{"id":"064a3bad-e7de-4ebb-a33c-a939750ed1ca","arxiv_id":"2501.02748","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A mock-data forecast shows that roughly forty LISA-detected SMBH merger events, combined with PTA background constraints, can recover galaxy and SMBH merger rates, while twelve events cannot.","lead":"This paper tests a framework that would use future LISA gravitational wave detections, combined with pulsar timing array data, to measure how often supermassive black holes and their host galaxies merge. The mock analysis shows that about forty detected events are enough to recover the merger rates, while fewer events create large biases at high redshift.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unity detection fraction in §4.1 is the load-bearing simplification: if LISA selection removes low-mass or high-redshift events, the inferred merger rates and the 'around forty events' threshold become optimistic.","rationale":"I read the paper in good faith and find the framework coherent: the hierarchical-Bayes machinery, the mock generation, and the internal recovery tests are appropriate for a forecasting study, and the qualitative claims that more LISA events improve constraints and that PTA data tighten parameters are well supported. The single most load-bearing simplification is the unity detection fraction in §4.1, because it enters directly into the Poisson likelihood and controls whether the observed joint mass–redshift distribution matches the true merger distribution. The reader's weakest-assumption analysis identified exactly this point, and I agree with it. I also considered two alternatives: the 'around forty events' threshold rests on only two realizations, and the constant-delay model is restrictive. Both are real limitations, but the selection-function issue is more fundamental because it affects the validity of the likelihood itself, not just the scatter of a particular mock realization. The proposed concrete test—computing SNRs and re-running the inference with a realistic ξ(Λ)—would settle whether the concern lands. Since the reader already assigned CONDITIONAL for this reason, no verdict adjustment is needed.","tokens_in":22565,"tokens_out":5620,"duration_ms":64673,"concrete_test":"Compute the LISA SNR for every event in the two mock catalogs of Fig. 2 and in the delay-time/occupation-fraction sets of §5.2–5.4 using a standard LISA sensitivity curve and the PhenomA waveform employed for the Fisher errors; apply a conservative detection threshold (e.g., SNR > 8), construct the effective selection function ξ(Λ), and re-run the hierarchical Bayesian inference with ξ(Λ) included in Eq. 13. If the injected merger-rate parameters remain inside the credible regions and the 36-vs-12 event conclusion survives, the concern is resolved; if not, the reported constraining power is overestimated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central forecast rests on the assumption in §4.1 that the LISA detection fraction ξ(Λ) is unity because 'the SMBH binary merger events in the considered mass and redshift ranges typically have large signal to noise ratios.' However, the mock catalogs in Fig. 2 extend down to M• ≈ 10^5 M⊙ and out to z ≈ 10; for the least massive, most distant events the SNR is not guaranteed to exceed a detection threshold. If a non-negligible fraction of events fall below threshold, the likelihood in Eq. 13 should include ξ(Λ) both in N_exp(Λ) and in the per-event normalization, otherwise the observed event count and the observed mass–redshift distribution are treated as identical to the true merger distribution. Selection would preferentially remove low-mass, high-redshift events—precisely the regime where the paper argues that limited event counts produce biases. Ignoring this effect therefore biases the recovered merger rates and makes the headline claim that 'around forty events' suffice optimistic. The paper provides no SNR distribution for the mock events and no recovery test under a realistic selection function.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a hierarchical Bayesian framework for inferring galaxy and SMBH merger rates from future LISA detections combined with current PTA constraints on the stochastic gravitational-wave background. The SMBH merger rate is constructed from the galaxy merger rate per galaxy, the galaxy stellar mass function, a redshift-dependent M_BH–M_* relation, and a delay time between galaxy and SMBH mergers. The authors generate mock LISA event catalogs with 36 and 12 events, and recover the hyperparameters of two galaxy-merger-rate parameterizations, with and without PTA data. They report that the event count and the joint mass–redshift distribution drive constraining power, that a 36-event catalog recovers merger rates consistent with galaxy-pair observations while a 12-event catalog produces high-redshift biases, that PTA data sharpen the constraints, that a constant delay time can be recovered, and that the SMBH occupation fraction at z>3 is poorly constrained.","tokens_in":22814,"tokens_out":7845,"duration_ms":80290,"significance":"If the result holds, the paper provides a useful forecast for LISA planning: roughly forty well-characterized SMBH merger events, combined with PTA data, may be sufficient to cross-check galaxy merger rates out to high redshift. The framework is standard and the injection–recovery tests are internally consistent; the authors explicitly report the imperfect recovery in the 12-event realization and in the tau=0.8 Gyr case, which is a sign of honesty. The paper also clearly labels its analysis as a mock-data self-test, so I see no circularity problem. The main significance is quantitative guidance on event counts and on the combination of LISA and PTA data, not a new astrophysical measurement.","major_comments":[{"comment":"The detection fraction xi(Lambda) is set to unity in Eq. (13) on the grounds that SMBH mergers in the considered mass and redshift ranges have large signal-to-noise ratios. However, the mock catalogs in Fig. 2 extend down to M_bullet about 1e5 M_sun and out to z about 10, and the paper provides no SNR distribution or detection-threshold calculation for these events. If a realistic LISA selection function removes a non-negligible fraction of low-mass or high-redshift events, the likelihood should use N_exp(Lambda) in both the Poisson count factor and the per-event normalization, and the effective event count would be smaller than the injected value. Because the paper's central claim is the contrast between 36 and 12 detected events, an unquantified selection function is load-bearing; the authors should add a recovery test with a fiducial LISA sensitivity and an explicit SNR threshold (for example SNR>8) and show how the posterior widths and the high-redshift biases change.","section":"Section 4.1, Eq. (13) and Fig. 2"},{"comment":"The headline statement that datasets with around forty events yield results consistent with galaxy-pair observations is based on a single 36-event realization and a single 12-event realization. Two realizations do not establish a threshold; the Poisson scatter in the 36-event result is not quantified, and the claimed contrast could be driven by the particular random draws. The repeated tests mentioned in Section 5.1.4 vary merger-rate parameters but not the event count. The authors should include several realizations at intermediate event counts (for example 20, 30, and 40) and report the distribution of recovery biases and credible-interval coverage.","section":"Section 5.1.3 and Abstract"},{"comment":"The delay-time inference assumes a delta-function delay distribution, and the recovery shown in Fig. 7 is a test of recovering a point delay under that same assumption. The abstract's statement that the method 'effectively constrains the delay time' is therefore stronger than what is demonstrated: the method has not been tested on, for example, a power-law or Gaussian delay distribution, which the authors themselves note in Section 5.2. The abstract and conclusion should be reworded to say that a constant delay time is assumed in the analysis.","section":"Section 5.2"}],"minor_comments":[{"comment":"The sentence 'According to our mock analysis, the models with delay times longer than 0.5Gyr (0.8Gyr), accretion becomes the primary driver of SMBH mass growth beyond z~6 (4)' is grammatically incomplete; it should read 'for models with delay times longer than 0.5 Gyr (0.8 Gyr), accretion becomes...'.","section":"Abstract"},{"comment":"The expression 'a= log10 kappa + 9 - 11 * b' uses a nonstandard asterisk and should be written as a = log10 kappa + 9 - 11 b. Also, the numerical values in Table A2 for the z<4 relation (for example a=-5.36, b=1.28 for GW eg1) differ from the central values quoted in the text (b=1.17 +/- 0.08); please clarify that the table entries are random draws from the observational priors rather than the central values.","section":"Section 5.1.2"},{"comment":"The default prior P_emptyset(theta) in Eq. (14) is described only as 'usually set to a uniform distribution'; please specify the explicit ranges used for the Monte Carlo averages over total mass and redshift so that the analysis is reproducible.","section":"Section 4.1, Eq. (14)"},{"comment":"In Section 5.3, the text says 'accretion generally dominants the mass assembly'; this should be 'accretion generally dominates the mass assembly'.","section":"Section 5.3"}],"recommendation":"major_revision","confidential_remarks":"This is a forecast and injection-recovery study rather than a data-driven measurement, which is appropriate for its stated scope. The two load-bearing issues are the unity detection fraction and the reliance of the 'around forty events' claim on only two realizations. Both can be addressed within the manuscript's scope by adding a selection-function test and additional intermediate-count realizations. I do not see a circularity problem because the mock nature of the data is explicit throughout."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid, carefully executed mock-data forecasting paper. It builds a hierarchical Bayesian framework that combines mock LISA detections with PTA SGWB constraints and simultaneously marginalizes over uncertainties in the stellar mass function and M_bulge-M_star relation. The main, credible takeaway is that the number of LISA events and their joint mass-redshift distribution drive the constraining power—36 events in their mock recover galaxy pair-observed merger rates, 12 events do not, and PTA data sharpen the posteriors. That claim holds as far as the two realizations go.\n\nWhat's actually new: the combination isn't in the cited literature, and the delay-time recovery tests (constant delay 0.2–1 Gyr) are a useful addition. The injection-recovery math is standard and self-consistent; the authors honestly report when injected values land outside the 1σ region (12-event case, tau=0.8 Gyr). The negative result on the occupation fraction f3 at z>3—flat posterior driven by degeneracy with delay time and galaxy merger rate—is a genuinely useful finding. The mock data are generated from a model and then inferred back; that's injection-recovery, not a derivation, and the paper labels it as such, so circularity isn't a concern.\n\nThe load-bearing simplification is section 4.1's assumption that the LISA detection fraction ξ is unity because SMBH mergers in the considered mass-redshift range 'typically have large SNR.' The mock catalogs extend down to M≈1e5 M_sun and out to z≈10; at the low-mass, high-z end, SNR is not guaranteed to clear threshold. A realistic selection function would preferentially remove those events, which is exactly the regime where the paper argues limited event counts bias the high-z recovery. So the 'around forty events' threshold should be read as an optimistic lower bound; the direction of the bias is known, and the qualitative conclusion survives, but the paper would be stronger with a selection-function robustness test. Secondary caveats: the 'around forty' threshold is inferred from one 36-event versus one 12-event realization, not a systematic event-count scan, and the delay-time analysis is restricted to a constant delay, which the authors acknowledge. No code is released, only data upon request; for a methods paper, that's a reproducibility gap, though the appendix tables give enough detail to reproduce the main analysis.\n\nWho this is for: the LISA galaxy/SMBH population forecasting community, and anyone planning mock studies for mHz GW observatories. It deserves a serious referee—the framework is coherent, the internal tests support the main claims, and the limitations are stated rather than hidden. My recommendation: send it out, with the request that the authors either add a selection-function test or at minimum discuss how a non-unity ξ would shift the event-count threshold. As it stands, this is a conditional rather than unconditional accept for me, but a relatively mild conditional.","headline":"Solid LISA-era mock-forecast paper; the unity detection fraction and a thin event-count basis are the main caveats, but the core results are credible.","tokens_in":23317,"tokens_out":3455,"would_cite":true,"duration_ms":31425,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Future LISA detections can recover galaxy and supermassive-black-hole merger rates, but only when the catalog reaches roughly forty events; with only a dozen events, high-redshift rates come out biased.","keywords":["gravitational waves","supermassive black holes","galaxy merger rate","pulsar timing arrays","LISA","stochastic gravitational wave background","hierarchical Bayesian inference","supermassive black hole scaling relations"],"falsifier":"Once real LISA data exist, compare the galaxy merger rate recovered from a catalog of roughly forty events against independent galaxy-pair measurements at $z>3$; if the forty-event catalog disagrees with galaxy-pair rates while a smaller catalog agrees, the central claim is falsified. A simulation with a realistic signal-to-noise-dependent LISA selection function would settle the same point before launch.","tokens_in":22350,"feed_emoji":"🔭","tokens_out":10296,"duration_ms":98094,"temperature":0.7,"pith_summary":"This paper argues that gravitational-wave detections of supermassive-black-hole mergers can be turned into direct measurements of how often galaxies and their central black holes merge, complementing surveys of galaxy pairs. The authors build a mock LISA catalog, add a PTA stochastic-background constraint, and fit a parameterized galaxy merger rate through the $M_\\bullet$-$M_*$ relation; they find that the number of detected events and the joint distribution of binary mass and redshift decide how well the rates are recovered. A catalog of roughly forty events reproduces the galaxy merger rate from galaxy-pair observations, while a catalog of twelve events biases the high-redshift behavior. The same data constrain the delay between galaxy and black-hole mergers, and they show that for delays above about 0.5–0.8 Gyr, accretion rather than mergers becomes the dominant channel of black-hole mass growth at high redshift.","feed_headline":"~40 LISA events recover galaxy and black-hole merger rates","feed_subtitle":"Mock LISA catalogs show ~40 detections recover merger rates; a dozen leave high-redshift bias.","key_machinery":"The load-bearing machinery is a convolution chain from galaxy mergers to black-hole mergers: the galaxy stellar mass function $\\Phi_{\\mathrm{GSMF}}(M_*, z)$ times a parameterized merger rate per galaxy $R_{\\mathrm{Gal}}(M_*, z)$ gives the galaxy merger rate, and a log-normal $M_\\bullet$-$M_*$ relation with redshift-dependent parameters maps it to a SMBH mass function. A delay-time kernel $P_{\\mathrm{delay}}(\\tau)$ shifts the SMBH merger events to later lookback times, and an occupation fraction $f_{\\mathrm{occ}}$ scales the number of host galaxies that actually contain a black hole. These rates enter an inhomogeneous Poisson hierarchical-Bayesian likelihood that compares predicted counts and the joint mass–redshift distribution to mock LISA events, while the same population feeds a power-law SGWB integral constrained by PTA strain. The paper sets LISA's detection fraction to unity, justified by the high signal-to-noise ratios of SMBH mergers in the considered range.","core_discovery":"On the paper's own terms, the central discovery is that the constraining power of future LISA observations resides in the number of detected SMBH mergers and in their joint distribution in binary mass and redshift. Using two mock realizations of 36 and 12 events, the hierarchical Bayesian recovery shows that the 36-event catalog reproduces the galaxy merger rate inferred from galaxy-pair observations, while the 12-event catalog misses the high-redshift behavior and biases parameters such as $n_0$, $\\alpha_1$, and $\\beta$ outside their $1\\sigma$ credible regions. Adding a PTA stochastic-background amplitude tightens the posteriors, especially for the normalization $n_0$ and the delay time. For a fixed constant delay, the reconstructed SMBH mass assembly from mergers is suppressed at high redshift as $\\tau$ grows, with accretion becoming the dominant growth channel beyond $z\\sim 6$ for $\\tau\\gtrsim 0.5$ Gyr and beyond $z\\sim 4$ for $\\tau\\gtrsim 0.8$ Gyr. The occupation fraction of SMBHs at $z>3$ is not recoverable, because it is degenerate with the delay time and the merger-rate parameters.","pith_inferences":["A straightforward extension would rerun the recovery with a mass- and redshift-dependent LISA selection function; the paper's unity-detection assumption means the effective event count, not the raw detection count, is what should be compared with the roughly forty-event threshold.","If the local $M_\\bullet$-$M_*$ relation holds at all redshifts rather than the near-infrared-motivated relation used here, the recovered high-redshift merger rates shift systematically upward, so independent constraints on that relation at $z>4$ will directly set the normalization of GW-inferred merger rates.","Relaxing the constant-delay assumption to a distribution of delay times, which the paper notes is computationally expensive, may widen the delay posteriors and make the occupation-fraction degeneracy even harder to break; testing this before LISA flies would clarify how much of the reported $f_3$ degeneracy is due to the constant-delay simplification.","The same hierarchical Poisson framework could be applied to a future catalog of individually resolved PTA binaries, giving a lower-redshift cross-check of the LISA-based rates."],"forward_implications":["A real LISA catalog of roughly forty SMBH mergers should reproduce the galaxy merger rate inferred from galaxy-pair observations, including the high-redshift end.","A catalog of twelve or fewer events will not resolve the high-redshift merger rate; using it as a definitive measurement would inject systematic bias.","PTA measurements of the stochastic gravitational-wave background will tighten the parameters inferred from LISA alone, even though PTAs sample a different mass and redshift window.","If the true delay between galaxy and black-hole merger is longer than roughly 0.5–0.8 Gyr, merger-driven mass growth of SMBHs falls below accretion beyond $z\\sim 6$ to $z\\sim 4$.","The SMBH occupation fraction at $z>3$ cannot be pinned down by this data combination on its own; it stays degenerate with delay time and the galaxy merger rate."],"supporting_citations":[{"why":"Supplies the measured PTA stochastic-background amplitude used as the external constraint in the joint LISA plus PTA fits.","marker":"Agazie et al. 2023"},{"why":"Sets the LISA detection assumption: SMBH mergers in the considered range have high signal-to-noise ratios, which the paper cites to set the detection fraction to unity.","marker":"Amaro-Seoane et al. 2017b"},{"why":"Provides the local $M_\\bullet$-$M_*$ relation used to assign SMBH masses to host galaxies at $z<4$.","marker":"Kormendy & Ho 2013"},{"why":"Provides the $z>4$ $M_\\bullet$-$M_*$ relation derived from near-infrared observations, replacing the local relation at high redshift.","marker":"Pacucci & Loeb 2024"},{"why":"Gives the galaxy pair fraction and merger rates out to $z=11.5$ against which the recovered gravitational-wave-based rates are compared.","marker":"Duan et al. 2024"},{"why":"Supplies cosmological simulation merger-rate predictions that the Case 2 parameterization approximates and against which recovered rates are compared.","marker":"Rodriguez-Gomez et al. 2015"},{"why":"Provides the PhenomA waveform used with a Fisher matrix to assign mock LISA event mass and redshift uncertainties.","marker":"Ajith et al. 2007"},{"why":"Provides the accretion-driven SMBH mass assembly curve used to determine when mergers or accretion dominate.","marker":"Pacucci & Loeb 2020"},{"why":"Supplies the occupation-fraction parameterization used to model $f_3$ at $z\\geq 3$.","marker":"Beckmann et al. 2023"}],"fun_headline_variants":["~40 LISA detections recover galaxy and black-hole merger rates","Only 12 LISA events miss high-redshift black-hole mergers","Long delay times shift black-hole growth to accretion by z~6","PTA data sharpen LISA's merger-rate constraints","Why 40 LISA events matter for cosmic merger history"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The forecast assumes LISA detects every SMBH merger in the mass and redshift range considered, so if real selection effects remove a significant fraction of low-mass or high-redshift events, the claimed constraining power of roughly forty detections would be optimistic.","fun_headline_variants_meta":{"raw":{"variants":["~40 LISA detections recover galaxy and black-hole merger rates","Only 12 LISA events miss high-redshift black-hole mergers","Long delay times shift black-hole growth to accretion by z~6","PTA data sharpen LISA's merger-rate constraints","Why 40 LISA events matter for cosmic merger history"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000326,"raw_usage":{"total_tokens":1910,"prompt_tokens":1118,"completion_tokens":792,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":734,"completion_tokens_details":{"reasoning_tokens":706}},"tokens_in":734,"tokens_out":792,"duration_ms":8469,"temperature":1.0,"reasoning_tokens":706,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:06:07.684452+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Once real LISA data exist, compare the galaxy merger rate recovered from a catalog of roughly forty events against independent galaxy-pair measurements at $z>3$; if the forty-event catalog disagrees with galaxy-pair rates while a smaller catalog agrees, the central claim is falsified. A simulation with a realistic signal-to-noise-dependent LISA selection function would settle the same point before launch.","supporting_citations":[],"review_version":1}