{"id":"953e9323-801c-4452-b35c-af378fe7bb1f","arxiv_id":"2504.16712","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"A Bayesian injection-recovery study forecasts that Taiji can detect phase-transition gravitational wave backgrounds with peak energy density above about 1.4e-11 over most of its band.","lead":"This paper simulates Taiji space observatory data with noise and foregrounds to test whether a gravitational wave background from early-universe phase transitions could be detected. It reports that Taiji could detect such signals above an energy density near 1.4e-11 and measure their peak frequency to better than 10 percent for stronger signals.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stated priors in Table I exclude the low-amplitude and low-frequency grid points that set the claimed threshold; without documentation of per-run priors, the 1.4e-11 sensitivity claim is not supported.","rationale":"The reader's verdict was CONDITIONAL with moderate confidence. My independent pass finds a more specific internal inconsistency that reinforces 'conditional': the stated benchmark priors cannot cover the grid values that define the sensitivity threshold. This is not a disagreement with the external physics; Taiji's basic sensitivity to Ω~1e-11 is plausible. The issue is that the paper's numerical evidence for its central number is not reproducible as described. The table of priors is presented as 'the prior ranges' for the Bayesian analysis, and no separate grid-specific priors are given. The threshold of 1.4e-11 sits below the lower edge of the ΩPT prior. A good-faith reader could assume the authors recentered priors per injection to make recovery feasible, but that assumption must be stated because it changes the meaning of the Bayes factors: a prior centered on the true peak frequency and amplitude removes the Occam penalty that a blind search would pay. Therefore the abstract's 'robustly detect' is not yet supported. The proposed rerun with a single global prior is a direct check: if the threshold survives, the concern is resolved; if not, the headline must be revised or the grid analysis must be rerun. I agree with the reader's overall conditional recommendation, though my identified weakest point differs from the reader's foreground-model assumption; the two concerns are related manifestations of the same injection-recovery design.","tokens_in":16905,"tokens_out":13596,"duration_ms":125328,"concrete_test":"Rerun the grid point ΩPT=1.4e-11 at the lowest and highest fPT with a single global prior spanning the full injection grid (e.g., log10 ΩPT ∈ [-11.5, -9], log10(fPT/Hz) ∈ [-3.5, -1.8]), keeping all other settings identical; recompute ln BF and ΔDIC. If ln BF falls below 8 at either point, or if the recovered uncertainties widen substantially, the headline threshold is an artifact of signal-centered priors. In addition, check the MCMC configuration files for all 100 runs to confirm whether priors were adapted per injection: if they were, report this and rerun the model-selection comparison with the same blind priors used for the null model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Table I reports log10 ΩPT prior U(-10.609,-10.209) and log10(fPT/Hz) prior U(-2.355,-1.955). The injection grid in §III spans ΩPT from 5e-12 to 5e-10 (log10 Ω from -11.30 to -9.30) and fPT from 4e-4 to 1e-2 Hz (log10 f from -3.40 to -2.00). The stated priors include only ΩPT ≳ 2.5e-11 and fPT ≳ 4.4e-3 Hz, so most of the 100 grid points, including the headline threshold ΩPT=1.4e-11 (log10=-10.85) and all but three high-frequency fPT values, lie outside the prior support. If the same priors were used for all runs, the recovered medians shown in Figs. 3-4 at those points are impossible; if the priors were recentered on each injection, the Bayes factors in Figs. 7-8 are computed with signal-informed priors, which inflate evidence relative to a blind search. The paper does not state which case applies, and no code or data are released. The central claim 'robustly detect... ΩPT ≳ 1.4e-11' therefore rests on an undocumented analysis choice, not on the described Bayesian procedure.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript presents a simulation-based forecast of Taiji's sensitivity to stochastic gravitational-wave backgrounds from first-order phase transitions. The analysis injects a two-parameter broken power-law spectrum (Ω_PT, f_PT) into simulated Taiji data that include analytic instrumental noise, a broken-power-law galactic double white dwarf foreground, and an extragalactic compact binary background. The authors use a hybrid Gaussian/log-normal likelihood and nested sampling to perform parameter estimation on a 100-point injection grid, and they report detection thresholds via Bayes factors and the Deviance Information Criterion, as well as parameter-precision curves. The headline claims are a detection threshold Ω_PT ≳ 1.4×10^-11 and frequency estimation better than 10% for Ω_PT ≳ 1.1×10^-10.","tokens_in":17247,"tokens_out":9219,"duration_ms":89115,"significance":"If correct, the result would provide a concrete, quantitative sensitivity target for electroweak-scale phase transitions with Taiji and a useful comparison with LISA forecasts. The paper has real strengths: the benchmark posterior in Table I and Fig. 2 is clean and physically sensible; the agreement between Bayes factors and DIC is a useful internal consistency check; and the noise and foreground modeling is explicit. The main caveats are that the headline threshold is an injection-recovery of the same template and foreground model used in the analysis, and that the documented priors are inconsistent with a large fraction of the injection grid. Because no code or data are released, the central numbers are not independently reproducible from the manuscript as written.","major_comments":[{"comment":"Table I reports priors log10 Ω_PT ~ U(-10.609,-10.209) and log10(f_PT/Hz) ~ U(-2.355,-1.955), i.e. Ω_PT ∈ [2.46×10^-11, 6.18×10^-11] and f_PT ∈ [4.4×10^-3, 1.1×10^-2] Hz. The grid defined in Section III contains Ω_PT values from 5×10^-12 to 5×10^-10 and f_PT values from 4×10^-4 to 1×10^-2 Hz. Only two of the ten Ω_PT rows (3.9×10^-11 and 6.5×10^-11) and three of the ten f_PT columns (4.9×10^-3, 7×10^-3, and 1×10^-2 Hz) lie inside the stated priors. The weakest injections that set the headline threshold, including Ω_PT = 1.4×10^-11, lie outside the prior support. If the same priors were used for all 100 runs, the recovered medians in Figs. 3 and 4 at those points cannot be produced by the stated procedure; if the priors were recentered on each injection, the Bayes factors in Fig. 7 and Eq. (33) are computed with signal-informed priors and are not blind-search evidence. The manuscript does not state which case applies. Please document the per-run priors, or use a single prior covering the full grid, and recompute or re-derive the threshold accordingly.","section":"Section III, Table I and injection grid"},{"comment":"Equations (2), (15), and (17) define the signal and foreground models, and the same functional forms are used to generate the synthetic data and to fit them. The reported thresholds therefore measure the recoverability of the assumed broken power-law spectrum with fixed foreground shapes, not the detectability of FOPT signals in general. The abstract's claim that Taiji can 'robustly detect and characterize phase transition signals' should be qualified as 'within the template family and foreground model considered.' As a concrete test, the authors should vary either the high-frequency slope of the FOPT template or the DWD foreground parameters within their population-synthesis uncertainties and report how the Ω_PT threshold shifts. The conclusion already notes the need for more physically motivated spectral shapes, but the abstract and the threshold claims should carry this qualification.","section":"Section II.A-I.D and Section III"},{"comment":"The optimal binning weights in Eq. (24) are defined through D_th(f_j, θ, n), which depends on the very signal parameters being estimated. The manuscript does not explain how the weights are set during the MCMC runs: if the true injected parameters are used, the binned data are constructed with knowledge of the signal and the subsequent likelihood is partly circular; if a fixed reference model is used, the stated optimality is not realized. Please specify the binning protocol (e.g., an iterative scheme or a fixed fiducial model) and, if necessary, show that the thresholds are insensitive to the choice.","section":"Section III, Eqs. (22)-(24)"}],"minor_comments":[{"comment":"The axis labels in Figs. 3 and 4 appear to have missing negative exponents, e.g., 'PT = 5.0 × 10 12' should presumably be '5.0 × 10^-12'; please correct the rendering.","section":"Figures 3 and 4"},{"comment":"The introduction refers once to 'TinQin'; this should be 'TianQin'.","section":"Section I"},{"comment":"The comparison to 'current constraints [75,76]' is made without noting that NANOGrav and EPTA operate at nHz frequencies; a direct comparison with Taiji's mHz band is not meaningful without explaining the frequency extrapolation involved.","section":"Section IV"},{"comment":"No code or data are released; a reproducibility statement or a link to the injection and recovery scripts would help readers verify the grid results.","section":"Throughout"},{"comment":"The concluding paragraph on single-detector limitations is appropriate and should be reflected in the abstract's 'robustly detect' wording, which currently sounds stronger than the analysis supports.","section":"Section IV"}],"recommendation":"major_revision","confidential_remarks":"The prior/grid inconsistency is the main technical concern and should be resolved before publication. If the authors can show that the per-run priors are reasonable or recompute the thresholds with a prior covering the full grid, the paper may be publishable. I also note that the DWD and ECB foreground parameters are taken from refs. [67,68,69], which are previous works by the same authors; this is not improper because those works are fits to external population-synthesis models, but the provenance should be described explicitly in the text so that the foreground model is not presented as more external than it is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a read if you care about space-based SGWB forecasts. The genuinely new thing is the first dedicated Taiji sensitivity grid for first-order phase-transition backgrounds using current noise and foreground models—100 points, Bayes factors, DIC cross-check. The benchmark posterior (Omega_PT=3.9e-11, f_PT=7e-3 Hz) is clean and all parameters recover within 2-sigma. The internal consistency between BF and DIC is a nice, if modest, validation.\n\nThe soft spot is real and central. Table I reports log10 Omega_PT prior U(-10.609,-10.209) and log10 f_PT prior U(-2.355,-1.955). The injection grid spans Omega_PT from 5e-12 to 5e-10 and f_PT from 4e-4 to 1e-2 Hz. Only one amplitude (3.9e-11) and three high-frequency values (4.9e-3, 7.0e-3, 1.0e-2) lie inside those priors. Yet Figures 3-4 show recovered medians at grid points far outside the prior. Either the priors were widened for those runs or recentered on each injection. The paper never says which. If widened, fine-but then Table I is misleading. If recentered on the injected values, the Bayes factors in Figures 7-8 are signal-informed and inflated relative to a blind search. This is not an esoteric technicality; it is the foundation of the paper's headline claim that Taiji can 'robustly detect' signals with Omega_PT >= 1.4e-11. Without documentation, that claim is not supported.\n\nThe other limitation is more standard: this is an injection-recovery of the same broken power-law template used in the analysis, and the foreground parameters are fixed to values from external fits. That is acceptable for a first estimate, but it measures recovery of the assumed spectrum, not detection of arbitrary FOPT spectra. No code or data are released, which makes the prior ambiguity harder to resolve.\n\nWho is this for? Anybody planning a Taiji science case or comparing LISA/Taiji forecasts for cosmological SGWBs. It is a solid piece of workflow engineering, and the authors clearly know the LISA literature. It deserves a serious referee; the prior documentation must be fixed before the thresholds are used as mission targets. I would not desk-reject it.","headline":"Useful Taiji-specific FOPT sensitivity map, but the headline threshold rests on undocumented per-run priors that contradict Table I.","tokens_in":17758,"tokens_out":4347,"would_cite":true,"duration_ms":35360,"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 Taiji can detect first-order phase-transition gravitational-wave backgrounds with peak energy density above about 1.4e-11 across most of its band, and measure the peak frequency to better than 10 percent for stronger…","keywords":["Taiji","gravitational wave background","first-order phase transitions","Bayesian parameter estimation","double white dwarf confusion noise","space-based gravitational wave observatory","electroweak phase transition","stochastic gravitational wave background"],"falsifier":"Rerun the paper's 100-injection grid with the double-white-dwarf foreground amplitude $A_1$ raised by a factor of 3; if the $\\Omega_{\\rm PT}\\gtrsim1.4\\times10^{-11}$ detection threshold does not shift upward, the quoted threshold is an artifact of the adopted foreground model. The same grid with a sound-shell spectral template instead of the broken power law would test whether the detection and $f_{\\rm PT}$ precision claims are tied to the specific assumed shape.","tokens_in":16716,"feed_emoji":"📡","tokens_out":11304,"duration_ms":92989,"temperature":0.7,"pith_summary":"The paper asks whether Taiji, the Chinese space-based gravitational-wave observatory, can detect the stochastic background produced by first-order phase transitions in the early universe once realistic foregrounds are included. Using a simulated four-year mission (three years after a 75% duty cycle), it combines Taiji instrumental noise, galactic double-white-dwarf confusion noise, and the extragalactic compact-binary background, injects phase-transition signals with a broken power-law spectrum, and recovers them with a Bayesian pipeline. It finds a detection threshold of $\\Omega_{\\rm PT}\\gtrsim 1.4\\times10^{-11}$ across most of the band, with best sensitivity near $10^{-3}$ to $10^{-2}$ Hz, and peak-frequency precision better than 10% for $\\Omega_{\\rm PT}\\gtrsim 1.1\\times10^{-10}$. If correct, these numbers give concrete sensitivity targets for electroweak-scale new physics, including scenarios connected to baryogenesis and dark matter production.","feed_headline":"Taiji could detect phase-transition signals down to Ω ≈ 1.4×10^-11","feed_subtitle":"The space observatory would measure peak frequencies of electroweak-scale phase transitions to better than 10 percent.","key_machinery":"The load-bearing object is the two-parameter broken power-law spectral template $P(f)$ with amplitude $\\Omega_{\\rm PT}$ and peak frequency $f_{\\rm PT}$, which defines both the injected signals and the signal model the fit must recover. Around it the paper constructs a full simulation-inference chain: Taiji noise power spectral densities for optical measurement and test-mass acceleration; the orthogonal A/E/T time-delay-interferometry channels and their response functions; a fixed broken power-law double-white-dwarf foreground and a fixed power-law extragalactic compact-binary foreground; variance-minimizing frequency binning; and a hybrid Gaussian/lognormal likelihood. Bayes factors and Deviance Information Criterion differences carry the detection claim, while nested-sampling posteriors carry the parameter-estimation claim.","core_discovery":"On the paper's own terms, the claim is that Taiji can both detect and characterize a first-order phase-transition gravitational-wave background whose spectrum is the acoustic broken power law $P(f)=(f/f_{\\rm PT})^3[7/4+3(f/f_{\\rm PT})^2]^{-7/2}$. Across a grid of 100 injections spanning $\\Omega_{\\rm PT}$ from $5\\times10^{-12}$ to $5\\times10^{-10}$ and $f_{\\rm PT}$ from $4\\times10^{-4}$ Hz to $10^{-2}$ Hz, the Bayesian analysis recovers the injected amplitudes and peak frequencies with no significant bias; the relative amplitude uncertainty shrinks as signal strength grows, and $\\Delta f_{\\rm PT}/f_{\\rm PT}$ drops below 0.1 once $\\Omega_{\\rm PT}\\gtrsim 1.1\\times10^{-10}$. Model selection with both Bayes factors and the Deviance Information Criterion marks the phase-transition component decisively detected when the peak energy density exceeds roughly $1.4\\times10^{-11}$ over most of the band, and the two metrics agree across the grid.","pith_inferences":["Because the foreground parameters are fixed to one population-model estimate, I would read $1.4\\times10^{-11}$ as a floor: if the actual double-white-dwarf confusion noise is higher or spectrally different, the detection threshold moves upward.","The same pipeline could be rerun with sound-shell or turbulence templates; the resulting shift in the $\\Omega_{\\rm PT}$-$f_{\\rm PT}$ threshold map would show which spectral features Taiji is genuinely sensitive to rather than just the assumed broken power law.","Since the posteriors in Table I constrain the DWD foreground parameters to a few percent, treating them as free in the fit rather than fixed is a cheap robustness upgrade and would make the quoted thresholds more conservative.","The paper's own note that a single detector cannot cross-correlate indicates that joint operation with LISA or TianQin could push the threshold down; an order-of-magnitude improvement is plausible but needs a joint-simulation demonstration."],"forward_implications":["If the thresholds hold, Taiji alone can act as a discovery instrument for electroweak-scale phase transitions in strongly supercooled, composite-Higgs, and hidden-sector scenarios.","The better-than-10% peak-frequency measurement above $\\Omega_{\\rm PT}\\simeq 1.1\\times10^{-10}$ makes $f_{\\rm PT}$ a usable observable, which through $f_{\\rm PT}\\simeq 10^{-6}(H_*R_*)^{-1}(T_*/100\\,{\\rm GeV})$ Hz constrains the transition temperature and inverse mean bubble separation.","The agreement between the Bayes-factor and DIC heatmaps provides a cross-check that the detection claims are not an artifact of one model-selection statistic.","Below roughly $1.4\\times10^{-11}$, or for peak frequencies below about $1.2\\times10^{-3}$ Hz, the double-white-dwarf confusion noise masks the signal and parameter estimates degrade sharply, so those regions should be treated as foreground-limited."],"supporting_citations":[{"why":"Supplies the broken power-law spectral template $P(f)$ that defines the injected phase-transition signal and the fitted parameters $\\Omega_{\\rm PT}$, $f_{\\rm PT}$.","marker":"[11]"},{"why":"Supplies the Taiji optical-measurement and acceleration noise levels entering the noise power spectral densities.","marker":"[60]"},{"why":"Supplies the fitted parameters of the galactic double-white-dwarf broken power-law foreground.","marker":"[67]"},{"why":"Also supplies the double-white-dwarf foreground parameters used in the Taiji sensitivity convention.","marker":"[68]"},{"why":"Supplies the extragalactic compact-binary background amplitude and $2/3$ spectral index.","marker":"[69]"},{"why":"Provides the segmentation and frequency-binning reconstruction framework adapted to Taiji.","marker":"[70]"},{"why":"Provides the optimal binning weights and the hybrid Gaussian/lognormal likelihood used for parameter estimation.","marker":"[71]"},{"why":"Provides the analytical A/E/T channel response functions used to convert noise and signals into energy-density spectra.","marker":"[62]"}],"fun_headline_variants":["Taiji detects phase-transition GW background at Ω≈1.4e-11","Taiji probes early-universe phase transitions down to Ω=1.4e-11","Taiji pinpoints phase-transition peak frequencies to <10% for strong signals","Taiji's threshold for phase-transition detection: Ω≈1.4e-11","Taiji hears cosmic phase transitions from Ω≈1.4e-11"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the adopted broken power-law template for the phase-transition signal and the fixed double-white-dwarf foreground parameters describe the real signals well enough that recovering these simulated injections reflects true detection capability.","fun_headline_variants_meta":{"raw":{"variants":["Taiji detects phase-transition GW background at Ω≈1.4e-11","Taiji probes early-universe phase transitions down to Ω=1.4e-11","Taiji pinpoints phase-transition peak frequencies to <10% for strong signals","Taiji's threshold for phase-transition detection: Ω≈1.4e-11","Taiji hears cosmic phase transitions from Ω≈1.4e-11"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000954,"raw_usage":{"total_tokens":4099,"prompt_tokens":1004,"completion_tokens":3095,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":620,"completion_tokens_details":{"reasoning_tokens":2985}},"tokens_in":620,"tokens_out":3095,"duration_ms":23704,"temperature":1.0,"reasoning_tokens":2985,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:57:33.874577+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rerun the paper's 100-injection grid with the double-white-dwarf foreground amplitude $A_1$ raised by a factor of 3; if the $\\Omega_{\\rm PT}\\gtrsim1.4\\times10^{-11}$ detection threshold does not shift upward, the quoted threshold is an artifact of the adopted foreground model. The same grid with a sound-shell spectral template instead of the broken power law would test whether the detection and $f_{\\rm PT}$ precision claims are tied to the specific assumed shape.","supporting_citations":[{"cited_title":"A brief analysis to Taiji: Science and technology,","cited_arxiv_id":null,"evidence_quote":"Supplies the Taiji optical-measurement and acceleration noise levels entering the noise power spectral densities."},{"cited_title":"Detecting a Gravitational-Wave Background from Inflation with Null Energy Condition Violation: Prospects for Taiji","cited_arxiv_id":"2404.08375","evidence_quote":"Also supplies the double-white-dwarf foreground parameters used in the Taiji sensitivity convention."},{"cited_title":"Sensitivity functions of space- borne gravitational wave detectors for arbitrary time-delay interferometry combinations regarding non- tensorial polarizations,","cited_arxiv_id":null,"evidence_quote":"Provides the analytical A/E/T channel response functions used to convert noise and signals into energy-density spectra."}],"review_version":1}