Title: Analytical non-gaussian covariance matrix for power spectrum and bispectrum

Speaker: Dr. Zhiyu Lu ( Leiden University, Netherlands)

Time: 14:30pm, July 3, 2026

Location: 3-302, PMO Xianlin Campus

Abstract: The covariance matrix plays a central role in cosmological parameter inference by propagating statistical uncertainties of data points into uncertainties of cosmological parameters. The traditional approach to estimating the covariance matrix relies on a large number of simulations at a fixed cosmology, which makes the framework computationally expensive and inflexible in analyses. In this talk, I present our recent progress in analytically computing the covariance matrix within perturbation theory, including both Gaussian and non-Gaussian contributions. The analysis consistently incorporates higher-order correlation functions up to the trispectrum level (4 point correlation function), enabling a more complete description of sample variance and mode coupling. The theoretical predictions are validated against the Quijote simulations for dark matter, halos, and galaxy fields, and further compared with the large-volume PTChallenge simulations. I will show that non-Gaussian contributions to the covariance are essential for obtaining unbiased and accurate cosmological parameter inference, especially for upcoming high-precision surveys.