{"id":"3681bec5-b8c4-4d2b-927d-4904a13f5eb3","arxiv_id":"2501.13917","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"New high-resolution spectra of the substellar companion HD 206893 B yield Teff = 1634 K, log g = 4.55, a near-solar C/O of 0.57, and a mass of about 22.7 Jupiter masses.","lead":"Astronomers used the Keck Planet Imager and Characterizer to capture high-resolution infrared spectra of HD 206893 B, a massive young companion in a multiplanet system, and measured its temperature, gravity, spin, and carbon-to-oxygen ratio. The result adds a benchmark data point for how giant planets form, since the measured C/O ratio is near-solar and cannot rule out either major formation pathway.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline Teff/log g/C/O rest on a truncated temperature prior that discards a data-supported 1900 K solution; the 1600 K branch is selected via evolutionary-model radius, not by the KPIC data alone.","rationale":"The reader's weakest assumption targets the custom PHOENIX cloud parameters, which is a related and real concern. My stress-test focuses on a more specific, load-bearing step: the post-hoc truncation of the temperature prior. This step determines which of two data-supported solutions is reported as the headline result, and it is justified using the same evolutionary models that later produce the mass and age. That is a genuine soft spot, but it does not overturn the paper's primary empirical contribution: a robust, two-epoch high-resolution detection of HD 206893 B and a careful, transparent forward-modeling analysis. The paper also honestly reports the full-grid results and the retrieval inconsistencies, which is good scientific practice. The appropriate verdict remains CONDITIONAL: the detection can stand, but the quoted bulk parameters and C/O should not be taken at face value until the temperature-prior choice is justified by data or presented as one of two viable solutions. A Bayes-factor test on the KPIC data would settle this directly. The reader identified the cloud grid as the weakest assumption; my concern overlaps with that model-dependence but is specifically about the prior truncation, so I mark partial agreement.","tokens_in":37623,"tokens_out":5161,"duration_ms":50791,"concrete_test":"Run a model comparison on the KPIC data between the two posterior modes. Use breads with narrow priors centered on Teff = 1600 K, log g = 4.5 and on Teff = 1900 K, log g = 4.8 (e.g., +-50 K and +-0.1 dex), for both epochs, and compute the log-evidence difference via nested sampling or the Laplace approximation. If the 1900 K mode is not disfavored by Delta ln Z > 2.3 (roughly 10:1), the Section 5.5 temperature truncation is not supported by the high-resolution data, and the headline Teff/log g/C/O/mass values should be presented as one branch of a genuinely bimodal posterior or re-derived with an explicit joint prior on radius from evolutionary models.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The detection itself is strong and not the issue: two independent KPIC epochs give >8 sigma CCFs with a background trace at zero. The load-bearing weakness is in the parameter inference. In Section 5.3, the full-grid MCMC on the KPIC data produces a bimodal posterior in Teff, with peaks near 1600 K, log g = 4.6 and 1900 K, log g = 4.8. In Section 5.5, the authors truncate the Teff prior at 1750 K, explicitly because the higher-temperature, higher-gravity solution implies a radius of 0.69 R_Jup, which they judge unphysical using CBPD23 evolutionary models. The paper's own GPI low-resolution grid search (Section 5.4, Figure 4) lists 1900 K, log g = 5.0 and 2000 K, log g = 5.0 as fits equally as good as 1600 K, log g = 4.5, so the low-temperature branch is not uniquely preferred by either the high-resolution or low-resolution data. The truncation is therefore a prior choice grounded in evolutionary models, and those same evolutionary models are later used in Section 7 to derive the quoted mass and age. This circularity means the headline Teff = 1634 K, log g = 4.55, mass = 22.7 M_Jup, and the C/O fit at fixed Teff = 1600 K, log g = 4.5 (Section 5.6) are conditional on an assumption not tested by the data. The paper does report the full-grid alternative (27.4 M_Jup, Teff = 1774 K), which is consistent with the dynamical mass, but the abstract and conclusions emphasize the truncated solution. A related but secondary concern is that the paper's own petitRADTRANS free retrieval on GPI JHK (Table 3) prefers log g = 2.78, radius 2.16 R_Jup, and grain sizes 28-48 microns, all inconsistent with the fixed cloud parameters (1 micron grains, PGS = 10^6, Kzz = 10^8) of the custom PHOENIX grid, showing that unquantified model systematics remain.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Using two epochs of Keck/KPIC K-band spectroscopy (R ~ 35,000), the paper reports a >8 sigma detection of the substellar companion HD 206893 B via cross-correlation with custom PHOENIX atmospheric models. A forward-modeling MCMC analysis yields a bimodal posterior in effective temperature with peaks near 1600 K and 1900 K; after truncating the Teff prior at 1750 K based on an evolutionary-model radius argument, the authors quote Teff = 1634+72-38 K, log g = 4.55+0.17-0.22, C/O = 0.57 +/- 0.02, and, using CBPD23 evolutionary models, a mass of 22.7+2.5-1.7 M_Jup, age 112+36-22 Myr, and radius 1.11 +/- 0.03 R_Jup. They also fit the orbit including two KPIC radial velocities and analyze the system's long-term stability, concluding that stable 1 Gyr configurations are predominantly low-eccentricity and coplanar. The paper includes a petitRADTRANS free retrieval of archival GPI spectra, which returns a substantially lower log g and larger radii/grain sizes than the adopted grid.","tokens_in":37948,"tokens_out":8324,"duration_ms":64646,"significance":"The paper provides the first high-resolution spectroscopic characterization of HD 206893 B and a benchmark comparison for L/T-transition companions with an independent dynamical mass anchor. The two-epoch detection is robust (SNR 9.8 and 12.7) and the full-grid evolutionary-model mass (27.4+5.9-5.4 M_Jup) agrees with the dynamical mass from Hinkley et al. (2023), lending credibility to the modeling framework. The C/O measurement, if systematics were quantified, would be a valuable addition to the trend of >4 M_Jup companions having near-solar C/O. However, the headline parameter values are conditional on a post-hoc truncation of the temperature prior and on fixed cloud parameters that are not independently validated; the paper's own free retrieval suggests this assumption may be violated. The limited-prior retrieval in Section 6 therefore cannot serve as independent confirmation. The results are significant but require a presentation that fully exposes these model dependencies.","major_comments":[{"comment":"The truncation of the Teff prior to <1750 K is a post-hoc decision based on evolutionary-model radii, and it directly conditions the headline results. The full-grid MCMC (Figure 3) shows a clear second mode at Teff ~ 1900 K, log g ~ 4.8, and Section 5.4's GPI grid search finds 1900 K, log g = 5.0 and 2000 K, log g = 5.0 to be equally good fits to the low-resolution data. The authors reject this mode because the implied radius of 0.69 R_Jup is deemed unphysical using CBPD23 models, but those same models are later used in Section 7 to derive the quoted mass (22.7 M_Jup), age, and radius. This circularity is not merely cosmetic: the C/O grid in Section 5.6 is computed at the truncated solution (1600 K, log g = 4.5), so the C/O measurement is also conditional on this prior choice. I recommend that the abstract and conclusions present both modes and clearly label the 22.7 M_Jup solution as prior-dependent, or that a joint fit including a physically motivated mass-radius prior be performed.","section":"Section 5.5"},{"comment":"The petitRADTRANS free retrieval on the full GPI JHK spectrum returns log g = 2.78+0.22-0.23, radius = 2.16+0.02-0.04 R_Jup, and Fe/MgSiO3 grain sizes of ~28-48 um (Table 3), which are inconsistent with the adopted PHOENIX grid parameters (a0 = 1 um, PGS = 10^6 dyne/cm2, Kzz = 10^8; Section 5.2). The authors then run a second retrieval with priors confined to +/-2 sigma of the forward-model posteriors (Table 4), which by construction returns log g ~ 4.5 and radius ~ 1.2 R_Jup. This is a consistency check, not an independent validation, and it does not address the possibility that the fixed cloud prescription biases the forward-model Teff/log g/C/O. The paper should either propagate the systematic uncertainty from the cloud model choice into the quoted parameters or explicitly state that all reported values assume that specific cloud prescription.","section":"Section 6"},{"comment":"The C/O ratio is quoted as 0.57 +/- 0.02 with 'only accounting for statistical uncertainties,' as stated in the abstract. This is misleading because the value is derived at the fixed (and prior-dependent) Teff = 1600 K, log g = 4.5, and assumes solar metallicity and the fixed cloud parameters. Given the bimodal temperature posterior and the retrieval's preference for different cloud properties, the systematic uncertainty on C/O is likely to be much larger than 0.02. A quantitative systematic error budget, or at least a clear statement of which assumptions dominate, is required before this measurement can be used in the formation-trend analysis.","section":"Section 5.6"},{"comment":"The low-resolution GPI fit does not uniquely favor the low-temperature branch: the grid search lists Teff = 1900 K, log g = 5.0; 2000 K, log g = 5.0; and 1600 K, log g = 4.5 as equally good fits (Figure 4). The text in Section 5.5 says the higher-temperature solution is 'ruled out,' but the data alone do not rule it out; only the evolutionary-model radius argument does. This wording should be corrected to reflect that the choice between branches is a modeling prior, not a data-driven result.","section":"Section 5.4"}],"minor_comments":[{"comment":"In the sentence comparing the C/O-fit radial velocity to the full posterior, 'in 5.2' should refer to Section 5.3 or 5.5, since Section 5.2 describes the model grid rather than the MCMC posterior.","section":"Section 5.6"},{"comment":"The Abstract quotes a detection at '>8 sigma' while the Conclusion states '>10 sigma'; please make these consistent, or clarify that the former refers to each epoch and the latter to the combined dataset.","section":"Abstract / Conclusion"},{"comment":"The caption of Table 3 says 'Parameters used for GPI K-band petitRADTRANS free retrieval' but the table includes columns for the J-, H-, and K-band fit as well; please update the caption to describe all columns.","section":"Table 3"},{"comment":"The reference list contains duplicates: Finnerty et al. 2022 appears twice, and Hsu et al. 2021 appears twice; these should be merged.","section":"References"},{"comment":"The units 'cm s-2 dex' for log g are nonstandard; use 'dex' throughout the text and tables.","section":"General"},{"comment":"The 1 Gyr stability analysis finds only 0.8% of orbital draws stable, and the stable configurations have semi-major axes and eccentricities that are not the median values from the joint fit. The discussion should state more prominently that the current orbital solution is only marginally stable at the system age, rather than implying that coplanar low-eccentricity configurations are strongly favored by the data.","section":"Section 8.1"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a solid detection and a useful data release, but the central parameter claims are presented with a post-hoc prior truncation that is not adequately flagged in the abstract. The full-grid solution is consistent with the dynamical mass, which is good, but the paper should present that solution with equal weight and give a systematic error budget. I believe the manuscript can be made publishable with a major revision that re-frames the parameter inference around the bimodality and the model-dependent radius prior."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this paper delivers the first R~35,000 K-band detection of HD 206893 B, at >8σ in two independent epochs, and uses it to measure a C/O ratio, a spin upper limit, two radial-velocity points, and an orbit/stability analysis. That is a real step forward for a target that is one of only a handful with both an outer brown-dwarf-mass companion and an inner planet. I came away convinced the detection is real and the RV points are probably the most robust new products.\n\nThe paper also earns credit for transparency. The full-grid MCMC gives a bimodal posterior in Teff, with peaks near 1600 K/log g 4.6 and 1900 K/log g 4.8. The authors then truncate the temperature prior at 1750 K, explicitly because the hot branch implies a 0.69 R_Jup radius, which they call unphysical using CBPD23 evolutionary models. They report the full-grid alternative (27.4 M_Jup, Teff=1774 K) in Table 6 and Section 7, and that mass is consistent with the dynamical mass. So the paper is not hiding the problem.\n\nBut the stress-test concern is real, and it lands on the load-bearing claims. The truncation is motivated by evolutionary-model radii, and those same CBPD23 models are then used to derive the quoted mass and age. That is a circular step: evolutionary models select the 1600 K branch, then the truncated posterior feeds back into evolutionary models. The abstract and conclusions emphasize Teff=1634 K, log g=4.55, 22.7 M_Jup, and the C/O fit at fixed 1600 K/4.5, not the full-grid alternative. The GPI low-resolution grid search (Fig. 4) lists 1900 K, log g=5.0 and 2000 K, log g=5.0 as equally good fits, so the low-T branch is not independently preferred by the data.\n\nThe C/O value (0.57±0.02) is statistical-only, and the paper's own petitRADTRANS retrieval on GPI JHK prefers log g≈2.8, radius≈2.2 R_Jup, and 28–48 micron grains, all inconsistent with the fixed 1-micron-cloud custom PHOENIX grid. The authors dismiss that retrieval as unphysical, which is defensible, but it is also a flag that the model-systematic uncertainty is unquantified. Custom grids and reduced data are not released, so independent verification of the parameter inference is limited.\n\nNet: a solid, honest observational paper with a load-bearing prior choice that needs to be front-and-center. I would send it to peer review. A referee should ask for the abstract/conclusions to present the bimodality and the truncation honestly, or for a principled justification (e.g., independent radius constraint) that breaks the circularity. The detection and RV points will stand regardless.\n\nRecommendation: engage with it, cite the RV/detection, treat the bulk parameters as conditional.","headline":"Solid first high-res detection and C/O for HD 206893 B, but the headline Teff/mass/C/O are conditional on a post hoc prior truncation that the data alone don't justify.","tokens_in":38824,"tokens_out":3193,"would_cite":true,"duration_ms":27330,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The first high-resolution spectral characterization of HD 206893 B measures a 22.7-Jupiter-mass L/T companion with near-solar C/O.","keywords":["high-resolution spectroscopy","directly imaged companion","L/T transition","brown dwarf","carbon-to-oxygen ratio","orbital stability","Keck Planet Imager and Characterizer","PHOENIX atmosphere models"],"falsifier":"Re-run the same forward-model MCMC on the two KPIC epochs with a PHOENIX grid in which cloud particle size, cloud deck pressure, and $K_{zz}$ are free parameters, or with a high-resolution free retrieval that fits the CO line profiles without assuming the fixed grid; if $T_{\\rm eff}$, $\\log g$, or C/O move by more than the quoted $1\\sigma$ uncertainties, the central result is model-dependent. A second concrete check is an additional radial velocity point in 2024-2025, which should separate the moderately eccentric joint-fit orbit from the near-circular orbits required for 1 Gyr stability.","tokens_in":2196,"feed_emoji":"🪐","tokens_out":6023,"duration_ms":116814,"temperature":0.7,"pith_summary":"This paper claims to have made the first high-resolution spectral characterization of HD 206893 B, a dusty, exceptionally red L/T-transition substellar companion in a system that also hosts an inner planetary-mass companion. By fitting two epochs of KPIC K-band spectra (resolving power about 35,000) with PHOENIX atmosphere models in a forward model that handles companion light and diffracted starlight together, the authors detect the companion at more than $8\\sigma$ and derive an effective temperature of $1634^{+72}_{-38}$ K, a surface gravity of $\\log g = 4.55^{+0.17}_{-0.22}$, and a carbon-to-oxygen ratio of $0.57 \\pm 0.02$. Those values, fed into substellar evolution tracks, give a mass of $22.7^{+2.5}_{-1.7}$ Jupiter masses, an age of $112^{+36}_{-22}$ Myr, and a radius of $1.11 \\pm 0.03$ Jupiter radii. The new radial velocities also tighten the orbit, and stability calculations favor low-eccentricity, coplanar configurations for the two companions. If correct, this adds a resolved-line data point to the trend that massive directly imaged companions have near-solar C/O ratios, while leaving both core-accretion and disk-fragmentation formation paths open.","feed_headline":"First high-res spectrum of HD 206893 B gives a 22.7-Jupiter companion","feed_subtitle":"High-resolution KPIC spectra pin down its 1634 K atmosphere and near-solar C/O, and favor a coplanar orbit.","key_machinery":"The carrying mechanism is a forward-model likelihood, implemented in the breads code, that simultaneously fits the companion spectrum and the diffracted starlight, with continuum and stellar speckle terms marginalized analytically. The atmospheric templates come from a custom PHOENIX grid built for L/T-transition objects, with fixed cloud parameters: 1-micron grains, a cloud deck at $10^6$ dyne cm$^{-2}$, and vertical mixing $K_{zz}=10^8$, while effective temperature and surface gravity range from 1200 to 2000 K and $\\log g=3.5$ to 5.0. Detection is confirmed by cross-correlating the data against a broadened model template; the parameters are then mapped out with an MCMC over the interpolated grid. The same grid, plus a follow-up grid varying carbon and oxygen at fixed $T_{\\rm eff}=1600$ K and $\\log g=4.5$, produces the C/O measurement.","core_discovery":"The central claim is that resolved CO and H2O lines in the K band can be used to pin down the atmosphere and orbit of HD 206893 B, which previous low-resolution work could only loosely constrain. Detecting the companion at >8 $\\sigma$ in two epochs, the authors report $T_{\\rm eff}=1634^{+72}_{-38}$ K and $\\log g=4.55^{+0.17}_{-0.22}$ from a custom PHOENIX grid, and a C/O ratio of $0.57 \\pm 0.02$ from a grid computed at fixed temperature and gravity. Bulk properties from the Chabrier et al. (2023) evolution tracks give a $22.7^{+2.5}_{-1.7}$ Jupiter-mass companion with radius $1.11 \\pm 0.03$ Jupiter radii and age $112^{+36}_{-22}$ Myr. A joint orbit fit including two new KPIC radial velocities yields a moderately eccentric orbit for B, but the 1 Gyr stability analysis strongly prefers near-circular orbits for both B and c and co-planarity between them. The paper also shows that the high-resolution CO line shapes reject the low surface gravity and inflated radius favored by an unconstrained low-resolution retrieval.","pith_inferences":["If the fixed cloud prescription in the custom PHOENIX grid is the dominant systematic, the C/O value is the most vulnerable parameter because it was computed on a small grid at one temperature and gravity; re-fitting the same KPIC data with C/O, cloud particle size, and cloud deck free would reveal how much of the result is model-driven.","A single additional radial velocity point in 2024 or 2025 would have strong leverage: with a roughly 25-35 year orbital period, it should discriminate the moderately eccentric joint-fit solution from the near-circular orbits required for 1 Gyr stability.","The paper's own low-resolution retrievals place the cloud-model mismatch mainly in J- and H-band extinction; a testable extension is to fit the KPIC K-band lines and the GPI K-band continuum in one joint model with a single cloud prescription and check whether temperature and gravity remain consistent.","If the near-solar C/O of B is confirmed alongside the host star's reported super-solar C/O, the system may hint that B accreted its gas inside the carbon-rich ice line or that the host C/O measurement carries a systematic offset; the paper does not draw that conclusion itself."],"forward_implications":["HD 206893 B becomes one of the few directly imaged companions with resolved-line atmospheric parameters, with a measured C/O of $0.57 \\pm 0.02$ consistent with the tentative trend that companions above roughly four Jupiter masses have near-solar C/O.","The two KPIC radial velocities, combined with archival astrometry, update the orbit; the joint fit allows an eccentricity near 0.27 for B, while 1 Gyr stability selects eccentricities below 0.1 for both B and c and favors coplanarity between them.","The preferred temperature and gravity, together with Chabrier et al. (2023) evolution tracks, give a mass of $22.7^{+2.5}_{-1.7}$ Jupiter masses, consistent with the independently measured dynamical mass of about 26 Jupiter masses.","The high-resolution CO line shapes strongly reject, with a Bayes factor of 0.00028, the low-gravity, large-radius solution preferred by the unconstrained low-resolution retrieval, indicating that resolved lines break the gravity-radius degeneracy.","The measured C/O ratio alone cannot rule out either core accretion or disk fragmentation as the formation pathway; the paper identifies future 3-5 micron sulfur measurements as the discriminating test."],"supporting_citations":[{"why":"Supplies the custom PHOENIX L/T-transition grid prescription, including cloud particle size, cloud deck pressure, and disequilibrium chemistry, used for the spectral fits.","marker":"Brock et al. 2021"},{"why":"Provides the PHOENIX model framework and spin-broadening treatment used to fit the companion's resolved absorption lines.","marker":"Barman et al. 2011"},{"why":"Defines the forward-model likelihood that fits companion and diffracted starlight together and underlies the >8-sigma cross-correlation detections.","marker":"Ruffio et al. 2019"},{"why":"Supplies the archival GPI low-resolution JHK spectra used to validate the model grid and motivate the 1750 K temperature cap.","marker":"Ward-Duong et al. 2021"},{"why":"Provides the earlier GRAVITY low-resolution characterization and free retrieval results that this work compares against.","marker":"Kammerer et al. 2021"},{"why":"Gives the system architecture, orbit and mass of HD 206893 c, and the dynamical mass used to check the new mass estimate.","marker":"Hinkley et al. 2023"},{"why":"Provides the substellar evolution tracks used to convert effective temperature and surface gravity into mass, age, and radius.","marker":"Chabrier et al. 2023"},{"why":"Supports the premise that high-resolution K-band bulk parameters are largely insensitive to cloud model choice.","marker":"Xuan et al. 2022"}],"fun_headline_variants":["KPIC resolves HD 206893 B: 22.7 Jupiter-mass companion's atmosphere","High-res Keck data reveals near-solar C/O for HD 206893 B","HD 206893 B: 1634 K and 22.7 Jupiter masses from KPIC","First high-res spectrum of HD 206893 B shows 22.7-Jupiter mass"],"cache_read_input_tokens":40448,"weakest_assumption_plain":"The load-bearing assumption is that the custom PHOENIX grid, with its fixed 1-micron grains, cloud deck at $10^6$ dyne cm$^{-2}$, fixed $K_{zz}=10^8$, and solar metallicity, is a faithful enough description of HD 206893 B's cloudy L/T-transition atmosphere that the quoted temperature, gravity, and C/O are not biased by the cloud prescription.","fun_headline_variants_meta":{"raw":{"variants":["KPIC resolves HD 206893 B: 22.7 Jupiter-mass companion's atmosphere","High-res Keck data reveals near-solar C/O for HD 206893 B","HD 206893 B: 1634 K and 22.7 Jupiter masses from KPIC","First high-res spectrum of HD 206893 B shows 22.7-Jupiter mass"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00025,"raw_usage":{"total_tokens":1684,"prompt_tokens":1203,"completion_tokens":481,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":819,"completion_tokens_details":{"reasoning_tokens":385}},"tokens_in":819,"tokens_out":481,"duration_ms":4716,"temperature":1.0,"reasoning_tokens":385,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T15:28:33.081441+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same forward-model MCMC on the two KPIC epochs with a PHOENIX grid in which cloud particle size, cloud deck pressure, and $K_{zz}$ are free parameters, or with a high-resolution free retrieval that fits the CO line profiles without assuming the fixed grid; if $T_{\\rm eff}$, $\\log g$, or C/O move by more than the quoted $1\\sigma$ uncertainties, the central result is model-dependent. A second concrete check is an additional radial velocity point in 2024-2025, which should separate the moderately eccentric joint-fit orbit from the near-circular orbits required for 1 Gyr stability.","supporting_citations":[],"review_version":1}