{"id":"12b0ddf6-1dd5-4048-95b7-0b6bc891d2b5","arxiv_id":"1908.02158","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Recurrence network measures from RV and TTV time series can separate regular from chaotic exoplanet dynamics, with Kepler-36b flagged as irregular.","lead":"Recurrence network analysis, a graph-based time series tool, is shown to distinguish chaotic from regular motion in planetary systems using only observable signals such as radial velocity and transit timing variations. The method offers a fast, integration-free stability check that could be applied to large exoplanet surveys.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Kepler-36 conclusion rests on one network measure while the other contradicts it; the method is not validated at the short length and missing-data fraction of the real TTV series.","rationale":"The reader's weakest_assumption pointed to the unsupported inference from surrogate null rejection to chaos. That is a real concern, but the manuscript itself contains a more directly falsifying symptom: the two RNA measures, which agree in the synthetic validation, give opposite classifications for Kepler-36b. The paper explains this away by invoking 'stable chaos' and by noting agreement with earlier studies, but this is post-hoc selection of the measure that matches the prior literature. Because the central claim includes the Kepler-36 application as evidence, the method's reliability at the real-data sample size (N ~ 75) and missing-data fraction (up to 25%) must be validated. The proposed synthetic subsampling test would settle whether the disagreement is an artifact of short, gap-filled time series or a genuine physical signature. If the synthetic test reproduces the disagreement for known chaotic orbits, the real-data conclusion gains support; if not, the Kepler-36 inference is unsubstantiated. This reinforces the conditional verdict rather than changing it: the synthetic core is credible, but the real-data extension needs additional validation. The paper is clearly written and the methodology is a legitimate novel application, so no rejection is warranted, but the current evidence does not justify an unconditional acceptance.","tokens_in":9856,"tokens_out":3720,"duration_ms":41849,"concrete_test":"Generate synthetic TTV time series from the SJS model for one known regular and one known chaotic initial condition, then subsample each to 72-77 data points with the same missing-data pattern as Kepler-36 (random gaps, 13-25%) and apply spline interpolation, time-delay embedding with the same parameter-selection rules, PPTS surrogates, and the rank-based significance test used in Figures 8 and 9. Record whether T and L each classify the trajectory correctly and whether the two measures agree. Repeat for many realizations to obtain a confusion matrix. If at N ~ 75 the measures disagree or misclassify at rates comparable to the Kepler-36 result, the real-data interpretation is unreliable. As a secondary check, rerun the Kepler-36 analysis requiring that both T and L must reject the null for a chaos claim, and report whether Kepler-36b remains classified as irregular.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The real-data conclusion that Kepler-36b is dynamically irregular is supported by only one of the two RNA measures. In Figures 8 and 9, transitivity T rejects the quasi-periodic null for the inner planet (Tb = 0.430), but average path length L places both planets inside the surrogate ensemble and therefore is consistent with quasi-periodicity for both (Lb = 3.557, Lc = 5.749). The paper interprets this disagreement as 'stable chaos' and cites agreement with Deck et al. and Panichi et al., but that is effectively selecting the measure that matches the previously published result. No criterion is given for preferring T over L in this regime. The synthetic SJS validation demonstrates that L and T agree in classifying regular and chaotic trajectories at N = 950 and N = 3500 with up to 15% missing data, but the Kepler-36 TTV series contain only 72 and 77 usable points with 13% and 25% missing data. The paper does not show that either measure, let alone both, remains reliable at this length and missing-data fraction. Thus the central claim that RNA can determine exoplanetary dynamics from observed time series is not established for the real-data case; the load-bearing step is not the surrogate-to-chaos leap alone, but the unexamined transfer of the synthetic validation to a regime where the two measures already disagree.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes recurrence network analysis (RNA) as a method to classify exoplanetary dynamics from observed scalar time series (radial velocity and transit timing variations). After time-delay embedding and recurrence-network construction, two network measures—average path length L and transitivity T—are used as dynamical indicators. On a synthetic Sun–Jupiter–Saturn (SJS) system, these measures reproduce the MEGNO stability map on a 100×100 grid, with false-grouping errors below 10%. Hypothesis tests using pseudo-periodic twin surrogates at the 1% significance level correctly identify regular, chaotic, and resonant trajectories, including robustness to noise and 15% missing data. Applied to Kepler-36 TTV data, transitivity rejects the quasi-periodic null for the inner planet while average path length does not; the paper interprets this disagreement as 'stable chaos' and claims consistency with earlier studies. The central claim is that RNA can determine exoplanetary dynamics from observed data, but the real-data demonstration is compromised by the conflicting measures and by the untested transfer of synthetic validation to much shorter and more incomplete real series.","tokens_in":10183,"tokens_out":6808,"duration_ms":60546,"significance":"If the synthetic results are representative, RNA offers a fast, parameter-light complement to n-body stability analysis, using directly observed time series without orbital fitting or integration. The synthetic validation is a genuine strength: the network measures recover the MEGNO map, surrogate tests are calibrated at the 1% level, and the method is tested on regular, chaotic, and resonant regimes. However, the real-data inference is not yet established: the two RN measures contradict each other on Kepler-36, the leap from rejecting quasi-periodicity to asserting chaos is unjustified, and the validation was performed at longer series lengths and smaller missing fractions than the real data. The paper introduces an interesting framework but overstates its current applicability to observational data; the contribution is promising but requires additional statistical work to support the central claim.","major_comments":[{"comment":"The paper asserts that rejecting the PPTS null implies chaotic dynamics because planetary motion is deterministic, high-dimensional, and nonlinear. This is an assumption, not a demonstrated property. A non-quasi-periodic signal may also arise from measurement noise, systematic errors, non-stationarity, or an inappropriate null model; the surrogate test alone does not distinguish these possibilities. This assumption is load-bearing because it is used to convert the T-based rejection for Kepler-36b into a conclusion about chaos. A concrete test would be to apply the same procedure to synthetic non-chaotic but non-quasi-periodic signals (e.g., slowly drifting periodic signals, chirps, or noise-corrupted periodic signals) and show that the rejection rate remains at the nominal level. Without such a test, the inference from rejecting quasi-periodicity to chaos remains unsubstantiated for short, noisy, unevenly sampled real series.","section":"III.A, surrogate hypothesis paragraph"},{"comment":"The embedding parameters for Kepler-36 (d = 6, τ = 4 and 5, ε = 165.0 and 52.90) are stated without justification, whereas for the synthetic series the parameters are selected via standard criteria (false nearest neighbors and mutual information). Since RN measures depend on ε—the paper itself acknowledges this in Section III.A—the lack of a sensitivity analysis for the real data leaves open the possibility that the T/L contradiction is a parameter artifact. The authors should explain how the real-data embedding parameters were chosen and show that the hypothesis-test conclusions remain stable over a plausible range of d, τ, and ε.","section":"Table I and Section III.B"}],"minor_comments":[{"comment":"The phrase 'consistent with earlier studies' overstates the case because the two RN measures give opposite conclusions for the inner planet; suggest softening to 'partially consistent' or explicitly reporting the discrepancy.","section":"Abstract and Section IV"},{"comment":"The abbreviation 'CPD' appears to be a typo for 'CDF' (cumulative distribution function), as used in the surrounding text.","section":"Eq. (2)"},{"comment":"The caption says '(b) and (c) Two RN measures L, T are pictured ... taking into account two observables TTV of Jupiter and RV of the Sun,' but it does not say which measure appears in which panel; the text indicates panel (b) is L for TTV and panel (c) is T for RV, so the caption should be explicit.","section":"Fig. 1 caption"},{"comment":"The journal name 'Mothly Notices' should be 'Monthly Notices.'","section":"References 8 and 20"},{"comment":"The sentence 'First Refs. 15,1617 proposed the method...' has a formatting error in the reference list; it should read 'Refs. 15–17' and the verb should be 'proposed' (or 'have proposed').","section":"Section II"},{"comment":"The term 'stable chaos' is used without a definition or a supporting citation in the immediate context; define it or explicitly cite the earlier works that introduce it (e.g., stable chaos as bounded chaotic motion with positive short-time Lyapunov exponents).","section":"Section III.B"},{"comment":"The phrase 'the conclusions drawn from the analysis should be treated in place' is unclear; 'in place' is likely a typo for 'in plane' or 'with care,' and the sentence should be rephrased.","section":"Section IV"},{"comment":"The statement that the method 'can be generalized to more than two planets' cites a paper 'in preparation'; if that work is not yet available, the generalization remains unverified and the citation should be updated or the statement hedged.","section":"Section IV and reference [36]"}],"recommendation":"major_revision","confidential_remarks":"The synthetic part of the paper is solid and the exposition is clear, but the real-data section is not enough to support the central claim. The author should either add a validation study at the Kepler-36 data length and missing fraction, provide a principled rule for resolving conflicts between network measures, or carefully limit the real-data conclusions. The inconsistency between the text and Fig. 7 caption regarding missing-data fractions should be corrected, and the assumption that rejecting the PPTS null implies chaos needs a concrete justification. With these revisions, the paper could become acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a clean proof-of-concept that recurrence network measures can separate regular from chaotic dynamics in synthetic TTV and RV time series. That part deserves credit. The synthetic validation against MEGNO is genuine: the network measures reproduce the stability map, the surrogate tests behave correctly in all three regimes, and the classification error stays below 10% with noise and 15% missing data. The method is fast and the presentation is clear. This is a legitimate new application of an established tool.\n\nThe soft spots are in the real-data transfer. First, the paper's leap from rejecting the pseudo-periodic twin surrogate null to \"the dynamics is chaotic\" is too quick. The null is specifically quasi-periodicity; rejecting it can also happen for nonstationarity, measurement artifacts, or the short length of the series. The author acknowledges the question but then proceeds as if the leap were safe. Second, and more seriously, the Kepler-36 result is not consistent across the two network measures. Transitivity rejects the null for Kepler-36b, but average path length keeps it for both planets. The paper calls this \"stable chaos\" and cites Deck et al. and Panichi et al., but no criterion is given for preferring T over L. That is effectively selecting the measure that matches the previously published answer. Third, the synthetic validation was done at 950 and 3500 data points, while the Kepler-36 series contain only 72 and 77 usable points with 13% and 25% gaps. The paper never shows that either measure remains reliable at that length and missing-data fraction, and the fact that the two measures disagree in the real case suggests the transfer is not innocuous.\n\nThese are real flaws, but they are concentrated in one section. The synthetic methodology is sound and the failure mode is instructive: it is a good example of how carefully a surrogate rejection must be interpreted when moving from long clean simulations to short messy observations. The paper deserves a serious referee, not a desk rejection. A good referee could push the author to validate the method on synthetic series resampled to Kepler-36's length and gap structure, and to either give a principled reason for preferring T over L or present both results without claiming a chaotic detection. The citation pattern is fine; the paper engages with the relevant literature.\n\nWho is this for? Researchers working on TTV-based stability indicators and anyone interested in applying surrogate-testing methods to short exoplanet time series. The synthetic part is worth reading even if the real-data conclusion remains provisional.","headline":"Solid synthetic proof-of-concept for recurrence network analysis of exoplanet time series, but the Kepler-36 conclusion picks the measure that fits the prior result.","tokens_in":10610,"tokens_out":1672,"would_cite":false,"duration_ms":19856,"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":"Transit timings alone reveal a planet's chaotic dynamics.","keywords":["recurrence network analysis","exoplanetary dynamics","chaos detection","transit timing variations","radial velocity","surrogate data hypothesis testing","Kepler-36","time series embedding"],"falsifier":"Generate many synthetic, genuinely quasi-periodic RV/TTV time series from a stable two-planet model with the same length, noise level, and ~15% missing-data pattern as the Kepler-36 observations, run the PPTS surrogate test at the 99% level, and count the false rejections; a rejection rate clearly above 1% would show that rejecting quasi-periodicity does not reliably imply chaos.","tokens_in":9685,"feed_emoji":"🪐","tokens_out":8703,"duration_ms":81446,"temperature":0.7,"pith_summary":"This paper proposes that the dynamics of an exoplanetary system—whether the planets move regularly or chaotically—can be read directly from a single observed time series, such as radial velocities or transit timing variations, without integrating the equations of motion. The idea is to turn the time series into a recurrence network, a graph whose nodes are nearby points in the reconstructed phase space, and to use network measures such as average path length and transitivity as indicators of order and chaos. On synthetic Sun–Jupiter–Saturn data, the method recovers the same stability map that a standard chaos indicator produces, and hypothesis tests with pseudo-periodic twin surrogates distinguish regular, resonant, and chaotic regimes at 99% significance even when the data are noisy and up to 15% missing. Applied to the real Kepler-36 transit timing data, the analysis suggests the inner planet is dynamically irregular while the outer planet is regular, consistent with earlier studies. The result matters because it offers a fast, model-free complement to n-body stability analysis for the growing catalogs of exoplanet observations.","feed_headline":"Transit timings alone reveal a planet's chaotic dynamics","feed_subtitle":"Recurrence networks classify regular vs chaotic motion without n-body integration, even on noisy, gappy data.","key_machinery":"The central object is the recurrence network (RN), obtained from the recurrence plot by deleting the diagonal: the adjacency matrix is $A_{ij} = R_{ij} - \\delta_{ij}$, so that nodes are reconstructed phase-space vectors and edges connect vectors that lie within an $\\epsilon$-ball at different times. Two network statistics carry the classification: the average path length $L$, which grows with chaotic spreading, and the transitivity $T$, the fraction of closed triangles, which is high for coherent, regular motion. The test of significance is provided by pseudo-periodic twin surrogates (PPTS), which scramble the signal while preserving its quasi-periodic structure; comparing the network measures of the original series to the surrogate ensemble yields a rank-based one-sided hypothesis test at the 99% level.","core_discovery":"On the paper's own terms, the central discovery is that the topology of a recurrence network built from a scalar observable preserves the underlying dynamics of the planetary system, so that the network measures average path length L and transitivity T can serve as proxies for regular versus chaotic motion. In a 100×100 grid of initial conditions for a Sun–Jupiter–Saturn model, the maps of L and T reproduce the MEGNO stability map, with chaotic regions corresponding to high L and low T, and the same measures classify individual trajectories when tested against pseudo-periodic twin surrogates with a one-sided 99% significance level. The method survives the addition of Gaussian noise and systematic deletion of up to 15% of the data, after spline interpolation. For the real Kepler-36 TTV series, the transitivity measure rejects the quasi-periodic null hypothesis for the inner planet Kepler-36b while the outer planet Kepler-36c is regular; the average path length measure, however, does not reject the null for either planet. The paper interprets this as evidence of stable chaos near the edge of the 7:6 resonance, in line with earlier dynamical analyses.","pith_inferences":["The leap from 'rejecting quasi-periodicity' to 'the dynamics is chaotic' is only valid under the assumption that planetary signals are deterministic and that no other non-quasi-periodic process is at play; in other fields a rejected PPTS null could also be produced by a slowly varying frequency, so the diagnostic label should be validated against independent Lyapunov estimates on a few well-known ","The disagreement between transitivity and average path length for Kepler-36b hints that the two measures probe different geometric aspects of the recurrence network; a combined statistic or a measure that is independent of the embedding parameters might sharpen the classification for short, gappy series.","The method is presented for RV and TTV, but the same construction applies to astrometric time series; it would be interesting to test whether astrometric data with similar noise levels yield comparable classification power."],"forward_implications":["RNA can serve as a computationally cheap screening tool for planetary stability, since it needs only the measured time series and avoids n-body integration; a 950-point series with its 100 surrogates is analyzed in under ten minutes on a desktop machine.","The method works with both radial-velocity and transit-timing observables and remains reliable when the data are noisy and up to 15% of the points are missing, so it is applicable to current ground- and space-based surveys.","For Kepler-36, the RNA result supports the picture of the inner planet being dynamically irregular, with stable chaos near the 7:6 resonance, and the outer planet regular.","The scheme generalizes to systems with more than two planets, so it can be applied to the longer TTV and RV time series expected from future surveys."],"supporting_citations":[{"why":"introduces recurrence networks as a nonlinear time series paradigm, the basis of the whole analysis.","marker":"[11]"},{"why":"Takens' theorem justifies reconstructing the phase space from a scalar time series, the first step of the method.","marker":"[12]"},{"why":"shows how recurrence-network measures reflect the geometry of chaotic dynamics, grounding the use of L and T as order/chaos indicators.","marker":"[18]"},{"why":"supplies the standard techniques for time delay embedding and for rank-based surrogate hypothesis tests.","marker":"[10]"},{"why":"demonstrates in the standard map that complex network measures disentangle regular and chaotic motion, the analogue the paper extends to planetary systems.","marker":"[25]"},{"why":"introduces pseudo-periodic twin surrogates, whose null hypothesis of quasi-periodicity is the basis for the significance tests.","marker":"[30]"},{"why":"provides the phase-space reconstruction method for non-uniformly sampled noisy time series, used after interpolation of the gappy observational data.","marker":"[33]"},{"why":"identified the rapid dynamical chaos in Kepler-36, the real-world result the RNA analysis is compared against.","marker":"[7]"},{"why":"further analyzed Kepler-36 with the reversibility error method, finding stickiness near the 7:6 resonance, which supports the stable-chaos interpretation.","marker":"[8]"}],"fun_headline_variants":["Network topology exposes chaos in transit timing data","Recurrence networks classify exoplanet dynamics from TTVs","Chaos in planets read off transit timings via networks","Stable or chaotic? Recurrence networks answer from data","No n-body runs: network measures reveal planetary chaos"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that, for planetary signals, rejecting the quasi-periodic null hypothesis in the surrogate test can be taken to mean the underlying dynamics is chaotic.","fun_headline_variants_meta":{"raw":{"variants":["Network topology exposes chaos in transit timing data","Recurrence networks classify exoplanet dynamics from TTVs","Chaos in planets read off transit timings via networks","Stable or chaotic? Recurrence networks answer from data","No n-body runs: network measures reveal planetary chaos"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000179,"raw_usage":{"total_tokens":1244,"prompt_tokens":833,"completion_tokens":411,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":449,"completion_tokens_details":{"reasoning_tokens":331}},"tokens_in":449,"tokens_out":411,"duration_ms":4563,"temperature":1.0,"reasoning_tokens":331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:07:51.426874+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate many synthetic, genuinely quasi-periodic RV/TTV time series from a stable two-planet model with the same length, noise level, and ~15% missing-data pattern as the Kepler-36 observations, run the PPTS surrogate test at the 99% level, and count the false rejections; a rejection rate clearly above 1% would show that rejecting quasi-periodicity does not reliably imply chaos.","supporting_citations":[{"cited_title":"Kantz \\ and\\ author T","cited_arxiv_id":null,"evidence_quote":"introduces recurrence networks as a nonlinear time series paradigm, the basis of the whole analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"shows how recurrence-network measures reflect the geometry of chaotic dynamics, grounding the use of L and T as order/chaos indicators."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the standard techniques for time delay embedding and for rank-based surrogate hypothesis tests."},{"cited_title":"Unified functional network and nonlinear time series analysis for complex systems science: The pyunicorn package","cited_arxiv_id":"1507.01571","evidence_quote":"demonstrates in the standard map that complex network measures disentangle regular and chaotic motion, the analogue the paper extends to planetary systems."},{"cited_title":"Surrogate Test to Distinguish between Chaotic and Pseudoperiodic Time Series","cited_arxiv_id":"nlin/0404054","evidence_quote":"introduces pseudo-periodic twin surrogates, whose null hypothesis of quasi-periodicity is the basis for the significance tests."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"further analyzed Kepler-36 with the reversibility error method, finding stickiness near the 7:6 resonance, which supports the stable-chaos interpretation."}],"review_version":1}