{"id":"7c79fb2e-d10c-4e22-89fd-2eac9ce3c38d","arxiv_id":"2504.18745","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A log-normal model of FRB dispersion measure, with variance computed from the electron power spectrum in hydrodynamical simulations, connects FRB observations to baryonic feedback and matter power spectrum suppression.","lead":"Fast radio bursts can map the electrons between galaxies, and the spread in their dispersion measures reveals how much energy feedback from black holes and stars has pushed gas out of halos. The authors build a simulation-based model that turns this spread into constraints on the suppression of the matter power spectrum, a key uncertainty for weak lensing surveys.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The forecasts and cross-simulation validation are generated from the same log-normal likelihood being assumed, so the log-normal shape is never tested against actual sightline DM distributions; if p(DMcosmic|z) has heavier tails, feedback constraints and Pmm suppression forecasts could be biased.","rationale":"The reader's weakest-assumption analysis identifies the log-normal p(DMcosmic|z) as the key unverified ingredient, and my reading of the paper converges on the same point. The paper explicitly states in Section 5.5.5 that no rigorous derivation exists for log-normality, and the validation and forecasting pipeline does not expose the assumption to falsification: mock FRB samples are generated from the same analytic log-normal distribution that defines the likelihood, so the MCMC recovery in Table 1 and Figure 10 can only test internal consistency and interpolation accuracy, not the shape of the DM distribution. The comparison in Figure 2 is suggestive but qualitative, and it is not carried through the inference for Astrid and SIMBA. This concern is load-bearing because the headline claim is a percent-level forecast for Pmm suppression; if the true p(DMcosmic|z) has a heavier tail or different skewness, the log-normal likelihood will assign incorrect weights to high-DM sightlines and bias the constrained feedback parameters and derived suppression. A concrete test using actual simulated sightlines would settle the issue directly. I do not see a more serious internal inconsistency: the analytic variance from Pee is standard, the large-scale variance caveat is acknowledged in Section 5.5.1, and the cross-simulation recovery on SIMBA and Astrid is a genuine strength. The paper is an honest methodological study, so the conditional verdict is appropriate and no verdict change is needed.","tokens_in":31663,"tokens_out":6313,"duration_ms":75823,"concrete_test":"Generate mock FRB samples by tracing actual sightlines through the IllustrisTNG, SIMBA, and Astrid CAMELS snapshots, or by resampling the true simulated p(DMcosmic|z) histograms rather than the analytic log-normal distributions. Then run the same MCMC analysis from Section 5.3 and compare the recovered Phydro/Pgravity-only values at k = 0.5, 1, 10, and 50 h/Mpc with the true simulation values. If any recovery bias exceeds the quoted 68% intervals from Table 1 and Figure 10, the log-normal likelihood is inadequate and the forecast precision should be downgraded; if not, the log-normality concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires p(DMcosmic|z) to be log-normal, and the paper's own Section 5.5.5 admits there is no rigorous derivation of log-normality. More importantly, the validation path does not test this shape against actual simulated sightline DM distributions: mock samples in Section 5.2 are drawn from the analytically constructed log-normal p(DMcosmic|z) shown in Figure 9, and the likelihood in Equations 15 and 16 uses the same log-normal form. The MCMC tests therefore verify only the nearest-neighbor interpolation of Pee and parameter recovery under the exact assumed likelihood, not whether the likelihood is adequate. Figure 2 supports log-normal over the Macquart F-form for IllustrisTNG sightlines, but no quantitative residual or tail comparison is given, and no such check is shown for Astrid or SIMBA. If the true distribution has heavier tails or different skewness, the inferred feedback parameters, and hence the reported Pmm suppression constraints in Table 1 and Figure 10, could be biased by more than the quoted 68% widths because the likelihood upweights high-DM outliers. Since the 10^4-FRB precision forecast is the paper's main result, this untested shape assumption is load-bearing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a simulation-based formalism for connecting the FRB dispersion-measure–redshift relation to baryonic feedback and the suppression of the matter power spectrum. The authors parameterize p(DMcosmic|z) as a log-normal whose mean is computed from the cosmological mean electron density (Eq. 5) and whose variance is computed from the electron power spectrum Pee(k,z) via Eq. (11). They calibrate Pee(k,z) and the matter-power-spectrum suppression Phydro/Pgravity-only to the CAMELS IllustrisTNG Latin-hypercube suite as functions of four feedback parameters and two cosmological parameters, using nearest-neighbor interpolation. They compare the log-normal form against the Macquart F-parameterization, argue that the F-parameter misestimates the DM variance and introduces biases in Pmm, and then perform MCMC forecasts with mock FRB samples drawn from the model itself. They report that 10^4 FRBs can constrain Phydro/Pgravity-only to percent-level precision at large scales and about 10% precision at k ~ 10 h/Mpc, with a prior-to-posterior width ratio of about 20. They also apply the IllustrisTNG-calibrated model to Astrid and SIMBA fiducial runs and claim that it recovers the true Pmm suppression. The paper includes a detailed caveats section covering large-scale variance, parameter degeneracies, sparse interpolation, selection effects, and the lack of a rigorous derivation of log-normality.","tokens_in":31910,"tokens_out":3976,"duration_ms":40233,"significance":"If the central claims hold, this is a valuable contribution: it provides a physically interpretable replacement for the ad hoc F-parameter, exploits the redshift-dependent DM variance that is automatically built into simulations, and links FRB observables to baryonic feedback at scales k ~ 10 h/Mpc, which are relevant for upcoming weak-lensing surveys. The analytic moment formalism in Eqs. (5) and (11) is standard and is applied in a novel way, and the use of the CAMELS suite to span diverse feedback scenarios is appropriate. The paper also gives a clear and honest caveats section, and it demonstrates concrete gains over the Macquart parameterization, which is a useful result in itself. However, as detailed below, the main forecast precision is currently a self-consistency result, and the log-normal shape assumption is not directly validated against simulated sightlines. The significance of the paper will be substantially higher if the authors add a quantitative distributional validation and reframe or robustify the forecast claims.","major_comments":[{"comment":"The log-normal shape of p(DMcosmic|z) is load-bearing for the central forecast, but the validation path is circular. In §5.2 and Fig. 9, mock DMcosmic samples are drawn from the analytically constructed log-normal p(DMcosmic|z), and the likelihood in Eqs. (15)–(16) uses exactly the same log-normal form. The MCMC recovery tests in §5.3 and §5.4 therefore verify only that parameters are recovered under the assumed likelihood, not that the log-normal assumption is adequate. The only direct comparison with simulated sightline DM distributions is Fig. 2, which is qualitative, limited to IllustrisTNG, and does not include a quantitative goodness-of-fit or tail comparison; no such check is shown for Astrid or SIMBA. Since the reported Pmm-suppression constraints in Table 1 and Fig. 10 could be biased if the true distribution has heavier tails or different skewness, the paper should add a quantitative distributional validation (e.g., quantile–quantile plots, Kolmogorov–Smirnov or Anderson–Darling tests, or tail-index comparisons) against actual sightline DM in IllustrisTNG, Astrid, and SIMBA at several redshifts, and/or a robustness test with an alternative likelihood form to bound the potential bias.","section":null},{"comment":"The forecast precision quoted in the abstract and Table 1 is presented as the expected constraining power with 10^4 FRBs, but it is a self-consistency check that does not include the systematic uncertainties acknowledged in §5.5. The 25 h^-1 Mpc CAMELS box misses about 4–10% of the DM variance from large scales (§5.5.1), and nearest-neighbor interpolation in a 6-dimensional space with 1000 calibration points is sparse (§5.5.3). Because the mock catalogs are generated from the same calibrated model that is fitted, the 68% widths in Table 1 and Fig. 10 do not propagate these errors. The claims such as 'percent-level precision at large scales and ~10% precision at k ≳ 10 h/Mpc' should either be explicitly reframed as conditional on the model and its known limitations, or the identified systematic uncertainties should be propagated into the quoted errors. As written, the forecast overstates the current state of the model.","section":"§5.5.1, §5.5.3, Table 1"},{"comment":"The cross-simulation validation for Astrid and SIMBA is a useful transfer test, but it inherits the same log-normal assumption: the mock samples for these simulations are drawn from the analytically constructed p(DMcosmic|z) in Fig. 9, not from actual sightline distributions. The test therefore demonstrates that the IllustrisTNG-calibrated Pee interpolation can reproduce the Pmm suppression of other feedback implementations under the assumed likelihood, but it does not test whether the log-normal shape holds for Astrid or SIMBA. In addition, the reported recovery accuracy for SIMBA at large scales (2.5% at k=0.5 h/Mpc) and Astrid at k=50 h/Mpc (32%) is noticeably worse than the IllustrisTNG values; the attribution of the large-scale discrepancy to the CAMELS box size is plausible but should be quantified, e.g., by comparing against a larger-volume simulation or by estimating the variance contributed by modes larger than the box.","section":"§5.4, Fig. 10"}],"minor_comments":[{"comment":"The abstract contains a typo: 'upto' should be 'up to'.","section":"Abstract"},{"comment":"The caption contains a typo: 'feedack strength' should be 'feedback strength'.","section":"Fig. 3 caption"},{"comment":"The discussion of Zhang et al. (2021) is slightly confusing: the text says 'in contrast to the findings of Zhang et al. (2021), we find that this functional form provides a poor fit', but then says 'we reanalyze their data and provide the updated fits'. Please clarify whether the reanalysis reproduces their data or uses a different fitting procedure, and state the normalization condition explicitly relative to their Table 1 and Figure 2.","section":"§3.1.1"},{"comment":"The 68% intervals for SP(k=50) and for logASN2 at N_FRB=10^4 are asymmetric and in some cases the upper or lower error bar is much smaller than the other; a note on how the percentiles are computed (highest posterior density interval versus equal-tailed interval) would aid interpretation.","section":"Table 1"},{"comment":"In the bottom panels of Fig. 10, the legend distinguishes 'Best Fit' from 'Posterior', but the text does not define what 'Best Fit' refers to (maximum a posteriori value?); please define this in the caption or text.","section":"Fig. 10"}],"recommendation":"major_revision","confidential_remarks":"This is a promising and timely contribution, and the authors are unusually transparent about limitations. The main issue is that the central forecast precision is currently a self-consistency check: the log-normal shape that defines the likelihood is also the shape used to generate the mock data, and no quantitative validation against actual simulated sightlines is provided. I believe this can be fixed within the manuscript's scope by adding direct distributional comparisons and a robustness test with a non-log-normal likelihood, and by reframing the forecast claims as conditional on the model. The paper should not be rejected on these grounds, but the load-bearing nature of the untested shape assumption makes major revision appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe paper gives the FRB community something it needed: a replacement for the ad hoc F-parameter that is tied to the electron power spectrum and calibrated to a large suite of hydro simulations. The log-normal p(DMcosmic|z) with analytic mean and variance is a clean construction, and the authors credit McQuinn and Reischke & Hagstotz for the variance formula. What is new is the calibration of Pee(k,z) to CAMELS IllustrisTNG as a function of four feedback parameters plus cosmology, and the demonstration that the resulting model recovers the matter power spectrum suppression in simulations with very different feedback physics (SIMBA, Astrid). The critique of the F-parameter in Section 3 is well-argued and timely.\n\nThe soft spots are real but not fatal. The headline forecasts in Table 1 and Figure 10 are self-consistency checks: mocks are drawn from the same log-normal likelihood being fitted, so the quoted percent-level constraints are idealized. The log-normal shape itself is only tested against actual sightline distributions for IllustrisTNG (Figure 2, using Zhang et al. 2021 data); for SIMBA and Astrid, the validation uses samples drawn from the analytic log-normal, not from simulated sightline DM distributions. If the true distribution has heavier tails or different skewness, the inferred feedback parameters and hence Pmm suppression could be biased. The 25 Mpc/h box misses roughly 10% of the variance from large scales, as the authors note. Nearest-neighbor interpolation in six dimensions with 1000 points is crude.\n\nNone of this undermines the central methodological contribution. The paper is careful, cites prior work properly, and flags its own limitations in Section 5.5. The cross-simulation recovery of Pmm suppression is a genuine plus, even if the mock generation inherits the log-normal assumption.\n\nWho is it for: people building likelihoods for FRB cosmological analyses, and anyone wanting a physically motivated bridge from FRB DM scatter to baryonic feedback. The limitations mean the forecasts should not be taken at face value for survey design, but the framework is a solid basis for future work with larger simulations and a proper test of log-normality.\n\nFor peer review: worth a serious referee. Send it. I would ask the authors to add a direct comparison of the log-normal against actual simulated sightline DM distributions for SIMBA and Astrid, and to temper the forecast language accordingly.","headline":"A useful, honest replacement for the F-parameter that ties FRB DM scatter to the electron power spectrum; the headline forecasts are self-consistency checks, but the framework is worth refereeing.","tokens_in":32495,"tokens_out":2273,"would_cite":true,"duration_ms":21971,"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":"This paper argues that the sightline-to-sightline scatter in fast radio burst dispersion measures, modeled as a log-normal distribution whose moments are computed from the feedback-dependent electron power spectrum, can constrain baryonic…","keywords":["fast radio bursts","dispersion measure","baryonic feedback","matter power spectrum suppression","log-normal distribution","electron power spectrum","cosmological inference","hydrodynamical simulations"],"falsifier":"Take many mock FRB sightlines from a strong-feedback hydrodynamical simulation and measure the third and fourth moments of $\\mathrm{DM}_{\\rm cosmic}$ directly; if they disagree with the log-normal values implied by the fitted mean and variance, or if a simulation-based likelihood gives different posteriors on $P_{\\rm hydro}/P_{\\rm gravity-only}$ than the log-normal likelihood, the central claim is falsified.","tokens_in":31446,"feed_emoji":"📡","tokens_out":8692,"duration_ms":81784,"temperature":0.7,"pith_summary":"The paper tries to establish that fast radio bursts (FRBs) can be turned into a precision probe of how baryonic feedback reshapes cosmic structure. It replaces the standard ad hoc feedback parameter $F$ with a log-normal model of the cosmic dispersion-measure distribution, $p(\\mathrm{DM}_{\\rm cosmic}|z)$, whose mean and variance are computed from cosmology and from the feedback-dependent electron power spectrum $P_{ee}(k,z)$. The electron power spectrum is calibrated against a suite of 1,000 hydrodynamical simulations spanning a wide range of feedback strengths, which also builds in the correct redshift evolution of the DM scatter. If the model is right, a sample of $10^4$ FRBs would pin the baryonic suppression of the matter power spectrum to percent-level accuracy on large scales and roughly ten percent accuracy at $k\\sim10\\,h\\,\\mathrm{Mpc}^{-1}$, the small scales that upcoming weak-lensing surveys need to understand.","feed_headline":"Fast radio bursts can map baryonic feedback to 1% precision","feed_subtitle":"A log-normal DM model turns FRB sightline scatter into a percent-level constraint on small-scale matter clustering.","key_machinery":"The load-bearing object is the log-normal ansatz for $p(\\mathrm{DM}_{\\rm cosmic}|z)$, specified by two analytically computed moments. The mean is $\\langle\\mathrm{DM}_{\\rm cosmic}(z_s)\\rangle = \\int_0^{z_s} \\frac{3c\\chi_e\\Omega_b H_0}{8\\pi G m_p} \\frac{f_d(z)(1+z)\\,dz}{\\sqrt{\\Omega_m(1+z)^3+\\Omega_\\Lambda}}$, and the variance is $\\sigma^2[\\mathrm{DM}_{\\rm cosmic}(z_s)] = \\int_0^{\\chi_s} d\\chi\\, W_{\\rm DM}^2(\\chi) \\int_0^\\infty \\frac{k\\,dk}{2\\pi} P_{ee}(k,z(\\chi))$, where $P_{ee}$ is the electron power spectrum, the Fourier-space clustering of the free-electron density contrast. The paper calibrates $P_{ee}(k,z)$ as a function of two cosmological and four feedback parameters using nearest-neighbor interpolation on 1,000 hydrodynamical simulations, then feeds the moments into a per-FRB likelihood that is sampled with MCMC to constrain the parameters and the suppression ratio.","core_discovery":"The central claim is that the log-normal parameterization $p(\\mathrm{DM}_{\\rm cosmic}|z)$, fully defined by its mean and variance, describes simulated FRB dispersion measures better than the standard $F$-parameterization, and that the variance computed from $P_{ee}(k,z)$ carries the feedback signal. Because the variance integral is dominated by scales out to $k\\sim10\\,h\\,\\mathrm{Mpc}^{-1}$, the scatter between FRB sightlines at fixed redshift responds to the same baryonic feedback that suppresses the matter power spectrum. Using the calibrated model, the paper reports that 10,000 FRBs would recover the true suppression ratio $P_{\\rm hydro}/P_{\\rm gravity-only}$ within 0.6, 2, 8, and 15 percent at $k=0.5,1,10,$ and $50\\,h\\,\\mathrm{Mpc}^{-1}$, and that the same model recovers the suppression in simulations with substantially different feedback physics.","pith_inferences":["Because the DM variance is an integral over $P_{ee}(k,z)$, it carries limited scale information; the paper's shape recovery relies on the correlation between feedback parameters and $P_{ee}$. An extension the authors do not make is to use the angular correlation of DM across neighboring sightlines, which would measure $P_{ee}(k,z)$ by scale and could break remaining degeneracies.","The paper reports strong degeneracies between $H_0$, $\\Omega_m$, and the feedback parameters; a natural next step, not taken here, is a joint FRB plus lensing plus kinematic/thermal Sunyaev-Zel'dovich analysis that could separate cosmological from feedback effects.","The authors note that roughly 10% of the variance comes from scales larger than their simulation box; re-running the calibration on a larger-volume suite would likely tighten the large-scale suppression constraints and test whether the reported percent-level numbers are optimistic."],"forward_implications":["With $10^4$ localized FRBs, the baryonic suppression ratio $P_{\\rm hydro}/P_{\\rm gravity-only}$ can be constrained to 0.6% at $k=0.5\\,h\\,\\mathrm{Mpc}^{-1}$ and about 8% at $k=10\\,h\\,\\mathrm{Mpc}^{-1}$, reaching the accuracy needed for upcoming weak-lensing cosmology.","Even today's roughly $10^2$ localized FRBs would constrain the suppression to 1-3% at large scales and 12-20% at $k\\sim10$-$50\\,h\\,\\mathrm{Mpc}^{-1}$.","The redshift evolution of the DM variance is included by construction, so the model avoids the assumptions behind the $F$-parameter (Poisson halo statistics and $\\sigma_{\\rm DM}=F z^{-1/2}$) that can bias feedback estimates by 20-40%.","The same calibrated model recovers the suppression in simulations with substantially different subgrid feedback physics, so it is not tied to one feedback prescription.","Constrained feedback parameters translate directly into predictions for baryon fractions in galaxy groups and clusters and for halo gas profiles within standard halo-model analyses."],"supporting_citations":[{"why":"Introduced the $F$-parameter DM distribution that this paper tests and replaces as the baseline feedback parameterization.","marker":"Macquart et al. (2020)"},{"why":"Provides the derivation of the DM variance from the electron power spectrum used in Equations 7-11.","marker":"Reischke & Hagstotz (2023)"},{"why":"Provides the 1,000-simulation calibration suite with varied cosmological and feedback parameters.","marker":"Villaescusa-Navarro et al. (2023)"},{"why":"Describes the Latin-hypercube and one-parameter-at-a-time datasets used for calibration.","marker":"Ni et al. (2023)"},{"why":"Supplies the simulated $p(\\mathrm{DM}_{\\rm cosmic}|z)$ distributions used to compare the $F$-parameter and log-normal fits.","marker":"Zhang et al. (2021)"},{"why":"Supplies the log-normal host-galaxy DM parameterization ($\\mu_{\\rm host}=5$, $\\sigma_{\\rm host}=0.5$) used to generate mock FRB samples.","marker":"Connor et al. (2024)"},{"why":"Provides the halo-model emulator used to quantify the $F$-parameter variance error and to evaluate halo-mass sensitivity.","marker":"Mead et al. (2021)"},{"why":"Establishes the theoretical connection between FRB dispersion-measure variance and the baryon distribution that underlies the whole approach.","marker":"McQuinn (2014)"}],"fun_headline_variants":["FRB sightline scatter constrains baryonic feedback to ~1%","Baryonic feedback measured from FRB DM scatter","FRB DM variance maps matter power suppression","Log-normal model turns FRB scatter into feedback probe","10,000 FRBs pin feedback to percent-level precision"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The inference depends on the assumption that the cosmic dispersion measure at a fixed redshift is exactly log-normal, which the paper concedes has no rigorous derivation; if the true distribution has different skewness or a heavier tail, the feedback constraints could be biased.","fun_headline_variants_meta":{"raw":{"variants":["FRB sightline scatter constrains baryonic feedback to ~1%","Baryonic feedback measured from FRB DM scatter","FRB DM variance maps matter power suppression","Log-normal model turns FRB scatter into feedback probe","10,000 FRBs pin feedback to percent-level precision"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000869,"raw_usage":{"total_tokens":3862,"prompt_tokens":1141,"completion_tokens":2721,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":757,"completion_tokens_details":{"reasoning_tokens":2642}},"tokens_in":757,"tokens_out":2721,"duration_ms":22173,"temperature":1.0,"reasoning_tokens":2642,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:10:53.457318+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take many mock FRB sightlines from a strong-feedback hydrodynamical simulation and measure the third and fourth moments of $\\mathrm{DM}_{\\rm cosmic}$ directly; if they disagree with the log-normal values implied by the fitted mean and variance, or if a simulation-based likelihood gives different posteriors on $P_{\\rm hydro}/P_{\\rm gravity-only}$ than the log-normal likelihood, the central claim is falsified.","supporting_citations":[],"review_version":1}