Speaker
Description
As CMB polarization experiments target the primordial B-mode signal, addressing complex foregrounds and systematics requires deploying an unprecedentedly large number of detectors in their focal planes. The resulting data volume will be enormous, posing a critical data reduction challenge. To address this, we present BrahMap: a modular, scalable, and HPC-ready map-making framework designed for next-generation CMB polarization experiments. BrahMap leverages a linear operator based architecture with multiple levels of hierarchical abstraction, enabling the extensible, modular, and intuitive development of custom data reduction pipelines. We demonstrate BrahMap's capabilities by implementing a generalized least squares (GLS) map-maker with multiple noise covariance operator models, including circulant approximations and the full Toeplitz covariance. With realistic LiteBIRD simulations, we further demonstrate that the exact Toeplitz covariance in map-making reduces correlated $1/f$ noise residual power at large-scales by two orders of magnitude compared to the optimal circulant approximation. In the most pessimistic scenario, this translates to an approximately four-fold reduction in the bias on the tensor-to-scalar ratio induced by noise residuals.