{"id":"f14ae076-70c3-4af5-8192-f58b2bda526c","arxiv_id":"2607.04909","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.5,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"Cumulative log10 Bayes factor of 1.38±0.79 favors the full (2,0) mode in GWTC-4.0; decisive evidence is projected to need ~166 events under optimistic assumptions.","lead":"Researchers stacked Bayesian evidence for the full (2,0) gravitational-wave mode (oscillatory ringdown plus displacement memory) across O4 binary-black-hole events and found only moderate cumulative support. The result maps how many more detections are needed before memory itself becomes detectable with current detectors.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"The N_events forecast rests on an untested assumption that the empirical log10 B distribution is free of systematic bias from missing multipoles and remains representative of future detections.","rationale":"The Reader correctly isolates the two soft spots that actually underwrite the headline number: (i) the empirical distribution of log10 B measured on the present sample is assumed to remain representative, and (ii) residual waveform systematics (especially the missing (3,±2) multipole) are assumed not to bias that distribution. Both assumptions are stated by the authors themselves and are not yet quantified. The concrete multipole-complete re-analysis proposed above would settle whether the concern lands without requiring a full re-run of the entire catalog. Because the paper already labels the forecast “optimistic” and calls for systematics studies, the appropriate verdict remains CONDITIONAL; the stress-test merely sharpens the same condition the Reader already identified.","tokens_in":25628,"tokens_out":597,"duration_ms":5124,"concrete_test":"Re-analyze the nine events with log10 B − 1σ > 10^{-3} (Table III) using IMRPhenomXHM (or SEOBNRv5PHM) both with and without the (2,0) mode under identical sampler settings; recompute the stacked log10 B and re-fit the Student-t. If the new median N_events shifts outside the original 68 % interval [111,248], the forecast is systematics-dominated.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper’s strongest quantitative claim is the optimistic forecast N_events = 166^{+82}_{-55} for decisive evidence (log10 B − 1σ ≥ 2). That number is obtained by fitting a Student-t (2.85 dof) to the 84 GWTC-4.0 log10 B values and a skew-normal to their uncertainties, then drawing 500 Monte-Carlo trajectories (Sec. V D). The same section and the abstract explicitly flag residual waveform systematics—most clearly the missing (3,±2) multipole that already shifts the single loudest event GW230814_230901 from log10 B = 0.63 ± 0.09 (vs THM) to 0.18 ± 0.10 (vs XHM). Because the stacking formula (Eqs. 4.3–4.4) treats every event as independent and identically distributed, any coherent bias of even ~0.1–0.2 in the high-SNR tail would move the median N_events by tens of events and could push the lower edge of the 68 % interval above the size of the current O4 catalog. The independence and representativeness assumptions are therefore load-bearing for the forecast, yet they are not stress-tested against a multipole-complete model.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript reanalyzes the 84 BBH events of GWTC-4.0 (plus six selected high-SNR GWTC-5.0 events) with the quasi-circular, non-precessing IMRPhenomTHM_20 model that includes both the oscillatory ringdown and displacement-memory pieces of the (2,0) mode. Bayesian model comparison against IMRPhenomTHM yields a stacked log10 B = 1.38 ± 0.79 for the GWTC-4.0 sample; the corresponding stacked memory SNR after ~7.5 months of O4a is 0.89^{+0.29}_{-0.11}. A single precessing candidate is also examined with IMRPhenomTPHM_20 and shows no additional support. From the empirical distribution of the per-event Bayes factors the authors forecast that decisive evidence (log10 B - 1σ ≥ 2) would require N_events = 166^{+82}_{-55} under the assumptions of the present analysis.","tokens_in":25922,"tokens_out":1222,"duration_ms":9284,"significance":"The work supplies the first systematic, catalog-level Bayesian assessment of the full (2,0) multipole (oscillatory + memory) rather than memory alone, using public O4 data, measured PSDs, calibration envelopes, and a carefully documented nested-sampling pipeline. The zero-noise NR and model injections that recover the expected high-SNR scaling of the Bayes factor, the explicit quadrature propagation of evidence uncertainties, and the transparent stacking of both Bayes factors and SNRs are concrete strengths. The cumulative numbers and the N_events forecast give the community a concrete, if optimistic, benchmark for when the (2,0) mode (and ultimately memory) may become detectable with second-generation detectors.","major_comments":[{"comment":"Sec. V D and the abstract: the optimistic forecast N_events = 166^{+82}_{-55} is obtained by fitting a Student-t (2.85 dof) and a skew-normal to the 84 GWTC-4.0 log10 B values and drawing 500 Monte-Carlo trajectories. The same section notes that the loudest event, GW230814_230901, drops from log10 B = 0.63 ± 0.09 (vs THM) to 0.18 ± 0.10 when compared with the multipole-complete IMRPhenomXHM. Because the stacking formulae (Eqs. 4.3–4.4) treat every event as i.i.d., a coherent bias of even ~0.1–0.2 in the high-SNR tail would shift the median N_events by tens of events. The independence and representativeness assumptions are therefore load-bearing for the central forecast, yet they are not stress-tested against a multipole-complete model for more than one event. A short additional comparison (or an explicit statement that the forecast is conditional on the present mode content) is needed be","section":null},{"comment":"Sec. V A / Fig. 1 and Tab. III: the cumulative log10 B = 1.38 ± 0.79 is dominated by a handful of high-SNR events, most notably GW230814_230901. The paper already shows that this event’s Bayes factor is sensitive to the missing (3,±2) multipole. Without a quantitative assessment of how residual waveform systematics propagate into the stacked evidence, the claim that the catalog provides “moderate support” for the (2,0) mode remains provisional. At minimum the authors should recompute the stack after replacing or down-weighting the systematics-sensitive events, or clearly label the result as model-dependent.","section":null}],"minor_comments":[{"comment":"Fig. 1: the dual color-coding (fill = full SNR, edge = (2,0) SNR) is hard to read once printed in grayscale; a second panel or a different visual encoding would help.","section":null},{"comment":"Eq. (4.7)–(4.8): the high-SNR scaling is validated only with zero-noise injections; a brief remark that real noise can flip the sign of O(1) Bayes factors (already noted in Sec. IV E) would strengthen the interpretation of the small measured values.","section":null},{"comment":"Appendix A: the relative-runtime plot is useful but the absolute wall-clock times (color bar) are given only for ΔlnZ = 0.10; quoting the corresponding numbers for ΔlnZ = 0.05 would make the cost of the tighter threshold clearer.","section":null},{"comment":"Throughout: a few typographical inconsistencies remain (“W A VEFORM”, occasional missing spaces around math mode). A final copy-edit pass would remove them.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The central scientific result (stacked log10 B and the associated SNR) is solid and publishable; the N_events forecast is the only load-bearing claim that currently rests on an untested representativeness assumption. Once the authors either recompute a few high-SNR events with a multipole-complete model or explicitly condition the forecast on the present waveform family, the paper will meet the standard of a solid observational methods contribution. I see no reason to reject or to demand a full re-analysis of the entire catalog."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The new piece is the first stacked Bayesian comparison that keeps the full (2,0) mode (oscillatory ringdown + displacement memory) rather than memory alone, run on the public GWTC-4.0 BBH sample with a calibrated time-domain model. They get log10 B = 1.38 ± 0.79 and a memory-only stacked SNR of 0.89 after ~7.5 months of O4a; both numbers sit comfortably with earlier forecasts that used only the memory piece.\n\nWhat they do well is straightforward and careful. Nested sampling with a tighter ΔlnZ, public PSDs and calibration, proper Fourier handling of the non-periodic memory offset, zero-noise NR and model injections that recover the expected SNR scaling of the Bayes factor, and explicit uncertainty propagation. The per-event posteriors barely move when the mode is added, which is the right sanity check at current SNRs. They also flag the loudest event (GW230814) against a multipole-complete model and correctly call the N_events = 166^{+82}_{-55} forecast optimistic.\n\nThe soft spot is exactly the one the stress-test flags, and the authors already own it: the bootstrap assumes the empirical log10 B distribution (Student-t with 2.85 dof) remains representative and free of coherent bias from missing multipoles. A ~0.1–0.2 shift in the high-SNR tail would move the median by tens of events. That is a real limitation on the forecast, not on the measured stack itself. The independence assumption for stacking is standard and not load-bearing for the central result.\n\nThis is for people who already work on higher-mode PE or memory forecasts; it gives a concrete intermediate target and a clean methodology. The math, data handling, and citation pattern look solid. I would send it to peer review without hesitation; the systematics caveat is already written into the abstract and conclusions, so a referee can simply ask for a clearer statement of how large a bias would break the forecast. Worth reading and worth citing for the stacked numbers.","headline":"Solid cumulative search for the full (2,0) mode on O4 data; the stacked log10 B = 1.38 ± 0.79 and memory SNR ~0.9 are clean, while the N_events ~166 forecast is explicitly optimistic and systematics-limited.","tokens_in":26571,"tokens_out":552,"would_cite":true,"duration_ms":5369,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Stacking current binary-black-hole events gives only moderate evidence for the full (2,0) gravitational-wave mode; decisive detection is projected after roughly 166 events.","keywords":["gravitational-wave memory","(2,0) mode","Bayesian model selection","IMRPhenomTHM_20","GWTC-4.0","SNR stacking","binary black holes"],"falsifier":"Accumulate a catalog of ~170 binary-black-hole events at O4/O5 sensitivity, recompute the stacked log-10 Bayes factor with the same two models (and with models that include the (3,±2) multipole), and check whether log-10 B − 1σ reaches or exceeds 2.","tokens_in":26540,"feed_emoji":"🌊","tokens_out":1012,"duration_ms":7356,"temperature":0.7,"pith_summary":"The paper tests whether the subdominant (ℓ=2,m=0) spherical-harmonic mode—containing both its oscillatory ringdown piece and the permanent displacement memory—is already present in the binary-black-hole signals of GWTC-4.0 and a few loud GWTC-5.0 events. By comparing two otherwise identical waveform models, one with and one without that mode, the authors obtain a cumulative log-10 Bayes factor of 1.38 ± 0.79 for the full O4a catalog—moderate support that falls short of decisive. Stacking the mode’s signal-to-noise ratio yields a memory-only value near 0.9 after 7.5 months of O4a data, consistent with earlier forecasts. The authors argue that including the oscillatory contribution first strengthens the evidence relative to memory-only searches and therefore serves as a practical intermediate step toward a memory detection. Under the observed distribution of per-event Bayes factors they estimate that decisive evidence (log-10 B − 1σ ≥ 2) would require roughly 166 events, while cautioning that waveform systematics remain uncontrolled.","feed_headline":"Full (2,0) mode needs ~166 events for decisive evidence","feed_subtitle":"Stacking O4a black-hole mergers yields only moderate support; memory alone is still weaker","key_machinery":"Paired Bayesian model comparison of IMRPhenomTHM_20 (full (2,0) mode = oscillatory ringdown + displacement memory) versus IMRPhenomTHM (identical model without the mode), with per-event log-10 Bayes factors stacked under independence and validated by SNR stacking and high-SNR scaling relations.","core_discovery":"For the 84 binary-black-hole events of GWTC-4.0 the stacked log-10 Bayes factor between the full-(2,0) model IMRPhenomTHM_20 and the model without that mode is 1.38 ± 0.79; under the fitted empirical distributions of the individual Bayes factors, decisive evidence is expected after N_events = 166^{+82}_{-55} detections at current sensitivity.","pith_inferences":["Once the (2,0) mode is established, the same stacking pipeline can be re-run with the oscillatory piece subtracted to isolate the pure memory contribution and quantify how much extra catalog size it requires.","The mild preference already seen for the loudest events (especially GW230814_230901) suggests that a handful of future high-SNR, edge-on systems could accelerate the climb to decisive evidence faster than the average-event forecast.","Any future claim of memory detection that does not first demonstrate control of the full (2,0) mode will be vulnerable to the same systematics the authors flag for the missing higher multipoles."],"forward_implications":["Decisive statistical evidence for the full (2,0) mode is expected before a clean detection of the memory component alone.","Controlling the presence of the full (2,0) mode supplies a practical intermediate step that also helps isolate systematics before memory-only searches.","Current individual events remain statistically consistent with either model; the (2,0) contribution is still too weak to shift recovered source parameters appreciably.","An optimistic total of ~166 events at present sensitivity would suffice for decisive evidence under the measured Bayes-factor distribution."],"fun_headline_variants":["Stacked (2,0) support log10 B=1.38 needs ~166 events for decisive evidence","Full (2,0) mode stacks to moderate Bayes factor; decisive near 166 events","GWTC-4.0 yields log10 B=1.38 for (2,0); ~166 events projected for decisive","Memory SNR stacks to 0.89; full (2,0) needs ~166 events for decisive support","Catalog stack finds modest (2,0) evidence; decisive after roughly 166 detections"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"That the distribution of Bayes factors measured on the present GWTC-4.0 sample will continue to describe future detections, and that residual waveform modelling errors (especially the missing (3,±2) multipole) do not systematically bias those Bayes factors.","fun_headline_variants_meta":{"raw":{"variants":["Stacked (2,0) support log10 B=1.38 needs ~166 events for decisive evidence","Full (2,0) mode stacks to moderate Bayes factor; decisive near 166 events","GWTC-4.0 yields log10 B=1.38 for (2,0); ~166 events projected for decisive","Memory SNR stacks to 0.89; full (2,0) needs ~166 events for decisive support","Catalog stack finds modest (2,0) evidence; decisive after roughly 166 detections"]},"model":"grok-4.5","effort":"low","cost_usd":0.003692,"raw_usage":{"total_tokens":1306,"prompt_tokens":988,"num_sources_used":0,"completion_tokens":132,"cost_in_usd_ticks":36920000,"prompt_tokens_details":{"text_tokens":988,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":186,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":988,"tokens_out":132,"duration_ms":2536,"temperature":1.0,"reasoning_tokens":186,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-11T11:35:38.923213+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Accumulate a catalog of ~170 binary-black-hole events at O4/O5 sensitivity, recompute the stacked log-10 Bayes factor with the same two models (and with models that include the (3,±2) multipole), and check whether log-10 B − 1σ reaches or exceeds 2.","supporting_citations":[],"review_version":1}