{"id":"6d5b72fb-8f87-404b-8c85-bc07ddaf82bb","arxiv_id":"2504.18625","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Forecasted SKAO 21cm intensity mapping plus Planck 2018 data could push the 95% upper limit on the summed neutrino mass from 0.285 eV to about 0.105 eV (fixed astrophysical nuisances).","lead":"This paper forecasts how well the future SKAO 21cm hydrogen survey could measure the sum of neutrino masses, using simulated observations and MCMC analysis. It finds that combining SKAO 21cm data with Planck CMB data could reach a 95% upper limit near 0.105 eV, better than Planck alone.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline 0.105 eV limit is a fixed-nuisance, foreground-free projection; the paper's own marginalized limit is 0.126 eV and residual foregrounds are acknowledged to add uncertainty, so the advertised sensitivity is optimistic.","rationale":"The reader's weakest assumption (self-consistent mocks, fixed covariance, no foregrounds) is essentially the one we identify. We partially agree because we put more weight on the fixed-nuisance headline and foreground cleaning than on covariance and code availability. The paper reports both fixed and marginalized results, so the internal analysis is not flawed; the issue is which number is advertised. The conditional verdict already captures this: the forecast is useful under its stated assumptions but should not be quoted as 0.105 eV without the fixed-nuisance caveat. Our proposed test would settle whether the headline limit survives a more realistic treatment of astrophysical uncertainties and foregrounds, and if it fails, the claim would need to be weakened or re-aimed at the marginalized result.","tokens_in":26395,"tokens_out":18288,"duration_ms":207591,"concrete_test":"Generate the Sigma_mnu = 0.06 eV auto-spectrum mock using an independent pipeline (e.g., lognormal or hydro-simulation light cones with a different HI bias and brightness-temperature evolution), add a foreground residual component at the level of current MeerKLASS cleaned maps, and analyze it with the topk likelihood using the marginalized-nuisance model. If the recovered 95% upper limit on Sigma_mnu exceeds about 0.2 eV, or drifts by more than the statistical error from the input value, the central 0.105 eV claim is an artifact of the self-consistent mock and omitted foregrounds.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper is coherent as an idealized forecast, but the central number is the least robust corner of its own analysis. Table 4 shows that the 0.105 eV (95%) limit for the Sigma_mnu = 0.06 eV auto-spectrum case is obtained with the HI nuisance combinations Tb bHI sigma8(z) and Tb f sigma8(z) fixed; marginalizing them raises the limit to 0.126 eV, and for the Sigma_mnu = 0.1 eV case the result changes from a measurement (0.098 +/- 0.022 eV) to an upper limit (< 0.151 eV). The improvement over Planck is driven by the sub-percent BAO distance errors reported in Section 3.1 (sigma_DA ~ 0.1-0.3%, sigma_H ~ 0.1-0.5%), which come from mocks generated with the same analytical power-spectrum model, the same polynomial nuisance parametrization, and a covariance fixed to the fiducial cosmology. No foreground-cleaning penalty appears in the likelihood; Appendix A explicitly says residual foregrounds and systematics are expected to increase uncertainties. Thus the advertised sensitivity is not robust to the two things a real SKA-Mid measurement must face: unknown HI astrophysics and foreground removal. This is a limitation, not an internal inconsistency, but it means the central claim should be read as an idealized projection rather than a prediction for SKAO.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper forecasts the sensitivity of SKAO 21cm intensity mapping (IM) observations to the sum of neutrino masses, Σmν, using synthetic data sets. The authors construct mock observations of the 21cm auto-power spectrum and of the cross-correlation between 21cm IM and DESI-like/Euclid-like galaxy surveys, for three fiducial values of Σmν (0.06, 0.1, 0.4 eV). They build a Gaussian likelihood with an analytic covariance, sample the posterior with MCMC (Cobaya), include Alcock-Paczynski distortions, a Gaussian beam, and redshift-dependent polynomial nuisance parameters, and combine the synthetic data with Planck 2018 CMB likelihoods. The main results are that the 21cm auto-spectrum alone yields a 95% upper limit of Σmν < 0.287 eV for the 0.06 eV fiducial, comparable to Planck alone, and that combining with Planck tightens the limit to Σmν < 0.105 eV when nuisance parameters are fixed (or < 0.126 eV when marginalized). The cross-correlation forecasts, when combined with Planck, give Σmν < 0.116 eV (DESI-like) and < 0.117 eV (Euclid-like). The paper presents a public likelihood code (topk) and investigates the scale dependence of the growth rate in an appendix.","tokens_in":26602,"tokens_out":8933,"duration_ms":84726,"significance":"If the forecast is realized, SKAO 21cm IM combined with CMB data would substantially tighten neutrino mass constraints, breaking the H0–Σmν degeneracy that limits CMB-only analyses, and would motivate early cross-correlation detections. The paper's strengths include a internally consistent MCMC pipeline with a public code, tests over multiple fiducial neutrino masses, explicit treatment of the AP effect and multipoles, and an appendix quantifying the scale dependence of the growth rate. However, the advertised headline sensitivity is conditional on the assumed analytical model and noise model, and it is not robust to the main systematic of 21cm IM, namely foreground residuals, which the paper itself acknowledges in Appendix A. The reported precision also suffers from internal inconsistencies between the summary table and the appendix tables. As an idealized forecast the work is sound, but the central claim needs to be reframed and verified before it can be taken as a prediction for SKAO.","major_comments":[{"comment":"There is an internal inconsistency in the reported 68% constraints on Σmν for the 0.4 eV fiducial between the summary table and the full tables. For example, Table 4 gives Σmν = 0.398 ± 0.018 eV for Planck + auto-spectrum with fixed nuisances, whereas Table B.5 gives 0.398 ± 0.036 eV. Similarly, Table 4 gives Σmν = 0.396+0.023−0.026 for the DESI cross-correlation, while Table B.6 gives 0.396+0.052−0.046. The same factor-of-two difference appears for the Euclid case and for the auto-only 0.4 eV results. Since Table 4 is the primary summary of the paper's findings, this is a load-bearing error that must be resolved before publication, and the correct values need to be identified and propagated consistently through the abstract, conclusions, and tables.","section":"Table 4 vs Appendix B"},{"comment":"The headline constraint Σmν < 0.105 eV is obtained with nuisance parameters fixed, i.e., holding Tb bHI σ8(z) and Tb f σ8(z) at their fiducial values. When these nuisance parameters are marginalized over, the limit loosens to Σmν < 0.126 eV, as shown in Table 4. The abstract and the conclusions present the fixed-nuisance number as the primary result without this qualification. Given that the paper's own analysis shows that marginalizing over HI astrophysics degrades the constraint by about 20%, the abstract should either lead with the marginalized value or explicitly state that the quoted limit assumes perfect knowledge of the HI bias and brightness temperature.","section":"Abstract, Section 3.2, Section 4"},{"comment":"The likelihood in Eq. (2.25) assumes no residual foreground contamination, and all quoted constraints are derived from synthetic data generated with that assumption. However, Appendix A states that \"residual foregrounds and systematic contamination are expected to increase the measurement uncertainties\" and that a comprehensive investigation is beyond the scope of the work. Foreground removal is one of the central challenges for SKA-Mid 21cm IM, and the sub-percent BAO distance errors reported in Section 3.1 (σDA ~ 0.1–0.3%, σH ~ 0.1–0.5%), which drive the H0–Σmν degeneracy breaking, come from this foreground-free model. To make the central claim robust, the authors should add an explicit foreground residual systematic term to the likelihood, or alternatively quantify how the Σmν constraints degrade for a plausible range of residual foreground levels.","section":"Appendix A and Section 2.3.1"}],"minor_comments":[{"comment":"The cross-correlation coherence r is fixed to 1, and the text asserts that possible variations of r are absorbed by the nuisance parameters. This is not demonstrated: r multiplies the cross-spectrum signal, while the noise term in Eq. (2.22) depends separately on the auto-spectrum P21 and the galaxy spectrum Pg, so the degeneracy is not exact. The authors should either sample over r or justify the assumption more carefully.","section":"Section 2.2.2, Eq. (2.22)"},{"comment":"There is a typo in the heading: \"syntethic data\" should be \"synthetic data\". A similar typo appears in the conclusions.","section":"Section 3.3"},{"comment":"The caption reads \"left and right panel\" but should be \"left and right panels\".","section":"Figure 1 caption"},{"comment":"The notation for the HI bias alternates between bHI and b HI in the text; the notation should be unified.","section":"Section 2.1, Eq. (2.1)"},{"comment":"The claim that fixing the covariance to the fiducial cosmology introduces less than one percent systematic error is only sketched. A brief quantitative demonstration, even for one representative redshift bin, would make the argument more convincing.","section":"Section 2.3.1, footnote 4"}],"recommendation":"major_revision","confidential_remarks":"The internal inconsistency between Table 4 and the appendix tables is the most urgent issue; it directly affects the advertised precision and must be corrected. The overstatement of the headline result in the abstract (fixed-nuisance limit without qualification) is also likely to draw criticism from readers familiar with 21cm IM systematics, and it should be addressed in revision. The paper is otherwise a coherent idealized forecast."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a competent forecast paper, worth engaging, but the 0.105 eV number in the abstract is the least robust number in the paper. It is a fixed-nuisance, foreground-free projection. The paper itself shows the marginalized limit is 0.126 eV, and residual foregrounds are acknowledged but not in the likelihood.\n\nWhat's new: the authors extend their own topk pipeline to massive neutrinos and to 21cm-galaxy cross-correlation with DESI/Euclid-like surveys. The specific forecasted limits (0.105, 0.116, 0.117 eV combined with Planck) are new; previous SKA neutrino forecasts used Fisher matrices and simulations. Doing full MCMC is a step up, and the paper honestly reports both fixed and marginalized constraints in Table 4 and the appendix tables.\n\nWhat's good: internal consistency is fine. Synthetic data are generated from the same analytical power spectrum model with the same nuisance parametrization, and the MCMC recovers the fiducial cosmology. The AP effect treatment is explicit, and the paper discusses the scale dependence of f(z,k) in Appendix A, showing the error from neglecting it is <1% on the scales used. The comparison with Villaescusa-Navarro et al. 2015 is fair.\n\nThe soft spots are real but proportionate. The advertised sensitivity depends on exactly the two things a real SKA-Mid measurement must face: unknown HI astrophysics and foreground removal. The fixed-nuisance limit is 0.105, but marginalizing over the polynomial nuisances worsens it to 0.126. For the 0.1 eV fiducial, the fixed-nuisance result is a 0.098±0.022 measurement, but marginalizing turns it into an upper limit <0.151. The sub-percent BAO distance errors that drive the H0–Σmν degeneracy breaking come from the same model that is being fitted, so they are best-case. And Appendix A says residual foregrounds and systematics are expected to increase uncertainties. None of this invalidates the central claim as an idealized forecast; it does mean the headline should be read as a projection, not a prediction.\n\nThe main missing piece for the field is reproducibility: the neutrino-enabled code and synthetic data are not yet public, so the specific numbers cannot be checked without emailing the authors. That is a fixable issue.\n\nWho this is for: people doing forecasts for SKAO science, neutrino-mass projections, and 21cm cross-correlation strategy. A serious referee should spend time on it; it deserves peer review rather than desk rejection. I would probably cite the numbers as forecast results under stated assumptions, not as evidence about the real SKAO sensitivity.","headline":"Solid, honest forecast paper; the 0.105 eV headline is a fixed-nuisance, foreground-free projection, so read it as an idealized sensitivity estimate rather than a prediction for SKAO.","tokens_in":27280,"tokens_out":2458,"would_cite":true,"duration_ms":24807,"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":"Future 21cm intensity mapping with SKAO could push the summed neutrino mass below 0.105 eV at 95% confidence.","keywords":["sum of neutrino masses","21cm intensity mapping","SKAO forecasts","cosmological parameter constraints","baryon acoustic oscillations","Alcock-Paczynski effect","galaxy cross-correlation","Markov chain Monte Carlo"],"falsifier":"Run the same likelihood pipeline on an end-to-end simulation in which the 21cm sky is generated from an independent hydrodynamical model and processed through foreground removal, with the covariance re-estimated from realizations rather than fixed to the fiducial cosmology; if the resulting 21cm-plus-Planck 95% upper limit on $\\Sigma m_\\nu$ comes out substantially above 0.105 eV, or the auto-spectrum alone fails to reach $\\sim 0.29$ eV, the central claim would be refuted.","tokens_in":26078,"feed_emoji":"📡","tokens_out":9878,"duration_ms":92396,"temperature":0.7,"pith_summary":"This paper forecasts how tightly the SKAO 21cm intensity-mapping survey could weigh the sum of neutrino masses, $\\Sigma m_\\nu$, using synthetic observations built from the SKA-Mid survey design. Its central claim is that the 21cm auto-power spectrum alone would give a 95% upper limit of $\\Sigma m_\\nu < 0.287$ eV, comparable to Planck CMB alone, and that adding Planck 2018 data tightens this to $\\Sigma m_\\nu < 0.105$ eV by breaking the CMB's degeneracy between $H_0$ and $\\Sigma m_\\nu$. The same combination with a DESI-like or Euclid-like galaxy survey in cross-correlation gives $\\Sigma m_\\nu < 0.116$ eV and $\\Sigma m_\\nu < 0.117$ eV, respectively, so the easier-to-detect cross-correlation route is competitive. The analysis also shows that at a fiducial mass of 0.1 eV the combined data would yield a measurement, not just an upper limit, of about $0.098 \\pm 0.022$ eV.","feed_headline":"21cm+Planck could weigh neutrinos at 0.105 eV","feed_subtitle":"Forecast: SKAO intensity mapping breaks the CMB's mass-Hubble degeneracy, more than halving the neutrino-mass ceiling.","key_machinery":"The argument runs on an extended 21cm intensity-mapping model: the anisotropic power spectrum is written in terms of the CDM and baryon power spectrum and a scale-dependent growth rate, $f(z,k)$, with beam smoothing and the Alcock-Paczynski (AP) effect folded into the observed monopole and quadrupole. The AP effect is the geometric distortion that lets the same survey measure the angular diameter distance and the Hubble parameter; it converts the 21cm baryon acoustic oscillation features into sub-percent constraints on $H_0$ and $\\Omega_m$ across six redshift bins from $0<z<3$. The second carrier is the analytical mode-count covariance from the survey volume and telescope noise, which sets the error bars on each multipole. Feed the 21cm likelihood into a Markov chain Monte Carlo analysis, combine with Planck, and the sharp $H_0$ measurement breaks the CMB's $H_0$-$\\Sigma m_\\nu$ anti-correlation, turning a $0.285$ eV limit into $0.105$ eV.","core_discovery":"The discovery claimed is a path to sub-0.1-eV neutrino-mass sensitivity from future 21cm observations. Working in a $\\Lambda$CDM plus massive-neutrino cosmology, the authors build synthetic SKA-Mid auto-spectrum and 21cm-galaxy cross-correlation data sets for three fiducial masses (0.06, 0.1, 0.4 eV), adding Planck 2018 CMB likelihoods. They report that the 21cm auto-spectrum alone gives $\\Sigma m_\\nu < 0.287$ eV at 95% confidence for a fiducial mass of 0.06 eV, matching Planck's $0.285$ eV; combined, the limit becomes $\\Sigma m_\\nu < 0.105$ eV, while the cross-correlation routes with DESI-like and Euclid-like surveys combined with Planck reach $0.116$ and $0.117$ eV. The mechanism is that the 21cm multipoles, through the Alcock-Paczynski effect, measure $H_0$ and $\\Omega_m$ at sub-percent precision, which removes the strong anti-correlation between $H_0$ and $\\Sigma m_\\nu$ in the CMB. For a fiducial mass of 0.4 eV the combined data would measure $\\Sigma m_\\nu = 0.398 \\pm 0.018$ eV, roughly a sixfold tightening relative to Planck alone.","pith_inferences":["The practical case for neutrino weighing may rest on the cross-correlation limits of 0.116-0.117 eV rather than the 0.105 eV auto-spectrum limit, because cross-correlations suppress foregrounds and are expected to deliver detections sooner.","The sub-percent Alcock-Paczynski distances are the most fragile ingredient; any residual redshift-space distortion, beam error, or scale-dependent bias that mimics them would widen $H_0$ and loosen the neutrino limit.","Because the synthetic data are generated from the same model, the pipeline, and the same fixed covariance that the fit later assumes, these forecasts are a best case; running the same likelihood on independent hydrodynamical mocks with foreground removal would quantify the degradation."],"forward_implications":["SKAO 21cm auto-spectrum plus Planck 2018 CMB data would set a 95% upper limit $\\Sigma m_\\nu < 0.105$ eV for a fiducial 0.06 eV mass, about 2.7 times tighter than Planck alone.","At a true mass near 0.1 eV the combined data would measure $\\Sigma m_\\nu = 0.098 \\pm 0.022$ eV, turning a bound into a detection.","21cm cross-correlation with a DESI-like or Euclid-like survey plus Planck would give limits of 0.116 and 0.117 eV, nearly matching the auto-spectrum route despite having no constraining power on $\\Sigma m_\\nu$ on their own.","The same 21cm data alone would improve the $H_0$ error by an order of magnitude and $\\Omega_m$ by a factor of roughly 2 relative to CMB-only constraints, so the neutrino-mass forecast rides on a broader late-Universe gain.","If the true sum were 0.4 eV, the combined data would measure it as $\\Sigma m_\\nu = 0.398 \\pm 0.018$ eV, roughly a sixfold improvement over Planck alone."],"supporting_citations":[{"why":"Supplies the SKA-Mid survey configuration (bands, sky areas, noise parameters) that defines the synthetic 21cm data sets.","marker":"[39]"},{"why":"Provides the forecasting likelihood and power-spectrum formalism that this work extends to massive-neutrino cosmologies.","marker":"[48–50]"},{"why":"Supplies the Planck 2018 CMB likelihoods and the fiducial cosmology used for the synthetic data and for the comparison baseline.","marker":"[19]"},{"why":"Gives the earlier hydrodynamical-simulation-based SKA neutrino-mass forecast that this paper contrasts with its analytic MCMC approach.","marker":"[51]"},{"why":"Provides the multipole expansion for 21cm intensity mapping whose monopole and quadrupole error model and signal-to-noise estimates are reused.","marker":"[49]"},{"why":"Supplies the cross-correlation forecasting pipeline for SKAO with galaxy surveys that the synthetic 21cm-galaxy data sets extend.","marker":"[50]"},{"why":"Provides the analytic mode-count and variance formulas that set the forecast error bars on the power-spectrum multipoles.","marker":"[62]"},{"why":"Provides the hydrodynamical simulation results interpolated for the neutral-hydrogen bias and shot-noise terms in the model.","marker":"[56, 57]"}],"fun_headline_variants":["21cm intensity mapping at SKAO could lower neutrino mass limit","SKAO 21cm plus Planck to cut neutrino mass ceiling to 0.105 eV","21cm auto-spectrum alone rivals CMB for neutrino mass constraints","Forecast: 21cm cross-correlations with galaxies weigh neutrinos","SKAO 21cm maps combined with CMB best neutrino mass bounds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The forecast assumes the synthetic observations are drawn from exactly the same analytical power-spectrum model, with the same polynomial nuisance parameters and a covariance fixed to the reference cosmology, that the MCMC later fits, and that foregrounds leave no residual imprint; if real SKA-Mid noise, beam response, or neutral-hydrogen astrophysics differ, the tight error bars, especially the sub-percent distances that break the $H_0$-$\\Sigma m_\\nu$ degeneracy, will be optimistic.","fun_headline_variants_meta":{"raw":{"variants":["21cm intensity mapping at SKAO could lower neutrino mass limit","SKAO 21cm plus Planck to cut neutrino mass ceiling to 0.105 eV","21cm auto-spectrum alone rivals CMB for neutrino mass constraints","Forecast: 21cm cross-correlations with galaxies weigh neutrinos","SKAO 21cm maps combined with CMB best neutrino mass bounds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000244,"raw_usage":{"total_tokens":1693,"prompt_tokens":1268,"completion_tokens":425,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":884,"completion_tokens_details":{"reasoning_tokens":326}},"tokens_in":884,"tokens_out":425,"duration_ms":4265,"temperature":1.0,"reasoning_tokens":326,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:14:07.183222+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same likelihood pipeline on an end-to-end simulation in which the 21cm sky is generated from an independent hydrodynamical model and processed through foreground removal, with the covariance re-estimated from realizations rather than fixed to the fiducial cosmology; if the resulting 21cm-plus-Planck 95% upper limit on $\\Sigma m_\\nu$ comes out substantially above 0.105 eV, or the auto-spectrum alone fails to reach $\\sim 0.29$ eV, the central claim would be refuted.","supporting_citations":[{"cited_title":"Weighing neutrinos with cosmic neutral hydrogen","cited_arxiv_id":"1507.05102","evidence_quote":"Gives the earlier hydrodynamical-simulation-based SKA neutrino-mass forecast that this paper contrasts with its analytic MCMC approach."}],"review_version":1}