{"id":"3004f7af-f89d-4d9b-8c4d-003d326c9da9","arxiv_id":"2501.08718","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Pixelization biases in pseudo-C_ell cosmic shear estimators are modeled from first principles, including a spin-2 HEALPix window derivation and empty-pixel corrections.","lead":"The paper derives and tests analytic models for how pixelizing galaxy shear catalogues into HEALPix maps biases measured cosmic shear power spectra. It proposes corrections, including new window functions and an interlacing scheme, to keep residual biases at the percent level for upcoming surveys.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Source-clustering assumption undermines direct applicability of the claimed first derivation; paper's own §3.6 defers this.","rationale":"The paper's strongest claim is the first derivation of the HEALPix pixel window approximation for discrete sources and spin-2 fields on the curved sky. That derivation is mathematically coherent under its stated assumptions: unclustered sources, weights independent of shear and density, and axisymmetric pixels on average. The simulation tests in §4 support the resulting model for Poisson-distributed source positions. The load-bearing weakness is not an internal inconsistency but a scope mismatch: the assumptions that make Eq. (3.12) exact are known to fail for real cosmic shear catalogues. Section 3.4 explicitly states that the weights are not independent of shear or density in reality, and §3.6 defers source clustering and source-lens clustering to a forthcoming work. The reader's CONDITIONAL verdict already captures this limitation, and the paper itself flags it rather than hiding it. The recommended fixed-position estimator of §6 is a plausible route around some of these issues, but it is introduced as an ansatz rather than derived from the central Poisson-based calculation, and its validation again uses unclustered sources. Therefore the central claim should be read as a derivation for Poisson catalogues; its extension to realistic clustered catalogues is not yet supported. No change to the reader's verdict is needed.","tokens_in":43926,"tokens_out":11170,"duration_ms":121844,"concrete_test":"Run the §4 simulation pipeline with source positions drawn from a clustered density field (e.g., a lognormal realization with galaxy bias b≈1.5, or subhalo/galaxy catalog from an N-body simulation) at the same mean occupancy λ=2 on Nside=2048, keeping unit weights and zero shape noise. Apply the standard estimator, subtract the model of Eqs. (3.28)–(3.29) with the HEALPix window, and measure the mean fractional residual for l ≤ Nside. If residuals are consistent with the Poisson-simulation residuals (Figure 6), the assumption is benign; if they exceed the claimed sub-percent budget, the Poisson-based derivation does not extend to clustered sources and the first-derivation claim must be restricted accordingly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central derivation of the HEALPix window approximation in §3.3 rests on Eq. (3.12): after averaging over source positions, the weighted source density is replaced by its uniform per-pixel expectation, N_p w^{(1)}(p)/Ω. This is exact only for unclustered (Poisson) sources with weights independent of position and of the shear. For real cosmic shear catalogues, source positions are strongly clustered and correlated with the shear through source-lens clustering; hence ⟨n_w(hat n)⟩ is not uniform, the azimuthal simplification that yields W^p_l from d^l_{2,2} no longer follows, and the 'first derivation' does not describe the actual transfer function of the standard estimator. The paper explicitly acknowledges this in §3.6, deferring clustered sources and SLC to a forthcoming work. The fixed-source estimator of §6 is a separate ansatz (Eq. 6.2), not derived from Eq. (3.12), and its validation in §6.2 still uses Poisson-distributed source positions. Thus the quantitative claims—sub-percent residual biases and the f(λ), g(λ), h(λ) corrections of Eqs. (3.28)–(3.29)—are not established for clustered catalogues, which limits the central claim's relevance to actual survey data.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper analyzes biases in the standard pseudo-Cℓ estimator of cosmic shear power spectra, caused by pixelization of discrete galaxy catalogues into HEALPix maps. The authors derive analytic expressions for the action of pixelization on spin-2 shear fields, including the HEALPix pixel window approximation, the effects of finite pixel occupancy and shot noise, empty-pixel masks, and a phase factor from parallel transport. They validate their models against 1000 Gaussian shear simulations with Poisson-distributed source positions, and propose improvements: unnormalized or globally normalized shear maps, a fixed-source estimator, interlaced HEALPix grids, and footprint-adapted pixel window functions. The central claim is a first derivation of the HEALPix pixel window function approximation for discrete, spin-2 sources on the curved sky.","tokens_in":44147,"tokens_out":8602,"duration_ms":85053,"significance":"If correct, the paper fills a real gap in the weak-lensing literature: the standard HEALPix window function for spin-2 fields is widely applied but its derivation in the discrete-catalogue setting was not previously published. The analytic models are derived from first principles, with no parameters fitted to the simulation outcomes; the interlacing rotation is fixed by symmetry, and the HEALPix window is an external input. Validation with 1000 simulations gives sub-percent agreement to ℓ≈Nside for the occupation-number corrections. The paper is also honest about its limits: §3.6 explicitly defers clustered sources and source-lens clustering, and §7 cautions that interlacing is spectrum-dependent. These scoping statements are commendable and should be preserved in any revision.","major_comments":[{"comment":"The abstract claims that the paper 'derive[s] models that can bring residual biases to the percent level on small scales', and §3.3 states that Eq. (3.23) provides 'for the first time' the HEALPix pixel window approximation for discrete sources and spin-2 fields. Both statements rest on Eq. (3.12), which replaces the weighted source density with its uniform per-pixel expectation and is exact only for unclustered (Poisson) sources with weights independent of position and of the shear. As the paper itself states in §3.6, clustered sources and source-lens clustering are deferred to a forthcoming work, and the fixed-source tests in §6.2 also use Poisson-distributed positions. The quantitative sub-percent claims in Figs. 6 and 8 are therefore not established for real survey catalogues, whose source positions are strongly clustered and correlated with the shear through source-lens clustering. I request an explicit scope statement in the abstract and in the introduction to §6.2, and a qualifier in §3.3 clarifying that the first derivation is for unclustered sources; with that qualification the derivation appears sound.","section":"Abstract; §3.3; §3.6; §6.2"}],"minor_comments":[{"comment":"The word 'prediciton' should be 'prediction'.","section":"Figure 4 caption"},{"comment":"The heading 'F ull sky , unit shear weights' contains stray spaces and should be 'Full sky, unit shear weights'.","section":"§3.4.2"},{"comment":"The symbol γ is used both for the Euler–Mascheroni constant and for the lower incomplete Gamma function γ(α, x); please disambiguate this notation.","section":"Eq. (3.26)"},{"comment":"The legend entries 'HEALPix window' and 'masked window' are not self-explanatory; please state in the caption that 'masked window' is the footprint-averaged (bespoke) pixel window, i.e. Eq. (8.1) restricted to the equatorial or polar pixels.","section":"§8, Fig. 12"}],"recommendation":"minor_revision","confidential_remarks":"To the editor: The manuscript is a thorough and well-scoped methods paper. The only substantive issue is the abstract's generality on source clustering; the authors already acknowledge the limitation in §3.6, so the required revision is local. I see no concerns about novelty or fit with JCAP. The paper will be a useful reference for ongoing and forthcoming lensing surveys."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hall and Tessore have written a legit reference paper for pixelization effects in cosmic shear. The new thing is the spin-2 derivation of the HEALPix pixel window function for discrete sources on the curved sky, plus several practical estimator improvements. The derivation in §3 is self-contained and the 1000-simulation tests back the Poisson-source models to sub-percent on large scales. They also give a clean treatment of empty-pixel occupancy corrections, shape noise subtraction, and a sensible fixed-source estimator that deconvolves the source density field.\n\nWhere are the soft spots? The stress-test note is right, and the paper is honest about it: §3.6 explicitly defers clustered sources and source-lens clustering to a forthcoming work. Equation (3.12) assumes the weighted source density is uniform in each pixel, which only holds for Poisson sources with weights independent of position and shear. Real lensing catalogues violate that badly. So the 'first derivation' is exact for an idealized catalogue, not for the actual transfer function of the standard estimator on survey data. The quantitative claims—sub-percent residuals, the f, g, h corrections—are therefore not established for real data. That is a serious limitation, but it is stated clearly, so it does not mislead. The interlacing scheme is honestly presented as spectrum-dependent, and the footprint-adapted windows only beat the isotropic window in the band-limited case, which shear is not. No code or data are shipped, so exact reproduction requires reimplementing the expressions.\n\nThe paper does what it says: it sorts out the Poisson case rigorously and gives the community a common set of formulae. The cited companion paper [20] takes the fixed-source estimator into the Euclid pipeline, so this one can be read as the analytic foundation. The fixed-source ansatz in §6 is not fully derived from first principles—it is a model—but their test against fixed-position simulations is a reasonable check.\n\nWho is this for? Anyone building a pseudo-C_l pipeline for Stage-IV lensing surveys, and anyone who wants to understand what the HEALPix polarization window is actually doing when applied to discretely sampled spin-2 fields. It deserves a serious referee; my own verdict would be accept after the authors make the scope of the Poisson assumption even more prominent and perhaps upgrade the fixed-source estimator from ansatz to derivation. I would cite it.","headline":"A genuinely new spin-2 HEALPix window derivation, rigorously tested on Poisson sources, but its own §3.6 concedes the clustering assumptions that keep it from being directly survey-ready; still worth serious refereeing.","tokens_in":44658,"tokens_out":2195,"would_cite":true,"duration_ms":24113,"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 gives the first derivation of the HEALPix pixel window function for spin-2 cosmic shear spectra from discrete galaxy catalogues, and supplies correction models that hold pixelization biases below one percent up to roughly the…","keywords":["cosmic shear","weak lensing","pseudo-Cℓ power spectra","HEALPix pixel window function","aliasing","galaxy survey systematics","spin-2 fields","parallel transport"],"falsifier":"Build mock shear catalogues from clustered source galaxies with known two-point clustering at Euclid-like depth, construct shear maps at $N_{\\rm side} = 2048$ (mean occupancy $\\lambda \\approx 2$), and compare the measured $EE$ and $BB$ spectra against the model of Equation (3.28) including $f(\\lambda)$, $g(\\lambda)$, $h(\\lambda)$. If the residuals exceed the claimed sub-percent level for $\\ell < N_{\\rm side}$, or if the predicted $B$-mode amplitude fails to track the data, the Poisson-source assumption carrying the derivation is falsified.","tokens_in":43721,"feed_emoji":"🔭","tokens_out":18746,"duration_ms":163376,"temperature":0.7,"pith_summary":"Cosmic shear surveys turn millions of galaxy shape measurements into pixelized maps, then read cosmological parameters off the angular power spectra of those maps. This paper argues that the pixelization step injects several distinct biases — the pixel window smooths the signal, empty pixels behave like a stochastic mask, sub-pixel shear modes alias into the measured multipoles, and pixel shapes and sizes vary across the sky — and that these effects matter precisely on the small scales ($\\ell \\sim N_{\\rm side}$ and above) where Stage-IV lensing surveys carry most of their information. Its central contribution is the first derivation of how the HEALPix pixel window function approximation arises for spin-2 fields sampled by discrete source catalogues on the curved sky, showing both when the standard multiplicative smoothing factor is justified and what it leaves behind. On that foundation the paper builds correction factors $f(\\lambda)$, $g(\\lambda)$, $h(\\lambda)$ for empty pixels and weight stochasticity, tests the full model on 1000 mock shear catalogues, and proposes estimator changes — fixed source positions and weights, globally normalized maps, interlaced grids, footprint-adapted window functions — that hold residual biases at the sub-percent level for $\\ell \\lesssim N_{\\rm side}$. If the models are right, these are exactly the corrections ongoing and forthcoming lensing surveys need before trusting their small-scale spectra.","feed_headline":"Cut to sub-percent: cosmic shear's pixel-size biases modeled","feed_subtitle":"Empty pixels, aliasing, and pixel shape all bias shear spectra; new corrections hold them below one percent for Euclid.","key_machinery":"The load-bearing object is the pixel-averaged spin-weighted spherical harmonic $\\pm2\\Upsilon^p_{\\ell m}$, obtained by integrating the shear over a pixel window weighted by the source catalogue. The argument rotates this integral into a frame aligned with the pixel centre, where axisymmetric pixels kill all but the $m=\\pm 2$ azimuthal terms, so that the source-averaged pixel window reduces to $W^p_\\ell = (2\\pi/\\Omega)\\int_0^\\pi d\\theta\\,\\sin\\theta\\, W_p(\\theta)\\, d^\\ell_{22}(\\theta)$, and the sky average of $(W^p_\\ell)^2$ is the familiar HEALPix window $W^2_\\ell$. The same source-averaging procedure produces the occupancy functions $f(\\lambda)$, $g(\\lambda)$, $h(\\lambda)$ that rescale signal, shot noise, and shape noise when pixels can be empty. Alongside this, the recommended estimator treats galaxy positions and weights as fixed, folding the binary occupancy map $\\Theta(N_p)$ into the pseudo-$C_\\ell$ mixing matrix, and the appendix on parallel transport supplies the phase factor $e^{-2i\\beta}$ that must rotate each galaxy's shear to its pixel centre before averaging.","core_discovery":"On the paper's own terms, the central discovery is that the HEALPix pixel window function for cosmic shear is a derived consequence rather than an assumed kernel: writing the pixelized spin-weighted spherical harmonics exactly, and then averaging over unclustered source positions while treating pixels as axisymmetric on average, collapses pixelization to a single scalar window $W_\\ell$ multiplying each multipole, with the residual sub-pixel shear variance appearing as a renormalized white-noise term. Equation (3.23), built from the per-pixel window $W^p_\\ell = (2\\pi/\\Omega)\\int_0^\\pi d\\theta\\,\\sin\\theta\\, W_p(\\theta)\\, d^\\ell_{22}(\\theta)$ sky-averaged as $W^2_\\ell$, is the first statement of this approximation for discrete spin-2 catalogues on the curved sky. The same machinery yields the occupancy corrections $f(\\lambda)$, $g(\\lambda)$, $h(\\lambda)$ of Equation (3.29), which turn the empty-pixel effect into an effective sky fraction and a rescaled shot noise, and simulations with 1000 mock catalogues show the corrected model reproduces measured $E$- and $B$-mode spectra at the sub-percent level up to $\\ell \\approx N_{\\rm side}$, with the residual dominated by aliasing. A further new result quantifies the bias from omitting the parallel-transport phase $e^{-2i\\beta}$ when averaging shears within pixels, finding a spatially varying multiplicative and additive bias that is acceptable for Stage-IV requirements at $N_{\\rm side} \\geq 128$.","pith_inferences":["The Poisson-source assumption is the boundary of the paper's quantitative claims: with clustered galaxies the corrections acquire extra $\\ell$-dependence and extra $B$-mode power, so a natural next test is whether the $f(\\lambda)$, $g(\\lambda)$, $h(\\lambda)$ model still holds at $\\lambda \\approx 2$ for Euclid-like clustering — I would expect new scale-dependent terms to be required.","The interlacing scheme is inherently equatorial, because it exploits the constant azimuthal spacing of pixels within HEALPix rings; its benefit should therefore scale with the fraction of a survey's area in equatorial rings, a testable prediction for surveys with different footprints that the paper leaves implicit.","The fixed-source 'visibility map' viewpoint connects pixelization bias directly to source-lens clustering: the same mixing-matrix machinery that absorbs the occupancy map could in principle absorb the clustered-source terms the paper defers, so the promised follow-up work may be a direct extension of Equation (6.2).","The sub-pixel shear-gradient variance identified in appendix B injects white noise that depends on the cosmology through the shear spectrum; the distinction between noise estimators such as Equations (3.39) and (3.40) is therefore not merely practical, and getting this term into models will matter more as survey depth and resolution grow."],"forward_implications":["Residual biases in standard pseudo-$C_\\ell$ shear spectra can be modelled to the sub-percent level for $\\ell \\lesssim N_{\\rm side}$, with the remaining bias from aliasing growing to several percent by $\\ell \\simeq 2N_{\\rm side}$.","Dividing the shear map by the per-pixel weight sum is the main source of analytic difficulty; unnormalized or globally normalized maps make the leading empty-pixel bias exactly $\\lambda^2$ and the noise bias a simple $\\sigma^2_\\gamma/2\\bar{n}$ term on the full sky.","Holding source positions and weights fixed converts the empty-pixel effect into a mixing matrix that can be computed from the weight map, and this realization-dependent model performs at least as well as the ensemble-averaged one in the paper's tests.","Rotating the HEALPix grid by $\\Delta\\phi = \\pi/4N_{\\rm side}$ and averaging the two maps suppresses most of the aliasing bias for blue spectra such as $C_\\ell \\propto \\ell^{-1}$, at the cost of oversmoothing redder spectra.","Averaging the pixel window over the survey footprint instead of the whole sky removes the percent-level window-function bias for band-limited spectra; for non-band-limited shear, aliasing must be mitigated first for this fix to pay off."],"supporting_citations":[{"why":"Supplies the HEALPix grid, the quadrature scheme, and the spin-0 pixel window function whose spin-2 analogue the paper derives.","marker":"[28]"},{"why":"Defines the standard pseudo-$C_\\ell$ estimator for shear catalogues and the dense-versus-sparse pixel regimes that frame the bias analysis.","marker":"[14]"},{"why":"Companion paper that implements and tests the recommended fixed-source, globally normalized estimator in the Euclid analysis pipeline.","marker":"[20]"},{"why":"Provides the simulation-based effective pixel window showing that near-future surveys will be sensitive to pixelization, the motivation the new models must answer.","marker":"[25]"},{"why":"Provides the tensor-spherical-harmonic rotation relations used to move the pixel-window integral into the pixel-centred frame.","marker":"[34]"},{"why":"Pseudo-$C_\\ell$ mixing-matrix formalism used to absorb empty-pixel and weight-fluctuation effects into the survey window.","marker":"[50]"},{"why":"Cartesian interlacing technique whose averaging of shifted grids the approximate HEALPix interlacing scheme transplants to the sphere.","marker":"[52]"},{"why":"Establishes the source-lens clustering effect that the paper's Poisson-source modelling omits and defers for later work.","marker":"[37]"}],"fun_headline_variants":["Derived pixel window fixes shear spectra to sub-percent","Pixelization biases in shear: model cuts to <1%","HEALPix shear: empty pixels and aliasing tamed","New window function for cosmic shear pixelization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Every quantitative correction in the paper assumes source galaxies are unclustered (Poisson-distributed) positions whose measurement weights are independent of the shear field and of source density; real catalogues violate both, and the paper itself defers a clustered-source treatment to a forthcoming work.","fun_headline_variants_meta":{"raw":{"variants":["Derived pixel window fixes shear spectra to sub-percent","Pixelization biases in shear: model cuts to <1%","HEALPix shear: empty pixels and aliasing tamed","New window function for cosmic shear pixelization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000191,"raw_usage":{"total_tokens":1413,"prompt_tokens":1083,"completion_tokens":330,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":699,"completion_tokens_details":{"reasoning_tokens":263}},"tokens_in":699,"tokens_out":330,"duration_ms":4024,"temperature":1.0,"reasoning_tokens":263,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:18:46.970607+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build mock shear catalogues from clustered source galaxies with known two-point clustering at Euclid-like depth, construct shear maps at $N_{\\rm side} = 2048$ (mean occupancy $\\lambda \\approx 2$), and compare the measured $EE$ and $BB$ spectra against the model of Equation (3.28) including $f(\\lambda)$, $g(\\lambda)$, $h(\\lambda)$. If the residuals exceed the claimed sub-percent level for $\\ell < N_{\\rm side}$, or if the predicted $B$-mode amplitude fails to track the data, the Poisson-source assumption carrying the derivation is falsified.","supporting_citations":[{"cited_title":"All-sky convolution for polarimetry experiments","cited_arxiv_id":"astro-ph/0008228","evidence_quote":"Provides the tensor-spherical-harmonic rotation relations used to move the pixel-window integral into the pixel-centred frame."},{"cited_title":"CMB temperature and polarisation pseudo-Cl estimators and covariances","cited_arxiv_id":"astro-ph/0410394","evidence_quote":"Pseudo-$C_\\ell$ mixing-matrix formalism used to absorb empty-pixel and weight-fluctuation effects into the survey window."}],"review_version":1}