{"id":"f60349ed-5c59-4ca3-9933-204d68a38c94","arxiv_id":"2501.09057","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"With a correctly modeled correlated noise, Einstein Telescope's triangular configuration can still measure the stochastic gravitational wave background; neglecting that noise biases the inferred signal parameters.","lead":"This paper simulates how correlated seismic 'Newtonian' noise affects the search for a stochastic gravitational wave background in the Einstein Telescope's triangular design. It shows that if this noise is modeled correctly, the background can still be measured, while ignoring the noise would bias the results.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central recovery claim is established only under a matched-model simulation: the same power-law correlated-noise CSD is used to generate and analyze the data.","rationale":"The reader identified the same load-bearing assumption as the weakest point, and I agree. The paper is a careful, internally consistent simulation study: the likelihood derivation is explicit, the AET diagonalization is standard, the PP plot provides calibration evidence under the assumed model, and the bias demonstration when correlated noise is ignored is compelling. Independent support includes the agreement between the analytic likelihood in Appendix C and the Gaussian estimator likelihood used in the main text. The concern is not that the mathematics is wrong, but that the central claim is conditioned on an untested model. Since both the signal and the correlated noise are modeled as power laws, the clean separation demonstrated in Figures 1 and 2 may be an artifact of the very steep -8 tilt assigned to the CSD. A realistic Newtonian-noise spectrum with a different frequency dependence could leak into the SGWB reconstruction, and the paper's own Figure 3 shows how sensitive the SGWB parameters are to unmodeled correlated power. The fixed-PSD choice is a secondary but real idealization: quoting percent-level accuracy while assuming the auto-spectrum is perfectly known is optimistic, not conservative, for a real analysis. These points do not refute the conditional claim, but they do mean the headline practical conclusion about ET is not yet supported. The appropriate verdict remains CONDITIONAL, so no adjustment to the reader's verdict is needed.","tokens_in":21455,"tokens_out":10741,"duration_ms":124072,"concrete_test":"Run the same pipeline with simulated data whose correlated Newtonian noise is not the single power law of Eq. (11), for example a power law with tilt nnoise=-6 (still inside the adopted prior) or a broken power law loosely following the realistic body-wave spectra of Refs. [7,12], while keeping the same injected SGWB parameters. Carry out the four-parameter Bayesian analysis with the original power-law CSD model and check whether the injected AGW and nGW fall inside the 90% credible intervals. If they do, the matched-model caveat is not fatal; if they do not, the central claim must be weakened to hold only when the CSD model is essentially exact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's positive result rests on a closed simulation loop: data are generated with the power-law CSD of Eq. (11), and the same power-law model is used in the likelihood to estimate the correlated-noise parameters r and nnoise alongside the SGWB parameters. Section V explicitly states that this is a key assumption and that it may not hold for real data. This assumption is load-bearing because the demonstrated separation between the SGWB and the correlated noise relies on very different spectral tilts (2/3 versus -8) and on well-separated pivots (25 Hz versus 2.75 Hz). If a realistic Newtonian-noise CSD is shallower, has a spectral break, or contains additional structure, the four-parameter power-law fit can absorb that mismatch into AGW and nGW, reintroducing the sort of bias the paper itself demonstrates in Figure 3 when correlated noise is neglected. The paper's own conclusions call for more sophisticated noise models, but no mismatch test is performed, so the practical claim that the triangular configuration remains competitive for ET is not yet established. A secondary idealization is that the auto-PSD is fixed to the reference value in Section IV A; the statement that this is conservative is not convincing, since fixing a parameter removes a source of uncertainty rather than inflating it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the impact of correlated Newtonian noise on stochastic gravitational-wave background (SGWB) parameter estimation for the Einstein Telescope (ET) in its triangular configuration, and compares this with a two-L-shaped-detector (2L) configuration. The authors derive a Gaussian likelihood for the SGWB estimator in the presence of correlated noise, validate it with a Wishart-likelihood equivalence in Appendix C, and perform Bayesian parameter estimation on simulated data using Bilby and Dynesty. Their main results are (i) when the correlated-noise CSD is modeled as a power law matching the injection, the SGWB parameters (amplitude and tilt) and the correlated-noise parameters (amplitude ratio r and tilt n_noise) are reconstructed with high precision from one day of observation; (ii) neglecting correlated noise in the likelihood produces strongly biased SGWB parameters; and (iii) the 2L configuration yields somewhat tighter constraints on the SGWB parameters than the triangular configuration. The statistical calibration is checked with a 100-realization PP plot.","tokens_in":21596,"tokens_out":7470,"duration_ms":87139,"significance":"If the results hold, the paper provides a useful validation that a simple joint estimation of SGWB and correlated-noise power-law parameters can, in principle, protect the SGWB measurement from correlated Newtonian noise in the triangular ET configuration. The bias demonstration in Figure 3 is a clear and potentially important warning for future ET analyses. The analytic likelihood derivation and the explicit PP-plot calibration are careful and constitute genuine strengths. The main limitation is that the central recovery claim is established only in a matched-model simulation: the data are generated with the same power-law CSD used in the analysis, and the auto-PSD is fixed to a reference value. The paper itself acknowledges the first caveat, but the abstract and conclusions still state the practical claim that the triangular configuration remains competitive and that percent-level accuracy is achievable without the caveat being fully reflected in the headline statements.","major_comments":[{"comment":"The central recovery claim rests on a matched-model simulation: the simulated data are generated with the power-law correlated-noise CSD of Eq. (11), and the same power-law model is used in the likelihood to estimate r and n_noise. The paper explicitly notes in Section V that this is a key assumption that may not hold for real data. This is load-bearing because the demonstrated separation between the SGWB and the correlated noise relies on the large spectral-tilt difference (n_GW = 2/3 versus n_noise = -8) and on the well-separated pivot frequencies (25 Hz versus 2.75 Hz). If the real Newtonian-noise CSD is shallower, has a spectral break, or contains additional structure, the four-parameter power-law fit can absorb the mismatch into A_GW and n_GW, reintroducing the sort of bias the paper itself shows in Figure 3 when correlated noise is neglected. No mismatch or robustness test is performed. I therefore recommend that the authors either add injection-recovery tests with misspecified noise CSDs (e.g., broken power laws, different tilts, or a smooth non-power-law term) or explicitly restrict the quantitative claims in the abstract and conclusions to the assumed power-law model.","section":"Section V, Eqs. (3) and (11)"},{"comment":"The text states that fixing the auto-PSD to the reference value 'should give conservative estimates.' This is not correct: fixing a parameter removes a source of uncertainty and cannot make the posterior widths conservative; it can only make them narrower than they would be if the PSD were estimated jointly. The reported 'percent-level accuracy' and the relative widths of the triangular versus 2L posteriors in Figure 4 depend on this choice. The authors should either relax this assumption (for example, by jointly estimating PSD amplitudes or by using the T-channel information they exclude) or remove the word 'conservative' and explicitly state that all precision statements assume a perfectly known auto-PSD.","section":"Section IV A, fixed PSD assumption"}],"minor_comments":[{"comment":"The expression 'e^{-ψI(f)}' should read 'e^{-iψI(f)}' to be dimensionally and notationally consistent with the phase factor in Eq. (B1).","section":"Appendix B, equation after (B3)"},{"comment":"The phrase 'percent-level accuracy' should be qualified as 'for the injected power-law correlated-noise model and with the auto-PSD fixed'; otherwise readers may overinterpret the claim as applying to realistic ET noise.","section":"Abstract and Conclusions"},{"comment":"The positive-definiteness constraint on No(f)/Nd(f) is mentioned only briefly; please state explicitly how the constraint is implemented in the sampler and verify that all injected values, including r = 0.8 at the lowest analyzed frequencies, satisfy the condition -1/2 ≤ No(f)/Nd(f) ≤ 1.","section":"Section V, Table I footnote and Figure 2"},{"comment":"The SNR of the correlated noise is quoted as 135 'according to the standard definition'; please give the explicit formula or a precise citation, since the SNR definition for a noise contribution interpreted as a signal is not standard.","section":"Section V, footnote 7"},{"comment":"Typo: 'arm-lenghts' should be 'arm lengths'.","section":"Introduction, paragraph 1"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is a careful, honest simulation study. The headline result—that the ET triangular configuration can still recover SGWB parameters in the presence of correlated Newtonian noise—holds under a matched-model assumption: data are generated with the same power-law CSD used in the likelihood. The authors say this plainly. That limitation is load-bearing for the practical claim, and the paper would be stronger with a mismatch test.\n\nWhat is genuinely new: it is the first SGWB-specific Bayesian study of correlated noise in the ET triangle that jointly estimates signal and noise parameters. The likelihood derivation is sound, the PP-plot calibration looks correct, and the demonstration that neglecting correlated noise biases the SGWB parameters (Figure 3) is clean and compelling. The appendix proof that the complex phase of the CSD is irrelevant for SGWB parameter recovery is a nice formal point, correctly limited to zero-mean signals.\n\nWhere it is soft: the central recovery claim is a self-consistency test. Real Newtonian noise from body waves is unlikely to be a single power law with tilt -8; if it has a break or additional structure, the four-parameter fit can absorb that into the SGWB amplitude and tilt, reintroducing bias. The authors acknowledge this but do not test it. A mismatch test with a more realistic spectrum is the natural next step and should be asked for in revision. Minor points: fixing the PSD to a reference value is called 'conservative' but it actually removes a source of uncertainty, so that label is misleading. And the 2L comparison uses 15 km arms versus 10 km, so the 1.5x narrower posteriors are mostly arm-length, not configuration; the authors do attribute the difference to arm length.\n\nWho this is for: ET design studies and SGWB data analysts. It deserves a serious referee; the framework is a solid foundation even if the strong claim is not yet established. My recommendation: send it to peer review, with a request for a robustness test against noise-model misspecification, or a revised framing that the result is a proof-of-principle under matched-model assumptions.","headline":"A careful and honest simulation study: joint recovery of the SGWB and correlated noise works under the matched-model assumption, but the practical claim still needs a mismatch test.","tokens_in":22218,"tokens_out":4180,"would_cite":true,"duration_ms":38347,"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":"The paper claims that correlated Newtonian noise in the triangular Einstein Telescope layout does not prevent accurate reconstruction of the stochastic gravitational-wave background, provided the noise's frequency dependence is modeled…","keywords":["stochastic gravitational wave background","Einstein Telescope","correlated noise","Newtonian noise","Bayesian parameter estimation","AET basis","gravitational wave detectors","power-law spectral model"],"falsifier":"Run the same Bayesian pipeline on simulated data whose correlated noise is generated from a more realistic spectrum, for example a body-wave Newtonian noise model with a spectral peak or turnover away from a pure $f^{-8}$ power law, while keeping the analysis template as a simple power law; if the recovered background amplitude or tilt shifts beyond the quoted percent level, the central claim would fail for realistic noise.","tokens_in":21169,"feed_emoji":"🔭","tokens_out":7679,"duration_ms":76189,"temperature":0.7,"pith_summary":"This paper asks whether the proposed triangular Einstein Telescope can measure the stochastic gravitational-wave background despite correlated Newtonian noise among its three nested interferometers. It argues that with a Bayesian analysis that models the correlated noise as a power law and estimates it alongside the background, one day of simulated data reconstructs both signal and noise parameters at percent-level accuracy. It also shows that ignoring the correlated noise biases the recovered background parameters, and that a two-L-shaped-detector layout remains slightly more precise, mainly because of longer arms. The paper concludes that the triangular configuration stays competitive for stochastic-background science when the correlated noise spectrum is correctly modeled.","feed_headline":"Correlated noise need not blind ET to the gravitational-wave background","feed_subtitle":"Fitting the noise's shape lets one day of ET data recover signal and noise with percent-level accuracy.","key_machinery":"The load-bearing object is the frequency-domain Gaussian likelihood for the time-averaged cross-power estimator $\\hat C_{IJ}(f)$, whose mean is $\\gamma_{IJ}(f)\\Omega_{\\rm GW}(f)+N_{IJ}(f)/S_0(f)$ and whose variance is set by the auto- and cross-power spectra divided by the number of segments. For the triangle this is evaluated in the AET basis, which diagonalizes the correlated-noise covariance because the three interferometers are assumed identical. The signal and noise are both modeled as power laws, with the noise correlation written as $N_o(f)=N_d(2.75\\,{\\rm Hz})\\,r\\,(f/2.75\\,{\\rm Hz})^{n_{\\rm noise}}$, and the four parameters are sampled jointly. A supporting result is the proof that a complex phase in the cross-spectral density can be absorbed into a redefinition of the zero-mean stochastic background, so only the real correlation amplitude matters for background reconstruction.","core_discovery":"In a triangular three-interferometer detector, seismic Newtonian noise correlates the channels at low frequency; the authors show that this does not spoil stochastic gravitational-wave background measurement if the correlation is included in the model. They derive a Gaussian likelihood for the cross-power estimator in the AET basis, add a power-law cross-spectral-density template for the correlated noise, and jointly estimate four parameters: background amplitude and tilt at 25 Hz, and noise correlation amplitude and tilt at 2.75 Hz. On one day of simulated data with injected background amplitude $A_{\\rm GW}=10^{-9}$ and tilt $n_{\\rm GW}=2/3$, all four parameters are reconstructed at percent-level accuracy for injected correlations between $-0.5$ and $0.8$; the noise parameters widen as $r\\to 0$, while the background parameters remain stable. If the correlated noise is omitted from the likelihood, the recovered background parameters are significantly biased. The two-L-shaped 15-km layout gives credible regions roughly 1.5 times narrower, which the authors attribute mainly to arm length and to having two fewer nuisance parameters.","pith_inferences":["The paper's own caveat implies the percent-level accuracy is not yet established for real data: a realistic Newtonian-noise spectrum with bumps, turnovers, or site-specific structure would test whether a single power-law template is enough to keep the background unbiased.","If the 2L advantage is mainly arm length, then a triangular design with longer arms or a different site could close or reverse the gap; nothing in the paper shows the L-shape geometry itself is intrinsically superior.","The same likelihood framework could be stress-tested on non-power-law backgrounds, such as cosmic strings or first-order phase transitions, where the signal and correlated noise may overlap in frequency differently than in the power-law case.","Because the phase of the correlated noise can be absorbed into the background definition for a zero-mean signal, real-data analyses may only need to model the modulus of the cross-spectral density when estimating the stochastic background, leaving phase modeling to resolved-source studies."],"forward_implications":["If ET is built in the triangular 10-km layout, stochastic-background searches do not have to treat correlated Newtonian noise as a showstopper; the noise can be estimated jointly with the signal.","Future ET pipelines that omit a correlated-noise term will misestimate the background amplitude and tilt, so a cross-spectral-density model must be included.","With one day of data, both astrophysical background and noise parameters can be constrained to percent level, so early science runs could already measure the background.","The triangular layout is competitive with two separated L-shaped detectors, which are only about 1.5 times tighter on background parameters, mostly because of longer arms.","When the injected correlation is weak, the noise parameters become harder to measure, but the background parameters stay stable across the full allowed range of correlation amplitudes."],"supporting_citations":[{"why":"Identifies correlated seismic and Newtonian noise in the triangular ET layout and supplies the motivation and template shape for the correlated noise.","marker":"[12]"},{"why":"Provides the ET design comparison, the PSD reference values, and the 2L configuration used as the baseline.","marker":"[11]"},{"why":"Supplies the SGWB estimator, its variance, and the unified detection formalism on which the likelihood is built.","marker":"[14]"},{"why":"Introduces the AET basis that diagonalizes the noise covariance for equal-arm triangular detectors.","marker":"[20]"},{"why":"Gives the expected correlated seismic and Newtonian noise spectrum with tilt -8, used for the injected noise model.","marker":"[7]"},{"why":"Derives the likelihood for a network of detectors with correlated noise, extended here from resolved sources to the stochastic background.","marker":"[13]"},{"why":"Provides the LVK stochastic-background upper limit and estimator conventions used to set the injected background amplitude and analysis range.","marker":"[90]"}],"fun_headline_variants":["For ET, modeling correlated noise keeps the gravitational-wave background clear","Ignore correlated noise and ET's gravitational-wave background goes biased","ET's triangular noise: model it or lose the stochastic background","Modeling correlated noise lets ET see the gravitational-wave background","Modeled correlated noise keeps ET triangular competitive for the background"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on simulated correlated noise being drawn from exactly the same power-law model that the analysis fits; real Newtonian noise with a different spectral shape could invalidate the percent-level accuracy claim.","fun_headline_variants_meta":{"raw":{"variants":["For ET, modeling correlated noise keeps the gravitational-wave background clear","Ignore correlated noise and ET's gravitational-wave background goes biased","ET's triangular noise: model it or lose the stochastic background","Modeling correlated noise lets ET see the gravitational-wave background","Modeled correlated noise keeps ET triangular competitive for the background"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001399,"raw_usage":{"total_tokens":5657,"prompt_tokens":946,"completion_tokens":4711,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":562,"completion_tokens_details":{"reasoning_tokens":4628}},"tokens_in":562,"tokens_out":4711,"duration_ms":31159,"temperature":1.0,"reasoning_tokens":4628,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:11:52.749251+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same Bayesian pipeline on simulated data whose correlated noise is generated from a more realistic spectrum, for example a body-wave Newtonian noise model with a spectral peak or turnover away from a pure $f^{-8}$ power law, while keeping the analysis template as a simple power law; if the recovered background amplitude or tilt shifts beyond the quoted percent level, the central claim would fail for realistic noise.","supporting_citations":[{"cited_title":"Upper limits on the isotropic gravitational-wave background from Advanced LIGO and Advanced Virgo’s third observing run,","cited_arxiv_id":null,"evidence_quote":"Provides the LVK stochastic-background upper limit and estimator conventions used to set the injected background amplitude and analysis range."}],"review_version":1}