Analytical method for computing the covariance matrix of cosmic shear two-point correlation function

Masahiro Takada · 2026

Read Analytical method for computing the covariance matrix of cosmic shear two-point correlation function by Masahiro Takada (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-16186.txt.

Introduction

Nagura, Terasawa, Terawaki, and Takada compare analytic methods for the Gaussian covariance of the cosmic shear two-point correlation function, allowing for finite survey geometry. With HSC Year 3 survey-mask convergence fields as a test case, they find the improved narrow-kernel approximation fails to match simulation covariances. Leftover errors in off-diagonal harmonic-space parts leak through the Legendre transform and distort real-space results. The weighted quartic-counts method matches better, showing off-diagonal modelling is crucial for sound estimates.

Catalog record

CREATORMasahiro Takada
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE26068
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.16186

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Analytical method for computing the covariance matrix of cosmic shear two-point correlation function Masahiro Takada Abstract Accurate estimation of the covariance matrix of cosmic shear statistics is essential for cosmological analyses using current and upcoming wide-area weak lensing surveys. In this work, we investigate analytical methods for computing the Gaussian covariance matrix of the cosmic shear two-point correlation function (2PCF), taking into account the effects of finite survey geometry. We compute the covariance of 2PCF based on the improved Narrow Kernel Approximation (iNKA), with a projection using the Legendre transformation. We also consider other analytical covariance estimators, the f sky f_{\mathrm{sky}} approximation and the weighted quartic-counts method. We evaluate the accuracy of those analytical methods using the convergence fields with the HSC Year 3 survey mask as a test case. We find that the covariance of the 2PCF obtained by using the iNKA does not reproduce the covariance measured directly from Gaussian simulations. Although the iNKA accurately models the diagonal structure of the harmonic-space covariance, residual inaccuracies in the off-diagonal components propagate through the Legendre transformation and significantly affect the real-space covariance. In contrast, the weighted quartic-counts method shows better agreement with the simulations. Our results demonstrate that accurate modeling of the off-diagonal structure of the harmonic-space covariance is crucial for obtaining reliable covariance estimates of real-space weak lensing statistics in the presence of survey window effects. 1 Introduction Current photometric surveys have given a wealth of information on the structure formation. When performing the cosmology analysis of such data, it is important to estimate the covariance accurately. In this work, we focus on the covariance of angular two-point correlation functions (2PCFs) and angular power spectrum, which are widely used in the literature. Roughly speaking, there are two ways to compute the covariance: to measure sample covariance from the many simulation realizations, or to model the covariance analytically. The former one is used in the Hyper Suprime-Cam (HSC) analyses (e.g., [ 24 , 18 , 3 , 27 , 7 ] ). By cutting out the survey region from full-sky simulations [ 26 ] , we can take into account for the effect of the survey geometry including the Super-Sample Covariance (SSC) (e.g., [ 25 , 28 , 29 ] ). Being made off of N-body simulations, covariance from the full-sky simulations incorporate non-Gaussian Covariance which arise from connected trispectrum. For the surveys with relatively small sky area such as HSC, we can cut out multiple pseudo-independent realizations from one full-sky simulation, but for the Stage-IV surveys with area \gtrsim 10,000~{\rm deg}^{2} , only few realizations can be extracted from a full-sky simulation and making many full-sky simulations will be computationally expensive. To circumvent the expensive simulations, modeling covariance analytically (e.g., [ 21 , 16 , 20 , 23 , 9 , 8 , 15 , 12 , 6 ] ) would be realistic option for the Stage-IV surveys. In the analytic model, the covariance is calculated as a sum of Gaussian Covariance term and non-Gaussian Covariance term. In this work, we focus on the Gaussian term which is dominant on the large scales. In the upcoming survey with wide area, the large-scale data becomes important [ 31 , 4 ] and so is its covariance. Survey masks introduce mode-coupling and affect covariance. The Gaussian Covariance for pseudo- C_{\ell} estimator in the presence of mask is studied in Refs [ 5 , 10 , 19 ] . For the two-point correlation functions, the effect of window can be taken into account by accurately calculating number of pairs of galaxies in angular bin for the shot/shape noise term [ 30 ] . For the mixed contributions from the sample variance and shot/shape noise (“mixed term”), triplet counting gives accurate estimate [ 20 ] . For the sample variance-only term, corresponding calculation involves with quartic counts [ 22 , 23 ] which can be computationally expensive. Hence, in this paper, focusing on the sample variance-only term, we explore the other approach where the real-space covariance is calculated as a double Legendre transform of harmonic space covariance with survey window effect. We will compare the both approach with the sample covariance measured from Gaussian realizations. We organize this paper as follows. In Section 2 , we summarize how to calculate the covariance of 2PCFs and power spectra analytically. In Section 3 , we introduce Gaussian simulations of cosmic shear fields to measure the sample covariance with which we compare our analytical covariance estimation. In Section 4 , we present the accuracy of analytic covariance estimate compared to the sample covariance. Section 5 is devoted to discussion and conclusion. 2 Formulation of the covariance matrix of cosmic shear two-point correlation function 2.1 Power spectrum with and without survey window effects In this work, we consider the Gaussian covariance of weak gravitational lensing observables in a finite survey area. Specifically, we focus on the power spectrum, C_{\ell} , of the spin-0 convergence field \kappa(\hat{\bm{n}}) , defined on the celestial sphere. In practice, a survey is limited to a partial sky coverage and has complex geometry involved with masks. These affect the measured power spectrum and its covariance. We refer to this effect as the survey window effect and denote physical quantities affected by this effect with a tilde symbol, e.g., \tilde{C}_{\ell} . We also consider the real-space counterpart of C_{\ell} , the two-point correlation function \xi(\theta) . For simplicity, we restrict our analysis to a single source redshift. On the celestial sphere, the convergence field \kappa in the direction \hat{\bm{n}} is expanded in spherical harmonics Y_{\ell m} as \displaystyle\kappa(\hat{\bm{n}})=\sum_{\ell,m}a_{\ell m}Y_{\ell m}(\hat{\bm{n}}), (2.1) and the power spectrum is defined by \displaystyle\expectationvalue{a_{\ell m}a_{\ell' m'}^*}=\delta^{K}_{\ell\ell^{\prime}}\delta^{K}_{mm^{\prime}}C_{\ell}, (2.2) where \delta^{K}_{\ell\ell^{\prime}} is the Kronecker delta function. The observed field including the survey window effect can be expressed as \displaystyle\tilde{\kappa}(\hat{\bm{n}})=w(\hat{\bm{n}})\kappa(\hat{\bm{n}}), (2.3) where w(\hat{\bm{n}}) is the survey mask field, defined as w(\hat{\bm{n}})=1 if \hat{\bm{n}} lies inside the survey region, and w(\hat{\bm{n}})=0 otherwise. When weights are also taken into account, the following discussion can be straightforwardly extended. The resulting power spectrum can be expressed using the mode-coupling matrix {\cal M}_{\ell\ell^{\prime}}(w,w) [ 13 , 1 , 19 ] as \displaystyle\tilde{C}_{\ell}=\sum_{\ell^{\prime}}{\cal M}_{\ell\ell^{\prime}}(w,w)C_{\ell^{\prime}}. (2.4) From this equation, we can obtain the unwindowed (i.e., window-corrected) power spectrum via C_{\ell}={\cal M}^{-1}_{\ell\ell^{\prime}}\tilde{C}_{\ell^{\prime}} . This is the so-called pseudo- C_{\ell} method for estimating C_{\ell} . 2.2 Relation between \xi(\theta) and \tilde{C}_{\ell} The two-point correlation function of the convergence field is given by the Legendre transform of the power spectrum as \displaystyle\xi(\theta)=\sum_{\ell}\frac{2\ell+1}{4\pi}C_{\ell}P_{\ell}(\cos\theta), (2.5) where P_{\ell}(x) is the Legendre polynomial. The measured correlation function is an unbiased estimator of the underlying correlation function, in the sense that it is not affected by the survey window. Nevertheless, Ref. [ 9 ] derived the relation between the windowed (measured) power spectrum and the correlation function (see Appendix C of Ref. [ 9 ] ): \displaystyle\xi_{i}=\frac{n^{2}}{N_{\mathrm{pair},i}}\tilde{\xi}_{i}, (2.6) where \xi_{i}\equiv\xi[\theta_{i,-},\theta_{i,+}] denotes the correlation function averaged over the i -th angular bin, n is the mean number density of source galaxies per steradian, and N_{\mathrm{pair},i} is the number of galaxy pairs in the i -th bin within the finite survey region. The quantity \tilde{\xi}_{i} is the “window-affected” correlation function, defined in terms of the windowed power spectrum \tilde{C}_{\ell} as \displaystyle\tilde{\xi}_{i}=2\pi\sum_{\ell}\left[P_{\ell+1}(x)-P_{\ell-1}(x)\right]_{\cos\theta_{i,+}}^{\cos\theta_{i,-}}\tilde{C}_{\ell}, (2.7) where we took average of P_{\ell}(\cos\theta) over

Read the complete text