{"id":"a223f47e-c6a4-441a-9fd7-9a887cc03c58","arxiv_id":"2509.02339","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Persistent homology Betti curves and persistence distributions for Trajectum Pb-Pb and O-O events are robust and reflect known flow and multiplicity phenomenology, with no enhanced parameter sensitivity over standard observables.","lead":"Scientists applied topological data analysis to simulated heavy-ion collisions, building 'Betti curves' that track how particle patterns connect at different momentum scales. The new observables work and match known physics, but they do not appear to be more sensitive to the model's key parameters than standard measurements.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Small-n Pearson correlations (n=20) and an unquantified comparison to [46] leave the central negative claim statistically unsupported; the 'no enhanced sensitivity' conclusion may be an artifact of noise.","rationale":"The reader's weakest_assumption correctly identifies the n=20 posterior draws as the core statistical limitation. My stress-test pass did not find a more fundamental flaw: the TDA construction (alpha complexes, Betti curves, persistence distributions) is mathematically standard, the computations use GUDHI, and the dictionary linking TDA observables to multiplicity, radial flow, and elliptic flow is qualitative but plausible. The paper is honest about its limitations, explicitly noting in Sec. V that a mock-data Bayesian analysis is left to future work. The single most load-bearing concern is therefore not the viability of TDA, but the empirical basis for the negative sensitivity claim. With n=20, the Pearson correlations in Sec. IV.D are too noisy to support a firm quantitative comparison, and the comparison to [46] is not demonstrated. This does not invalidate the paper's positive contributions, but it does mean the abstract's 'do not show enhanced sensitivity' statement should be treated as provisional. The reader's CONDITIONAL verdict already captures this; my analysis agrees and recommends no change. I also considered whether the post hoc selection of the homology radius r=0.47 GeV/c (where the spread is widest) could bias the result; that would bias toward finding larger correlations, not smaller, so it does not weaken the negative claim. The comparison to standard observables is the more direct gap, but it is secondary to the small-n statistical power issue. My concrete test would settle the concern by requiring a matched, higher-statistics comparison with error bars on the correlation coefficients.","tokens_in":19288,"tokens_out":4890,"duration_ms":60137,"concrete_test":"Recompute the Sec. IV.D correlation analysis using N=200 posterior draws (or bootstrap the existing 20 draws to estimate sampling errors) and, crucially, compute Pearson r for the conventional observables of [46] using the exact same parameter draws and the same Trajectum runs. Report 95% confidence intervals (e.g., via Fisher z-transform or bootstrap) for both TDA and conventional observables. If the TDA |r| intervals broadly overlap the conventional-observable intervals, the 'no enhanced sensitivity' claim is supported; if TDA intervals are shifted higher, or if the current |r|>0.4 values collapse under resampling, the negative conclusion is an artifact and should be revised to 'inconclusive pending larger ensembles.'","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's central negative claim—'persistent homology observables ... do not show enhanced sensitivity ... compared to conventional observables'—rests on the correlation analysis of Sec. IV.D. There, Pearson coefficients are computed from only 20 posterior draws (Sec. III). For n=20, the standard error of r is about 0.24 (1/sqrt(17)), so the observed values in the range (-0.6, 0.6) are mostly within ~2.5 standard deviations of zero. The paper reports no confidence intervals, p-values, or a null model for r, so the spread in Fig. 8 could be dominated by sampling noise rather than genuine parameter sensitivity. Furthermore, the comparison to conventional observables is only asserted via reference [46]; no quantitative side-by-side values, error bars, or matched ensembles are provided. If the same 20 draws were used for standard observables, the comparison would be meaningful; as it stands, differences could reflect mismatched ensembles or metrics. This is load-bearing because the paper's headline conclusion that TDA offers no sensitivity gain is a statement about the relative size of correlations. With n=20, the analysis lacks the statistical resolution to distinguish 'no enhanced sensitivity' from 'not enough draws to tell.' The positive viability and robustness claims are not threatened by this, but the negative sensitivity claim is not robustly established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper implements persistent homology observables for alpha complexes built from transverse-momentum point clouds in the Trajectum hydrodynamic model, presenting Betti curves and persistence distributions for Pb–Pb at 5.02 TeV and O–O at 7 TeV across several centralities and particle species. The model parameters are taken as 20 random posterior draws from an earlier Bayesian analysis [1], which allows the authors to quote statistical and systematic uncertainties. A dictionary is constructed linking the topological observables to multiplicity, mass ordering from radial flow, and elliptic-flow-induced azimuthal correlations, using azimuthal randomization to isolate the latter. The paper's headline conclusion is that persistent homology observables are viable and robust but do not show enhanced sensitivity to model parameters compared with conventional observables. This conclusion is based on Pearson correlation coefficients computed from the 20 posterior draws and on a qualitative comparison with a previous correlation study [46].","tokens_in":19664,"tokens_out":3987,"duration_ms":50726,"significance":"If the main claims are established, the paper would provide a useful benchmark for TDA in heavy-ion phenomenology: it demonstrates that Betti curves and persistence distributions can be computed in a state-of-the-art event generator, that the results are stable under posterior parameter uncertainty, and that they encode familiar physics (multiplicity ordering, mass ordering from radial flow, and elliptic-flow effects). The O–O predictions and the explicit systematic-error treatment are valuable for future experimental comparisons. The authors are also careful to note that no persistent-homology observable entered the Bayesian fit of Ref. [1], so the predictions are genuine extrapolations rather than in-sample fits. However, the central negative claim—that TDA observables do not show enhanced parameter sensitivity—is not established with the statistical rigor currently presented, because it rests on a very small number of posterior draws and an unquantified comparison to Ref. [46].","major_comments":[{"comment":"The Pearson coefficients in Fig. 8 are computed from only n=20 posterior parameter draws, as stated in Sec. III. For n=20, the sampling uncertainty of r under the null hypothesis is approximately 1/sqrt(17) ≈ 0.24, so most of the reported values in the range (-0.6, 0.6) are within roughly 2.5 standard errors of zero. No confidence intervals, p-values, permutation nulls, or bootstrap estimates are given. This is load-bearing because Sec. V uses these correlations to conclude that the persistent-homology observables 'do not show enhanced sensitivity ... compared to conventional observables.' With n=20, the analysis cannot currently distinguish a genuine absence of enhanced sensitivity from insufficient sampling of the posterior. Please provide uncertainty estimates for the correlation coefficients, or increase the number of posterior draws, or both.","section":"Sec. IV.D and Sec. III"},{"comment":"The statement 'most of them have not been larger than the correlations observed for standard observables [46]' is not supported by any quantitative side-by-side comparison in this manuscript. Ref. [46] may involve different observables, a different parameter set, and possibly different posterior statistics, so the comparison is not apples-to-apples. To make the headline claim robust, the authors should compute Pearson (or rank) correlations for the same set of conventional observables using the same 20 posterior draws and present the two sets of coefficients in a common table or figure. Without this, the 'no enhanced sensitivity' conclusion is an assertion rather than a demonstrated result.","section":"Sec. V, comparison to [46]"},{"comment":"The correlation analysis is performed at radii chosen by 'qualitatively looking at the spread of the 20 individual calculations ... and selecting the interval in which the spread is the widest.' This data-driven selection, combined with scanning 25 (or 23) parameters and several observables, introduces a multiple-comparisons bias: large |rho| values are more likely to be found when both the radius and the parameter axis are scanned. The same issue applies to the persistence-window choice at P = 0.10 ± 0.04 GeV/c. The authors should either report the correlations as a function of r over a dense grid, or account for the selection effect, or present the analysis as exploratory rather than as evidence for the absence of enhanced sensitivity.","section":"Sec. IV.D, selection of homology radii"},{"comment":"Pearson's r measures only linear association and is highly sensitive to outliers. Fig. 9 explicitly demonstrates this: a strong Pearson correlation between nc and beta0 is driven by outliers, with no visible monotonic ordering in nc. Yet the later comparison with standard observables and the discussion of 'strong correlations' (e.g., with (zeta/s)_T0 in Fig. 10) continue to rely on Pearson coefficients without a linearity diagnostic or robust alternative. I recommend adding rank-based correlation measures (e.g., Spearman rho), scatter plots for the parameters highlighted in the text, and a statement of how many of the apparent correlations survive after removing outliers or using robust estimators.","section":"Sec. IV.D, Figs. 9 and 10"}],"minor_comments":[{"comment":"The sentence 'By performing around 400k calculations for each of these different choices' is ambiguous: does 'calculations' mean simulated events, hydrodynamic runs, or individual observable computations? Please clarify the computational cost and the number of events per posterior draw.","section":"Sec. III"},{"comment":"The text refers to 'saddle points in the distributions dN0/dP and dN0^rand/dP' when discussing the dimension-1 persistence distributions shown in Fig. 7. The subscripts appear to be typos and should be dN1/dP and dN1^rand/dP.","section":"Sec. IV.C, Fig. 7"},{"comment":"The introduction and conclusions mention 'n-point connected correlation functions' as traditional observables, but the dictionary built in this paper mainly connects Betti curves to multiplicity, radial flow, and elliptic flow. The connection to n-point connected correlators is not made explicit; either remove the phrase or spell out the correspondence.","section":"Sec. I and Sec. V"},{"comment":"The left and right panels use different vertical-axis ranges and different collision energies, which makes direct visual comparison of Pb–Pb and O–O somewhat difficult. Consider adding an inset or a ratio panel if this comparison is central.","section":"Fig. 2"},{"comment":"The list of parameters is clear, but there is an inconsistency in the text: the main text says 26 parameters in the Bayesian analysis, while Appendix C explains that 25 are varied for Pb–Pb and 23 for O–O because Norm is doubled and Woods-Saxon parameters are absent for O–O. Please make this count explicit at the first mention in Sec. III to avoid confusion.","section":"Appendix C"},{"comment":"There are several minor typographical issues, including missing spaces before 'Cech' in the footnote of Sec. II A, inconsistent notation for beta^rand_l, and the use of 'an' as a parameter name in Appendix C, which is easily misread as the English article. A careful proofreading pass would help.","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"The paper's positive contributions—the TDA implementation, the dictionary, the O–O predictions, and the systematic-error treatment—are solid and would be of interest to the heavy-ion community. The main concern is the statistical support for the negative headline claim. This is not a matter of internal inconsistency; it is a matter of the analysis not yet having the resolution to distinguish 'no enhanced sensitivity' from 'not enough posterior draws.' I would therefore encourage a revision in which the correlation analysis is strengthened (e.g., more draws, confidence intervals, matched comparison with standard observables) or the conclusion is explicitly softened to an exploratory statement. The current manuscript is close to publishable once this is addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful part of this paper is the construction itself: alpha-complex persistent homology on transverse momentum clouds in Trajectum, with Betti curves and persistence distributions for Pb-Pb and O-O, per species and centrality, with posterior parameter uncertainty bands. No TDA observable entered the Bayesian fit, so the predictions are genuine extrapolations, and the uncertainty propagation is a real step beyond earlier persistent-homology papers. The qualitative dictionary is convincing: multiplicity shows up in beta0, mass ordering in the beta1 peaks tracks radial flow, azimuthal randomization isolates elliptic-flow effects, and the Pb-Pb versus O-O comparison is sensible. They also catch one apparent correlation with nc and say so. The appendix is coherent on the math. Credit where due.\n\nThe soft spot is exactly the one in the stress-test note. The central negative claim rests on Pearson correlation coefficients computed from 20 posterior draws. For n=20, r values around 0.5 have standard errors near 0.2, and the plots show r between -0.6 and 0.6 across 25 parameters, so much of that spread could be sampling noise. No confidence intervals, p-values, or null model are given. The comparison to standard observables is a sentence referencing [46], not a matched side-by-side. That makes 'do not show enhanced sensitivity' under-supported: it could be true, but this analysis cannot distinguish no-sensitivity from not-enough-draws. The radius interval (0.47 +/- 0.04) is chosen post hoc by looking at spread, which is fine for exploratory correlators but should not be dressed as definitive.\n\nI also think the paper's own conclusions are more careful than the abstract. Section V says they cannot conclude persistent homology observables carry more information, and that constraining power can only be fully assessed with experimental measurements or a mock Bayesian fit. That is an honest limitation statement. But the abstract's phrasing commits to the negative sensitivity claim, and that is what should be revised.\n\nBottom line: worth refereeing. The viability/robustness half is solid and useful as a benchmark. The negative claim needs either more posterior draws, confidence intervals or a bootstrap, or a matched comparison to standard observables; alternatively, the conclusion should be scaled back to 'these observables are practical, and we do not yet see enhanced sensitivity.' I would send it to peer review with that demand, not desk-reject.","headline":"Useful, honest TDA benchmark for heavy-ion simulations, but the 'no enhanced sensitivity' claim is built on 20 posterior draws and an asserted comparison, so it is underpowered as stated.","tokens_in":20079,"tokens_out":1963,"would_cite":true,"duration_ms":24415,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["25.75.-q"],"model":"deepseek-v4-flash","headline":"Persistent homology is a workable, stable observable family for heavy-ion collisions, but it does not beat conventional observables in sensitivity to model parameters.","keywords":["persistent homology","heavy-ion collisions","quark-gluon plasma","Betti curves","alpha complexes","Trajectum","Bayesian parameter estimation","oxygen collisions"],"falsifier":"Adopt the paper's own suggested test: generate mock Betti-curve data from Trajectum, then run Bayesian fits with and without the topological observables and compare how much the posteriors of parameters such as (zeta/s)T0 shrink. If fits including Betti curves tighten constraints beyond the baseline, the 'no enhanced sensitivity' conclusion is false; if not, it stands.","tokens_in":1560,"feed_emoji":"🕳️","tokens_out":2063,"duration_ms":81926,"temperature":0.7,"pith_summary":"This paper establishes that persistent homology—a toolset that tracks how many connected clusters and holes appear in a point cloud as a length scale grows—can be applied end-to-end to simulated heavy-ion collisions. Working inside a full relativistic hydrodynamics model, the authors compute Betti curves and persistence distributions for the transverse-momentum point clouds of final-state hadrons in Pb-Pb and O-O collisions, with statistical and systematic uncertainties propagated from a Bayesian parameter fit. The observables reproduce known physics: centrality and mass ordering track particle multiplicities, and comparing real events with azimuthally randomized ones isolates the imprint of elliptic flow. The quantitative message is that these topological observables are practical and stable, but they do not carry more sensitivity to the model's tunable parameters than conventional observables do. The paper therefore positions persistent homology as an alternative perspective and a benchmark tool, with its real constraining power deferred until experimental measurements exist.","feed_headline":"Persistent homology proves viable for heavy-ion collisions","feed_subtitle":"Betti curves reproduce flow and multiplicity physics, with no sensitivity gain over standard observables.","key_machinery":"Alpha complexes: the family of simplicial complexes formed by growing disks around each final-state hadron's transverse-momentum position; sweeping the radius r and counting connected components (Betti number beta_0) and holes (beta_1) produces Betti curves whose birth-death structure is summarized by persistence distributions. The paper's discriminating device is the randomized-azimuth counterpart beta^rand: rebuilding the same point clouds with azimuthal angles drawn uniformly from [0,2pi) and subtracting recovers the contribution of angular correlations (elliptic flow) to the topology. Sensitivity to model parameters is quantified by Pearson correlation coefficients between each observabl","core_discovery":"The central claim is that topological data analysis, concretely the Betti curves beta_0(r) and beta_1(r) and the persistence distributions dN_l/dP of alpha complexes built from final-state hadron momenta, constitutes a workable and informative observable family for heavy-ion collisions. In Pb-Pb collisions at 5.02 TeV and O-O collisions at 7 TeV, these observables encode, in a single curve family, features normally separated into multiplicity, mass-ordered radial flow, and anisotropic flow; the azimuthal-randomization ratio beta^Delta/beta exposes the anisotropic-flow component. Correlating the observables with the parameters of the Trajectum Bayesian posterior, the authors find genuine corr","pith_inferences":["If experimental Betti curves become available and disagree with Trajectum, the same observables would switch roles: from parameter constraint to model-discrimination diagnostics, potentially revealing 3D collective structure that n-point correlators miss.","The paper's correlation analysis uses 20 posterior draws; with that sample size, Pearson coefficients below roughly 0.6 are only marginally significant, so the negative sensitivity result should be re-checked with a larger posterior sample or a mock-data Bayesian fit before being treated as definitive.","Extending the point cloud from the transverse plane to include longitudinal momentum—a future direction the paper names—is a natural test: TDA sensitivity to long-range 3D correlations could differ from the transverse-only case studied here.","Because the paper's dictionary ties Betti curves to known observables, the same pipeline could be ported to smaller collision systems such as Ne-Ne or to high-multiplicity proton-proton events, where conventional flow analyses are contested."],"forward_implications":["Pb-Pb and O-O Betti curves with full statistical and systematic uncertainties are now available as benchmarks for future experimental measurements at the LHC.","beta_0(r=0) equals the charged-hadron multiplicity, making the dimension-0 Betti curve a topological embedding of multiplicity plus clustering information.","The ratios beta^Delta_0/beta_0 and beta^Delta_1/beta_1 isolate anisotropic-flow effects, with stronger signals in semicentral collisions and in dimension 1, linking topology to flow phenomenology.","The correlation between dN/dP_1 and the bulk-viscosity peak temperature (zeta/s)T0 suggests Betti-curve data could help constrain this fluid parameter if the correlation survives larger statistics.","The absence of enhanced sensitivity relative to standard observables means the near-term role of these observables is cross-validation and model benchmarking rather than parameter constraint."],"supporting_citations":[{"why":"Supplies the Bayesian posterior parameter draws used for event generation and for propagating systematic uncertainties into the topological observables.","marker":"[1]"},{"why":"Provides the Trajectum relativistic hydrodynamics framework that generates the simulated final-state particle point clouds.","marker":"[6]"},{"why":"Provides the explicit oxygen nuclear configurations used for O-O collision simulations.","marker":"[17]"},{"why":"Supplies the mathematical background and computational conventions for persistent homology and alpha complexes.","marker":"[19]"},{"why":"Establishes the earlier application of persistent homology to heavy-ion flow correlations that this work complements with a different complex construction.","marker":"[21]"},{"why":"Supplies the GUDHI computational topology library used to evaluate homology and birth-death radii.","marker":"[37]"},{"why":"Provides the T RENTo initial-condition model generalized by Trajectum.","marker":"[42]"},{"why":"Provides the baseline Pearson correlations for standard observables against which the topological observables' sensitivity is compared.","marker":"[46]"}],"fun_headline_variants":["Topological data analysis scores in heavy-ion collisions","Persistent homology curves decode heavy-ion flow","Heavy-ion collisions get a topological map, no sensitivity gain","Betti curves viable for quark-gluon plasma, not sharper","Topology joins heavy-ion toolkit, mirrors standard observables"],"cache_read_input_tokens":21888,"weakest_assumption_plain":"The comparison that supports the paper's main negative conclusion—that topological observables are not more sensitive than conventional ones—uses Pearson correlations computed from just 20 parameter draws from the Bayesian posterior, a sample too small for correlation estimates below roughly 0.6 to be reliable.","fun_headline_variants_meta":{"raw":{"variants":["Topological data analysis scores in heavy-ion collisions","Persistent homology curves decode heavy-ion flow","Heavy-ion collisions get a topological map, no sensitivity gain","Betti curves viable for quark-gluon plasma, not sharper","Topology joins heavy-ion toolkit, mirrors standard observables"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000138,"raw_usage":{"total_tokens":981,"prompt_tokens":728,"completion_tokens":253,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":472,"completion_tokens_details":{"reasoning_tokens":174}},"tokens_in":472,"tokens_out":253,"duration_ms":3904,"temperature":1.0,"reasoning_tokens":174,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T11:35:22.436736+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Adopt the paper's own suggested test: generate mock Betti-curve data from Trajectum, then run Bayesian fits with and without the topological observables and compare how much the posteriors of parameters such as (zeta/s)T0 shrink. If fits including Betti curves tighten constraints beyond the baseline, the 'no enhanced sensitivity' conclusion is false; if not, it stands.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Bayesian posterior parameter draws used for event generation and for propagating systematic uncertainties into the topological observables."},{"cited_title":"Hensel, M","cited_arxiv_id":null,"evidence_quote":"Supplies the GUDHI computational topology library used to evaluate homology and birth-death radii."},{"cited_title":"Cohen-Steiner, H","cited_arxiv_id":null,"evidence_quote":"Provides the T RENTo initial-condition model generalized by Trajectum."}],"review_version":1}