{"id":"25ecf24c-7c9e-4820-881e-08e50328414d","arxiv_id":"2412.16931","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Mutual information between two nearby particle trajectories distinguishes regular from chaotic orbits in Schwarzschild and Kerr spacetimes, matching the fast Lyapunov indicator.","lead":"The authors propose a new way to detect chaotic motion near black holes by measuring how much information two nearby particle paths share, instead of measuring how fast they separate. This could make it easier to classify orbits in curved spacetime, with possible relevance to gravitational-wave modeling.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"MIPP's core 1-vs-0 dichotomy is neither derived nor fully specified; because the estimator's binning, window, and pairing variable are undefined, the claimed transition-detection advantage over FLI is not verifiable as stated.","rationale":"The reader identified essentially the same weakness: the MIPP heuristic is unproven, the initial separation is calibrated, and the estimator is under-specified. My concern adds a concrete physical mechanism why the regular=1 assumption can fail: nearby regular orbits on distinct tori have slightly different frequencies, so over sufficiently long times their phases drift and the joint distribution broadens, lowering mutual information. That mechanism is not mentioned in the reader's rationale. The paper does contain real supporting evidence: a high-precision symplectic integrator with energy errors around 1e-9, and broad qualitative agreement between MIPP and FLI across the parameter scans in Figs. 4-8. Those comparisons show the idea is worth pursuing, but they do not establish the transition-detection advantage claimed in the conclusions, especially because FLI is used as the reference while the paper simultaneously argues MIPP is better than FLI at the ambiguous critical orbits. Since the concern is about under-specification and lack of robustness rather than a demonstrated contradiction, the appropriate disposition remains CONDITIONAL: the paper should not be fully accepted until the algorithm and robustness checks are provided.","tokens_in":14108,"tokens_out":11937,"duration_ms":125077,"concrete_test":"Ask the authors for the exact MIPP algorithm used to produce Fig. 4, specifically the chosen observable, histogram bin count and width, and total integration time. Then recompute MIPP for the regular and chaotic reference orbits of that scan while varying the bin count over 8, 16, 32, and 64 and the integration window over T, 10T, and 100T. If the regular orbit's MIPP falls below about 0.5 or the chaotic orbit's rises above 0.5 for any reasonable setting, the assumed 1/0 separation is an artifact of the unspecified implementation and the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that MIPP is a chaos indicator and can detect orbital-state transitions rests on the assumed dichotomy in Sec. IV.C: normalized mutual information approaches 1 for regular orbits and 0 for chaotic orbits. That dichotomy is not derived from Eqs. (25)-(26), and the estimator is never fully defined. The text does not specify which phase-space coordinate or time series serves as X and Y, how the joint distribution is accumulated, the histogram bin count or bin width, or the total integration window. The only calibration parameter mentioned, initial separation δr=10^(-8), is justified circularly: it is chosen because 'for order orbits, the MIPP calculation results are close to one, while for chaotic orbits, they are close to zero.' Mutual information estimates from finite time series are strongly binning- and window-dependent, so the reported 1/0 separation could be produced by the unspecified implementation rather than by the dynamics. Moreover, even for genuinely regular orbits, two nearby trajectories generally lie on different invariant tori and have slightly different frequencies; over long times their phases drift and the joint distribution can fill a two-dimensional region, causing normalized mutual information to drop below 1. This is a concrete mechanism by which the assumed regular=1 behavior fails, independent of binning. The paper itself concedes in Sec. V that MIPP values near 0.5 'may indicate a failure of MIPP' and require other indicators, which undermines the stronger claim that MIPP can accurately indicate transitions on its own.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a new chaos indicator named MIPP (Mutual Information for Particle Pair), defined as the normalized mutual information between two nearby trajectories in curved spacetime. The authors integrate charged-particle motion in Schwarzschild and magnetized Kerr spacetimes with an explicit symplectic PRK integrator, and compare MIPP scans over initial radius, energy, angular momentum, and black-hole spin against FLI scans. They report that MIPP values near 1 identify regular orbits and values near 0 identify chaotic orbits, and they claim that MIPP can read orbital-state transitions more directly than FLI. The central claim is that MIPP is a proper chaos indicator.","tokens_in":14433,"tokens_out":4454,"duration_ms":39554,"significance":"If the MIPP dichotomy (regular = 1, chaotic = 0) were established with a fully specified estimator, the method would be a useful complement to FLI, with the advertised advantages of avoiding renormalization and of indicating transitions directly from scan plots. The manuscript has genuine strengths: it uses a high-order explicit symplectic integrator with small energy error (about 1e-9), tests multiple parameter scans in both Kerr and Schwarzschild spacetimes, and reports CPU costs for the comparative scans. The underlying idea of using mutual information between a particle pair as a dynamical probe is interesting and potentially transferable to other relativistic systems. However, the current evidence is largely qualitative: agreement with FLI is shown through visual comparison of scan plots, the estimator is not completely specified, and the key 1/0 classification is calibrated rather than derived.","major_comments":[{"comment":"The estimator is underdefined. The manuscript does not specify which time series or phase-space coordinates serve as X and Y for the two particles, how the joint distribution rho(x,y) is accumulated during the integration, the number or width of histogram bins, the total integration window, or the sampling rate. Normalized mutual-information estimates from finite time series are known to depend strongly on binning and window choice, so the reported separation between 1 and 0 cannot be reproduced or assessed from the text as it stands. Please provide a complete algorithmic specification, including a pseudocode box, the exact binning rule, the sample count, the integration time, and ideally make the code available.","section":"Section IV.C, Eqs. (25)-(26)"},{"comment":"The dichotomy 'regular orbit gives normalized MI = 1, chaotic orbit gives normalized MI = 0' is asserted rather than derived. Equations (25)-(26) are only the definitions of mutual information and its normalization; the claim that for ordered systems H(X,Y) = H(X) or H(Y) is not generally valid in the two-particle setup. Two nearby trajectories on different invariant tori generally have slightly different frequencies and will dephase over a long time window, so their joint distribution can approach the product of the marginal distributions, driving normalized MI below 1 even for regular orbits. This is a concrete failure mechanism independent of binning. The authors should either derive the asymptotic behavior from the dynamics (for example using action-angle variables) or demonstrate numerically that MIPP stays close to 1 for regular orbits over a range of bin counts and window lengths.","section":"Section IV.C, paragraph after Eq. (26)"},{"comment":"The choice of initial separation delta r = 1e-8 is justified circularly. The text states that this value is chosen because 'for order orbits, the MIPP calculation results are close to one, while for chaotic orbits, they are close to zero'; in other words, the parameter is tuned to reproduce the classification that the method is supposed to predict. Because the same scan families are later used to claim agreement with FLI, the validation is only partially independent. Please replace this with an objective selection criterion (for example, a convergence test as a function of delta r on a held-out set of orbits, or a scale set by the geodesic-deviation scale) and show that the classification is stable over a range of delta r and over the threshold value 0.5.","section":"Section IV.C, last paragraph"},{"comment":"The agreement between MIPP and FLI is reported only qualitatively ('align well', 'completely consistent'), with no quantitative measure. The FLI threshold value of 10 is cited as 'experimentally verified' but no reference or criterion is given for that value. To support the claim that MIPP is a proper chaos indicator, the paper should provide a quantitative comparison, for example a classification agreement rate or confusion matrix against FLI over the scanned orbits, and explicitly state how near-boundary orbits with MIPP near 0.5 are counted. This is load-bearing because the paper itself concedes that MIPP values near 0.5 'may indicate a failure of MIPP', leaving the transition-detection claim dependent on the very ambiguity it aims to resolve.","section":"Section V, Figs. 4-8"}],"minor_comments":[{"comment":"The sentence 'Our result show that information theory significantly deepen our understanding' should read 'Our results show that information theory significantly deepens our understanding'.","section":"Abstract"},{"comment":"The equation numbering is inconsistent: the Hamiltonian in Eq. (2) is later referred to as 'Eq. (1)' and in Section III the text refers to 'the Hamiltonian (20)' and 'each sub-Hamiltonian in Eq. (20)', while Eq. (20) is actually the time-transformation d tau = T(r,theta) dw. Please renumber the equations consistently.","section":"Section II / Section III"},{"comment":"The phrase 'the well-known Wlad potential' should read 'the well-known Wald potential', and the spelling 'Wald' should be used consistently.","section":"Section II"},{"comment":"The captions of Figures 3-8 do not explicitly define the vertical axis label for the MIPP panels; please state that the ordinate is the normalized mutual information (bar-I) and the abscissa is the scanned parameter.","section":"Figure captions"}],"recommendation":"major_revision","confidential_remarks":"The manuscript falls within the scope of the journal and the proposed method is potentially interesting, but the current version has a load-bearing specification gap and a circular calibration issue that can likely be fixed with additional algorithmic details and robustness tests. No concerns about citation or novelty beyond those expressed in the report."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know before you read this. First, the paper genuinely proposes a new diagnostic — normalized mutual information between two nearby particle trajectories — and its parameter scans across r, E, L, and spin show good qualitative agreement with FLI. Second, the core regular=1 / chaotic=0 dichotomy is not derived from Eqs. (25)–(26); it is a heuristic, and the estimator itself is never fully specified. That second point is the difference between a useful suggestion and a verified method.\n\nThe symplectic integrator section is solid and follows the Wu et al. line; the energy-error checks look careful. The parameter scans cover a reasonable range, including Schwarzschild, and MIPP tracks FLI well except in the admitted gray zone near 0.5. Acknowledging that gray zone is honest, though it undercuts the stronger \"MIPP can accurately indicate transitions\" claim.\n\nThe soft spots are real and mostly in Section IV.C. The paper never says which time series or phase-space coordinate plays the role of X and Y, how the joint distribution is accumulated, what bin count or width is used, or over what integration window. Mutual information from finite series is sensitive to exactly those choices. The initial separation δr=10^-8 is justified by saying it gives values close to 1 or 0 — that is circular. And the stress-test point about tori is not idle: two nearby regular orbits generally have slightly different frequencies; over long times their phases drift and the joint distribution can fill a region, pushing normalized MI below 1. That mechanism is independent of binning and undermines the assumed clean dichotomy.\n\nThe CPU-time claims are overstated. The reported costs differ by a few percent (736 vs 740 s, etc.), so \"significantly improving computational efficiency\" is not supported. No code, no error bars, and no robustness checks for binning or window are provided.\n\nWho is this for? Practitioners testing chaos in geodesic motion around black holes might find MIPP a convenient quick screen, but only after the estimator is pinned down. The EMRI application is speculative. It deserves a serious referee — the idea is plausible and the FLI agreement is real — but the manuscript in this form should not be accepted; it needs a full specification, robustness analysis, and preferably code.\n\nRecommendation: send to review, with a clear request for those additions.","headline":"A plausibly useful but under-specified chaos indicator: the 1-vs-0 dichotomy is a tuned heuristic, not a derived result, and the estimator needs full specification before the claims are verifiable.","tokens_in":14919,"tokens_out":2476,"would_cite":false,"duration_ms":22246,"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":"Mutual information between two nearby orbits can tell chaotic from regular motion around black holes.","keywords":["mutual information","chaos detection","Kerr black hole","Schwarzschild spacetime","charged particle motion","fast Lyapunov indicator","symplectic integrator","Wald magnetic field"],"falsifier":"A specific disproof would be an orbit that a high-resolution Poincaré section or spectrum clearly identifies as regular but whose MIPP value falls at or below 0.5 over long integration, or conversely a chaotic orbit with MIPP pinned near 1; a second, easier check is to show that the classification flips when the histogram bin count or the integration window is changed while the trajectory is held fixed.","tokens_in":13923,"feed_emoji":"🕳️","tokens_out":6111,"duration_ms":52122,"temperature":0.7,"pith_summary":"The paper proposes a new chaos indicator called mutual information for particle pair (MIPP), defined for two nearby test-particle orbits in curved spacetime. It claims that the normalized mutual information between the two trajectories sits near 1 for regular orbits and near 0 for chaotic orbits, and tests this in Schwarzschild and Kerr spacetimes with an external magnetic field. Compared with the fast Lyapunov indicator (FLI), MIPP matches FLI's classifications across parameter scans and, the authors argue, shows the moment an orbit changes state directly from the scan plot. The practical payoff would be a faster, renormalization-free numerical probe that could screen large ensembles of relativistic orbits without visual inspection.","feed_headline":"Two-orbit mutual information flags chaos around black holes","feed_subtitle":"A normalized mutual-information value near 1 marks regular orbits, near 0 marks chaos, matching FLI at lower cost.","key_machinery":"The carrying object is MIPP itself: the normalized mutual information $\\bar{I}(X;Y) = I(X;Y)/H(X,Y)$ computed from the time series of two nearby trajectories, where $X$ and $Y$ are binned state samples (here the orbital coordinate $r$ along each trajectory). The paper pairs this with the two-particle method, in which the second trajectory is launched at a small separation (chosen as $10^{-8}$), and with a fourth-order partitioned Runge-Kutta symplectic integrator that keeps energy errors near $10^{-9}$. The normalization makes the indicator dimensionless with a nominal 1/0 separation between regular and chaotic motion; the boundary near 0.5 is used to flag intermediate or unreliable cases.","core_discovery":"The central claim is that mutual information, a standard information-theoretic measure, can serve as a dynamical chaos indicator in general relativity. The authors define MIPP as the normalized mutual information between two trajectories started with a small initial separation, so that a value near 1 means the second trajectory's states are statistically determined by the first, and a value near 0 means the trajectories have decorrelated. Using explicit symplectic integration of charged-particle motion around Schwarzschild and Kerr black holes, they report that MIPP reproduces FLI classifications for scans over launch radius, energy, angular momentum, and black hole spin, and that MIPP detects regular-to-chaotic transitions without the case-by-case inspection that FLI's final value demands. The authors conclude that MIPP can accurately indicate transitions in orbital states while FLI requires careful judgment of critical orbits, that MIPP needs no renormalization, and that it uses slightly less CPU time.","pith_inferences":["A natural next test would replace the ad hoc histogram binning with a data-driven partition, such as fixed quantiles, to see whether MIPP's 1/0 separation survives, since the present estimator's bin parameters are not specified.","The 0.5 plateau could be repurposed as a quantitative ambiguity score rather than a failure mode, giving a built-in confidence warning for borderline orbits.","The same two-trajectory mutual information could be applied to unbounded dynamics such as three-body escape, where ordinary Lyapunov exponents vanish and MIPP's finite-time normalization might still separate regular from irregular regimes.","Because the initial separation of $10^{-8}$ is calibrated by the desired output, an independent test on analytically known integrable and nonintegrable toy systems would clarify whether MIPP inherits its power from the two-particle divergence rate or from the binning procedure."],"forward_implications":["If MIPP is reliable, a single scan plot over a parameter such as energy or launch radius shows where order breaks into chaos, removing the need to inspect each orbit's final FLI value.","The indicator inherits the two-particle method's advantage of avoiding variational equations and adds the elimination of renormalization, so large parameter-space surveys become cheaper.","Because MIPP agrees with FLI in both Schwarzschild and Kerr with an external magnetic field, it is a candidate generic probe for geodesic and charged-particle chaos in other stationary spacetimes.","The observed MIPP plateau near 0.5 for high energies in Schwarzschild gives an explicit warning zone where the indicator alone should not be trusted and other methods must be consulted.","For gravitational-wave sources such as extreme-mass-ratio inspirals, a fast orbit-state indicator could help identify chaotic phase transitions in waveform modeling."],"supporting_citations":[{"why":"It supplies the definition of mutual information and the information-thermodynamics relation that motivates MIPP's use for orbital states.","marker":"[24]"},{"why":"It introduces the two-nearby-trajectory Lyapunov indices in curved spacetime that MIPP extends and compares against as FLI.","marker":"[19]"},{"why":"It provides the Wald potential describing the external magnetic field around the Kerr black hole used in the Hamiltonian.","marker":"[28]"},{"why":"It gives an explicit symplectic integrator for charged particles around a magnetized Kerr black hole that underpins the numerical orbits.","marker":"[11]"},{"why":"It is the authors' prior work defining Shannon entropy for a single orbit, the starting point for the mutual-information construction.","marker":"[2]"},{"why":"It provides the baseline comparison of methods for computing Lyapunov exponents with two nearby trajectories.","marker":"[16]"},{"why":"It is the optimized fourth-order partitioned Runge-Kutta symplectic algorithm used for the numerical integration.","marker":"[34]"}],"fun_headline_variants":["Mutual information reveals chaos in black hole orbits","Information theory detects chaos near black holes","Pairwise mutual information signals chaos around black holes","Chaos indicator from information theory for black holes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method's power rests on the unproven heuristic that mutual information between two nearby trajectories approaches 1 for regular orbits and 0 for chaotic orbits, an expectation calibrated by choosing the initial separation so that test orbits give these values; if that heuristic fails for a class of orbits, MIPP's classification boundary loses its meaning.","fun_headline_variants_meta":{"raw":{"variants":["Mutual information reveals chaos in black hole orbits","Information theory detects chaos near black holes","Pairwise mutual information signals chaos around black holes","Chaos indicator from information theory for black holes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000543,"raw_usage":{"total_tokens":2526,"prompt_tokens":800,"completion_tokens":1726,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":416,"completion_tokens_details":{"reasoning_tokens":1669}},"tokens_in":416,"tokens_out":1726,"duration_ms":24353,"temperature":1.0,"reasoning_tokens":1669,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T05:57:16.675479+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A specific disproof would be an orbit that a high-resolution Poincaré section or spectrum clearly identifies as regular but whose MIPP value falls at or below 0.5 over long integration, or conversely a chaotic orbit with MIPP pinned near 1; a second, easier check is to show that the classification flips when the histogram bin count or the integration window is changed while the trajectory is held fixed.","supporting_citations":[{"cited_title":"Applying explicit symp lectic integrator to study chaos of charged particles around magnetized Kerr black hole,","cited_arxiv_id":null,"evidence_quote":"It gives an explicit symplectic integrator for charged particles around a magnetized Kerr black hole that underpins the numerical orbits."},{"cited_title":"Screen chaotic motion by Shannon entropy in curved spacetimes","cited_arxiv_id":"2410.20870","evidence_quote":"It is the authors' prior work defining Shannon entropy for a single orbit, the starting point for the mutual-information construction."},{"cited_title":"A Note on the Construc tion of Explicit Symplectic Integrators for Schwarzschild Spacetimes,","cited_arxiv_id":null,"evidence_quote":"It is the optimized fourth-order partitioned Runge-Kutta symplectic algorithm used for the numerical integration."}],"review_version":1}