{"id":"34460cb6-5eb2-4344-ba59-7c4f9c92898c","arxiv_id":"2411.08407","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"An HHT-based pipeline estimates protoneutron star mass/radius-squared from simulated core-collapse supernova gravitational waves with accuracy comparable to STFT-based methods.","lead":"This paper applies the Hilbert-Huang transform to extract oscillation frequencies from simulated core-collapse supernova gravitational wave signals, then uses these frequencies to estimate the mass-to-radius-squared ratio of the protoneutron star. The method matches the accuracy of a standard short-time Fourier transform approach, suggesting it could serve as a complementary analysis tool for future detections.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The comparable-accuracy claim rests on an untested, post-hoc IMF-summation rule (sum IMF1-3, drop IMF4); if that rule is biased for other waveforms or after cWB reconstruction, the HHT frequency and hence the MPNS/R2PNS estimate is unreliable.","rationale":"The reader's weakest-assumption identification is correct. The paper's own evidence (pure-signal RMSE ratios of 0.66 for he3.5 and 0.959 for y20, plus the Fig. 12 box plots) supports the comparable-accuracy claim for the two tested waveforms, and the stress-test found no internal inconsistency that would overturn the headline result. The root-selection issue mentioned by the reader is actually benign: for the 2g2 coefficients in Table II, f'(x)=b+2cx+3dx^2 has discriminant 4c^2-12bd < 0, so the cubic is strictly monotonic and the real root is unique; no selection ambiguity remains. The load-bearing concern is instead the IMF-summation heuristic. It is post-hoc, adopted after inspecting the he3.5 decomposition, and not independently validated; it is exactly the step that connects the HHT output to the universal relation. Without it, the cWB+HHT estimates have no well-defined frequency. The proposed concrete test would settle whether the heuristic is robust or whether the reported accuracy is conditional on this tuning. Given this, the appropriate verdict remains CONDITIONAL with moderate confidence; no adjustment is needed.","tokens_in":13410,"tokens_out":7552,"duration_ms":78796,"concrete_test":"Run a controlled validation on the same two waveforms plus at least two additional CCSN models (or synthetic chirps with a known 2g2 instantaneous-frequency law). For each, inject into ET noise at several distances, apply cWB reconstruction, then CEEMD with the Table I parameters. Measure the bias and scatter of the IF of c1+c2+c3 against the known 2g2 frequency, and compare the RMSE of MPNS/R2PNS under alternative combination rules such as c1+c2, c1..4, or IA-weighted sums. If the RMSE is insensitive to the rule and the bias is consistent with zero, the heuristic is not load-bearing; if the RMSE changes materially or bias appears, the reported comparable accuracy is an artifact of the tuned rule.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that cWB+HHT matches cWB+STFT accuracy is supported only if the instantaneous frequency of the summed IMFs c1+c2+c3, with IMF4 excluded, faithfully tracks the 2g2 mode. This rule is adopted after observing mode splitting in the he3.5 signal (Sec. IV A, Fig. 9); no model-based criterion fixes the number of IMFs to sum or the IMF4 cutoff (small IA, IF below 200 Hz), and no independent validation is given for other waveforms, cWB-reconstructed signals, or different noise realizations. The HHT IF of a sum of components with overlapping time-frequency support is not guaranteed to equal the 2g2 IF; it can be an amplitude-weighted mixture, especially when 2g1 and 2g2 are close, as Fig. 8 shows IF1 lying between the two modes near -0.2 to -0.1 s. Since the estimated frequency is inserted into the universal relation Eq. (10), any systematic bias in the summed IF propagates directly into a biased MPNS/R2PNS estimate. The authors themselves state in Sec. V that further validation with additional simulations is necessary and that improved mode-splitting techniques are needed. Thus the headline comparison is conditional on a heuristic that is the least-secure link in the analysis chain.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a parameter-estimation pipeline for protoneutron stars (PNSs) in core-collapse supernova gravitational-wave signals. The method combines coherent WaveBurst (cWB) reconstruction with the Hilbert-Huang transform (HHT) based on complementary ensemble empirical mode decomposition (CEEMD) to extract the 2g2 g-mode frequency, which is then inserted into empirical universal relations to infer MPNS/R2PNS. The authors compare this approach with a short-time Fourier transform (STFT) baseline, first on pure simulated signals and then on signals injected into simulated Einstein Telescope noise over 100 realizations. They report comparable or slightly better RMSE for the HHT-based method in the pure-signal case and comparable accuracy in the noise-injected case.","tokens_in":13673,"tokens_out":3790,"duration_ms":39671,"significance":"If the reported accuracy is robust, the method provides a complementary time-frequency tool for CCSN asteroseismology that does not suffer from the time-frequency resolution tradeoff of STFT. The paper includes Monte Carlo simulations with 100 noise realizations and uses publicly available waveform models, which is a strength. However, the central claim rests on a heuristic IMF-summation rule and an unspecified root-selection procedure for the cubic universal relation, and the pure-signal comparison lacks uncertainty quantification. These issues currently limit the reproducibility and generalizability of the result.","major_comments":[{"comment":"The procedure for solving the cubic universal relation Eq. (10) is incomplete. The text states that roots are computed via the eigenvalues of the companion matrix, but a cubic equation generally has up to three real roots, and the paper does not specify how the physically relevant root for MPNS/R2PNS is selected. Without this selection rule, the RMSE results in Fig. 10 and Fig. 12 cannot be reproduced by an independent reader.","section":"Sec. III C"},{"comment":"The rule of summing IMF1 through IMF3 and excluding IMF4 is adopted after observing mode splitting in the he3.5 pure signal, with the justification that IMF4 has small IA and IF below 200 Hz. This is a post-hoc heuristic with no quantitative threshold or model-based criterion, and it is not validated for other waveforms, for cWB-reconstructed signals, or across noise realizations. The authors' own observation in Fig. 8 that IF1 lies between the 2g1 and 2g2 modes near -0.2 to -0.1 s indicates that mode mixing can bias the instantaneous frequency, so the summed IF may not faithfully represent the 2g2 mode. The paper needs either a principled rule for selecting the IMFs to sum or an independent validation that the chosen summation tracks the 2g2 frequency without bias.","section":"Sec. IV A, Fig. 9"},{"comment":"The pure-signal RMSE ratios (0.66 for he3.5 and, implicitly, 0.959 for y20) are computed from a single noise-free realization for each model. No uncertainty is associated with these numbers, so the reported 33.9% and 4.1% improvements of HHT over STFT cannot be distinguished from random variability. A bootstrap over time samples or multiple independent realizations would be needed to support the claim of higher accuracy in the noiseless case.","section":"Sec. IV A"},{"comment":"Although the box plots in Fig. 12 show that the cWB+HHT and cWB+STFT methods produce similar RMSE distributions, the paper does not report any statistical test comparing the two methods across the 100 noise realizations (e.g., a paired test at each distance). The conclusion of comparable accuracy is visually plausible but would be strengthened by a quantitative comparison that accounts for the pairing of noise realizations.","section":"Sec. IV B"}],"minor_comments":[{"comment":"The sentence 'In this simulation, the GW signal comprises two polarization states' should use plural form: 'In these simulations'.","section":"Sec. II A"},{"comment":"The paper reports the HHT/STFT RMSE ratio for he3.5 as 0.66 but does not explicitly give the corresponding ratio for y20; stating both values would make the comparison clearer.","section":"Sec. IV A"},{"comment":"The caption of Fig. 12 contains a typo: 'Each box plot shows its an interquartile range' should be 'shows its interquartile range'.","section":"Sec. IV B"},{"comment":"The CEEMD parameters in Table I (number of IMFs, ensemble size, noise standard deviation, stoppage criterion) are fixed without a sensitivity study or a reference justifying these specific values; a brief discussion of their robustness would help the reader assess the method's reliability.","section":"Sec. III B"},{"comment":"The definition of the analytic signal is given for an IMF ck(t), but the text later says it applies to 'any combination of IMFs'; it would be clearer to define the Hilbert transform for an arbitrary time series or to explicitly state that the analytic signal of the summed IMFs is used in the frequency extraction.","section":"Sec. III B, Eq. (9)"},{"comment":"The STFT reference method uses the frequency bin of maximum amplitude at each time; this estimator is known to be biased in the presence of noise, which may explain the increasing RMSE of the STFT-based method at large distances. The authors could mention this as a possible reason for the observed gap and as a caveat when interpreting the comparison.","section":"Sec. IV B"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses a timely topic and the Monte Carlo framework over 100 noise realizations is a plus. The main concerns are methodological: the missing root-selection rule for the cubic universal relation and the post-hoc, unvalidated IMF-summation heuristic are load-bearing for the central claim of comparable accuracy. I would encourage the authors to provide the missing details and to add a statistical comparison of the box-plot distributions. If these points are addressed, the paper could be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says: it applies the Hilbert-Huang transform to extract the 2g2 frequency and estimate MPNS/R2PNS, then compares head-to-head with STFT on two CCSN waveforms. The comparison is fair, the noise-injection results are reported as box plots over 100 realizations, and the central conclusion—comparable accuracy—is supported for the tested cases. In the pure-signal case the HHT RMSE is lower, but the honest bottom line is that the two methods end up roughly equivalent once cWB reconstruction and noise are in the loop.\n\nWhat is new is modest but real: this is the first application of HHT to PNS parameter estimation from g-mode GWs, and the explicit cWB+HHT versus cWB+STFT comparison is useful practical information. The paper inherits an established HHT pipeline from the same group and applies it to a new problem, which is a legitimate incremental contribution. The citation pattern is appropriate; the universal-relation literature and prior HHT-GW work are all cited, and the comparison against STFT is not circular.\n\nThe soft spots are proportionate. The weakest link is the IMF summation rule: summing c1+c2+c3 and dropping c4 is justified after observing mode splitting in he3.5, with no independent criterion or validation on other waveforms. If that rule biases the instantaneous frequency, the bias propagates directly into MPNS/R2PNS through the universal relation. This does not overturn the comparison claim for these two waveforms, but it does mean \"comparable accuracy\" is demonstrated rather than guaranteed. The paper itself acknowledges the need for further validation and better mode-splitting techniques, which is the right level of honesty. Two smaller gaps: the selection of the cubic root is not described, and no code or data are provided. The pure-signal RMSE without uncertainties is a minor issue because the noise-injection results carry the main weight.\n\nWho is this for: anyone working on CCSN GW parameter estimation who wants a complementary frequency-extraction tool. It deserves a serious referee. The revision should ask for more waveforms, an explicit IMF-selection procedure, and the root-selection details.\n\nRecommendation: send to peer review. The paper is not groundbreaking, but it is a solid, honest methodological comparison with a clear limitation that is openly stated.","headline":"A clean, honest head-to-head showing HHT matches STFT for PNS g-mode parameter estimation, with a post-hoc IMF-summing rule that needs independent validation before the method generalizes.","tokens_in":14254,"tokens_out":1545,"would_cite":false,"duration_ms":17154,"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 claims that a Hilbert-Huang transform pipeline estimates protoneutron-star mass-radius ratios from supernova gravitational waves with accuracy comparable to STFT, and equivalent after cWB reconstruction.","keywords":["Hilbert-Huang transform","empirical mode decomposition","gravitational waves","core-collapse supernovae","protoneutron star asteroseismology","universal relations","coherent WaveBurst"],"falsifier":"Inject a synthetic waveform with an artificially prescribed g-mode chirp and known mass-radius relation into detector noise, run the cWB+HHT pipeline, and check whether the summed IF1..3 follows the injected frequency at every post-bounce time; a systematic deviation at some frequencies would falsify the summing rule.","tokens_in":13208,"feed_emoji":"🌊","tokens_out":11764,"duration_ms":97243,"temperature":0.7,"pith_summary":"The paper tries to establish that the Hilbert-Huang transform (HHT) can extract the g-mode frequency of a protoneutron star from a core-collapse supernova gravitational-wave signal and use it to estimate the star's mass-to-radius-squared ratio $M_{\\rm PNS}/R_{\\rm PNS}^{2}$. When applied to two simulated waveforms, the HHT route matches or beats a short-time Fourier transform (STFT) baseline: the root-mean-square error is 33.9% lower for the he3.5 signal and 4.1% lower for the y20 signal in the noiseless case. In simulated third-generation detector noise, the combination of coherent WaveBurst reconstruction with the HHT gives parameter estimates statistically comparable to cWB with STFT. A second independent frequency-extraction method with similar accuracy is useful because it can cross-check the physical parameters inferred from a future single supernova event.","feed_headline":"HHT estimates protoneutron star properties as well as STFT does","feed_subtitle":"A new frequency-extraction pipeline could cross-check mass-radius ratios from a future core-collapse supernova signal.","key_machinery":"The pipeline runs in three stages. Coherent WaveBurst (cWB) reconstructs the transient signal from the three-detector network using a time-frequency excess-power search with Wilson-Daubechies-Meyer wavelets. Complementary ensemble empirical mode decomposition (CEEMD) then splits the reconstructed time series into four intrinsic mode functions (IMFs) plus a residual, adding and subtracting Gaussian noise in 1000 realisations to stabilise the decomposition. Finally, the Hilbert transform of each IMF gives an instantaneous frequency ${\\rm IF}_k(t)$; because the 2g2 mode is split across the first three IMFs, the sum ${\\rm IF}_{1..3}={\\rm IF}_1+{\\rm IF}_2+{\\rm IF}_3$ (with IMF4 discarded) is taken as the mode frequency and inserted into the cubic universal relation $f = a + bx + cx^{2} + dx^{3}$, whose solution $x = M_{\\rm PNS}/R_{\\rm PNS}^{2}$ is evaluated at each time step.","core_discovery":"The central claim is that the g-mode frequency needed for parameter estimation can be recovered by summing the instantaneous frequencies of the first three intrinsic mode functions produced by complementary ensemble empirical mode decomposition, and that solving the universal relation for the 2g2 mode with this summed frequency yields estimates of $M_{\\rm PNS}/R_{\\rm PNS}^{2}$ with accuracy comparable to or better than STFT. In the noiseless case the HHT root-mean-square error is 33.9% lower for the he3.5 waveform and 4.1% lower for y20. In detector-noise simulations with coherent WaveBurst reconstruction, the RMSE of cWB+HHT and cWB+STFT are statistically equivalent, supporting the paper's conclusion that the two frequency-extraction methods have comparable accuracy.","pith_inferences":["The sum-of-first-three-IMFs rule has only been checked on two waveforms; testing it on a larger set of progenitors and equations of state would reveal whether it introduces a systematic bias in the recovered $M/R^2$.","Injecting synthetic chirps with a known frequency evolution and known $M/R$ would isolate the frequency extractor from the universal-relation uncertainty and directly test the decomposition heuristic.","The HHT's ability to localise features in time-frequency space could be turned on other CCSN signatures (SASI, convection, f/p-modes), giving independent constraints on the same protoneutron-star parameters."],"forward_implications":["A future core-collapse supernova gravitational-wave detection could have its protoneutron-star mass-radius ratio estimated by two time-frequency methods that do not share the STFT's resolution limitations.","The HHT's smoother, continuous frequency tracks allow the g-mode evolution to be traced without the time-discontinuities that appear in the STFT maximum-amplitude track.","In the noise-injected runs, the estimation error grows with source distance in the same way for both methods, and the distance at which it degrades tracks the cWB network match.","Because the pipeline does not assume a waveform model, it can be paired with other reconstruction methods, such as the BayesWave algorithm suggested by the authors as future work."],"supporting_citations":[{"why":"Supplies the Hilbert-Huang transform (empirical mode decomposition plus Hilbert spectral analysis) that the paper adapts for frequency extraction.","marker":"[40]"},{"why":"Provide the g-mode universal relation and its erratum, with the 2g2 coefficients linking frequency to $M_{\\rm PNS}/R_{\\rm PNS}^{2}$.","marker":"[33, 34]"},{"why":"The coherent WaveBurst algorithm used to reconstruct the gravitational-wave transient from the three-detector noise.","marker":"[17–19]"},{"why":"Supplies the he3.5 waveform, one of the two simulated supernova signals tested.","marker":"[46]"},{"why":"Supplies the y20 waveform, the second simulated supernova signal tested.","marker":"[47]"},{"why":"Defines complementary ensemble empirical mode decomposition, the specific variant of EMD used to obtain the IMFs.","marker":"[59]"},{"why":"Identifies the mode-splitting problem in empirical mode decomposition that motivates summing IMFs 1-3.","marker":"[58]"}],"fun_headline_variants":["HHT matches STFT for protoneutron star parameter estimation","Gravitational wave analysis: HHT vs STFT for supernova cores","Hilbert-Huang transform offers comparable core-collapse GW estimates","New frequency tool could cross-check neutron star mass-radius","HHT achieves STFT-level accuracy on protoneutron star signals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method assumes that the 2g2 g-mode frequency is recovered by summing the instantaneous frequencies of the first three IMFs and discarding the fourth, a rule chosen after observing the mode split in these two waveforms.","fun_headline_variants_meta":{"raw":{"variants":["HHT matches STFT for protoneutron star parameter estimation","Gravitational wave analysis: HHT vs STFT for supernova cores","Hilbert-Huang transform offers comparable core-collapse GW estimates","New frequency tool could cross-check neutron star mass-radius","HHT achieves STFT-level accuracy on protoneutron star signals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000163,"raw_usage":{"total_tokens":1199,"prompt_tokens":860,"completion_tokens":339,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":248}},"tokens_in":476,"tokens_out":339,"duration_ms":3890,"temperature":1.0,"reasoning_tokens":248,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:37:06.892350+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inject a synthetic waveform with an artificially prescribed g-mode chirp and known mass-radius relation into detector noise, run the cWB+HHT pipeline, and check whether the summed IF1..3 follows the injected frequency at every post-bounce time; a systematic deviation at some frequencies would falsify the summing rule.","supporting_citations":[{"cited_title":"Bizouard, P","cited_arxiv_id":null,"evidence_quote":"Supplies the Hilbert-Huang transform (empirical mode decomposition plus Hilbert spectral analysis) that the paper adapts for frequency extraction."},{"cited_title":"Kaneyama, K","cited_arxiv_id":null,"evidence_quote":"Supplies the he3.5 waveform, one of the two simulated supernova signals tested."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the y20 waveform, the second simulated supernova signal tested."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines complementary ensemble empirical mode decomposition, the specific variant of EMD used to obtain the IMFs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Identifies the mode-splitting problem in empirical mode decomposition that motivates summing IMFs 1-3."}],"review_version":1}