Speaker
Description
The angular power spectrum is a key observable for upcoming high-precision cosmological surveys. However, modelling them typically involves the numerical evaluation of highly oscillatory 3-dimensional integrals that are computationally expensive using brute-force methods. In recent years, there has been increased interest in numerical methods that go beyond the well-known Limber approximation to robustly calculate angular power spectra in a wide parameter regime. Building on Picard-Lefshetz theory, we propose a new method to evaluate the angular power spectrum in the complex plane. Using a saddle point approximation, the computation is reduced to a one-dimensional integral, resulting in a significant increase in computation speed. This method elegantly complements the Limber approximation in the low-multipole regime and can be straightforwardly extended to include general integrands and redshift-space distortions. An application to typical survey kernels is presented. A publicly available code is under development.
