XSci

Transit-geometry consistency and noise sensitivity in the archival Rossiter-McLaughlin signal of WASP-117 b

Chandan Kumar Shukla

Submitted by Chandan Kumar Shukla

Published November 14, 2026 · Version v1, November 14, 2026 · DOI 10.66977/xsci.2611.002c

Astronomy

Abstract

Photometric constraints used in a Rossiter-McLaughlin (RM) analysis must describe the same eccentric orbit as the spectroscopic model. We examine this consistency for WASP-117 b using 23 public two-minute TESS transits from ten sectors and 98 public ESPRESSO radial velocities. A documented geometry-prior center, a/R = 13.12 and i = 88.15 deg, predicts a 7.223-hour transit when combined with e = 0.302 and omega = 242 deg, whereas its circular interpretation predicts 5.844 hours. The TESS light curves independently favor a/R = 16.87 (-0.45, +0.41) and impact parameter b = 0.446 (-0.029, +0.024), including uncertainty in the external eccentricity constraint. In a controlled RM comparison with identical white-noise assumptions, replacing the documented geometry priors with the TESS constraint shifts the posterior median projected angle from 34.4 deg to 49.3 deg. A fitted velocity trend preserves a comparable shift. A flexible correlated-noise model substantially weakens the angular constraint, so we do not identify the latter value as a definitive corrected obliquity. The result is a reproducible geometry-consistency diagnostic and a conditional, same-data sensitivity measurement, rather than a new detection of misalignment or orbital evolution.

View PDF

Review reports

xPeerd (AI review)

November 14, 2026

AI Review (xPeerd — DAReview)

Summary

This manuscript studies whether transit-geometry constraints transferred into an eccentric Rossiter–McLaughlin analysis of WASP-117 b are internally consistent. Using 23 public TESS transits and 98 public ESPRESSO radial velocities, the author argues that the imported geometry centered on a/R = 13.12 and i = 88.15° is photometrically inconsistent when combined with the eccentric spectroscopic model, and that replacing it with a TESS-derived geometry shifts the inferred projected spin-orbit angle by about 15° under matched white-noise assumptions. The paper is strongest where it is explicit about scope, limitations, and reproducibility; it repeatedly avoids overstating its contribution and carefully distinguishes a conditional sensitivity analysis from a definitive new obliquity measurement . The methods are generally transparent, with documented priors, selection rules, robustness checks, and a reproducibility archive described in detail . The main weaknesses are methodological rather than rhetorical: the argument depends on approximate transfer of photometric information into the RV model rather than a joint likelihood, relies on archive pipeline RVs rather than re-extracted spectra, uses an approximate Gaussian-line RM prescription, and includes GP stress tests that the author admits are not fully converged in all parameters . I do not see strong paper-mill or fabrication signals; the manuscript is unusually cautious and technically self-limiting, though the declared use of generative AI for drafting and code preparation means especially careful human verification is warranted .

Potential Major Revisions

  1. The central inference would be methodologically stronger if the paper either implemented, or more forcefully justified not implementing, a full joint photometry–spectroscopy model. The manuscript itself acknowledges that “This approximate transfer retains the main photometric covariance… It is not a substitute for a full simultaneous photometry–spectroscopy likelihood” on page 2, Section 3.1 . Because the headline claim is that changing geometry information alone materially alters the inferred projected angle, the approximation used to transfer the TESS constraint into RV space becomes structurally important. At present, the paper shows that the effect is robust within its chosen pipeline, but not necessarily that the magnitude of the effect would persist under a unified generative model. A major revision should either add a simultaneous fit, or provide a stronger sensitivity analysis demonstrating that the conclusion is not an artifact of the Gaussian transfer and added systematic floors.

  2. The manuscript should make clearer, earlier, and more quantitatively, how much of the angle shift can truly be attributed to geometry rather than to model mismatch with earlier literature. The author writes on page 2, Section 2.2, that “the present comparison keeps all 98 DRS 2.2.1 velocities. This difference, our limb-darkening and jitter priors, and the approximate RM prescription prevent a direct attribution of every difference from a published angle to geometry alone” . This is an important caveat, but it currently undercuts the rhetorical force of the comparison unless elevated into the main interpretive logic. The paper should explicitly separate two claims: first, the imported geometry is photometrically inconsistent; second, within this specific matched reanalysis, substituting geometry shifts λ. Those are both defensible, but they are not equivalent to correcting prior published obliquity work.

  3. The Gaussian-process section presently weakens interpretability because the paper includes results that are explicitly not fully converged. On page 4, Section 3.3/4 transition, the author states that “Of the four 16,000-step GP stress tests, 3 do not reach that threshold in every parameter. Their slowly sampled angular tails are explicitly flagged… and their displayed intervals are exploratory” . This honesty is commendable, but if the GP analyses are displayed prominently, readers may still over-read them. A major revision should either improve GP convergence materially, simplify the GP presentation to non-quantitative diagnostics, or move the non-converged outputs to supplement-only status. Otherwise the paper risks mixing solid inferences with exploratory ones in the main results stream.

  4. The treatment of the photometric discrepancy should be strengthened statistically. On page 5, Section 4.1, the manuscript reports large objective-function differences such as “χ² = 1762.9, versus 251.3 when geometry is free,” but immediately notes that these are “descriptive objective values only” and not calibrated significance statements because of correlated bins and unavailable prior posterior structure . That caution is correct, but the result then sits in an awkward epistemic state: visually persuasive, numerically striking, yet not formally quantified. A major revision should add a more principled goodness-of-fit or posterior predictive assessment that respects the dependence structure and uncertainty propagation. Without this, the core transit-shape inconsistency remains plausible but not optimally demonstrated.

  5. The paper’s reproducibility standards are high, but the evidentiary base remains partially indirect because the RV analysis uses pipeline products rather than re-derived measurements from spectra. The manuscript states on page 2, Section 2.2, “These are public pipeline RV measurements: we did not independently re-extract velocities from the spectra” . That is acceptable for an archival reanalysis, but the manuscript’s critique of prior geometry transfer would be more compelling if accompanied by a short discussion of whether CCF-level choices, mask choice, or extraction details could interact with RM inference in ways relevant to the reported sensitivity. As written, the paper is careful, but it still leaves a methodological gap between public-product reanalysis and stronger claims about inferential correction.

  6. The novelty claim should be framed even more tightly to avoid ambiguity. The paper appropriately says in the abstract that the work is “a reproducible geometry-consistency diagnostic and a conditional, same-data sensitivity measurement, rather than a new detection of misalignment or orbital evolution” . It later states, on page 8, that “The contribution is this documented consistency and sensitivity analysis. It is neither a newly discovered planet nor a new measurement of orbital precession” . This is admirably restrained. However, because the manuscript still presents shifted angle posteriors in a way that may attract attention, the title, abstract, and discussion should all consistently foreground “consistency test” over “revised angle implication.” That would better align the messaging with the actual evidentiary strength.

  7. Anti-fraud / coherence assessment: I do not detect classic paper-mill traits such as inflated novelty, incoherent terminology, unverifiable datasets, or fabricated precision. In fact, the manuscript repeatedly narrows its claims and discloses weaknesses. However, one integrity-related issue requires explicit editorial attention: the preparation note says “Generative-AI assistance was used for literature retrieval, analysis-code preparation and manuscript drafting” . Given the technical nature of the work, the author should provide a concise statement clarifying how all mathematical derivations, code outputs, table values, and citations were independently verified by the author. This is not evidence of misconduct; it is a transparency requirement.

Potential Minor Revisions

  1. Title and framing consistency: The title emphasizes “noise sensitivity,” but the strongest result is arguably the transit-geometry inconsistency. Consider rebalancing the title or subtitle so that the geometry-consistency test is foregrounded more clearly than the downstream angle sensitivity .

  2. Page 1, Abstract: The sentence “A documented geometry-prior center, a/R = 13.12 and i = 88.15°, predicts a 7.223-hour transit when combined with e = 0.302 and ω = 242°, whereas its circular interpretation predicts 5.844 hours” is precise but dense. Consider defining in one phrase why this contrast matters for readers outside the immediate RM niche .

  3. Page 1, Section 1: The sentence “The motivating question is narrower” is good, but the paragraph could benefit from a one-sentence explicit statement of hypothesis. At present, the motivation is inferable rather than formally posed .

  4. Page 1: The phrase “The table audit is explicitly version-specific” could be clarified by briefly stating why the distinction between arXiv versions and final published tables may matter for the imported priors .

  5. Page 2, Section 2.1: The event-selection rule is transparent, but the five interval constraints are hard to parse in prose. A small schematic or compact table would improve readability .

  6. Page 2, Section 3.1: Notation shifts between a/R, ã, and ãcirc/ãecc may challenge some readers. Standardize notation presentation and define all symbols once in a compact notation box or table .

  7. Page 2, Section 3.1: The sentence “The reported photometric intervals are the latter 16th–84th percentiles, not a joint Bayesian posterior” is useful; however, it would help to explain in one more sentence what inferential limitation follows from that distinction .

  8. Page 2–3: The transition from transit fitting to controlled RM inference is abrupt. Add a bridging sentence explaining exactly what information is carried forward and what is intentionally held fixed .

  9. Page 3, Table 1: The phrase “documented geometry” could be misread as “endorsed geometry” or “correct geometry.” Consider “imported literature-table geometry” or similarly neutral terminology, since the manuscript later critiques its consistency .

  10. Page 3: The sentence “These are disclosed sensitivity analyses, not independent confirmatory datasets or a retrospectively chosen discovery sample” is strong, but stylistically a little defensive. It could be simplified for smoother reading while preserving transparency .

  11. Page 4, Table 2: A note already says that marginal medians need not define a representative joint draw. Consider moving this warning into the main text as well, since it is essential to interpretation .

  12. Page 5, Figure 1 caption: The caption is informative, but readers may benefit from a direct reminder that orange and blue-green correspond to the two geometry scenarios used later in the RV comparison .

  13. Page 6, Table 4: Very useful table, but adding one derived column for median shift relative to the documented-geometry counterpart would make the sensitivity pattern visually clearer .

  14. Page 6–7: The sentence “The RV curves themselves fit comparably… illustrating why an apparently acceptable RV fit does not validate an imported transit-geometry prior” is important and could be highlighted more prominently in the discussion, possibly as a concluding sentence to Section 4.3 .

  15. Page 7, Section 4.4: The phrase “These outcomes are not adequately summarized as one universal inflation of white-noise error bars” is strong but somewhat abstract. Consider rephrasing with a more concrete explanation of what changes qualitatively in the posterior shape .

  16. Page 8, Conclusion: The sentence “scaling a single parameter is insufficient” is good, but it may help to explicitly name the parameter and the practical mistake being cautioned against, for clarity to non-specialists .

  17. References: The reference list is generally clean. Check consistency in punctuation and styling for software citations, especially rmfit, where the formatting differs somewhat from journal references .

  18. Preparation note: “Research manuscript for author review • 14 November 2026 • Not submitted” is clear, but the manuscript would benefit from a brief version number or revision identifier for traceability, especially since a reproducibility archive is supplied .

  19. Typographic/style issues: I do not see extensive obvious grammar problems in the retrieved text. The prose is generally polished. The main minor issues are density, occasional defensiveness, and notation burden rather than outright grammatical error .

  20. AI-content analysis: Because the manuscript openly declares AI assistance in “literature retrieval, analysis-code preparation and manuscript drafting,” the style should be reviewed for over-compressed summary sentences and formulaic caveat phrasing. I do not treat this as a defect by itself, but it increases the need for manual checks of every equation statement, reference claim, and numerical value .

Recommendations

  1. Preserve the paper’s current restraint. Its strongest quality is that it does not oversell the claim. Keep the contribution framed as a reproducible consistency test plus conditional sensitivity analysis, not a definitive revised obliquity measurement .

  2. Strengthen the central evidence by adding either:

    • a full joint photometry+RV model, or
    • a more extensive validation that the Gaussian-transferred photometric constraint does not materially distort the λ shift .
  3. Reduce interpretive risk in the GP section by improving convergence or relocating exploratory GP outputs out of the main inferential pathway .

  4. Add a statistically better-supported demonstration of transit-shape inconsistency than descriptive χ² contrasts alone .

  5. Make the comparison to prior literature more disciplined: distinguish “this imported geometry is inconsistent with the eccentric transit shape under our forward model” from “earlier published inferences were wrong.” The manuscript already hints at this distinction; it should become explicit throughout .

  6. Add a short verification statement covering AI-assisted drafting/code preparation, confirming author-side checking of derivations, implementation, and citations .

Overall, my assessment is that this is a serious and unusually transparent archival methods paper with a coherent scientific rationale and no obvious fabrication markers, but it still needs substantial revision before the main methodological conclusion can be considered fully secure.

Reviewer3 (AI review)

November 14, 2026

AI Review (reviewer3.com)

Overview

Transit Geometry Inconsistency Identification

Contribution

The author demonstrates that previously published transit-geometry prior centers for WASP-117 b yield light-curve shapes inconsistent with TESS photometry when combined with the planet's eccentric orbital parameters.

Geometry Sensitivity in Obliquity Modeling

Contribution

This study establishes that incorporating TESS-derived transit geometry constraints into Rossiter–McLaughlin radial velocity modeling shifts the inferred median projected stellar obliquity by 14° to 17° under matched noise assumptions.

Comments (4)

1. Missing uncertainties for photometric subset controls in Table 3 — Minor

Table 3. Reduction and subset controls. Entries are conditional best fits, not independent precision measurements. The four-event 2018 bootstrap is less informative and less nearly Gaussian than the full-sample constraint.

Table 3 reports only point estimates for the conditional best fits across the 2018 subset (a/R★ = 17.574, b = 0.352) and the sectors 96–106 subset (a/R★ = 16.842, b = 0.443), even though a 2018-only bootstrap was computed for use in Table 4. Including the bootstrap uncertainty intervals for the subset fits directly in Table 3 would allow readers to evaluate the stated lack of significant secular change in impact parameter between 2018 and 2025–2026.

2. Unjustified ad hoc systematic floors in photometric prior — Major

The Gaussian photometric constraint used in the RV fits is constructed in coordinates (ãg, b, k), using the mean and full covariance of the conditional event bootstrap. We add independent systematic floors of 0.05, 0.015 and 0.0003 in these respective coordinates. Sampling the RV model

In Section 3.1, independent systematic floors of 0.05, 0.015, and 0.0003 are added to the bootstrap covariance of (ãg, b, k) when constructing the Gaussian photometric constraint for the RV fits, and their numerical values are not reported inside the manuscript text. Please provide a quantitative justification for the magnitudes of these systematic floors and explicitly report the numerical mean vector and covariance matrix of the (ãg, b, k) prior in Table 1 or the text so the analysis is self-contained.

3. Unreported posterior distributions for nuisance RM parameters — Major

Projected rotation; line width v sin i★ ∈ (0.1, 6) km/s; β ∈ (0.5, 30) km/s, uniform

The Rossiter–McLaughlin anomaly depends strongly on stellar rotation (v sin i★) and line width (β), which are given broad uniform priors up to 6 km/s and 30 km/s in Table 1, as well as the posterior values of ã and i. Reporting the posterior medians and credible intervals for v sin i★, β, ã, i, and jitter across the two geometry scenarios in Table 4 would clarify whether the ~15° shift in λ is accompanied by unphysical shifts in stellar line broadening or rotation speed.

4. Missing numerical summary of GP hyperparameter posteriors — Major

For all 98 velocities, allowing covariance on tens of minutes to hours substantially broadens or redistributes the angle posterior (Figure 3). The corresponding GP analysis of the 92 best-precision velocities is less diffuse and closer to the robust white-noise solutions.

While Figure 3 displays the projected-angle intervals for the four Gaussian-process runs, the manuscript does not report the numerical posterior medians or intervals for λ or for the fitted Matérn-3/2 covariance amplitude and timescale parameters. Providing these numerical summaries in the text or a table would clarify what correlation timescales and amplitudes the GP models adopt when absorbing or broadening the RM anomaly in the 98-RV versus 92-RV datasets.

XSci (AI review)

November 14, 2026

Summary

WASP-117 b is an eccentric, roughly ten-day-period Saturn whose projected spin-orbit angle has been measured more than once, with results that do not agree. The authors ask whether transit-geometry constraints taken from the photometric literature can be transferred consistently into an eccentric Rossiter-McLaughlin (RM) likelihood. Using 23 public two-minute TESS transits and 98 archival ESPRESSO velocities, they show that the documented prior centre a/R* = 13.12, i = 88.15 deg forward-predicts a 7.223-hour transit when combined with e = 0.302 and omega = 242 deg, against 5.844 hours for its circular interpretation and an independently fitted a/R* = 16.87 (-0.45, +0.41), b = 0.446 (-0.029, +0.024). In a matched radial-velocity experiment changing only the geometry, the median projected angle moves from 34.4 deg to 49.3 deg; a correlated-noise model then broadens that posterior, and the authors decline to call 49.3 deg a corrected obliquity.

Contribution and Significance

The core contribution is a diagnostic and a quantified consequence. The diagnostic is arithmetically simple and therefore useful: with g = (1 + e sin omega)/sqrt(1 - e^2) = 0.769269 for this system, 13.12/g = 17.0552, which is what an eccentric fit to the TESS light curves returns. That correspondence identifies a circular-equivalent scaled separation standing in for an eccentric one, and the authors price it correctly: holding the data, the likelihood and all other priors fixed, they change only the geometry and measure the displacement in lambda. It is 14.2-17.0 deg across four matched pairs (Table 4), two to three times the formal uncertainty on any published angle for this star; since such angles feed obliquity catalogues and population studies, an error of that size is worth correcting.

What is genuinely new is narrower than the manuscript's length implies, and the authors are largely candid about it. The misalignment itself is not new (Lendl et al. 2014: beta = -44 +/- 11 deg; Carone et al. 2021: lambda = -46.9 (+5.5, -4.8) deg); the eccentric scaled separation near 17 is not new, and the value here is no more precise than Patel & Espinoza (2022); and the methodological lesson is established, Dawson & Johnson (2012) being cited and Kipping (2010), which treats exactly this error, not. What remains is the identification that one published RM analysis appears to have imported a circular-equivalent prior into an eccentric model, plus a same-data measure of the bias: real, but one object, one prior table, and no replacement obliquity.

Strengths

  • The controlled design is the right one, cleanly executed: Table 4 reuses the identical 98 velocities and shared priors across nine fits, isolating the geometry effect instead of inferring it from published numbers.
  • The photometric result survives every control that matters: unbinned fitting, SAP and PDCSAP, linear and quadratic baselines and leave-one-sector-out all return a/R* within 16.73-16.96 and b within 0.432-0.460 (Table 3).
  • The displacement survives the likelihood changes that matter: the Student-t and 92-best-precision variants move the absolute angles but preserve it (16.2 and 17.0 deg; Table 4).
  • The scoping is disciplined and the archive unusually complete: the notes to Tables 2-4 and Section 4.4 say what the numbers do not establish, lambda ~ 49 deg is never called a measurement, and header hashes, a pinned rmfit commit and fixed seeds are supplied.

Recommendation

Major Revision for Astronomy & Astrophysics. The central diagnostic is correct and checkable and the experiment that prices it is well designed; what is missing is the closure of the audit itself and the evidence behind the third headline result. All three concerns below can be met within the manuscript's own approach, with no new observations. Given the single-object scope, the A&A Research Note may be the right article type.

Major Concerns

  1. The audit of the cited work is not carried to the point where its central claim is established (Sections 1 and 5; Table 2). The manuscript states that "the publisher-formatted final tables were not independently retrieved in this analysis", yet its whole contribution is a claim that specific published values are internally inconsistent; the version of record must be checked. More consequentially, it audits a prior centre but never reports what the audited analysis obtained: the arXiv v1 cited gives ESPRESSO posterior medians a/R* = 13.486 (+0.22, -0.21), i = 87.574 (+0.108, -0.105) deg and lambda = 27.31 (+4.53, -4.30) deg, none of which appears here. That matters, because the headline comparison (7.223 h against 5.844 h) uses the prior centre, whereas Table 2 row 3 shows the published posterior geometry predicts 6.0747 h - a far smaller discrepancy whose inputs are never given. Report the audited posterior and its lambda, say which configuration each chi^2 in Section 4.1 refers to, and quantify the significance of 6.07 h against 5.856 h, the comparison that bears on the published result.

    The provenance of 13.12 is likewise left as a coincidence when it is checkable: Section 3.1 says the correspondence "does not prove how an upstream code obtained that value", but reference [5] fits (T0, P, b, R*/a, Rp/R*, C1, C2) with no eccentricity term, and its own comparison table already contrasts that a/R* with the earlier TESS value of 16.93. Establishing from its methods whether a circular orbit was assumed would turn the central diagnostic from conjecture into demonstration.

  2. The 14.9 deg shift is not attributed to a cause (Section 4.3; Tables 1 and 4). Under the documented geometry the RM model must describe a 7.22-hour transit while the event lasts about 5.86 hours, with the centre offset truncated at +/-10 min (Table 1), so it is misspecified in duration before chord position enters. Only the aggregate effect of changing a/R*, i and k together is reported, so a reader cannot tell whether lambda moves because the chord is too long or because the impact parameter is wrong (0.705 under the published posterior geometry, against 0.446 from TESS). Two more matched fits, one changing a/R* alone and one changing i alone, would identify the responsible parameter and turn a single-system result into a transferable rule. And since lambda = 34.4 deg is the output of a deliberately misspecified model, not the published 27.31 deg, state what the 14.9 deg displacement does and does not estimate.

  3. The correlated-noise conclusion rests on chains the authors report as unconverged (Sections 3.3 and 4.4; abstract; Figure 3). Section 3.3 states that "Of the four 16,000-step GP stress tests, 3 do not reach that threshold in every parameter", yet Section 4.4 concludes from them that the posterior is "substantially" broadened, and the abstract asserts that "A flexible correlated-noise model substantially weakens the angular constraint". This is one of the three results claimed in Section 5 and the reason given for not delivering a corrected obliquity, so it must be supported rather than flagged: run these fits to the standard met by the nine Gaussian fits, and tabulate the intervals, since no GP number appears anywhere. A GP whose amplitude reaches 20 m/s and whose timescale reaches six hours, against a 5.86-hour transit, can also absorb the RM signal by construction; constraining its hyperparameters from the out-of-transit baseline would make the test informative rather than near-certain.

Minor Comments

  1. Table 1 adopts omega ~ N(242 deg, 2.5 deg) where the audited table gives N(242, 2.3); the u2 and jitter departures are flagged but not this one.
  2. Give the a/R*, i, k, e and omega behind Table 2 row 3, so that calculation can be reproduced.
  3. The systematic floors of 0.05, 0.015 and 0.0003 (Section 3.1) are unjustified, and the k floor rivals the statistical uncertainty of (-0.00038, +0.00034).
  4. The symbol "ag" (Section 3.1) is never defined; state whether it is a/R* times g, and unify the notation, which appears as a/R_0, a-hat and a-tilde.
  5. Table 3's 2018-only subset (a/R* = 17.574, b = 0.352) carries no interval, yet Section 4.2 infers no significant secular change in b; give that bootstrap interval.
  6. Report the posterior v sin i* for each fit in Table 4, since it is strongly correlated with lambda here and should move diagnosably with it.
  7. Place the two flagged deviations (about 42 and 21 m/s at 09:21:58 and 09:10:19 UTC) in transit phase, which sets their leverage on lambda.
  8. Section 3.2 carries a stray Omega in "sigma_rot = v sin i Omega / 1.31" and leaves X undefined in "v_p = (v sin i) X"; also state the rmfit amplitude constant.
  9. Add Kipping, D. M. (2010), "Investigations of approximate expressions for the transit duration", MNRAS, 407, 301.
  10. Add Albrecht, S. H., Dawson, R. I., & Winn, J. N. (2022), "Stellar Obliquities in Exoplanetary Systems", PASP, 134, 082001, for why a 15-20 deg catalogue bias matters.
  11. State in Section 1 that Carone et al. fixed their other transit parameters to Lendl et al. (2014), and give that a/R*, which plausibly explains their agreement with the TESS-constrained magnitude.
  12. The audited ESPRESSO inclination posterior, 87.574 (+0.108, -0.105) deg, sits roughly 2.9 sigma from its own prior centre of 88.15 +/- 0.20 deg - independent support the authors do not use.