{"id":"7cdc9f85-95d2-47c5-8cc2-368d7219d440","arxiv_id":"2507.06413","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A joint RV and transit reanalysis confirms habitable-zone planet L 98-59f (P=23.07 d, m sin i=3.0 Mearth) and finds a 2.9 to 4.2 sigma candidate at P=1.736 d.","lead":"This paper reanalyzes five years of HARPS, ESPRESSO, TESS, and HST data on the nearby M dwarf L 98-59 and confirms a non-transiting planet in the habitable zone at a 23.07-day orbit. It also reports a marginal new planet candidate at 1.736 days, potentially one of the lightest planets found by radial velocity, and argues the star rotates every 76 days.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Planet f confirmation is not robust to the GP activity model: under the sSHO kernel with a wide period prior, Delta ln Z for adding a fifth planet drops to 0.4, so the claim depends on a prior choice rather than the data alone.","rationale":"The reader's CONDITIONAL verdict is well founded. The load-bearing assumption is indeed that the celerite SHO kernel family, with the adopted period priors, correctly separates stellar activity from a dynamical 23.07 d signal. Table 4 demonstrates the fragility: only the sSHO U(25,500) configuration reduces the evidence for planet f to Delta ln Z = 0.4, and Fig. 2b shows the GP absorbing that signal. The paper is transparent about this spread, which is creditworthy, but the summary 'average Delta ln Z ~ +5.7' is not a robustness statement; it is a selective average over configurations that exclude the least favorable model. The independent L1-periodogram (log10 FAP -5.2) and the dSHO N(78,5) model provide genuine supporting evidence, so a flat rejection would be unfair. However, the rotation period is not independently established by the data; the paper's own GP period posterior shifts between ~28 and ~77 d depending on the prior, and H21's photometric rotation period is 39.6 +/- 2.2 d. Accepting the 76 +/- 4 d rotation period is necessary to secure the strong evidence for f. The candidate 06 section contains an explicit self-contradiction: Sec. 5.3 states the wide-prior evidence increase is insignificant but the fit is called 'highly significant' because the log-likelihood increase is significant, which is exactly the error highlighted by Rajpaul et al. (2024), a citation the authors themselves invoke. A concrete test with an activity-indicator-driven GP would settle the degeneracy. For these reasons the reader's CONDITIONAL verdict, and the central concern it identifies, remain unchanged.","tokens_in":27187,"tokens_out":10051,"duration_ms":112038,"concrete_test":"Re-model the HARPS + ESPRESSO RVs with a physically-motivated GP conditioned on the measured activity indicators (e.g., H alpha and BIS as simultaneous GP regressors, following Rajpaul et al. 2015), with no restrictive period prior on the GP, and compare the four-planet and five-planet models. If the 23.07 d Keplerian remains favored with Delta ln Z > 5, planet f is robust to the activity-model choice; if not, the celerite-based confirmation is model dependent and should be downgraded.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the 23.07 d signal is planetary rests on model selection among GP kernels and period priors. Table 4 shows the evidence increase for a fifth planet ranges from Delta ln Z = 11.3 with no GP and 10.4 with sSHO N(78,5) down to 0.4 with sSHO U(25,500), where the GP absorbs the 23 d signal (Fig. 2b). The paper's 'average Delta ln Z ~ +5.7' is an arithmetic mean over selected configurations, not a marginalization over the activity-model uncertainty; excluding the wide-prior run is a post-hoc choice. The rotation period is itself disputed (H21: 39.6 +/- 2.2 d from TESS+Evryscope; this paper claims 76 +/- 4 d partly using the same GP family whose period posterior changes with the prior). If the flexible sSHO U(25,500) GP is the appropriate default, the data do not require planet f. Independent support from the L1-periodogram and the dSHO N(78,5) model prevents a rejection, but the confirmation is conditional on a prior preference. The same defect is explicit in Sec. 5.3 for candidate 06, where the paper admits the wide-prior evidence increase is insignificant yet calls the fit 'highly significant' based on Delta log L.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reanalyzes HARPS and ESPRESSO radial velocities together with TESS and HST photometry to confirm a fifth non-transiting planet, L 98-59f, at an orbital period of 23.07 ± 0.08 d and minimum mass of 3.0 ± 0.5 M⊕, and to report an additional non-transiting candidate, L 98-59.06, at 1.7361 d with minimum mass 0.58 ± 0.12 M⊕. The analysis uses dynesty nested sampling, three celerite GP kernels with different period priors, GLS and L1 periodograms, a 10,000-sample bootstrap FAP (0.46%), and an injection-retrieval test (90.4% recovery). The authors also present evidence for a stellar rotation period of 76 ± 4 d and compare the system's architecture with other multi-planet systems.","tokens_in":27412,"tokens_out":9733,"duration_ms":101660,"significance":"If the central claim holds, the paper materially strengthens the case for a rare five-planet M-dwarf system and places a non-transiting habitable-zone planet on firmer observational footing. The paper is transparent in reporting full evidence tables (Table C1), priors and posteriors (Table C2), and a numerical FAP and injection-retrieval test. It also uses an independent SERVAL extraction for the HARPS data and adds twelve new TESS sectors and five HST transits. However, the confirmation of planet f is not robust across the paper's own GP activity models: Table 4 shows that the evidence increase for adding a fifth planet ranges from Δln Z = 11.3 without a GP down to 0.4 with the sSHO kernel and a wide period prior, where the GP absorbs the 23.1-d signal. The paper's 'average' evidence and the conclusion's quoted Δln Z range exclude that configuration, making the central claim conditional on a prior choice rather than on the data alone.","major_comments":[{"comment":"The evidence for planet f is strongly dependent on the GP activity model. Adding a fifth planet gives Δln Z = 11.3 without a GP, 10.4 with sSHO N(78,5), 7.9 with sSHO N(39,5), but only 0.4 (σ = 0.4) with sSHO U(25,500), where the GP absorbs the 23.1-d signal (Fig. 2b). The text's 'average Δln Z ∼ +5.7' and the conclusion's range '3.3 ≤ Δln Z ≤ 11.3' exclude the wide-prior run and are not a principled model average. Please replace this with an evidence estimate marginalized over kernel and prior choices, or provide a data-driven justification for restricting the GP period prior; otherwise the confirmation claim is not robust.","section":"Table 4, Sec. 4.2"},{"comment":"The claimed rotation period of 76 ± 4 d is largely prior-driven. The best six-planet model uses a dSHO kernel with prior N(78,5) and returns 76.6 ± 4.2 d; the sSHO with the wide prior returns 28.0+2.4−1.7 d in the four-planet model; and H21 report 39.6 ± 2.2 d from TESS and Evryscope. The Perger et al. (2021) argument that celerite SHO kernels prefer the second harmonic is qualitative and does not quantify how the prior shifts the posterior. Because the GP is the component that can absorb the 23-d signal, the rotation-period determination must be made prior-independent or explicitly folded into the uncertainty of the planet-f confirmation. Also, Table C2 lists the GP prior as N(39,5) while the posterior is 76.6 ± 4.2, which appears inconsistent with the stated N(78,5) model in Table 4 and needs correction.","section":"Sec. 5.1, Table C2"},{"comment":"The significance statement for L 98-59.06 is internally inconsistent. The text says that with wide period priors U(1.1, 1.9) the evidence increase is insignificant, yet in the same paragraph calls the candidate 'highly significant' because the log-likelihood increase is significant. The log-likelihood difference does not include the Occam penalty and is not a valid model-comparison statistic in this context; the Bayesian evidence or a calibrated FAP should be used. Please either report the wide-prior evidence honestly and downgrade the candidate, or reconcile the RV-only and combined-fit evidence values so that the paper does not make contradictory statements about the same signal.","section":"Sec. 5.3, Sec. 4.3.4"},{"comment":"The conclusion states '3.3 ≤ Δln Z ≤ 11.3, 2.1 ≤ σ ≤ 4.4' for the fifth planet, but Table 4 contains Δln Z = 0.4 and σ = 0.4 for the sSHO U(25,500) configuration. This discrepancy is not a minor typo: it directly affects the paper's central confirmation claim. Please correct the summary statistics and replace the ad hoc average with a clearly defined model-averaged evidence or a sensitivity statement that explicitly includes the wide-prior case.","section":"Conclusion vs. Table 4"}],"minor_comments":[{"comment":"The abstract calls 289 K an 'effective temperature' for planet f, but the paper actually computes an equilibrium temperature assuming a Bond albedo of 0 and isotropic heat redistribution. Table C2 lists 285 K for planet f. Please use 'equilibrium temperature' and specify the assumed albedo and redistribution, or reconcile the two values.","section":"Abstract"},{"comment":"Six HARPS points are removed by iterative 4σ clipping, but the paper notes these outliers are not the same as those found by C19 or D21. Please describe the clipping algorithm precisely and test the sensitivity of the planet parameters and evidence to this choice.","section":"Sec. 2.1.1, Table 2"},{"comment":"The numerical FAP of 0.46% is much lower than the analytical GLS FAP of 5.3% for the same 1.736-d signal. Please clarify the cause of this difference, for example the different null model or window-function effects, so readers can compare the two numbers.","section":"Sec. 4.3.3"},{"comment":"The statement that the absence of a transit for L 98-59.06 implies an inclination offset of ±4.2° is not derived anywhere in the paper. Please provide the calculation or a reference for this constraint.","section":"Sec. 5.3"},{"comment":"The axis label 'Freqency [1/d]' contains a typo and should read 'Frequency [1/d]'.","section":"Figure 1, Figure 2"},{"comment":"The text notes that the 06-b separation is Δ(RH) = 11.9, below the Δ(RH) = 13 stability limit mentioned in the same section, yet the system is described as stable. Please clarify whether the SPOCK stability prediction gives a probability or confidence for this configuration and how the 11.9 value should be interpreted.","section":"Sec. 5.4, Figure 7"}],"recommendation":"major_revision","confidential_remarks":"The paper deserves a chance after substantial revision. The main issue is not circularity but prior sensitivity: the planet-f confirmation and the rotation-period claim both depend on which GP kernel and period prior are used, and the paper's own Table 4 contains a configuration that destroys the evidence for planet f. A model-averaged evidence calculation or a clear, data-driven justification of the GP prior choice is needed. The contradictory significance statement for L 98-59.06 in Sec. 5.3 should also be fixed. If the authors address these points, the paper could become a solid contribution to the L 98-59 system."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a competent and unusually transparent reanalysis of L 98-59. The independent SERVAL extraction, the added TESS/HST data, and the full evidence tables make it a credible piece of work. But the abstract's confirmation of planet f is weaker than it appears: the Bayesian evidence for f collapses from Delta ln Z = 11.3 (no GP) to 0.4 under the sSHO kernel with a wide period prior (Table 4). The paper's 'average Delta ln Z ~ +5.7' averages over model configurations rather than marginalizing over the activity-model uncertainty, and the wide-prior run is excluded without a strong justification. If that GP is the right default, these data do not require planet f. There is independent support from the L1-periodogram and from D21's earlier detection with different data, so this is not fatal, but the confirmation should be presented as conditional on the activity model.\n\nThe new 1.736 d candidate is real but marginal: it is always 2.9-4.2 sigma better than the five-planet model, with a bootstrap FAP of 0.46% and an injection-retrieval rate of 90.4%. However, the paper admits the evidence increase is insignificant with wide period priors, yet still calls the fit 'highly significant' based on Delta log L. That is overstatement. The candidate deserves follow-up, not confirmation.\n\nThe rotation period claim of 76±4 d is also softer than stated: it depends on the dSHO/cdSHO kernel and the second-harmonic argument, while H21 found 39.6 d. The paper's own L1-periodogram gives both ~34.7 and ~68.3 d with similar cross-validation scores.\n\nWhat is genuinely new: the candidate, the independent extraction, the 13 new TESS sectors and 5 HST transits, and the refined ephemerides with errors reduced by up to an order of magnitude. The paper is well-structured and honest: it labels 06 as a candidate, quotes the 2.9-4.2 sigma range, and cites Rajpaul et al. about the danger of simple RV fits.\n\nWho it's for: anyone working on M dwarf planet systems or on GP modeling of stellar activity in RV data. It deserves a serious referee; the referee should ask for a fairer summary of the GP sensitivity and more cautious wording for the candidate. I'd send it to review.","headline":"A transparent, well-documented reanalysis of L 98-59 that confirms planet f only under some GP priors, adds a marginal 1.74 d candidate, and should be sent to peer review with a request for more honest summaries.","tokens_in":28149,"tokens_out":3193,"would_cite":true,"duration_ms":34551,"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 paper confirms L 98-59f as a non-transiting planet in the habitable zone of the nearby M dwarf L 98-59, with strong Bayesian evidence in most activity models.","keywords":["L 98-59","exoplanets","radial velocity","M dwarf","habitable zone","Gaussian process","Bayesian evidence","sub-Earth candidate"],"falsifier":"Collect new radial-velocity measurements over at least one full 76-day rotation cycle and test whether the 23.07-day signal keeps a constant phase and amplitude while the activity indicators vary; if the signal's phase drifts or its amplitude tracks the stellar rotation cycle, the planetary interpretation is wrong, and if an independent dataset shows no 23.07-day signal after equivalent activity modeling, the confirmation does not stand.","tokens_in":26887,"feed_emoji":"🪐","tokens_out":12491,"duration_ms":116305,"temperature":0.7,"pith_summary":"This paper tries to settle whether the 23-day radial-velocity signal in the nearby M dwarf system L 98-59 is a real planet rather than a symptom of stellar activity. Combining new extracts of HARPS velocities, published ESPRESSO velocities, 21 TESS sectors, and five HST transits in a joint Bayesian model, it reports strong evidence for the fifth planet, L 98-59f, with an orbital period of $23.07\\pm0.08$ d, a minimum mass of $3.0\\pm0.5\\,M_\\oplus$, and an equilibrium temperature of 289 K. The same analysis uncovers an inner candidate at $1.736$ d with a minimum mass of $0.58\\pm0.12\\,M_\\oplus$; its statistical significance ranges from $2.9\\sigma$ to $4.2\\sigma$ depending on how stellar activity is modeled. If both signals hold, L 98-59 becomes a six-planet system with a habitable-zone world, a benchmark for formation and future atmospheric studies.","feed_headline":"Confirmed: habitable-zone planet orbits nearby M dwarf L 98-59","feed_subtitle":"Reanalysis of RV and transit data strongly supports a 3-Earth-mass world at 289 K, plus a sub-Earth candidate.","key_machinery":"The load-bearing element is the comparison of Bayesian evidences ($\\Delta\\ln\\mathcal{Z}$) among four-, five-, and six-planet models, with stellar activity absorbed by Gaussian-process kernels from the damped simple-harmonic-oscillator family (single, double, and coupled-double). These kernels are fitted together with Keplerian planet orbits to the combined radial velocities and the TESS/HST light curves, and the parameter space is integrated with nested sampling. The confirmation of planet f rests on $\\Delta\\ln\\mathcal{Z}\\approx +5.7$ on average across activity models, up to $+11.3$ without a GP; the candidate's significance range of $2.9\\sigma$ to $4.2\\sigma$ is likewise tied to which kernel and period prior is chosen.","core_discovery":"The paper establishes, on its own terms, that the 23.07-day radial-velocity signal previously reported as a candidate survives a joint reanalysis of the HARPS and ESPRESSO velocities, TESS and HST transit photometry, and activity indicators. In most noise-model configurations the five-planet interpretation beats the four-planet interpretation by an average log-evidence margin of about $+5.7$, reaching $+11.3$ when no Gaussian-process activity model is included; the authors classify this as strong Bayesian evidence. The planet, L 98-59f, is non-transiting, with a minimum mass of $3.0\\pm0.5\\,M_\\oplus$ and an effective temperature of 289 K, placing it in the system's habitable zone. The paper also reports a sixth planet candidate, L 98-59.06, with period $1.7361^{+0.0007}_{-0.0008}$ d and minimum mass $0.58\\pm0.12\\,M_\\oplus$, whose significance is between $2.9\\sigma$ and $4.2\\sigma$ depending on the activity model; the authors do not claim confirmation for this candidate, but bootstrap and injection-retrieval tests lead them to argue it is not white noise. They further favor a stellar rotation period of $76\\pm4$ d.","pith_inferences":["A decisive test the paper does not perform: extend the radial-velocity baseline over several 76-day rotation cycles and compare the 23.07-day signal's phase stability with the behavior of the activity indicators; that would separate a true Keplerian from a flexible activity model.","If the 1.736-day candidate is real, it would be among the lightest planets found by radial velocities, and its near-2:1 commensurability with planet b would make the system a useful testbed for dynamical and formation models.","The habitable-zone classification of L 98-59f depends on the adopted equilibrium-temperature and habitable-zone model; a different albedo or heat-redistribution assumption could move the planet outside the conservative habitable zone, so that label is model-dependent."],"forward_implications":["L 98-59f is confirmed as a non-transiting planet at $23.07\\pm0.08$ d with a minimum mass of $3.0\\pm0.5\\,M_\\oplus$ and an equilibrium temperature of 289 K, in the habitable zone.","The stellar rotation period is constrained to $76\\pm4$ d, favoring the longer of the previously debated values and explaining the shorter period as a harmonic of the spot pattern.","A sixth planet candidate, L 98-59.06, orbits at $1.736$ d with a minimum mass of $0.58\\pm0.12\\,M_\\oplus$; its $2.9\\text{--}4.2\\sigma$ significance leaves it unconfirmed but unlikely to be pure white noise.","No transit is found for the candidate, implying an inclination offset of about $\\pm4.2^\\circ$ from an edge-on orbit if it is real.","The refined ephemerides reduce the period uncertainties of the confirmed planets by up to an order of magnitude and preserve a compact, stable architecture with adjacent separations above about 12 mutual Hill radii."],"supporting_citations":[{"why":"First proposed the fifth planet at 23.15 d and supplied the ESPRESSO radial velocities and activity indicators used here.","marker":"D21"},{"why":"Provided the public HARPS spectra that this paper re-extracts independently and an earlier Gaussian-process rotation analysis.","marker":"C19"},{"why":"Discovered the three transiting planets L 98-59b, c, d that anchor the transit fit.","marker":"K19"},{"why":"Supplied the stellar parameters adopted in Table 1 and proposed the shorter 39.6 d rotation period that this work argues against.","marker":"H21"},{"why":"Supplied the Gaussian-process kernel machinery used to model stellar activity in the radial velocities.","marker":"Foreman-Mackey, Agol, Ambikasaran, & Angus (2017)"},{"why":"Provided the five HST white-light transits of planet b that are newly incorporated into the joint fit.","marker":"Zhou et al. (2022)"},{"why":"Provides the interpretation that simple-harmonic-oscillator kernels tend to lock onto the second harmonic of the rotation period, supporting the 76 d value.","marker":"Perger et al. (2021)"},{"why":"Supplies the cautionary point that a non-zero radial-velocity fit alone is not enough to confirm a planet, shaping how the candidate is presented.","marker":"Rajpaul, Barragán, and Zicher (2024)"},{"why":"Defines the Bayesian evidence thresholds (Delta ln Z > 2.5 moderate, > 5 strong) used to interpret the model comparisons.","marker":"Trotta (2008)"}],"fun_headline_variants":["Nearby M dwarf L 98-59 confirmed with non-transiting HZ planet","Confirmed habitable-zone planet found around nearby M dwarf L 98-59","L 98-59: non-transiting planet in habitable zone confirmed","3-Earth-mass non-transiting planet confirmed in L 98-59's HZ"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire confirmation of planet f rests on the assumption that the chosen Gaussian-process activity models, with their specific period priors, fully capture the star's variability without being flexible enough to also absorb the 23-day signal; if that assumption fails, these data do not confirm the planet.","fun_headline_variants_meta":{"raw":{"variants":["Nearby M dwarf L 98-59 confirmed with non-transiting HZ planet","Confirmed habitable-zone planet found around nearby M dwarf L 98-59","L 98-59: non-transiting planet in habitable zone confirmed","3-Earth-mass non-transiting planet confirmed in L 98-59's HZ"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00126,"raw_usage":{"total_tokens":5243,"prompt_tokens":1112,"completion_tokens":4131,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":728,"completion_tokens_details":{"reasoning_tokens":4042}},"tokens_in":728,"tokens_out":4131,"duration_ms":31342,"temperature":1.0,"reasoning_tokens":4042,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:07:01.375457+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Collect new radial-velocity measurements over at least one full 76-day rotation cycle and test whether the 23.07-day signal keeps a constant phase and amplitude while the activity indicators vary; if the signal's phase drifts or its amplitude tracks the stellar rotation cycle, the planetary interpretation is wrong, and if an independent dataset shows no 23.07-day signal after equivalent activity modeling, the confirmation does not stand.","supporting_citations":[{"cited_title":"(2017, December), AJ, 154(6),","cited_arxiv_id":null,"evidence_quote":"Supplied the Gaussian-process kernel machinery used to model stellar activity in the radial velocities."}],"review_version":1}