Pith. Sign in

REVIEW

Extracting non-Gaussian information of line-intensity maps with weighted skew spectra

PaperPDF from arXiv

data files

ReviewThe objections and verdict, as JSON Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
Eleonora Vanzan, Azadeh Moradinezhad Dizgah, Ely D. Kovetz

Cosmology and Nongalactic Astrophysics · Submitted 29 Oct 2026 · 130 references

Pith's verdict: Careful, honest extension of weighted skew spectra to line-intensity maps; the compression claim is internally solid, but the per-bin bias-ratio headline is conditional on tree-level model fidelity that no simulation validates.

T0 · AI review

How this review was made: An AI referee (deepseek/deepseek-v4-flash) read the paper on October 30, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

  1. T0 AI review An AI read the paper and answered a fixed list of questions.
  2. T1 Audit Automatic checks that the references and files the paper relies on exist and match.
  3. T2 Compute The paper's code or calculations re-run with the same results.
  4. T3 Formal statement The claims written out in a formal math language a computer can read.
  5. T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 2 major objections · 6 minor · Ask about this paper · The compiled record · Read the full review →
Challenge this review Comment (0) Read on arXiv ↗
Cite ▾
arXiv ID2610.40063 (version 1)
Pith IDZLDQJU7F, Pith's permanent record of this paper and its review
CitationAzadeh Moradinezhad Dizgah, Eleonora Vanzan, Ely D. Kovetz. "Extracting non-Gaussian information of line-intensity maps with weighted skew spectra." arXiv:2610.40063 (2026). https://pith.science/paper/ZLDQJU7F
Cite this reviewBibTeX for Pith's review of this paper
Claim paper

Why Pith says this

Highlight any text to comment on it.

Referee report

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

On its own terms, this paper establishes that the full Fisher information of the tree-level bispectrum of a line-intensity map — the luminosity-weighted biases, the shot noise, and the cosmological parameters — is carried by just four…

The reading

This paper argues that the information content of the full tree-level bispectrum of a line-intensity map can be captured by just four skew spectra — power-spectrum-like cross-correlations of the observed intensity field with filtered squares of itself. For a SPHEREx-like survey of H$\alpha$ emission, the authors show that three clustering-matched kernels plus one kernel tuned to the shot noise recover the bispectrum's Fisher constraints on the luminosity-weighted bias parameters, the shot noise, and the cosmological parameters to within a few percent. The second, more original claim is astrophysical: combined with the power spectrum, the skew spectra measure the ratios of the second-order biases to the linear bias ($b_2/b_1$ and $b_{G_2}/b_1$) in each redshift bin, quantities that are invariant under the unknown mean intensity and so require no model for the emitters. Because those ratios are not accessible to the tree-level power spectrum alone, they offer a direct test of whether line emission depends on halo properties beyond mass — assembly history or environment. An all-sky SPHEREx-like configuration could detect departures of order 0.1 in these ratios at $3\sigma$ for $z \lesssim 3$.

What carries the argument

The central object is the weighted skew spectrum, the cross-spectrum of the observed intensity-fluctuation field with a quadratic composite field built from two filtered copies of itself, $S_n(k)=\int \frac{d^3q}{(2\pi)^3} D_2^{(n)}(\mathbf{q},\mathbf{k}-\mathbf{q})\,\delta_{\mathrm{obs}}(\mathbf{q})\delta_{\mathrm{obs}}(\mathbf{k}-\mathbf{q})$. The mode-coupling kernels $D_2^{(n)}$ mirror the three separable templates of the tree-level bispectrum of a biased tracer — $L_0$, $F_{12}L_1$, and $L_2$ — and are inverse-variance weighted by the observed power spectrum so each skew spectrum approximates the maximum-likelihood estimator of the corresponding template amplitude. The LIM extension adds a fourth kernel matched to the shot-noise templates, orthogonalized against the clustering set, whose distinct $\alpha$-to-$b_1$ response ratio (larger by a factor of about 1.7) supplies the response direction the clustering kernels lack. The argument's leverage comes from the two bias ratios $b_2/b_1$ and $b_{G_2}/b_1$, which are invariant under the mean-intensity–bias rescaling $\langle I\rangle\to\lambda\langle I\rangle$, $\{b_1,b_2,b_{G_2}\}\to\{b_1,b_2,b_{G_2}\}/\lambda$, so they are the per-bin observables that probe the halo–emission connection without a model for the emitters.

Extended reading notes

Core claim

On its own terms, the paper establishes two results. First, the compression of the bispectrum into weighted skew spectra survives the transition from galaxy clustering to line-intensity mapping: with four kernels — three matched to the $L_0$, $F_{12}L_1$, and $L_2$ templates of the tree-level biased bispectrum, plus a fourth shot-noise-matched kernel — four skew spectra reproduce the full-bispectrum Fisher constraints on the luminosity-weighted biases to within about 3%, on the shot noise to within about 1%, and on cosmological parameters to better than about 2%. The fourth kernel is needed because a single SPHEREx deep-field redshift bin is too small (fundamental mode $k_f \simeq 0.01\,h/\mathrm{Mpc}$) to separate the shot-noise amplitude from the biases through the $k$-dependence of three kernels alone; the deficit is volume-set, not LIM-specific, and closes when the shot-sector kernel is added. Second, the joint power-spectrum plus skew-spectra analysis measures the ratios $b_2/b_1$ and $b_{G_2}/b_1$ per redshift bin without any model for the emitters, because the ratios are exactly invariant under the rescaling that leaves the unknown mean intensity $\langle I\rangle$ and the linear bias $b_1$ degenerate. In the all-sky configuration, the evolution of these ratios locates the sign change of $b_2$ and distinguishes the published tidal-bias relations at high significance for $z \lesssim 3$, providing a test of whether the line luminosity depends on secondary halo properties such as assembly history or environment. These ratios are inaccessible to the tree-level real-space power spectrum alone.

Load-bearing premise

The headline sensitivities and the 'within a few percent' compression comparison presuppose that the observed field is exactly the modeled signal — tree-level real-space perturbation theory up to $k_{\mathrm{max}} = 0.15\,h/\mathrm{Mpc}$, Gaussian instrumental noise of known power, and no beam smoothing, survey window, interlopers, or continuum residuals — and the authors flag these idealizations in Sections IIIA, IIIC, and V; if any fails, the quoted bias-ratio sensitivities and the recovery percentages could change.

Editorial extensions

If this is right

  • Four skew spectra, three clustering kernels plus the shot-noise-matched orthogonalized kernel, replace the full tree-level bispectrum in a line-intensity-map Fisher analysis, recovering its constraints on biases, shot noise and cosmological parameters to within a few percent.
  • The joint power-spectrum plus skew-spectra analysis measures the emitter bias ratios $b_2/b_1$ and $b_{G_2}/b_1$ per redshift bin, free of the unknown mean intensity and of any model for the emitters' redshift evolution.
  • An all-sky SPHEREx-like configuration detects order-0.1 departures in these ratios at $3\sigma$ below $z \approx 3$, distinguishing the published halo bias relations and locating the sign change of $b_2$ — a test of whether line emission depends on secondary halo properties beyond mass.
  • The same per-bin joint analysis yields a real-space, model-free measurement of $\sigma_8(z)$ across $0.4 \lesssim z \lesssim 4.6$, complementary to redshift-space galaxy surveys.
  • The compression holds when cosmological parameters are varied and when luminosity moments are left free (within 4.5% for all parameters and 2% for the moments), so the result is not tied to a fixed emission model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A practical rule for survey design follows: the required number of kernels tracks the number of distinct response directions in the parameter space, which grows as the survey volume shrinks; a simulation-based check on larger-volume maps (or wider redshift bins) would test whether the three-kernel set suffices there, as it does for galaxy surveys.
  • Because the forecast is deliberately conservative — the regularized four-kernel covariance can only discard information, so the true compression loss is bracketed between the bispectrum bound and the reported few-percent gap — a full-covariance simulation would likely tighten, not weaken, the equality between skew spectra and bispectrum.
  • The $b_{G_2}/b_1$ ratio, which the paper notes is a sharp discriminant near $b_{1,h}\approx 1$ where competing tidal-bias fits disagree, suggests a targeted observing design: concentrating an all-sky survey's depth in the redshift bins where $b_1$ crosses unity could arbitrate between conflicting Lagrangian tidal-bias prescriptions more decisively than a uniform scan.
  • Comparing H$\alpha$, [OII], [OIII] and [CII] bias ratios in the same volume converts the halo–emission hypothesis into a null test: lines with different metallicity and ionization sensitivity should depart from the halo relations by different amounts, and any differential departure would directly signal environment-dependent emission.
Original abstract

The bispectrum of line-intensity maps carries cosmological and astrophysical information beyond the power spectrum, but its large data vector hampers the analysis of survey data. Skew spectra address this by compressing the bispectrum into a handful of power-spectrum-like observables, cross-spectra of the observed field with filtered squares of it. We present the first application of optimally weighted skew spectra to line-intensity mapping of galaxy-hosted emission, where all clustering and shot-noise terms are luminosity weighted and the mean intensity is unknown. We perform idealized real-space Fisher forecasts for a SPHEREx-like survey of H$\alpha$ emission at $0.2 \lesssim z \lesssim 4.6$, in its deep fields and in a noisier all-sky configuration. First, the compression carries over to intensity maps: on perturbative scales and for a fixed redshift evolution of the emitters, four skew spectra, one of them matched to the shot noise, recover the bispectrum constraints on biases, shot noise and cosmological parameters to within a few percent. Second, the joint analysis of the power spectrum and the skew spectra measures the ratios of the second-order biases to the linear bias in each redshift bin. Within a bin the biases are degenerate with the unknown mean intensity, but their ratios are not, so their redshift evolution is measured without a model for the emitters. For emission that depends on halo mass alone, these ratios are determined by luminosity-weighted averages of the halo biases, so their evolution can test the halo-emission connection, including dependence on assembly history or environment. The bias ratios are not accessible to the tree-level real-space power spectrum alone. At $z \lesssim 3$ an all-sky SPHEREx-like configuration can detect departures of order $0.1$ in the bias ratios at $3\sigma$, while deeper fields perform better in the highest-redshift bins.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-30: 1 request, 7,938 input tokens.

What the abstract itself states, as probabilities: claims first 0.94 · states limitation 0.77 · human participants 0.01 · controlled comparison 0.18 · randomized assignment 0.01 · states quantitative result 0.97 · compares with prior results 0.76 · mentions code or data availability 0.02.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; .

The bispectrum of line-intensity maps carries cosmological and astrophysical information beyond the power spectrum, but its large data vector hampers the analysis of survey data. Skew spectra address this by compressing the bispectrum into a handful of power-spectrum-like observables, cross-spectra of the observed field with filtered squares of it. We present the first application of optimally weighted skew spectra to line-intensity mapping of galaxy-hosted emission, where all clustering and shot-noise terms are luminosity weighted and the mean intensity is unknown. We perform idealized real-space Fisher forecasts for a SPHEREx-like survey of H$\alpha$ emission at $0.2 \lesssim z \lesssim 4.6$, in its deep fields and in a noisier all-sky configuration. First, the compression carries over to intensity maps: on perturbative scales and for a fixed redshift evolution of the emitters, four skew spectra, one of them matched to the shot noise, recover the bispectrum constraints on biases, shot noise and cosmological parameters to within a few percent. Second, the joint analysis of the power spectrum and the skew spectra measures the ratios of the second-order biases to the linear bias in each redshift bin. Within a bin the biases are degenerate with the unknown mean intensity, but their ratios are not, so their redshift evolution is measured without a model for the emitters. For emission that depends on halo mass alone, these ratios are determined by luminosity-weighted averages of the halo biases, so their evolution can test the halo-emission connection, including dependence on assembly history or environment. The bias ratios are not accessible to the tree-level real-space power spectrum alone. At $z \lesssim 3$ an all-sky SPHEREx-like configuration can detect departures of order $0.1$ in the bias ratios at $3\sigma$, while deeper fields perform better in the highest-redshift bins.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 The bispectrum of line-intensity maps carries cosmological and astrophysical information beyond the power spectrum, but its large data vector hampers the analysis of survey data. background 0.97 0.06 0.03 asserted 1.00
s1 Skew spectra address this by compressing the bispectrum into a handful of power-spectrum-like observables, cross-spectra of the observed field with filtered squares of it. background 0.94 0.07 0.04 asserted 1.00
s2 We present the first application of optimally weighted skew spectra to line-intensity mapping of galaxy-hosted emission, where all clustering and shot-noise terms are luminosity weighted and the mean intensity is unknown. contribution 1.00 0.94 0.10 asserted 1.00
s3 We perform idealized real-space Fisher forecasts for a SPHEREx-like survey of H$\alpha$ emission at $0.2 \lesssim z \lesssim 4.6$, in its deep fields and in a noisier all-sky configuration. method 0.93 0.94 0.04 asserted 0.94
s4 First, the compression carries over to intensity maps: on perturbative scales and for a fixed redshift evolution of the emitters, four skew spectra, one of them matched to the shot noise, recover the bispectrum constraints on biases, shot noise and cosmological parameters to within a few percent. result 1.00 0.93 0.38 asserted 1.00
s5 Second, the joint analysis of the power spectrum and the skew spectra measures the ratios of the second-order biases to the linear bias in each redshift bin. result 0.98 0.86 0.20 asserted 1.00
s6 Within a bin the biases are degenerate with the unknown mean intensity, but their ratios are not, so their redshift evolution is measured without a model for the emitters. interpretation 0.56 0.77 0.13 asserted 1.00
s7 For emission that depends on halo mass alone, these ratios are determined by luminosity-weighted averages of the halo biases, so their evolution can test the halo-emission connection, including dependence on assembly history or environment. interpretation 0.99 0.36 0.13 asserted 0.77
s8 The bias ratios are not accessible to the tree-level real-space power spectrum alone. gap 0.37 0.50 0.12 asserted 1.00
s9 At $z \lesssim 3$ an all-sky SPHEREx-like configuration can detect departures of order $0.1$ in the bias ratios at $3\sigma$, while deeper fields perform better in the highest-redshift bins. result 1.00 0.88 0.18 asserted 0.95

What kind of paper the classifier reports this is: method or tool 0.92 · evidence from simulation or computation 0.96.

What was examined, and what was not

sources read
  • arXiv listing abstract v1
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures
Figure 1. Responses of the weighted skew spectra to the biases and the shot-noise parameter, ∂PSn,δobs /∂θ. Solid (dashed) curves denote positive (negative) response. Each kernel responds to all four parameters. The lower panels show the ratio of the α response to the b1 response, (∂PSn,δobs /∂α)/(∂PSn,δobs /∂b1), as a function of k, with the L0 ratio repeated in each panel as a reference (light curve). For the three clustering kernels, the ratio takes a common value ≃ 0.038 across the analysis range, while for the shot-sector kernels it is larger by a factor ≃ 1.7. free, the joint analysis…
Figure 2. Contributions to the total observed Hα power spectrum for the SPHEREx-like survey, as a function of redshift: clustering, shot-noise, and instrumental-noise terms. On the scales entering our analysis, k ≤ kmax = 0.15 h/Mpc, the intensity maps are clustering-dominated at all but the highest redshifts, while shot noise and instrumental noise control the inverse-variance weights of Eq. (9). bias b [L 2 ] 1 which depends on L 2 (M) would change in the presence of a mass-dependent scatter, which re-weights the halo mass kernel contributing to the integrand. The line-emission model,…
Figure 3. Contributions to the observed skew spectra for the SPHEREx-like survey in the second redshift bin, centered at z = 0.8, as an example. For each of the {dens, disp,tide, orth} kernels, we show the clustering terms of Eq. (11), the shot-noise terms of Eq. (20), and the total. The vertical line marks kmax = 0.15 h/Mpc. set augmented with the shot-sector kernel of Eq. (23), S (3) = {Sdens, Sdisp, Stide} , S (4) = S (3) ∪ {Sorth} ; (29) the power spectrum; and the combination of the power spectrum with S (4) . Since Sorth is a linear combina￾tion, with constant coefficients, of Sshot
Figure 4. Marginalized constraints on the luminosity￾weighted biases and the shot-noise deviation α for the SPHEREx-like Hα survey, with kmax = 0.15 h/Mpc and instrumental noise included: full bispectrum (solid) ver￾sus the standard three-kernel skew spectra set S (3) (dot￾ted) and the extended four-kernel skew spectra set S (4) (dashed). closely in both size and orientation, and the marginalized constraints agree to within 3% for the biases and 1% for α (Table I). The skew spectra capture the bispectrum in￾formation on the bias and shot-noise parameters (
Figure 5. and Table II report the results of the joint anal￾ysis, with b [L 2 ] 1 free and the parametrization of Sec. III D: the moments varied by a common factor, the biases by a common shift. With this parametrization the power spec￾trum alone already separates ⟨I⟩ and b1 in the summed analysis, to 1.1% and σ(b1) = 0.012, by the mechanism of Sec. IV B: a common shift of b1 cannot mimic a common rescaling of ⟨I⟩ across bins. Adding the skew spectra im￾proves both by about 20%, more than the inverse-variance combination of the two separate constraints would give (1.06% on ⟨I⟩): the two…
Figure 6. Redshift evolution of the clustering amplitude combinations A1 = ⟨I⟩ b1, A2 = ⟨I⟩ b2, AG2 = ⟨I⟩ bG2 , and the shot-noise amplitudes C = (1 + α1) ⟨I 2 ⟩ b [L 2 ] 1 , D = (1 + α2) ⟨I 3 ⟩, E = (1 + αP) ⟨I 2 ⟩, from a joint analysis of power spectrum and skew spectra, for the SPHEREx-like survey of the previous sections (upper panels: 200 deg2 ) and for an all-sky version (lower panels: 30 000 deg2 ). The error bars are the marginalized 1σ from the joint P + S (4) analysis (dashed bars would extend below zero and are clipped). of b2/b1 with redshift but do not test the relati
Figure 7. Redshift evolution of the ratios A2/A1 = b2/b1 and AG2 /A1 = bG2 /b1, from the higher-order statistics alone and from a joint analysis with the power spectrum, for the SPHEREx-like survey of the previous sections (upper panels: 200 deg2 ) and for an all-sky version (lower panels: 30 000 deg2 ). Axes are shared within columns to make the comparison easier between the two configurations; the higher-z all-sky bins suffer from instrumental noise. the bispectrum in the deep-field (all-sky) configuration, more than in the summed analysis of Sec. IV A: this is expected, since a single…
Figure 8. Variations of the ratios b2/b1 and bG2 /b1 within different choices of halo mass function and biases (red), lower mass cut of the emitters (blue), luminosity tilt (orange), b2(b1) relation (light gray), presence of a Lagrangian tidal bias (purple). See main text for details. The 1σ error bars from the joint P + S (4) analysis in
Figure 9. Marginalized 1σ constraints on ln σ8(z) for the SPHEREx-like survey of the previous sections from a joint power spectrum and four-kernel skew spectra analysis, with kmax = 0.15 h/Mpc. of current redshift-space analyses at redshifts probed by Stage-IV spectroscopic galaxy surveys (z ≤ 2). Never￾theless, this is a proof of principle that intensity maps provide a handle on the growth of structure across a wide redshift range beyond what is accessible by current galaxy surveys without any astrophysical model, and in a form complementary to RSD: combined with a measurement of fσ8(z), fro
Figure 10. Weighted (upper row) and unweighted (lower row) skew spectra as a function of redshift, at fixed k = 0.05 h/Mpc, for the SPHEREx-like deep-field survey of the main text. Dashed lines indicate negative values. In the unweighted case, only the clustering kernels are shown, since the definition of the shot noise ones relies on the weights. where χ is the comoving distance. The small-scale cutoff of the skew-spectrum estimators is implemented as a smoothed step in the internal momenta: each internal leg carries a window W(q) = 1 2 [1 − tanh(β ln q kmax )] , β = 50 , (A15) the value ado
Figure 11. Marginalized constraints on {Ωc, Ωb, h, ns, ln(1010As), b1, b2, bG2 , b[L 2 ] 1 , α} for the SPHEREx-like Hα survey, kmax = 0.15 h/Mpc, from the full bispectrum (solid) versus the standard three-kernel skew spectra set S (3) (dotted) and the extended four-kernel skew spectra set S (4) (dashed); the gray contours show the bispectrum with cosmology fixed. The agreement between bispectrum and skew-spectra constraints in the presence of non-amplitude-like parameters is a non-trivial result that was not guaranteed by the construction of the skew-spectra estimators. shot-noise…
Figure 12. Per-bin constraint ratios between the compressed statistics and the full bispectrum, marginalized (top rows) and un-marginalized (bottom rows), when varying {Ωc, Ωb, h, ns, ln(1010As), b1, b2, bG2 , b[L 2 ] 1 , α} for the SPHEREx-like Hα survey, kmax = 0.15 h/Mpc. The ±2% gray band marks the numerical error floor: ratios inside the band should be read as unity. The marginalized constraints need a prior on the cosmological block to regularize the matrix inversion, since a single redshift contains highly degenerate directions: we show marginalized constraints including a Planck prior
Figure 13. Marginalized constraints on the parameters of
Figure 14. Eigenvalue spectrum of the computed correlation matrix of the three-kernel (left) and four-kernel (right) skew-spectrum data vector, per redshift bin, before regularization. Exactly one collective eigen-direction per bin lies at or below the measured error floor ε = 2 × 10−3 of Eq. (G1), shown as a horizontal band; a second eigenvalue lies close to it; the rest of the spectrum is separated by an order of magnitude. The three-kernel spectrum does not fall below the floor in any bin. The dashed lines mark the clipping thresholds, λfloor for S (4) and ε for S (3) ; the negative eigenva

Figures reproduced from arXiv:2610.40063 by the authors.

References

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.