{"id":"e7da88c5-bb5e-41ae-90e6-7d516bafe7a6","arxiv_id":"2501.13575","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"New CHEOPS transits show AU Mic c's mid-transit times shifted ~80 minutes later by 2023, and a dynamical model attributes the timing variations to a 0.203 Earth-mass planet d on a ~12.6-day orbit.","lead":"Using CHEOPS space telescope data from 2022-2023, the team measured when two known planets around the young star AU Mic cross its face, and found their transit times are shifting by up to about 80 minutes. The shifts can be matched by adding a small third planet, roughly 0.2 Earth masses, on a 12.6-day orbit, though the exact configuration is not yet pinned down.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 12.617-d, 0.203-M_E planet d solution is a property of the BIC-selected configuration (2), which yields an AU Mic c mass conflicting with published RVs; alternative configurations (3)/(4) give different d parameters.","rationale":"The paper does several things well: the CHEOPS reductions are described, the TTV measurements from TESS, Spitzer, and CHEOPS are combined, the failure of the two-planet model is clearly shown in Fig. B.1, and the authors are explicit that the exact configuration is not determined. A conditional verdict is therefore appropriate, and this stress-test does not change it. The concern is specifically about the strongest_claim as phrased: a single planet d period and mass are given in the abstract and results, but Table B.1 shows that these values come only from configuration (2). That configuration yields M_c = 4.40 M_E, which the authors themselves note conflicts with several independent RV analyses, while configuration (3), with a c mass consistent with those analyses, is only modestly disfavored by BIC and gives different planet d parameters. Configuration (4), which actually includes the RVs, gives yet another solution. Therefore the headline d parameters are not robust to model choice, even before considering stellar activity. The reader's weakest_assumption identified sparse data and the possibility of activity-induced systematics; my concern is more specific and is already visible in the paper's own model comparison: the quoted d solution depends on an unconstrained or possibly incorrect c mass. This does not invalidate the existence of a third body, but it does invalidate presenting one configuration's parameters as the system's parameters. The correct response remains CONDITIONAL, so the recommendation is UNCHANGED.","tokens_in":25308,"tokens_out":8524,"duration_ms":78912,"concrete_test":"Re-run TRADES configuration (2) with an additional Gaussian prior on M_c centered on the RV-derived mass (e.g., 22.2 ± 6.7 M_E from Zicher et al. 2022) and recompute the MAP and HDI for P_d and M_d. If the posterior shifts away from 12.617 d and 0.203 M_E toward the ~13.6-13.9 d and ~0.24-0.98 M_E solutions of configurations (3) and (4), the headline planet d parameters are confirmed to depend on an unconstrained and likely incorrect c mass.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is configuration (2) in Table B.1: P_d = 12.6173 d and M_d = 0.203 M_E. For this to be the headline result, configuration (2) must be the correct dynamical interpretation. That is not established even under the pure-gravity assumption. Configuration (2) is the BIC-selected model, but its MAP mass for AU Mic c is M_c = 4.40 M_E, which the authors state disagrees with Cale et al. (2021), Zicher et al. (2022), and Donati et al. (2023). Configuration (3), which gives M_c = 14.24 M_E and is consistent with those RV-based masses, has a BIC only 16 points higher and yields P_d = 13.614 d and M_d = 0.242 M_E. Configuration (4), which includes the actual RV data, yields yet another solution: P_d = 13.928 d and M_d = 0.983 M_E. Thus the 12.6-day, 0.203-M_E planet d is not a robust inference from the TTV data; it is tied to a particular model choice that forces an implausibly low mass for c. The authors acknowledge the configuration uncertainty, and even state that the fit has more derived parameters than available transits for AU Mic c (Sect. 4.2), but the abstract and results nevertheless promote configuration (2) as the solution. A secondary hazard is that the activity-driven biases documented in Sect. 4.1 can affect transit shape and mid-transit times, but the primary, already-evident problem is that the quoted d parameters do not survive a change of configuration that is contained in the paper itself.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents CHEOPS transit photometry of AU Mic b and c from 2022–2023, combines these with earlier TESS, Spitzer, and CHEOPS measurements to construct O–C diagrams, and reports a ~1150-day TTV superperiod for planet b. Using the TRADES dynamical code, the authors fit four configurations of the system and claim that the observed TTVs of both planets require a third planet, AU Mic d, with a BIC-preferred solution (configuration 2) of orbital period 12.6173 days and mass 0.203 M_Earth, along with a misaligned, likely retrograde orbit for planet c. The paper explicitly acknowledges that the limited number of transits prevents a unique dynamical interpretation and that the exact configuration and parameters remain undetermined.","tokens_in":25732,"tokens_out":4067,"duration_ms":35014,"significance":"If confirmed, the presence and characterization of a third planet in the young, active AU Mic system would be a valuable contribution to our understanding of early planetary system evolution. The paper's strengths include careful handling of stellar activity (semi-automated flare removal, roll-angle correction), use of public fitting and dynamical tools (Allesfitter, TRADES), and transparent presentation of posterior distributions in appendices. However, the headline planet-d parameters are not robust to the model configuration choices: configurations (3) and (4), which are also reported in Table B.1, give period and mass values that differ by far more than the quoted uncertainties. The paper is therefore better read as a status report on possible AU Mic d solutions than as a definitive detection of a specific planet, and the significance of the central claim is accordingly limited.","major_comments":[{"comment":"The abstract and Sect. 4.2 promote configuration (2) as the main solution on the basis of the lowest BIC, but the BIC difference relative to configuration (3) is only about 16 (−536.6 vs −520.3), and configuration (3) yields P_d = 13.614 d and M_d = 0.242 M_E while configuration (4) yields P_d = 13.928 d and M_d = 0.983 M_E. Because these values differ by far more than the stated 1σ uncertainties of configuration (2), the quoted P_d = 12.6173 d and M_d = 0.203 M_E are not a robust inference; they are conditional on a weakly preferred model. Please present the full spread of solutions as the headline or add a model-averaging or cross-validation argument to justify the emphasis on configuration (2).","section":"Sect. 4.2 and Table B.1"},{"comment":"Configuration (2), the BIC-selected solution, yields M_c = 4.40 M_E, which the authors themselves state disagrees with published RV masses (Cale et al. 2021; Zicher et al. 2022; Donati et al. 2023), whereas configuration (3) yields M_c = 14.24 M_E and is consistent with those masses. Since radial velocities are included only in configuration (4), the BIC ranking predominantly reflects the TTV data. The claim that configuration (2) is the main solution is therefore not well supported unless the authors can explain why the preferred TTV-only solution drives planet c to an implausibly low mass.","section":"Table B.1 and Sect. 4.2"},{"comment":"The authors note that the dynamical fit has 12 derived parameters while AU Mic c has only 9 observed transits (Tables 3–4). This ratio indicates a high risk of overfitting, and the quoted posterior uncertainties may not reflect the model uncertainty inherent in the configuration choice. Please provide an explicit count of the independent constraints per planet and consider a leave-one-out or hold-out test (e.g., predicting the 2023 transit times from a fit to pre-2023 data only) to demonstrate that the reported configurations have predictive power rather than simply absorbing noise.","section":"Sect. 4.2"},{"comment":"The abstract states that the observed TTVs 'can be explained by a third planet' with a specific period of ~12.6 days and mass of 0.203 M_E, while also stating that the exact configuration and planetary parameters could not be determined. These two statements are in tension. Please reframe the abstract to emphasize the range of solutions across configurations (e.g., P_d ≈ 12.6–13.9 d, M_d ≈ 0.20–0.98 M_E) and to make clear that the quoted numbers are from the BIC-preferred configuration only, or state explicitly that the specific values are not yet secure.","section":"Abstract"}],"minor_comments":[{"comment":"The text 'the estimated 4.4 M ⊙ for AU Mic c' should read '4.4 M⊕'; the solar-mass symbol is a typo that obscures the meaning.","section":"Sect. 4.2"},{"comment":"The reference list contains two entries labeled 'Szabó et al. 2022' (Sz2021 and Sz2022 in Tables 5 and 6); please disambiguate them (e.g., Szabó et al. 2021 and Szabó et al. 2022) in the table notes to avoid confusion.","section":"References"},{"comment":"The caption describes the harmonic fit as 'without any dynamical interpretation'; please clarify that the quoted 1150-day superperiod is the period of this empirical fit and not a dynamically derived quantity.","section":"Fig. 3"},{"comment":"The symbols ⋆, ⋆⋆, ⋄, ‡, and • used in the column headers are defined in the notes, but the mapping to specific papers is easy to miss; please make the notes more explicit, e.g., '⋆ This work (2023)', '⋆⋆ Szabó et al. (2022, 2021 data)', etc.","section":"Tables 5 and 6"},{"comment":"The phrase 'distributed to 5–5 self-consistent groups of transits' is unclear; consider rewording to 'five independent groups of transits for each planet' or similar.","section":"Sect. 4.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of A&A and the observational work is careful, but the central claim about planet d's specific parameters is not supported by the internal evidence: the BIC-selected configuration disagrees with published RV masses, and alternative configurations in the same table give substantially different parameters. A revision that presents the configuration spread as the main result, or that adds a robustness test, would be needed before publication. The authors' own abstract caveat is actually the most accurate summary of the current state of the analysis."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper delivers a solid set of new CHEOPS transit times and a refined TTV superperiod for AU Mic b, but the specific \"planet d\" parameters in the abstract—P=12.6 d, M=0.203 M_E—are tied to a BIC-selected dynamical configuration that disagrees with published RV masses for AU Mic c. The alternative configurations in the paper's own Table B.1 give noticeably different d parameters, so the headline should be read as provisional, not established.\n\nThe observational work is careful. The flare removal and roll-angle correction are reasonable, the mid-transit times are clearly tabulated, and the 80-minute TTV of AU Mic c is robust. Revising the superperiod to ~1150 d from a longer baseline than Szabó et al. 2022 is a genuine improvement. The paper is transparent about its four dynamical configurations and even prints the caveat about 12 fitted parameters exceeding the number of c transits.\n\nThe problem is emphasis. Configuration (2) wins by BIC but forces Mc=4.4 M_E, which the authors themselves note conflicts with Cale et al. 2021, Zicher et al. 2022, and Donati et al. 2023. Config (3) is only ΔBIC~16 away, gives Mc=14.2 M_E consistent with those RV studies, and shifts planet d to P=13.6 d, M=0.24 M_E. Config (4), including RVs, gives P=13.9 d and M~1 M_E. So the abstract's \"~12.6 d, 0.203 M_E\" is not the robust inference; it is one of several allowed solutions, selected by a model that disagrees with independent mass constraints. The retrograde-orbit claim for c comes from the same configuration, so it carries the same caveat.\n\nNone of this is fatal—the paper does state that the exact configuration is undetermined. But the framing in the abstract and results leans on the BIC-selected solution without giving the spread across configurations equal weight. A revision that led with the allowed range rather than the single MAP solution would be more honest.\n\nWho it's for: people working on AU Mic and TTV modeling of active M dwarfs. The transit times and superperiod deserve to be in the literature. The planet d mass is a target for future observations, not a measurement.\n\nRecommendation: send this to peer review. A good referee can push the authors to sharpen the abstract and report the configuration spread. It is a useful, careful paper with a framing problem, not a flawed measurement.","headline":"Useful new CHEOPS transit times and a refined TTV superperiod, but the abstract's planet d mass and period are configuration-dependent and should be presented as provisional.","tokens_in":26790,"tokens_out":2528,"would_cite":true,"duration_ms":22664,"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":"AU Mic's transit-timing record from 2018 to 2023 points to a third planet, AU Mic d, a roughly 0.2-Earth-mass body on a 12.6-day orbit between the two known sub-Neptunes.","keywords":["AU Mic","transit-timing variations","CHEOPS","exoplanets","sub-Neptune planets","dynamical modeling","stellar activity","third planet"],"falsifier":"Observe the next several AU Mic b and c transits with CHEOPS and TESS: the configuration-(2) solution predicts that the 2025 TESS sector and 2026 CHEOPS transits will continue the TTV curve with the 12.6-day perturbation pattern. If the timing residuals instead follow only the smooth roughly-1150-day superperiod with no 12.6-day component, or if the c offset returns to the old ephemeris, then the inferred planet d does not exist.","tokens_in":25114,"feed_emoji":"🪐","tokens_out":12637,"duration_ms":94790,"temperature":0.7,"pith_summary":"AU Mic is a 22-million-year-old M dwarf whose two transiting sub-Neptunes, b and c, have been arriving at their transits minutes early or late, a signature of gravitational tugs from other bodies. The paper combines new CHEOPS photometry from 2022-2023 with earlier TESS, Spitzer, and ground-based timings to map these transit-timing variations over 2018-2023. It finds that AU Mic c's transits ran about 80 minutes late by 2023 and that the TTV superperiod of AU Mic b is about 1150 days. Dynamical modeling then shows that the observed timing curve can be explained by a third planet, AU Mic d, on a 12.6-day orbit between b and c, with a mass of about 0.2 Earth masses. If correct, the result would confirm a low-mass 'middle d' architecture for a young planetary system and demonstrate that TTV observations can expose sub-Earth planets that are too small to transit.","feed_headline":"A 0.2-Earth planet drives AU Mic's shifting transits","feed_subtitle":"CHEOPS timing data place a sub-Earth world between two young sub-Neptunes and hint AU Mic c orbits backward.","key_machinery":"The engine of the analysis is the transit-timing variation (TTV) signal, meaning the O-C residuals between observed mid-transit times and a linear ephemeris; these residuals encode the gravitational signature of any unseen companions. The transit times themselves come from Allesfitter fits to CHEOPS light curves, with flares removed semi-automatically and stellar variability absorbed by a Gaussian-process model. The TTVs are then fed to TRADES, an N-body dynamical code that integrates the full system and simultaneously fits planet-to-star mass ratios, periods, eccentricities, and orbital angles, producing the four configurations whose Bayesian information criterion values decide which solution is preferred. The load-bearing identity is the mapping from the shape and superperiod of the O-C curves to the mass and period of the perturbing planet, here yielding the 12.6-day, 0.2-Earth-mass solution for AU Mic d.","core_discovery":"On the paper's own account, the central discovery is that the combined 2018-2023 transit-timing record of AU Mic b and c cannot be reproduced by a two-planet system; the best-fitting model, configuration (2), requires a third body named AU Mic d with orbital period $12.6173^{+0.0050}_{-0.0114}$ days and mass $0.203^{+0.022}_{-0.024}$ $M_\\oplus$, placing it between the orbits of b and c. The same dynamical solution yields a TTV superperiod of roughly 1150 days for planet b and a peak-to-peak TTV amplitude of about 24.5 minutes. It also implies that AU Mic c moves on a misaligned, presumably retrograde orbit, with a projected spin-orbit angle near $171.9^\\circ$, while planet b remains well aligned. The paper emphasizes that the exact configuration is not uniquely pinned down because the number of fitted parameters (12) exceeds the number of transits available for AU Mic c, so the derived d parameters mark the preferred region of parameter space rather than a final measurement.","pith_inferences":["A natural extension is that the non-transiting geometry of d could be verified independently with high-precision radial velocities, since a 0.2-Earth-mass body on a 12.6-day orbit around a 0.6-solar-mass star would imprint a small but coherent periodic signal.","The 1150-day superperiod gives observers a scheduling tool: TTVs measured near the extrema of AU Mic b's O-C curve carry the most information about d, so future observations should target those phases.","Because no stellar activity model entered the dynamical fit, the cleanest test of the d claim is to repeat the fit with spot-induced timing offsets included; a surviving 12.6-day component would make the planet interpretation robust.","The same O-C inversion strategy could be exported to other active young M-dwarf systems, where stellar noise currently hides sub-Earth planets; AU Mic provides a proof of concept that high-cadence space photometry plus TTV modeling can reach sub-Earth masses."],"forward_implications":["If AU Mic d is real, it joins the small set of sub-Earth planets found through transit timing, with a mass just over twice that of Mars.","The 12.6-day period places d between b and c, confirming the 'middle d' family of solutions proposed in earlier work.","A misaligned, retrograde orbit for AU Mic c, if confirmed, would indicate that the outer planet experienced a scattering or inclination excitation event rather than forming in a perfectly coplanar disk.","Continued CHEOPS and TESS monitoring over the next few years should either converge on the configuration (2) parameters or reveal that the timing residuals have another source.","The year-to-year changes in transit depth and impact parameter of both planets are attributed to starspot distributions, so radius values from any single epoch are biased by stellar activity."],"supporting_citations":[{"why":"Predicted the existence of AU Mic d from earlier TTVs and defined the 12.6-12.7 day solutions that this paper's configuration (2) confirms and narrows to the 'middle d' family.","marker":"Wittrock et al. (2023)"},{"why":"Provided the prior CHEOPS transit timings, the TTV detection for AU Mic b and c, and the ephemerides used as priors in the Allesfitter fits.","marker":"Szabó et al. (2022)"},{"why":"Reported the discovery of AU Mic b and c and supplied the stellar mass and reference ephemeris used in the dynamical integrations.","marker":"Plavchan et al. (2020)"},{"why":"Introduced TRADES, the N-body integrator used to model all four dynamical configurations.","marker":"Borsato et al. (2014)"},{"why":"Provided Allesfitter, the software used to fit the CHEOPS transit light curves and extract mid-transit times.","marker":"Günther & Daylan (2021)"},{"why":"Supplied the PIPE PSF photometry extraction that produced the high-cadence CHEOPS light curves used in the analysis.","marker":"Brandeker et al. (2024)"},{"why":"Supplied the combined TESS ephemerides and mean orbital periods that anchor the O-C calculations for both planets.","marker":"Gilbert et al. (2022)"},{"why":"Provided the SERVAL radial-velocity measurements and published planetary masses used in configuration (4) and in the mass comparisons.","marker":"Zicher et al. (2022)"},{"why":"Reported independent Rossiter-McLaughlin evidence for a misaligned AU Mic c orbit, supporting the retrograde geometry inferred here.","marker":"Yu et al. (2025)"}],"fun_headline_variants":["Sub-Earth world shakes AU Mic's transit timings","Retrograde hint: AU Mic c may orbit backward","Third planet, 0.2 Earth masses, drives AU Mic TTVs","CHEOPS finds planetary interloper in AU Mic system","A 0.2-Earth planet between two sub-Neptunes explains TTVs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the observed shifts in transit times come entirely from gravity and that fewer than a dozen transit groups suffice to pin down a 12-parameter model; if starspots or instrumental quirks mimic those shifts, the 0.2-Earth-mass planet d may be an artifact.","fun_headline_variants_meta":{"raw":{"variants":["Sub-Earth world shakes AU Mic's transit timings","Retrograde hint: AU Mic c may orbit backward","Third planet, 0.2 Earth masses, drives AU Mic TTVs","CHEOPS finds planetary interloper in AU Mic system","A 0.2-Earth planet between two sub-Neptunes explains TTVs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000269,"raw_usage":{"total_tokens":1731,"prompt_tokens":1166,"completion_tokens":565,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":782,"completion_tokens_details":{"reasoning_tokens":472}},"tokens_in":782,"tokens_out":565,"duration_ms":5207,"temperature":1.0,"reasoning_tokens":472,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T15:48:42.825177+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe the next several AU Mic b and c transits with CHEOPS and TESS: the configuration-(2) solution predicts that the 2025 TESS sector and 2026 CHEOPS transits will continue the TTV curve with the 12.6-day perturbation pattern. If the timing residuals instead follow only the smooth roughly-1150-day superperiod with no 12.6-day component, or if the c offset returns to the old ephemeris, then the inferred planet d does not exist.","supporting_citations":[{"cited_title":"A., & Morris , B","cited_arxiv_id":null,"evidence_quote":"Supplied the PIPE PSF photometry extraction that produced the high-cadence CHEOPS light curves used in the analysis."},{"cited_title":"A., Barclay , T., Quintana , E","cited_arxiv_id":null,"evidence_quote":"Supplied the combined TESS ephemerides and mean orbital periods that anchor the O-C calculations for both planets."}],"review_version":1}