{"id":"34589545-8ea9-4a46-bd77-e3fd8cbe893b","arxiv_id":"2609.04488","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"The first analytic derivation of the super-sample and in-box non-Gaussian covariance of the line-intensity mapping power spectrum, validated against painted [CII] N-body simulations and implemented in the public SSLimPy code.","lead":"This paper derives, for the first time, the analytic super-sample covariance of line-intensity mapping power spectra, the extra uncertainty that comes from the surveyed patch sitting in a large-scale overdensity or underdensity of the Universe. It also derives the in-box non-Gaussian covariance, validates both against painted N-body simulations, and releases a public code, SSLimPy, for computing LIM covariances.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"Eq. (4.6) corrects the jackknife in-box covariance with a term built from the same response function being validated, so the Fig. 3 agreement may be substantially by construction until that correction is independently checked.","rationale":"The reader's verdict is CONDITIONAL, and my analysis supports that verdict while shifting the emphasis. The reader's stated weakest assumption is the halo-mass-luminosity model with no response of L(M) to the background mode, with the jackknife correction mentioned as a deferred secondary issue. I regard the validation circularity as more load-bearing: the empirical claim that the SSC prediction is 'validated against painted N-body simulations' is the main support for the paper's central assertion, and the validation pipeline subtracts from the jackknife estimate a term constructed from the very response function whose accuracy is under test. Without a report of the relative size of that subtraction, the agreement in Fig. 3 does not constitute an independent confirmation of the SSC amplitude. This concern does not overturn the derivation itself: the theory is internally consistent, the tree-level truncation and halo-model limitations are honestly stated, the theory-only response curve is shown and matches the shape, and the external comparison with Ref. [50] supports the scaling behavior. The malformed code URL is a reproducibility issue but not a scientific flaw. Because the reader already assigned CONDITIONAL, and because the identified concern is a reason to keep that condition rather than to reject or accept outright, the verdict is unchanged.","tokens_in":32614,"tokens_out":4993,"duration_ms":45903,"concrete_test":"Report, for every k bin in the z=1 [CII] validation, the magnitude of the subtracted term in Eq. (4.6) relative to both the raw jackknife covariance and the subbox covariance, and re-run the Fig. 3 comparison with the correction term set to zero: plot Cov_subbox - Cov_jackknife_uncorrected against the predicted SSC (using both theory and fitted responses). If the subtraction is comparable to or larger than the SSC contribution, or if the uncorrected residual disagrees with the SSC prediction by more than the claimed 25-50% accuracy, the validation is not independent. Varying NJK (e.g., 8, 27, 125) would further check that the correction, rather than the jackknife estimator itself, drives the agreement.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central empirical support for the first-principles SSC claim is the validation in Section 4. The 'corrected jackknife' estimate of the in-box covariance, Eq. (4.6), subtracts (NJK-1) Var(delta_b^j - delta_b^i) times (dP_TT/ddelta_b) tensor (dP_TT/ddelta_b), using the same model response dP_TT/ddelta_b from Eq. (3.15) whose accuracy the validation is meant to establish. The SSC prediction added to this corrected jackknife is also quadratic in that same response. Thus the comparison in Fig. 3 tests a combination in which an error in the response appears with a coefficient set by the difference between the jackknife-subtraction amplitude and the SSC normalization; the two uses of the response do not cancel, but the procedure no longer provides an independent measurement of the SSC amplitude. The paper explicitly defers a systematic study of this correction ('we defer a systematic study of this bias in the standard jackknife covariance estimation for future work'), and the magnitude of the subtracted term relative to the raw jackknife covariance and to the SSC contribution is never reported. If that term is large on the scales where the SSC is claimed to matter (k >~ 0.1 Mpc^-1), the subbox-versus-prediction agreement could be mostly a consistency relation rather than a genuine validation. The theory-only curve (without bias fitting) already degrades to about 50% in two intermediate bins and overestimates the response at small scales, which is consistent with the model being imperfect; the unquantified jackknife correction is the step that could mask or amplify exactly such an error.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"This paper presents an analytic derivation of the full real-space covariance of the line-intensity mapping (LIM) power spectrum, with emphasis on the super-sample covariance (SSC) and the in-box non-Gaussian covariance. The SSC is obtained as the shell-averaged product of the response of the LIM temperature power spectrum to a background density mode (Eq. 3.15) times the variance of the background mode over the survey window (Eq. 2.18); the in-box contribution is the collapsed-trispectrum term computed within a halo model merged with standard perturbation theory (Section 3 and Appendix B). The model is validated against painted [CII] N-body simulations (Mithra LIMSims) by measuring the field-to-field variance of 512 subboxes, with bias parameters used either at their predicted values or fitted to the same simulations: agreement is at the few-percent level for the fitted model over most scales, degrading to about 30 percent (fitted) and 50 percent (theory-only) in two intermediate k-bins. The framework is then applied to [CII] and 21-cm lines, compared with the empirical field-to-field variance fits of Gkogkou et al. (2023), and used to quantify covariance contributions for MeerKLASS and CCAT-prime-like surveys. The public code SSLimPy implements the derived expressions.","tokens_in":33074,"tokens_out":15558,"duration_ms":121950,"significance":"The derivation is a useful and, to my knowledge, novel analytic contribution: previous LIM covariance estimates relied on Gaussian approximations, mocks, or jackknife-based methods, none of which capture the effect of super-survey modes, and the halo-model-plus-SPT route is well matched to the halo-mass-luminosity structure of LIM models. The paper is notably transparent: the theory-only and bias-fitted predictions are shown separately, the validation discrepancies (30-50 percent in two intermediate bins) are reported honestly, the main modeling assumptions are stated explicitly (Sections 3.2-3.3), and the deferred jackknife systematics study is flagged. The public code release and the external comparison with the Gkogkou et al. fits (good agreement for the area scaling, qualitative for the frequency scaling) are strengths. If the validation concern below is addressed, the results would justify the paper's practical recommendation that low-noise LIM analyses include the SSC and in-box non-Gaussian terms at intermediate and small scales.","major_comments":[{"comment":"The validation of the SSC is partially circular. The 'corrected jackknife' in-box covariance of Eq. (4.6) is obtained by subtracting (N_JK - 1) Var(delta_b^j - delta_b^i) times the response-function outer product, where the response is the same dP_TT/ddelta_b from Eq. (3.15) whose accuracy the validation is meant to establish; the SSC prediction then added is also quadratic in that same response. The comparison in Fig. 3 therefore tests Cov_jk^raw + (B - A)(dP_TT/ddelta_b tensor dP_TT/ddelta_b) against the subbox covariance, and its discriminating power depends on the relative size of the subtraction amplitude and the SSC normalization, which is never reported. Because the paper defers a systematic study of the jackknife bias, the agreement shown in Fig. 3 does not by itself establish the SSC amplitude. I request that the authors report, per k-bin, the magnitude of the subtracted term relative to both the raw jackknife covariance and the SSC contribution, and ideally calibrate the jackknife correction independently, for example with simulations whose in-box covariance is known by construction.","section":"Section 4, Eq. (4.6) and Fig. 3"},{"comment":"The abstract's claim of a 'successful validation' at high precision overstates what the first-principles model alone delivers. The parameters <T b1>, <T^2>, <T(b2 - 4bG2/3)>, <T^2 b1>, and eta are fitted to the very same Mithra LIMSims that serve as the validation benchmark (Table 1), so the few-percent agreement of the fitted curves partly measures the fit quality; the theory-only prediction degrades to about 50 percent in two intermediate bins and overestimates the small-scale response (Fig. 2, right panel). A test of genuine predictive power would be to apply the theory-only response at a redshift, line, or simulation not used in the fitting, or to validate against the CO maps of the same simulation suite; short of that, the abstract and conclusions should state that the model reproduces the simulations once nuisance parameters are calibrated to them.","section":"Section 4, Table 1; Abstract"},{"comment":"The 'measured' response function in Fig. 2 is not purely empirical: the background mode assigned to each subvolume is inferred from Eq. (4.2), delta_T_b = <T b1> delta_b, using the theoretical value of <T b1> (0.27 micro-K), while the fitted value is 0.23 micro-K (Table 1). Because the regression slope of P_TT on the inferred delta_b scales inversely with the conversion factor, a roughly 17 percent overprediction of <T b1> inflates the measured response relative to the true one, and this systematic is not propagated into the quoted agreement levels. Please state which value of <T b1> enters Eq. (4.2) and propagate this uncertainty into the response comparison.","section":"Section 4, Eq. (4.2) and Fig. 2"},{"comment":"The response function neglects any response of the luminosity-mass relation L(M) to the background mode, with the justification that environmental effects are 'negligible and effectively absorbed into the stochastic scatter'. No quantitative bound is given, yet the SSC is quadratic in the response, and the response itself differs by up to a factor of two between the [CII] and HI models considered in Section 5.1 (Fig. 4). Since environmental or assembly effects on line luminosity at fixed mass are poorly constrained, this assumption should be tested, for example by comparing the halo-model response with the separate-universe response of a galaxy-formation model that includes environment-dependent star formation or quenching.","section":"Section 3.2, Eq. (3.15)"}],"minor_comments":[{"comment":"The two intermediate k-bins in which the fitted (theory-only) predictions reach 30 (50) percent deviations are not identified by their wavenumber range; please give the k values.","section":"Section 4"},{"comment":"In Fig. 3 the left panel is the covariance diagonal and the right panel is a slice at fixed k2; please state the k2 value in the panel labels and caption and check it against the text, which discusses k2 = 0.3 Mpc^-1.","section":"Fig. 3"},{"comment":"The quantitative comparison of the frequency-scaling exponent beta should restate the caveat that the Ref. [50] subfields are not fully independent and that Delta_nu_proj has different meanings in the two analyses, since these caveats directly affect the interpretation of the beta difference.","section":"Section 5.2"},{"comment":"The shorthand moments <T b1>, <T^2>, <T(b2 - 4/3 bG2)>, and <T^2 b1> are used throughout Section 4 before being defined; a single defining sentence at the end of Section 3.2 would help the reader.","section":"Section 3.2"},{"comment":"The forecasts for MeerKLASS and CCAT-prime rely on the approximate F_0 treatment of anisotropic resolution for the in-box non-Gaussian term (Eq. C.9); the main text should state clearly which plotted quantities use this approximation.","section":"Appendix C and Section 5.4"}],"recommendation":"major_revision","confidential_remarks":"The derivation itself is standard, internally consistent, and the code release is a genuine service to the community. The main risk is that the abstract's novelty and validation claims outrun the evidence: the 'first derivation' assertion should be vetted against the closest prior work, and the validation is partly self-referential (Eq. 4.6 and Table 1 both use the model being tested). If the authors quantify the jackknife subtraction and either calibrate it independently or add a theory-only test on a line or redshift not used for fitting, the paper would be publishable in my view."},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Quick take: this is the first analytic derivation of the LIM power-spectrum SSC, it is basically correct, and the LIM-specific parts are new. What is genuinely new is Eq (3.15): the standard separate-universe response recast with luminosity-weighted halo-model integrals, including the shot-noise response term I2_1. The derivation is internally consistent, and the argument that LIM differs from galaxy clustering because absolute intensities remove the beat-coupling/local-average cancellation is correct and matters for the field. The volume-scaling result (V sigma_V^2 varies by only a factor of two over four orders of magnitude in V) and the factor-of-two line dependence between [CII] and HI are useful for survey design. The external check against Gkogkou et al. field-to-field fits, area exponent gamma = -0.44 versus their -0.45, supports the mechanism. The paper is honestly written: the authors flag the tree-level truncation, the halo-model intermediate-scale deficit, and the assumption that L(M) does not respond to the background mode, and they explicitly defer a systematic study of the jackknife bias.\n\nNow the soft spots, in proportion. The Section 4 validation is partly circular. The bias parameters in Table 1 are fitted to the Mithra LIMSims that serve as the validation benchmark, so the few-percent agreement on most scales is partly a measure of the fit. More concerning, the jackknife correction in Eq (4.6) subtracts a term quadratic in the same model response function whose accuracy the validation is meant to establish, and its magnitude relative to the raw jackknife covariance is never reported. The stress-test note is right: if that term is large where the SSC is claimed to matter (k greater than about 0.1 Mpc^-1), the Fig 3 agreement could be substantially a consistency relation rather than an independent check. The theory-only curve degrading to about 50 percent in two intermediate bins and overestimating at small scales is consistent with a somewhat imperfect model, not a wrong one.\n\nNone of this is fatal. The derivation stands on its own, the shape of the response matches, and the external scaling comparison is genuinely independent. But the authors should report the theory-only SSC comparison prominently, quantify or drop the jackknife correction, and propagate model uncertainty into the Section 5.4 forecasts. Also, the code URL as printed looks malformed and should be corrected.\n\nThis paper is for anyone modelling LIM covariances for [CII], CO, or 21cm surveys, and the response-function and scaling results are citable. It deserves a serious referee; I would engage, with the validation questions front and centre.","headline":"First analytic LIM SSC derivation that is basically sound, but the few-percent validation is weaker than it looks because the same model response is subtracted in the jackknife correction and added as the SSC in the same comparison.","tokens_in":33663,"tokens_out":4528,"would_cite":true,"duration_ms":40341,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"First derivation of LIM power spectrum super-sample covariance.","keywords":["line-intensity mapping","super-sample covariance","halo model","standard perturbation theory","non-Gaussian covariance","CII","21cm","field-to-field variance"],"falsifier":"A direct test would compare the SSC predicted by this model for a given survey and line against the measured field-to-field variance from a large-volume lightcone simulation that includes realistic galaxy formation physics (e.g., hydrodynamic or semi-analytic models) allowing luminosity to vary with environment. If the simulated SSC deviates systematically from the analytic prediction by more than the validation margin (the paper finds ≤30% at intermediate scales), the assumption of no environmental luminosity response would be falsified. Additionally, a future high-signal-to-noise LIM observation (e.g., with CCAT-Prime) could measure the power spectrum in multiple independent fields and test whether the scatter between fields matches the total covariance including SSC.","tokens_in":39730,"feed_emoji":"📡","tokens_out":1671,"duration_ms":101623,"temperature":0.7,"pith_summary":"This paper provides the first analytical derivation of the super-sample covariance (SSC) for the line-intensity mapping (LIM) power spectrum, along with the non-Gaussian in-box covariance. The SSC arises from the uncertainty of whether the survey volume sits in a large-scale overdensity or underdensity, a contribution that cannot be estimated from the data alone. The authors combine the halo model with standard perturbation theory to compute the power-spectrum response to background modes, validating their predictions against painted N-body simulations. They find that the SSC is crucial at intermediate and small scales for low-noise surveys, and that its relative importance does not diminish with survey volume. If correct, LIM analyses that neglect the SSC risk underestimating errors and biasing inference of global cosmological and astrophysical parameters.","feed_headline":"First LIM super-sample covariance from first principles","feed_subtitle":"Neglecting it biases global parameter inference for low-noise surveys like CCAT-Prime and MeerKLASS.","key_machinery":"The central machinery is the LIM halo model extended with standard perturbation theory (SPT) to compute the nonlinear LIM power spectrum and its response to a large-scale background mode. The key object is the response function $dP_{\\rm TT}/d\\delta_b$ given in Eq. (3.15), which combines the growth-only response, the dilation term from the local expansion rate, and the shot-noise response from the one-halo term. This response function is built from luminosity-weighted bias integrals $I^\\alpha_\\beta$ that depend on the halo mass–luminosity relation and the halo profile. The SSC itself is then given by Eq. (2.18) as the variance of the background mode over the survey window times the product of response functions at each wavenumber.","core_discovery":"The paper claims to derive, for the first time, the real-space LIM power spectrum super-sample covariance (SSC) from first principles, using a hybrid halo model and standard perturbation theory framework. As a byproduct, it also derives the non-Gaussian in-box covariance. The SSC is expressed as a product of the power-spectrum response function, which quantifies the shift in the LIM power spectrum due to a smooth background density perturbation, and the variance of the background mode over the survey window. The response function is computed from luminosity-weighted halo bias integrals and SPT mode-coupling kernels. The model is validated against painted N-body simulations for [CII] at z=1, achieving agreement at the few-percent level over most scales, and degrading to ≤30% on two intermediate-scale bins when bias parameters are fitted to simulations. Forecasts for MeerKLASS and CCAT-Prime show that the SSC dominates the covariance at intermediate and small scales when instrumental noise is subdominant, and that it introduces sizeable off-diagonal correlations between Fourier modes.","pith_inferences":["The framework can be extended to include redshift-space distortions and anisotropic resolution, which may significantly alter the off-diagonal covariances and affect constraints on growth of structure.","The assumption that line luminosity does not respond to the background mode (no environmental dependence) could be tested by comparing SSC predictions from separate-universe simulations that allow luminosity to vary with environment; if environmental effects are non-negligible, the response function would change, altering the SSC.","The same formalism could be applied to cross-power spectra between different lines (e.g., [CII] x CO or [CII] x 21 cm) to build multi-tracer covariance matrices, where the SSC may correlate different line signals.","The jackknife correction used in the validation (Eq. 4.6) introduces a mild systematic that could be studied further to improve empirical estimation of in-box covariance from simulations."],"forward_implications":["LIM surveys that neglect the SSC will underestimate their covariance at intermediate and small scales, leading to overconfident parameter constraints and potentially biased inference of global quantities.","For low-noise experiments such as CCAT-Prime and MeerKLASS L-band, the SSC contributes significantly to the diagonal and off-diagonal covariances, requiring its inclusion for accurate likelihood analysis.","The relative importance of the SSC is roughly independent of survey volume because both Gaussian and SSC contributions scale similarly with volume, making the SSC relevant even for large surveys at small scales.","The SSC amplitude depends on the specific spectral line and its mass–luminosity relation: lines that trace more massive halos exhibit larger response functions and thus larger SSC.","The public code SSLimPy released with the paper allows the full LIM covariance including SSC to be computed for any survey geometry and spectral line, facilitating rigorous forecasts and data analyses."],"supporting_citations":[{"why":"Established the super-sample covariance formalism for galaxy clustering, which is extended here to LIM.","marker":"[41]"},{"why":"Provides the only prior empirical study of field-to-field variance in LIM, used for comparison in Section 5.2.","marker":"[50]"},{"why":"Presents the painted N-body simulations and halo model framework used for validation in Section 4.","marker":"[54]"},{"why":"Derives the galaxy power spectrum multipoles covariance including SSC in perturbation theory, providing the methodological template for this work.","marker":"[25]"},{"why":"Foundational halo model review used to describe the LIM power spectrum and its two- and four-point functions.","marker":"[51]"},{"why":"Provides the standard perturbation theory kernels (F_n, G_n) used to compute the bias expansion and response function.","marker":"[53]"}],"fun_headline_variants":["First-principles SSC for line-intensity maps","Neglecting field-to-field variance biases LIM surveys","LIM power spectrum covariance gains first SSC term","SSC dominates LIM covariance at small scales","First LIM super-sample covariance from halo model"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The load-bearing premise is that line luminosity depends only on halo mass, with no additional response to the large-scale background density (i.e., no environmental dependence of the mass–luminosity relation). If real or simulated line emissions vary with environment beyond what is captured by the halo mass function and bias, the response function and hence the SSC amplitude would be misestimated.","fun_headline_variants_meta":{"raw":{"variants":["First-principles SSC for line-intensity maps","Neglecting field-to-field variance biases LIM surveys","LIM power spectrum covariance gains first SSC term","SSC dominates LIM covariance at small scales","First LIM super-sample covariance from halo model"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000203,"raw_usage":{"cost":0.00020314,"is_byok":false,"cost_details":{"upstream_inference_cost":0.00020314,"upstream_inference_prompt_cost":0.00017262,"upstream_inference_completions_cost":3.052e-05},"total_tokens":1342,"prompt_tokens":1233,"completion_tokens":109,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":0,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":56}},"tokens_in":1233,"tokens_out":109,"cost_source":"openrouter_usage","duration_ms":2818,"temperature":1.0,"reasoning_tokens":56,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-09-08T18:49:22.311505+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"falsifier":"A direct test would compare the SSC predicted by this model for a given survey and line against the measured field-to-field variance from a large-volume lightcone simulation that includes realistic galaxy formation physics (e.g., hydrodynamic or semi-analytic models) allowing luminosity to vary with environment. If the simulated SSC deviates systematically from the analytic prediction by more than the validation margin (the paper finds ≤30% at intermediate scales), the assumption of no environmental luminosity response would be falsified. Additionally, a future high-signal-to-noise LIM observation (e.g., with CCAT-Prime) could measure the power spectrum in multiple independent fields and test whether the scatter between fields matches the total covariance including SSC.","supporting_citations":[{"cited_title":"CONCERTO: Simulating the CO, [CII], and [CI] line emission of galaxies in a 117 $\\rm deg^2$ field and the impact of field-to-field variance","cited_arxiv_id":"2212.02235","evidence_quote":"Provides the only prior empirical study of field-to-field variance in LIM, used for comparison in Section 5.2."},{"cited_title":"Precision Tests of CO and [CII] Power Spectra Models against Simulated Intensity Maps","cited_arxiv_id":"2111.03717","evidence_quote":"Presents the painted N-body simulations and halo model framework used for validation in Section 4."},{"cited_title":"Galaxy Power Spectrum Multipoles Covariance in Perturbation Theory","cited_arxiv_id":"1910.02914","evidence_quote":"Derives the galaxy power spectrum multipoles covariance including SSC in perturbation theory, providing the methodological template for this work."},{"cited_title":"Bernardeau, S","cited_arxiv_id":null,"evidence_quote":"Provides the standard perturbation theory kernels (F_n, G_n) used to compute the bias expansion and response function."}],"review_version":1}