced slope error could also bias the retrieved log Ptop. This is the weakest link in the quantitative claims.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents the first atmospheric retrieval from REACH/Subaru high-resolution spectra (R ~ 100,000 in Y, J, H) of the benchmark L4.5 brown dwarf HR 7672 B. The forward model is built on ExoJAX and simultaneously treats brown-dwarf molecular opacity (H2O, FeH, CIA), a simplified optically thick cloud, telluric transmission, host-star light leakage, and a low-order continuum correction. The authors run four HMC retrievals (mass-constrained/free x cloudy/clear) over narrow J- and H-band windows and report detections of H2O and FeH, a cloud-top pressure log10(Ptop/bar) ~ 1.16, a cloud-top temperature near 2112 K, and a strong preference for the cloudy model (Delta BIC ~ 22). The molecular identifications are supported by a null-test cross-correlation procedure (Appendix B), and the paper also characterizes periodic noise in REACH faint-target spectra.","tokens_in":31670,"tokens_out":4928,"duration_ms":64886,"significance":"If the results hold, this paper demonstrates that high-contrast, high-resolution YJH spectroscopy with REACH can recover molecular abundances and cloud properties for a faint companion at 0.7 arcsec separation, a useful step for the field. The work benefits from a benchmark target with a dynamical mass, a publicly archived retrieval code and data (Zenodo/GitHub), and an explicit null test for CCF-based molecular detection. The comparison with the KPIC/Keck retrieval of Wang et al. (2022) is appropriate and places the new abundance and T-P constraints in context. The central atmospheric conclusions, however, rest on two assumptions that need quantitative stress-testing: the fixed-temperature-pressure telluric model and the post-hoc selection of very narrow wavelength windows for the cloud-vs-clear comparison.","major_comments":[{"comment":"The telluric transmission model fixes T = 273 K and P = 0.6005 bar for all species and fits only column-scale factors beta_i. Section 3.2.1 reports residuals up to 0.1 in normalized flux around telluric lines, which is roughly 100 times the mean per-pixel uncertainties (7.8e-4 and 8.1e-4) quoted in Figure 2 and Section 2.1. Because the telluric model is multiplied into and fitted simultaneously with the companion model (Eq. 5), a coherent line-shape error cannot be absorbed by the beta_i and can directly bias the retrieved molecular abundances and pseudo-continuum, particularly in the H band where telluric and brown-dwarf H2O lines overlap at 1.48-1.54 um. The statement that these residuals have no significant impact because they are 'comparable to the spectral noise' is not quantitatively supported: the jitter term sigma = 0.06 (Table 2) inflates the noise but does not correct a coherent systematic. I request a robustness test that varies the telluric T-P profile, or uses per-species effective temperatures, or repeats the retrieval after masking or down-weighting the strongest telluric lines, and reports the resulting shifts in logH2O, logFeH, and logPtop. Note also that the telluric-model validation was performed on a June 6 standard-star observation, while the companion was observed on June 24; the residual budget on the science night is assumed by the analysis, not demonstrated.","section":"Section 3.2 / 3.2.1, Eq. (2)"},{"comment":"The cloud-formation claim rests on the wavelength dependence of the continuum, but the retrieval uses only two narrow windows (J orders 43-45, 1.265-1.3 um; H orders 57-60, 1.48-1.54 um) that were selected after inspecting the data. The Y band and the 1.57-1.67 um H-band region are excluded (Sections 2.1 and 3). With such limited coverage, the Delta BIC = 21.9 preference for the cloudy model may reflect the particular choice of windows rather than a real, robust continuum shape. Please demonstrate that the cloud-vs-clear conclusion is stable to the inclusion or removal of wavelength ranges, or use injected cloud models to calibrate the sensitivity of this BIC comparison under the present spectral coverage.","section":"Section 4.1.3 / Section 5.1"},{"comment":"The J-band host-star leakage is fixed at 63% based on a retrieval around the single Paschen-beta line, with the stated expectation that deviations are absorbed by the linear trend parameter a in Eq. (6). A single-line measurement, combined with the assumption that the telluric transmission is identical on the two nights (TA = TB), is a fragile anchor for a parameter that directly sets the J-band continuum level and therefore affects the retrieved FeH abundance and line-to-continuum ratios. I recommend running a sensitivity test in which log(scale_star_y) is a free parameter and reporting the resulting change in logFeH and logPtop; if the degeneracy with the linear correction is strong, the abundance uncertainties should be enlarged accordingly.","section":"Section 3.4.2"}],"minor_comments":[{"comment":"The notation T_1bar^0 and T^Ptop_0 is easy to confuse in the tables; please define the reference pressure explicitly in the table caption for each quantity.","section":"Table 1 / Table 2"},{"comment":"The pink wavelength ranges used in the retrieval are difficult to distinguish in the printed figure; consider adding shaded vertical bands spanning both panels in each wavelength window.","section":"Figure 2"},{"comment":"The overall CCF peak S/N = 11.4 is quoted in the main text, but the molecular S/N values of 3.6 (FeH) and 6.9 (H2O) appear only in Appendix B; the main text should state these values explicitly because they are the actual detection statistics for the two species.","section":"Section 4.1.1 / Appendix B"},{"comment":"The BIC formula uses the maximum of the likelihood, but the text does not specify how this maximum is obtained from the HMC posterior; please state whether it is the maximum over posterior samples or a separate optimization.","section":"Section 3.6, Eq. (10)"},{"comment":"The telluric absorber list includes O2 in the J band and H2O, CO2, and CH4 in the H band; please state explicitly whether O2 is also included in the H-band window or whether its opacity is negligible there.","section":"Section 3.2"},{"comment":"The long-period noise characterization is thorough, but the paper does not translate the measured 15-22% noise amplitude into a quantitative uncertainty on the excluded 1.57-1.67 um region or on the continuum shape; a one-sentence quantitative summary would help readers judge the impact of the wavelength selection.","section":"Section 2.2 / Appendix A"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid first-science demonstration for REACH, and the release of the retrieval code and data is commendable. My recommendation of major revision is driven by the load-bearing systematic assumptions: the fixed-TP telluric model and the narrow, post-hoc wavelength selection. Both can be addressed with sensitivity calculations rather than new observations, so I do not see a reason to reject at this stage."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What's actually new: this is the first atmospheric retrieval from REACH/Subaru spectra of a faint substellar companion, it adds a first FeH detection for HR 7672 B, and the characterization of the periodic fringing noise is a genuinely useful instrument result. The authors also ship their code and data, which makes the analysis reproducible in a way that is still not standard in this field.\n\nThe retrieval methodology is solid. The molecular detections rest on a proper null test: the CCF shows a peak for the residual against a template including the molecule, and no peak when the full model is subtracted. That is the right way to argue against circularity. The H2O detection at S/N 6.9 is convincing; FeH at 3.6 is marginal but the null test gives it credibility. The preference for a cloudy model over a clear one is large in BIC terms, and the authors checked the result against alternative initial cloud pressures. The comparison with Wang et al. (2022) is careful and appropriately hedged.\n\nThe main soft spot is the telluric model. Fixing pressure and temperature at 0.6005 bar and 273 K while absorbing shape errors with a single scalar per molecule is a strong assumption. The authors report residuals up to 0.1 in normalized flux, which is more than an order of magnitude above the per-pixel uncertainties, and they dismiss it too quickly. Because telluric lines are fitted simultaneously with the brown dwarf spectrum, a coherent shape error, especially in H2O lines over the 1.48–1.54 μm range, could shift the retrieved water abundance and the pseudo-continuum level. I would not call this fatal, because the jitter term and the null test buffer the qualitative conclusions, but it is a real worry for the quantitative abundance and cloud-top pressure. A sensitivity test with, say, a second telluric temperature/pressure or a fitted temperature parameter would settle it.\n\nThe wavelength masking (excluding Y band and the 1.57–1.67 μm H-band region) is post-hoc but well motivated by the noise characterization. The fixed stellar leakage fraction in J band is another approximation, but the H-band leakage is free and results are stable across setups.\n\nWho should read this: anyone using REACH or planning high-resolution spectroscopy of directly imaged companions. It is a strong instrument demonstration and a reasonable first retrieval for this object. The quantitative results are probably not final, but the qualitative story—H2O, FeH, and a cloudy atmosphere—is likely to hold. I would send this to peer review and would cite it myself.","headline":"A credible first REACH science paper whose telluric model is the one soft spot; the detections and cloudy preference are likely robust, and it deserves a serious referee.","tokens_in":32268,"tokens_out":2091,"would_cite":true,"duration_ms":29977,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"High-resolution REACH spectra of the L4.5 brown dwarf HR 7672 B show water and iron hydride absorption and require an optically thick cloud deck near 14 bar.","keywords":["brown dwarf atmospheres","high-resolution spectroscopy","atmospheric retrieval","cloud opacity","FeH","water absorption","directly imaged companions","REACH"],"falsifier":"Re-analyse the two nights of data after dividing out the telluric absorption using a rapidly rotating standard star observed on the same night; if the retrieved H2O and FeH mixing ratios or the cloud-top pressure shift by more than the quoted uncertainties, the fixed-T-P telluric model is falsified.","tokens_in":31132,"feed_emoji":"☁️","tokens_out":19289,"duration_ms":189123,"temperature":0.7,"pith_summary":"The paper reports the first atmospheric retrieval applied to a faint companion observed with REACH, the R~100,000 near-infrared spectrograph on the Subaru Telescope. Using J- and H-band spectra of the L-type brown dwarf HR 7672 B, the authors identify water and iron hydride as the main absorbers and find that the data strongly prefer a model with an optically thick cloud whose top sits near 14 bar at about 2110 K. The cloud-top temperature falls between the condensation temperatures of titanium dioxide, aluminium oxide, and iron, so those species are plausible cloud materials. The same retrieval simultaneously corrects for light leaking from the host star and for Earth's telluric absorption, and it measures a projected rotation speed of about 41 km/s. If the claim holds, it shows that high-dispersion spectroscopy of faint directly imaged substellar companions can recover molecular abundances and cloud structure with a relatively simple model.","feed_headline":"REACH spectra reveal water, FeH, and a thick cloud deck on HR 7672 B","feed_subtitle":"First science demonstration for REACH: water, FeH, and a ~14 bar cloud top are retrieved from R~100,000 spectra.","key_machinery":"The load-bearing machinery is a single autodifferentiable spectral model that simultaneously computes the brown dwarf spectrum, telluric transmission, and host-star leakage. The brown dwarf part uses a power-law temperature–pressure profile, constant volume-mixing-ratio opacities for water and iron hydride, collision-induced absorption, and a two-parameter cloud: a fixed, optically thick opacity ($\\tau_\\mathrm{cloud}=500$) inserted at pressures $P>P_\\mathrm{top}$. Telluric transmission is modelled as the exponential of summed cross-sections for $\\mathrm{H_2O}$, $\\mathrm{CO_2}$, $\\mathrm{CH_4}$, and $\\mathrm{O_2}$ at fixed summit pressure and temperature, with a free Doppler shift. Host-star light is treated as a scaled version of the observed HR 7672 A spectrum. The model is fit jointly to the J-band photometric magnitude and the high-resolution spectra using Hamiltonian Monte Carlo with a jitter term, and the molecular detections are cross-checked with cross-correlation functions built from the best-fit model.","core_discovery":"The central claim is that the J- and H-band spectrum of HR 7672 B is dominated by water ($\\mathrm{H_2O}$) and iron hydride ($\\mathrm{FeH}$) absorption, and that a wavelength-independent continuum source—an optically thick cloud at $P_\\mathrm{top} = 10^{1.16}\\,\\mathrm{bar}\\simeq14\\,\\mathrm{bar}$ with a cloud-top temperature near 2112 K—is required to reproduce the observed features. A cloud-free model can also fit the data, but only by letting collision-induced absorption supply the continuum, and the cloudy model is favoured by $\\Delta\\mathrm{BIC}\\simeq22$. The retrieved temperature–pressure profile places the cloud top between the condensation curves of $\\mathrm{TiO_2}$, $\\mathrm{Al_2O_3}$, and Fe, making those the likely cloud materials. Cross-correlation analysis independently confirms the molecular detections, with S/N of 3.6 for $\\mathrm{FeH}$ in the J band and 6.9 for $\\mathrm{H_2O}$ in the H band. The paper also presents the first characterization of the periodic noise systematics that currently limit REACH's usable wavelength range.","pith_inferences":["If the retrieved cloud is as optically thick as assumed, the J-band continuum should be nearly grey; a test would be to compare the continuum slope across the full J band with the cloudy and clear-sky models fit to future full-band data.","The paper's 3.3σ shortfall of retrieved water relative to thermochemical equilibrium could reflect vertical mixing or a non-constant abundance profile that the constant-VMR assumption cannot capture; a per-order or per-night jackknife could reveal whether residual telluric error drives the deficit.","An empirical check of the telluric model—dividing the companion spectra by a same-night rapidly rotating standard star instead of forward-modelling telluric lines—would directly test the paper's weakest assumption; agreement would strengthen the cloud claim, disagreement would implicate telluric systematics.","The same joint modelling of companion, host-star leakage, and telluric transmission is a natural template for future REACH observations of directly imaged exoplanets, provided the targets rotate slowly enough to resolve molecular line structure."],"forward_implications":["REACH can deliver molecular detections and cloud-top constraints for faint companions at $R\\sim100{,}000$, complementing K-band high-resolution instruments by probing deeper atmospheric layers near 4–15 bar.","The retrieved water abundance (log VMR $\\simeq -3.8$) is consistent within $3\\sigma$ with the earlier K-band retrieval, supporting a uniform composition across the altitudes probed by the two bands.","A cloud-top temperature near 2110 K provides a concrete benchmark for cloud-condensation models of L dwarfs, pointing to $\\mathrm{TiO_2}$, $\\mathrm{Al_2O_3}$, or Fe grains.","Characterizing the periodic noise systematics gives future REACH users a template for mitigating fringing-like signals in post-processing, which will extend the usable spectral range toward the longer H band.","In the mass-free cloudy retrieval, the mass upper limit is unconstrained, indicating that for cloudy atmospheres J/H-band spectra alone may not pin down surface gravity when cloud opacity hides the collision-induced continuum."],"supporting_citations":[{"why":"Provides the autodifferentiable spectral model that forms the basis of the forward model used for the retrieval.","marker":"Kawahara et al. (2022)"},{"why":"Supplies the accelerated opacity-computation scheme that makes high-resolution line-by-line opacities tractable for this analysis.","marker":"Kawahara et al. (2025)"},{"why":"The earlier K-band retrieval of HR 7672 B whose abundances, v sin i, C/O, and [Fe/H] are used for comparison and input values.","marker":"Wang et al. (2022)"},{"why":"Dynamical mass measurement (72.7 +/- 0.8 M_Jup) adopted as the strict prior in the mass-constrained retrieval.","marker":"Brandt et al. (2019)"},{"why":"Water line list used for both the brown dwarf molecular opacity and the telluric water transmission.","marker":"Polyansky et al. (2018)"},{"why":"Iron hydride line list that defines the FeH opacity detected in the J band.","marker":"Dulick et al. (2003); Bernath (2020)"},{"why":"Collision-induced absorption coefficients for H2-H2 and H2-He used as the continuum in the clear-sky comparison.","marker":"Karman et al. (2019)"},{"why":"Describes the REACH instrument whose first science demonstration this paper presents.","marker":"Kotani et al. (2020)"},{"why":"Source of the J-band magnitude (14.39 +/- 0.20) fitted simultaneously with the high-resolution spectra.","marker":"Boccaletti et al. (2003)"},{"why":"Molecular line database through which the water and iron hydride line lists are provided.","marker":"Tennyson et al. (2016)"}],"fun_headline_variants":["REACH finds H2O and FeH, and a thick ~14-bar cloud on HR 7672 B","HR 7672 B's atmosphere: water, iron hydride, and a thick cloud top","REACH's first light: H2O, FeH, and clouds on HR 7672 B","Cloud top on HR 7672 B may be TiO2 or Fe, says REACH","Water and FeH dominate, but a thick cloud is key to HR 7672 B's spectrum"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes the model of Earth's telluric absorption—computed at fixed temperature and pressure and fitted simultaneously with the brown dwarf spectrum—is accurate enough that its deviations of up to 0.1 in normalized flux around some telluric lines do not bias the retrieved molecular abundances or cloud-top pressure.","fun_headline_variants_meta":{"raw":{"variants":["REACH finds H2O and FeH, and a thick ~14-bar cloud on HR 7672 B","HR 7672 B's atmosphere: water, iron hydride, and a thick cloud top","REACH's first light: H2O, FeH, and clouds on HR 7672 B","Cloud top on HR 7672 B may be TiO2 or Fe, says REACH","Water and FeH dominate, but a thick cloud is key to HR 7672 B's spectrum"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001555,"raw_usage":{"total_tokens":6286,"prompt_tokens":1089,"completion_tokens":5197,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":705,"completion_tokens_details":{"reasoning_tokens":5070}},"tokens_in":705,"tokens_out":5197,"duration_ms":39324,"temperature":1.0,"reasoning_tokens":5070,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T05:42:35.086124+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-analyse the two nights of data after dividing out the telluric absorption using a rapidly rotating standard star observed on the same night; if the retrieved H2O and FeH mixing ratios or the cloud-top pressure shift by more than the quoted uncertainties, the fixed-T-P telluric model is falsified.","supporting_citations":[{"cited_title":"M., & Marchis, F","cited_arxiv_id":null,"evidence_quote":"Source of the J-band magnitude (14.39 +/- 0.20) fitted simultaneously with the high-resolution spectra."}],"review_version":1}