{"id":"7e9ea5a3-3fb4-4348-a59f-68deaed0dbc6","arxiv_id":"2504.13160","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":11,"one_line_summary":"A previously unknown non-transiting planet with a 355-day period and about 36 Earth masses is inferred in the Kepler-139 system from transit-timing and radial-velocity data.","lead":"Using transit-timing and radial-velocity data, the authors report evidence for a new non-transiting planet, Kepler-139f, situated between three smaller transiting planets and an outer giant planet in the Kepler-139 system. The result adds a rare example of a TTV-discovered hidden planet and supports the idea that outer giant planets can tilt inner planets out of view.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported 355-day period and 36-M⊕ mass rest on a ~3 m/s RV signal comparable to the adopted 2.9 m/s jitter; without activity/window checks, the RV selection of one of three TTV period families is not secure.","rationale":"The paper has genuine independent support: the five-planet model improves the TTV fit (Δχ2_TTV = 60) using publicly available Kepler photometry, and the predicted nontransiting geometry is consistent with the absence of detectable f transits. I do not see an internal contradiction that would force rejection. The load-bearing weak point is not the existence of some fifth body but the specific headline parameters. In Section 4.2 the authors themselves report three TTV-compatible period families, and the 354-day choice is made by a ~3 m/s RV signal. That amplitude is comparable to the 2.9 m/s jitter they fit, and no activity diagnostics are given. The paper even cites Weiss et al.'s periodogram showing the 350-day peak was not significant. The Bayes factor used to call the period choice 'decisive' is not a valid evidence ratio for a multimodal posterior. This is precisely the assumption the reader flagged: unmodeled RV systematics could remove one pillar and leave the period and mass unconstrained. Because the data and code are not released, the check cannot be performed from the paper alone. The appropriate verdict remains conditional: the discovery is plausible and valuable, but the specific period and mass should be treated as provisional pending activity, window-function, and proper evidence tests.","tokens_in":14429,"tokens_out":9436,"duration_ms":99414,"concrete_test":"Request the 38 Keck/HIRES epochs from the KGPS team and measure S-index, bisector span, and Hα from the same spectra; correlate these activity indicators with the RV residuals from the best four-planet joint fit. Then refit the five-planet model with a quasi-periodic Gaussian-process activity term and compare the 354/384/685-day period families using nested-sampling evidence rather than exp(Δχ2/2). If the ~355-day RV component is absorbed by the activity model or matches an annual window alias, and the 354-day family no longer wins the evidence comparison, the claimed P = 355 ± 2 days and M = 36 ± 10 M⊕ would not be supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.2 identifies three TTV-compatible period families (Pf ≈ 354, 384, and 685 days), and the RV data are what select the 354-day solution and set the planet's mass. For mf = 36 M⊕ at Pf = 355 days around a 1.078 M☉ star, the expected RV semiamplitude is K ≈ 3 m/s, which is equal to the fitted RV jitter (2.9 m/s) and only ~1.5 times the 2 m/s formal uncertainties. The paper presents no activity indicators for the 38 Keck RVs, and it itself notes that Weiss et al. found the ~350-day RV periodogram peak not significant, with seven taller peaks. Ground-based RV windows have strong annual aliases, so a 355-day signal is hard to distinguish from a one-year systematic without a window-function analysis. The quoted Bayes factor preferring the 354-day family, exp(Δχ2/2) ≳ 10^8, rests on approximating a posterior explicitly described as multimodal by a single multivariate Gaussian, which is not a valid evidence comparison. Thus, if the 3 m/s RV signal is stellar activity or an annual window artifact, the headline period and mass are not established even if a fifth planet is present, because the TTVs alone leave the three period families unresolved.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the discovery of Kepler-139f, a nontransiting planet with orbital period 355 ± 2 days and mass 36 ± 10 Earth masses, inferred from a joint fit of Kepler transit-timing variations (TTVs) and Keck/HIRES radial velocities (RVs) to a five-planet N-body model of the Kepler-139 system. The authors re-measure transit times for the three transiting planets, show that a four-planet model fails to reproduce the ~30-minute TTVs of Kepler-139c, and find that adding a fifth planet improves the fit by Δχ²_TTV=60 and Δχ²_RV=77, with ΔBIC=109. They then use Laplace-Lagrange secular simulations to argue that the outer giant planet Kepler-139e, if inclined by a few degrees relative to the inner system, can reduce the probability that Kepler-139f transits concurrently with planets d, b, and c, thereby explaining why the new planet is not seen in transit. The paper frames the result as the first TTV-discovered nontransiting planet in a system that also hosts an outer giant planet, with implications for the apparent truncation of compact multiplanet systems.","tokens_in":14817,"tokens_out":6450,"duration_ms":67033,"significance":"If the detection is secure, the paper reports a valuable and unusual object: a nontransiting, sub-Neptune-mass planet discovered through TTVs in a system with an outer giant planet, and it provides a concrete dynamical mechanism by which an outer giant can suppress the transit probability of an inner planet. The authors use public KGPS RV data and reanalyze Kepler photometry with standard tools, and they explicitly test the Laplace-Lagrange approximation against N-body integrations. The claimed Δχ² improvements are large and the TTV signal of planet c is visible by eye, which are genuine strengths. However, the headline period and mass rest on a weak RV signal whose interpretation is not yet demonstrated to be robust against stellar activity and annual window aliases, and the model comparison used to select among TTV-compatible period families relies on an invalid Gaussian approximation to a multimodal posterior. Because the RV data are the only discriminating evidence among the three period families, these issues are load-bearing for the central discovery claim.","major_comments":[{"comment":"The claim that the 354-day period family is 'decisive' rests on a Bayes factor computed by approximating the posterior as a multivariate Gaussian, but the posterior is explicitly described as multimodal. A Gaussian approximation around one mode ignores the other two period families and the prior volume, and exp(Δχ²/2) is a likelihood ratio, not a marginal likelihood ratio. The period degeneracy among Pf≈354, 384, and 685 days is therefore not resolved as strongly as claimed. I recommend computing model evidence with nested sampling or thermodynamic integration, or at least reporting a cross-validated predictive comparison that treats the three period families symmetrically.","section":"§4.2"},{"comment":"The selection of the 354-day solution and the mass of Kepler-139f rely on an RV signal with K≈3 m/s at P≈355 days, which is comparable to the adopted jitter σjit,RV=2.9 m/s and only about 1.5 times the ~2 m/s formal uncertainties. The paper does not present stellar activity indicators, a window-function analysis, or an independent RV reduction for the 38 Keck RVs, and footnote 5 reports that a previous periodogram search found the ~350-day peak not statistically significant, with seven taller peaks. Given the annual sampling window, the 355-day signal could be a systematic or activity artifact. To support the headline period and mass, the authors should add an activity-diagnostic and window-function analysis, and ideally fit the RVs alone to show that the 355-day signal persists without the TTV constraints.","section":"§4.2 and footnote 5"},{"comment":"The reported detection significance is based on a pipeline that excluded 7 transit times as >4σ outliers and added four jitter terms to force χ²/dof≈1, while the quoted Δχ²_TTV=60, Δχ²_RV=77, and ΔBIC=109 come from fits made before those adjustments. The footnote that 'inflating uncertainties during model comparison does not affect conclusions' addresses only Δχ², not a penalized model comparison; post hoc outlier rejection and jitter fitting can inflate the apparent significance. The authors should report model comparison with a consistent likelihood applied to the full dataset, or justify the outlier cuts a priori, and show how Pf and mf change under alternative jitter and outlier treatments.","section":"§4.2"}],"minor_comments":[{"comment":"In the introduction, 'The transiting plants' should be 'The transiting planets'.","section":"§1"},{"comment":"The median mass for planet f (36 M⊕) differs from the maximum-likelihood value (45 M⊕) by about 1σ; the text should identify which value is used for the dynamical simulations and explain the skewness of the posterior.","section":"Table 1 and §4.2"},{"comment":"The right panel of Figure 5 shows the conditional transit probability for planet f as a function of the outer giant's initial inclination, but the text does not define precisely whether this inclination is measured relative to the inner system's reference plane or relative to the line of sight; please state the definition explicitly.","section":"§5"},{"comment":"The transit-probability calculations are predictions under the fitted model and do not use the observed nontransit of planet f as a constraint on the giant's inclination; the text acknowledges this in §6, but the abstract and conclusion could state more explicitly that the dynamical mechanism is a plausibility argument rather than a measurement of the giant's inclination.","section":"§5"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline here is that this paper probably identifies a real nontransiting planet in Kepler-139, but the specific orbit and mass are less solid than the abstract implies. The TTVs of planet c show a ~30-minute wiggle that the four-planet model cannot explain; that part is convincing. Adding a fifth planet improves the TTV and RV fits dramatically (Δχ² = 137 total, ΔBIC = 109). So the existence of an additional perturbing body seems well motivated.\n\nWhat's genuinely new is that this would be the first TTV-discovered nontransiting planet in a system with an outer giant, and the authors use it as a concrete illustration of outer-giant-driven transit suppression. The dynamical simulations are clearly labeled as illustrative, and the paper is honest about the limits of the inclination argument, which I appreciate.\n\nThe soft spots are in the parameter inference. The TTVs alone allow three period families (354, 384, 685 days). The 354-day choice is made by the RV data, where the expected K is ~3 m/s, essentially equal to the fitted jitter of 2.9 m/s. No activity indicators or window-function analysis are shown, and the paper itself notes that Weiss et al. did not find the ~350-day RV peak significant. The Bayes factor quoted for the 354-day family rests on approximating a decidedly multimodal posterior as a single Gaussian, which is not a valid evidence comparison. The post hoc rejection of 7 TTV outliers and the addition of four jitter terms are common practice, but they do inflate the apparent significance. And while the TTV amplitude is large, the mass estimate of 36±10 M⊕ depends on the same RV signal.\n\nNone of this kills the paper. The central claim — there is something out there perturbing planet c — is probably right. But the specific period and mass of Kepler-139f should be viewed as conditional until the RV systematics are checked and the model comparison is done properly. The authors should release the transit times and code; that alone would strengthen the paper.\n\nBottom line: worth a serious referee, and I'd probably accept it with major revision. I just wouldn't put the 355-day/36-M⊕ numbers in a table without reading the RV analysis carefully.","headline":"Plausible but not secure: the TTVs strongly demand a fifth planet in Kepler-139, but the claimed 355-day, 36-M⊕ companion rests on a marginal RV signal and a questionable model comparison; still worth refereeing.","tokens_in":15368,"tokens_out":2199,"would_cite":false,"duration_ms":20498,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The Kepler-139 system contains a hidden nontransiting planet, Kepler-139f, with a period of 355 days and a mass of 36 Earth masses, inferred from the gravitational tugs it imprints on the transits and radial velocities of its neighbors.","keywords":["exoplanets","transit-timing variations","nontransiting planet","Kepler-139","radial velocities","secular dynamics","compact multiplanet systems","outer giant planets"],"falsifier":"Continue RV monitoring for several more years: the 355-day signal should maintain phase and amplitude if it is a planet, while a signal from stellar activity would drift or correlate with line-profile changes. Independently, a re-reduction of the Kepler photometry with a different transit-timing pipeline should reproduce the same 30-minute timing variations of planet c.","tokens_in":14140,"feed_emoji":"🪐","tokens_out":7221,"duration_ms":64035,"temperature":0.7,"pith_summary":"The paper reports a new planet, Kepler-139f, that never crosses the face of its star, yet is inferred to exist from the tiny changes it causes in the transit times of a neighboring planet and from a decade-long series of radial-velocity measurements. If the detection holds, Kepler-139f would be the first nontransiting planet found through transit-timing variations in a system that also hosts an outer giant planet. The paper argues that the outer giant tilts the inner planets' orbits on long timescales, making it genuinely less likely that such a planet transits, and uses simulations to show the conditional transit probability drops from 58% to 15% as the giant's inclination grows to 10 degrees. A fair reading of the central claim is that a planet with period $355\\pm2$ days and mass $36\\pm10$ Earth masses sits just outside the three transiting planets and inside the giant.","feed_headline":"A hidden 36-Earth-mass planet tugs Kepler-139's transits","feed_subtitle":"Timing wobbles plus a decade of RVs expose a nontransiting world hidden by an outer giant.","key_machinery":"The analysis rests on a joint N-body model of transit times and radial velocities, using the TTVFast integrator for the transit-timing predictions and parallel-tempered MCMC to explore a multimodal posterior with three possible periods for the unknown planet (354, 384, and 685 days). The 354-day solution wins decisively, with a Bayes factor $\\gtrsim 10^8$ relative to the others. For the dynamics, the paper uses Laplace-Lagrange secular theory, implemented through celmech, to evolve the system's inclinations over $10^5$-$10^6$ years and Monte Carlo sampling of viewing directions to compute conditional transit probabilities. The key identity is that the outer giant's inclination drives a 180-degree-out-of-phase oscillation of the c/f pair's inclination, which can suppress transits of planet f while planets d, b, and c still all transit.","core_discovery":"The central claim is that the Kepler-139 system contains a fifth planet, Kepler-139f, on a nontransiting orbit with period $355 \\pm 2$ days and mass $36 \\pm 10$ Earth masses, inferred from a joint fit to transit-timing variations of Kepler-139c and radial velocities from a twelve-year campaign. A model with only the four previously known planets fails to reproduce the roughly 30-minute timing variations of planet c; adding planet f improves the transit-timing fit by $\\Delta\\chi^2 = 60$ and the radial-velocity fit by $\\Delta\\chi^2 = 77$, with $\\Delta\\mathrm{BIC} = 109$. The authors also show through secular simulations that an outer giant inclined by even a few degrees lowers the probability that an observer sees all four inner planets transit, from 58% to 15% as the giant's inclination grows from 0 to 10 degrees. They conclude that the outer giant is a plausible cause of planet f's nontransiting orientation and that systems like this illustrate how outer giants can truncate apparent compact multiplanet systems.","pith_inferences":["An immediate test is to search for similar 355-day-period companions in other Kepler systems that host outer giants, using the same joint TTV+RV machinery; the paper's logic predicts such companions occur at rates comparable to the apparent truncation fraction.","The secular argument implies a testable correlation: the outermost transiting planet in compact systems should more often have a detectable outer giant than deeper planets do.","If the inclination-excitation mechanism is general, systems with high mutual-inclination outer giants should show lower multi-transit yields; population-level analysis of Kepler's multiplicity function could confirm this.","Future photometry might catch a rare transit of Kepler-139f during an epoch when the c/f inclination oscillation favors alignment, directly confirming the planet."],"forward_implications":["Kepler-139f becomes the first TTV-discovered nontransiting planet in a system known to host an outer giant planet.","The outer giant planet's inclination is a plausible mechanism for hiding the outermost inner planet, supporting the idea that outer giants truncate compact systems.","The masses of the three transiting planets are revised to typical sub-Neptune densities, removing the unusually dense planet c that appeared in an RV-only fit.","The joint TTV+RV approach can determine the period of a nontransiting planet even when RV data alone are ambiguous.","Systems with outer giants may host more nontransiting planets than transit surveys reveal, affecting multiplicity statistics."],"supporting_citations":[{"why":"Supplies the TTVFast N-body integrator used to predict transit times for the joint TTV+RV fits.","marker":"Deck et al. (2014)"},{"why":"Provides the original Kepler transit-time catalog that the paper re-analyzes to improve uncertainties and remove outliers.","marker":"Holczer et al. (2016)"},{"why":"Supplies the 38 radial velocities used to constrain the outer giant and the new planet.","marker":"Weiss et al. (2024)"},{"why":"Provides the Laplace-Lagrange secular model used to evolve inclinations and compute transit probabilities.","marker":"Hadden & Tamayo (2022)"},{"why":"Provides the parallel-tempered MCMC sampler used to explore the multimodal five-planet posterior.","marker":"Vousden et al. (2016)"},{"why":"Converts the inferred mass of Kepler-139f into a predicted radius, used to argue its transits would have been detected.","marker":"Chen & Kipping (2017)"},{"why":"Establishes the apparent truncation of compact systems that this paper's dynamical mechanism helps explain.","marker":"Millholland et al. (2022)"}],"fun_headline_variants":["Nontransiting 36-Earth-mass planet found via transit-timing wobbles","Kepler-139's hidden fifth planet: 355-day, 36-Earth-mass world","Outer giant tilts inner orbits, hiding Kepler-139f from view","Timing wobbles reveal a 36-Earth-mass nontransiting planet Kepler-139f","Kepler-139f: A 355-day, 36-Earth-mass world that never transits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The detection rests on the assumption that the nine transit-timing measurements of planet c and the residual ~3 m/s radial-velocity wiggle are caused by a real planet rather than by stellar activity, instrument drifts, or an additional planet; the paper does not check this with activity indicators.","fun_headline_variants_meta":{"raw":{"variants":["Nontransiting 36-Earth-mass planet found via transit-timing wobbles","Kepler-139's hidden fifth planet: 355-day, 36-Earth-mass world","Outer giant tilts inner orbits, hiding Kepler-139f from view","Timing wobbles reveal a 36-Earth-mass nontransiting planet Kepler-139f","Kepler-139f: A 355-day, 36-Earth-mass world that never transits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000703,"raw_usage":{"total_tokens":3168,"prompt_tokens":935,"completion_tokens":2233,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":551,"completion_tokens_details":{"reasoning_tokens":2130}},"tokens_in":551,"tokens_out":2233,"duration_ms":14712,"temperature":1.0,"reasoning_tokens":2130,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T12:14:16.528731+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Continue RV monitoring for several more years: the 355-day signal should maintain phase and amplitude if it is a planet, while a signal from stellar activity would drift or correlate with line-profile changes. Independently, a re-reduction of the Kepler photometry with a different transit-timing pipeline should reproduce the same 30-minute timing variations of planet c.","supporting_citations":[],"review_version":1}