{"id":"a3c81de3-092b-4663-8359-9efbc160eabb","arxiv_id":"2412.01472","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Time-resolved absorption trails of six species in TOI-1518 b favor a strong atmospheric drag, with Fe+ requiring stronger drag than Fe, suggesting magnetic effects.","lead":"Using two transits observed with the MAROON-X spectrograph, this paper measures the time-dependent absorption of six chemical species in the atmosphere of the ultra-hot Jupiter TOI-1518 b and shows that their Doppler trails are best matched by global circulation models with strong wind drag. The result sharpens the debate on what slows winds in these extreme planets and hints that magnetic effects are at work.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The quantitative drag-timescale and Fe+/Fe magnetic interpretation rest on single-Gaussian fits to double-peaked and non-Gaussian CCF trails; a full forward-model comparison is needed to confirm the inferred τ_drag range.","rationale":"The reader's weakest assumption identifies the Gaussian CCF parameterization as a key fragile premise, and I agree that this is the most load-bearing issue for the quantitative drag timescale and the magnetic interpretation. My stress-test reading adds specificity: the double-peaked model CCF structure is acknowledged in Sect. 4.3, the observed non-Gaussianity is acknowledged in Sect. 3.4/Fig. B.5, and for Fe+ the amplitude and FWHM diagnostics are explicitly non-constraining, so the entire Fe+/Fe magnetic claim hinges on the Gaussian centroid of an asymmetric CCF. A second, smaller contributor is the untreated Mg contamination of the Fe+ trail (Fig. B.3), which could bias the centroid in the same direction. I do not think these issues invalidate the qualitative strong-drag conclusion, because the no-drag and weak-drag models are strongly disfavored by the data regardless of the fitting detail. However, they justify the reader's conditional acceptance rather than full acceptance; the requested revision (full forward-model or double-Gaussian comparison, and significance quantification of the Fe+ versus Fe difference) is exactly what would settle the concern. The concrete test I propose is a direct CCF-level likelihood comparison, which removes the Gaussian summary step entirely and would show whether the inferred τ_drag range and the Fe+/Fe drag difference are robust.","tokens_in":27382,"tokens_out":7646,"duration_ms":76961,"concrete_test":"Re-run the model-data comparison without single-Gaussian summaries: for each of the five GCMs, inject the model into the actual data frames using the same PCA pipeline, compute the full phase-binned CCF maps, and compare observed and model CCFs bin-by-bin via a log-likelihood or chi-square statistic in the ±10 km/s window, marginalizing over small Kp/Vsys offsets. If Fe still prefers τ_drag = 10^4 s and Fe+ still prefers 10^3 s with Δχ² > 9 (or the Bayesian equivalent), the Gaussian parameterization is not the driver. Additionally, refit the Fe+ trail with the contaminating Mg feature masked out of the Fe+ template to check whether the Fe+ centroid shift and its drag preference persist.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central drag inference (Figs. 11-13, Sect. 4.3) is obtained by fitting a single Gaussian to each phase-binned CCF, but the paper itself reports that the model CCFs have a double-peak structure near 0 and -7 km/s and that a single Gaussian cannot capture this asymmetry (Sect. 4.3, Fig. 10), while Fig. B.5 shows the observed 1D CCFs also deviate from Gaussian profiles (Sect. 3.4). For an asymmetric, double-peaked profile, the fitted centroid, FWHM, and amplitude depend on the relative heights of the two peaks, which evolve with orbital phase and with τ_drag. The discrimination between τ_drag = 10^3 and 10^4 s for Fe, and the Fe+ preference for 10^3 s over 10^4 s, rests on exactly these summary statistics. The paper further notes that centering the Gaussian on the CCF maximum caused a misestimation of the FWHM and probably of the error bars, so the uncertainties used to distinguish models are themselves suspect. For Fe+, amplitude is explicitly non-discriminating and FWHM error bars are too large, leaving the position as the sole diagnostic; the position is exactly what a single-Gaussian centroid of a double-peaked CCF is most likely to bias. A related, uncorrected complication is the partial contamination of the Fe+ trail by Mg (Fig. B.3), which could shift the Fe+ centroid in a phase-dependent way. These issues do not overturn the qualitative strong-drag direction, because the no-drag model is far too blueshifted, but they make the quoted τ_drag range and the Fe+/Fe magnetic interpretation conditional on the Gaussian parameterization being unbiased.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"TOI-1518 b is observed with MAROON-X during two transits (2022-08-13 and 2023-10-19). Cross-correlation with single-species synthetic templates yields detections of 14 species, and phase-binned CCF trails are measured for six species (Fe, Fe+, Ca, Ca+, Na, Mg), all of which show a progressive blueshift during transit. The trails are compared with five SPARC/MITgcm GCMs that differ only in the Newtonian drag timescale (τ_drag = ∞, 10^6, 10^5, 10^4, 10^3 s); model spectra are injected into the real data and processed with the same PCA plus Gaussian-fitting pipeline. The no-drag and 10^6 s models are much more blueshifted than the data, and the data lie between the 10^4 and 10^3 s models. Fe+ appears to require stronger drag than Fe, which the authors interpret as a sign of magnetic (ohmic) drag; Ca+ shows a qualitatively different trail, attributed to probing layers up to 10^-8 bar near or above the Roche lobe. A CHIMERA retrieval using the newly derived SOPHIE mass (1.83 ± 0.47 MJup) gives log10 Fe = −4.88(+0.63/−0.76), consistent with stellar metallicity, and demonstrates Kp/Vsys-induced biases for the other species. Ancillary results include the first significant RV detection (K = 192(+48/−49) m/s), a refined ephemeris from 56 TESS transits, and a VO detection with the HyVO line list that is not recovered with the VOmyt line list.","tokens_in":27732,"tokens_out":21432,"duration_ms":169714,"significance":"If the strong-drag conclusion holds, this paper is a significant addition to the emerging picture, alongside WASP-121 b (Wardenier et al. 2024), that ultra-hot Jupiter wind speeds are strongly suppressed by some dissipative process, and it is the first such phase-resolved study for a very grazing geometry (b ≈ 0.88–0.90) where the limb-probing interpretation is cleaner. The claimed Fe+/Fe differential drag is a potentially novel diagnostic of altitude- or temperature-dependent (magnetic) drag, and the Ca+ above-Roche interpretation is a testable hypothesis. The methodology has real strengths: the GCM spectra are injected into the actual data and analyzed with the same PCA and CCF pipeline as the observations, giving a symmetric data-model comparison; the σPCA term is propagated into the trail uncertainties; the retrieval biases for species with offset Kp/Vsys are demonstrated rather than ignored; and the SOPHIE mass, refined ephemeris, and HyVO line-list comparison are useful independent contributions.","major_comments":[{"comment":"The quantitative drag-timescale inference rests on single-Gaussian fits to CCFs that the manuscript itself shows to be double-peaked and non-Gaussian. Section 4.3 states that the GCM CCF \"exhibits a double-peak structure with one centered near 0 km/s and another around −7 km/s\" and that a single-peaked function \"cannot accurately capture the asymmetric nature of the CCF,\" while Sect. 3.5 and Fig. B.5 report that centering the Gaussian on the CCF maximum caused \"a misestimation of the FWHM and probably of the error bars of the measured Vres.\" For a two-peak profile, the fitted centroid, FWHM, and amplitude are set by the relative heights of the two peaks, which evolve with orbital phase and with τ_drag; the data/model separation between τ_drag = 10^3 and 10^4 s for Fe (Fig. 11) is drawn from exactly these statistics, including the FWHM argument that the 10^3 s signal is \"too small and not wide enough.\" Applying the same estimator to data and models makes the comparison procedurally symmetric, but it does not establish that the derived τ_drag range applies to the underlying CCF shapes. A full forward-model comparison (two-Gaussian fits to both data and models, a direct likelihood comparison of the binned CCF profiles, or an injection-recovery calibration of the centroid and width bias) is needed to support the quoted τ_drag = 10^3–10^4 s range, particularly because the grid contains only five models and the range is inferred by interpolation between two adjacent grid points. The qualitative exclusion of the no-drag and 10^6 s models is not affected by this concern, since those trails are far more blueshifted than the data.","section":"§4.3, Figs. 10–13"},{"comment":"The Fe+/Fe differential-drag claim rests on a single fragile diagnostic. The paper states that for Fe+ the model amplitude is unaffected by drag and that the FWHM error bars are too large to differentiate between the two strong-drag models, leaving only the trail position to discriminate; position is precisely the summary statistic most vulnerable to the single-Gaussian bias described above. Additionally, the observed Fe+ trail is partially contaminated by a strong Fe+ line in the Mg triplet (Fig. B.3), while the model CCFs are computed for Fe+ alone, so the contamination can displace the data centroid relative to the models in a phase-dependent way. The ohmic-drag interpretation (Fe+ traces hotter dayside layers where magnetic drag is stronger, following Beltz et al. 2022) is plausible, but the abstract's statement that \"Fe+ favors a stronger drag than Fe\" is stronger than the current diagnostics support. The manuscript should either quantify the significance of the Fe+/Fe τ_drag difference including the Mg contamination and fit-systematic uncertainties, or present the interpretation as a hypothesis with a concrete test (e.g., an MHD GCM prediction of the differential Fe/Fe+ trail), and align the abstract and conclusions with that framing.","section":"§4.3, Fig. 12, Fig. B.3"}],"minor_comments":[{"comment":"There are numerous typos and spacing artifacts, e.g., \"atmosheres\" (Sect. 1), \"the S/N was was always\" (Sect. 2), \"TO-1518b\" (Appendix A), \"FHWM\" (Sect. 4.3), \"Fe+ seem\" (Sect. 6), and inconsistent spacing in \"di fferent\" and \"W ASP\" throughout; a careful language edit is needed.","section":"Throughout"},{"comment":"In the Ba+ row of Table 3, the amplitude uncertainty reads \"11.4 ± -0.6\"; the minus sign appears spurious and should be corrected.","section":"Table 3"},{"comment":"The injection and cross-correlation description in Sect. 4.2 mentions only the Fe template; please clarify that the model trails for Fe+, Ca, Ca+, Na, and Mg (Figs. 12–13) were produced with the corresponding species templates, since the species-dependent comparison is central to the argument.","section":"§4.2"},{"comment":"The GCM gravity of 10.56 m/s² implies an assumed planetary mass of about 1.5 MJup, while the new SOPHIE mass is 1.83 ± 0.47 MJup; state the mass assumed by the GCMs explicitly and comment on the sensitivity of the modeled trails to the updated mass.","section":"§4.1, Table 4"},{"comment":"For the Ca+ trail, the data were analyzed without PCA (Sect. 3.5), but the injected model spectra were PCA-processed with three components following the iron procedure of Sect. 4.2; this asymmetry should be acknowledged, even though the qualitative Ca+ conclusion (model trails do not reproduce the observed red-starting trail) is unaffected.","section":"§4.3, Fig. 13"},{"comment":"The detections with S/N ≈ 4 (Mn, Cr) should be described as tentative given the S/N systematics discussed in Sect. 3.5 (PCA affects low-velocity residuals); the phase-resolved analysis is unaffected because all six trailed species have single-map S/N ≥ 7.9.","section":"§3.3, Table 3"},{"comment":"The abstract states categorically that \"Fe+ favors a stronger drag than Fe,\" whereas the conclusions phrase this as \"Fe+ seem to need more substantial drag\"; align the abstract with the caveats given in Sect. 4.3, or strengthen the conclusions once the required analysis is done.","section":"Abstract and §6"}],"recommendation":"major_revision","confidential_remarks":"The qualitative strong-drag result is likely to survive revision, and the paper is honest about its own limitations (Gaussian fits, PCA effects, retrieval biases), which is to its credit. The main risk is that the quantitative claims (τ_drag = 10^3–10^4 s, Fe+ stronger drag than Fe) are presented with more certainty than the diagnostics support; I expect that a re-analysis with two-Gaussian or full-shape CCF comparison will confirm the former and clarify the latter. The paper fits the journal well and builds fairly on Wardenier et al. (2024) and the WASP-76b/WASP-121b trail literature. No concerns about data provenance or citation practices."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid, honest paper that delivers genuinely new data for TOI-1518 b—six time-resolved species trails, a first RV mass, a 14-species detection set including VO with the HyVO line list—and makes a plausible case for strong drag. I buy the qualitative conclusion: the no-drag model is far too blueshifted, and the strong-drag models sit close to the data across most species. That part holds up.\n\nThe quantitative claim is shakier, and the paper itself tells you why. The model CCFs are double-peaked and asymmetric (Sect 4.3, Fig 10), yet the comparison is done with single-Gaussian fits. For Fe+, the drag discrimination rests almost entirely on the fitted position, because amplitude does not discriminate and the FWHM errors are too large. The position is exactly the parameter a single-Gaussian centroid of a double-peaked profile will bias most. Add the Mg contamination in the Fe+ trail (Fig B.3) and the Fe+ prefers τ_drag=10^3 s over 10^4 s is a motivated hypothesis, not a measurement. The quoted 10^3–10^4 s range for Fe is likewise conditional on the Gaussian parameterization being unbiased; the FWHM misestimation they note in Fig B.5 makes the error bars suspect. None of this overturns the strong-drag direction, but it does mean the magnetic-drag interpretation needs a double-Gaussian or full forward-model comparison before I'd call it established.\n\nWhat the paper does well besides the new data: the retrieval work is careful, with the Kp/Vsys bias for weak species explicitly demonstrated for VO, and the iron abundance result looks robust to their assumptions. The PCA robustness checks are thorough, and the Ca+ probing above the Roche lobe is a nice, testable idea. The SOPHIE RV detection is a real add and tightens the system parameters usefully.\n\nFor a dynamics or high-resolution spectroscopy person, this is worth reading and citing. It deserves a serious referee. The main revision should be a more appropriate CCF shape model and a significance estimate for the Fe+/Fe drag difference; with those, the magnetic-drag claim would be on much firmer ground. I'd send it to review.","headline":"A careful, useful paper that likely gets the qualitative strong-drag direction right, but the quantitative drag timescale and the Fe+/Fe magnetic interpretation are softer than the abstract implies, and the authors know it.","tokens_in":28502,"tokens_out":1406,"would_cite":true,"duration_ms":14697,"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 paper reports that time-resolved absorption trails of six chemical species in the ultra-hot Jupiter TOI-1518 b require strong atmospheric drag, with a drag timescale of roughly 10^3 to 10^4 seconds, and that the larger drag needed for…","keywords":["ultra-hot Jupiter","atmospheric dynamics","cross-correlation spectroscopy","general circulation models","magnetic drag","atmospheric escape","transmission spectroscopy","TOI-1518 b"],"falsifier":"Fit each binned cross-correlation function with a two-Gaussian model that lets the morning and evening limb signals have independent positions and widths, then recompute the six trails. If the Fe and Fe+ trails both match the tau_drag = $10^{4}$ second model within the quoted errors, the claimed preference for $10^{3}$-second drag in Fe+ and the magnetic-drag interpretation would not survive; if the double-Gaussian fits change the phase-dependent positions by more than the quoted uncertainties, the inferred drag timescale itself would need revision.","tokens_in":27149,"feed_emoji":"🌪️","tokens_out":7587,"duration_ms":61426,"temperature":0.7,"pith_summary":"This paper tries to establish how fast winds move in the ultra-hot Jupiter TOI-1518 b and why they move that way. Using two transits of high-resolution optical spectra, it detects 14 species and tracks the time-varying absorption trail of six of them: Fe, Fe+, Ca, Ca+, Na, and Mg. Comparing those trails with general circulation model predictions, it concludes that a strong drag is slowing the winds, with a drag timescale of about $10^{3}$ to $10^{4}$ seconds. The species-dependent trails, with Fe+ needing stronger drag than Fe, are read as evidence that magnetic (ohmic) drag acts more strongly on the hotter dayside, while Ca+ appears to probe gas above the Roche lobe, likely outflowing material. The paper also uses a radial-velocity mass measurement to anchor a retrieval analysis that finds a roughly solar iron abundance.","feed_headline":"Strong drag slows winds on ultra-hot Jupiter TOI-1518 b","feed_subtitle":"Comparing six species' absorption trails against circulation models pins the drag timescale near 10^3–10^4 seconds.","key_machinery":"The carrying object is the time-resolved cross-correlation trail: for each of six species, the line-center position measured from a Gaussian fit to the cross-correlation function in each of nine transit phase bins is plotted against orbital phase. The comparison is made with the same trails extracted from general circulation models in which all unresolved dissipation, including turbulent mixing, shocks, and magnetic drag, is parameterized by a single Newtonian drag timescale spanning $10^{3}$ to $10^{6}$ seconds. The species-dependence of the trails carries the argument: Fe+ and Fe probe different thermal regions, and Ca+ probes much higher altitudes, so the differences among the six trails separate the drag signal from geometry and rotation.","core_discovery":"The central claim is that TOI-1518 b's atmosphere is strongly braked: the measured phase-dependent Doppler shifts of six species fall between the predictions of general circulation models with drag timescales of $10^{3}$ and $10^{4}$ seconds, and the amplitude and width of the Fe signal favor $10^{4}$ seconds, while the Fe+ signal favors the stronger $10^{3}$-second drag. The paper interprets the Fe+/Fe difference as a sign of magnetic drag, since ionized iron forms preferentially on the hotter dayside, where ohmic dissipation is expected to be strongest. It further claims that Ca+ traces layers up to $10^{-8}$ bar, corresponding to an effective radius of 1.97 ± 0.04 R_p, essentially the Roche lobe radius, so its qualitatively different trail reflects outflow rather than the day-to-night circulation that shapes the other species. The detection of 14 species includes VO only with the newer HyVO line list, and the refined planetary mass of 1.83 ± 0.47 M_Jup from radial-velocity data anchors the Kp-Vres and retrieval analysis.","pith_inferences":["If the single-Gaussian fits bias the phase-dependent trail positions, the inferred drag timescale could shift; a reanalysis that fits the model's double-peaked cross-correlation functions with two Gaussians, one per limb, would test whether the claimed 10^3 to 10^4 second range survives.","The magnetic-drag interpretation could be tested by measuring trails of other ion/neutral pairs, such as Ti+/Ti and Ca+/Ca, across a sample of ultra-hot Jupiters with different dayside temperatures and expected magnetic field strengths.","The grazing geometry of TOI-1518 b amplifies the leading/trailing limb asymmetry, so comparing its trails with a low-impact-parameter ultra-hot Jupiter could separate wind shifts from rotational and geometric contributions.","Since Ca+ seems to trace outflow above the Roche lobe, simultaneous observations of Ca+ with other escape tracers could constrain mass-loss rates and the geometry of the outflow in this and similar systems."],"forward_implications":["If the strong-drag conclusion is correct, the equatorial jet is suppressed and the day-to-night flow dominates, so TOI-1518 b's circulation is much more symmetric than drag-free models predict.","The Fe+/Fe difference, interpreted as magnetic drag, implies that ionized species in ultra-hot Jupiters can serve as diagnostics of the altitude- and temperature-dependent dissipation that neutral species cannot reach.","Because Ca+ appears to probe gas above the Roche lobe, its lines can be used as a transmission-spectroscopy probe of atmospheric escape and outflow geometry in similar planets.","The roughly solar iron abundance, measured with a mass refined by radial velocities, supports the picture that refractory elements remain gaseous in ultra-hot Jupiters and can trace the host star's metallicity.","The demonstrated Kp/Vres bias in multi-species retrievals implies that abundance estimates for species detected away from the dominant iron signal require per-species velocity offsets or single-species retrievals to be reliable."],"supporting_citations":[{"why":"Discovery and initial system parameters of TOI-1518 b; the baseline whose Kp, mass, and ephemerides the paper refines.","marker":"Cabot et al. (2021)"},{"why":"First time-resolved blueshifting iron trail in an ultra-hot Jupiter; the phenomenon this paper measures for TOI-1518 b.","marker":"Ehrenreich et al. (2020)"},{"why":"Model framework linking limb asymmetry and winds to the observed trail blueshift; basis for interpreting TOI-1518 b's trails.","marker":"Wardenier et al. (2021)"},{"why":"Similar strong-drag conclusion for WASP-121b using the same GCM-versus-trail comparison; a direct precedent for this work.","marker":"Wardenier et al. (2024)"},{"why":"Data-reduction and cross-correlation methodology for the high-resolution spectrograph used here.","marker":"Pelletier et al. (2023)"},{"why":"Radiative-transfer code that converts general circulation model outputs into phase-dependent transmission spectra.","marker":"Lee et al. (2022)"},{"why":"Shows ohmic drag is stronger on the dayside and at lower pressure; supports the Fe+/Fe magnetic-drag interpretation.","marker":"Beltz et al. (2022)"},{"why":"HyVO line list whose use enables the VO detection; load-bearing for the VO detection and its retrieval discussion.","marker":"Bowesman et al. (2024)"}],"fun_headline_variants":["Fe+ reveals magnetic drag on ultra-hot Jupiter TOI-1518 b","Six-species trails pin down strong atmospheric drag on TOI-1518 b","Magnetic drag confirmed in ultra-hot Jupiter TOI-1518 b","TOI-1518 b: strong winds braked by magnetic effects","Strong drag detected in ultra-hot Jupiter TOI-1518 b"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The inferred drag timescale assumes that each species' cross-correlation trail is faithfully summarized by a single Gaussian fit, even though the model trails are double-peaked and the observed one-dimensional profiles deviate from Gaussian shapes; a phase-dependent bias in those fits would move the inferred drag toward weaker or stronger values.","fun_headline_variants_meta":{"raw":{"variants":["Fe+ reveals magnetic drag on ultra-hot Jupiter TOI-1518 b","Six-species trails pin down strong atmospheric drag on TOI-1518 b","Magnetic drag confirmed in ultra-hot Jupiter TOI-1518 b","TOI-1518 b: strong winds braked by magnetic effects","Strong drag detected in ultra-hot Jupiter TOI-1518 b"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000334,"raw_usage":{"total_tokens":1947,"prompt_tokens":1134,"completion_tokens":813,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":750,"completion_tokens_details":{"reasoning_tokens":715}},"tokens_in":750,"tokens_out":813,"duration_ms":6888,"temperature":1.0,"reasoning_tokens":715,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T04:17:42.849643+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fit each binned cross-correlation function with a two-Gaussian model that lets the morning and evening limb signals have independent positions and widths, then recompute the six trails. If the Fe and Fe+ trails both match the tau_drag = $10^{4}$ second model within the quoted errors, the claimed preference for $10^{3}$-second drag in Fe+ and the magnetic-drag interpretation would not survive; if the double-Gaussian fits change the phase-dependent positions by more than the quoted uncertainties, the inferred drag timescale itself would need revision.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Discovery and initial system parameters of TOI-1518 b; the baseline whose Kp, mass, and ephemerides the paper refines."},{"cited_title":"P., Parmentier , V., Line , M","cited_arxiv_id":null,"evidence_quote":"Similar strong-drag conclusion for WASP-121b using the same GCM-versus-trail comparison; a direct precedent for this work."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Radiative-transfer code that converts general circulation model outputs into phase-dependent transmission spectra."},{"cited_title":"A., Qu , Q., McKemmish , L","cited_arxiv_id":null,"evidence_quote":"HyVO line list whose use enables the VO detection; load-bearing for the VO detection and its retrieval discussion."}],"review_version":1}