Compressed Likelihoods and Early Universe Constraints for Cosmic Microwave Background Experiments

Abstract

In this dissertation we present cosmic microwave background likelihood tools for the Planck and BICEP3 data, and constrain extensions to the Big Bang ΛCDM cosmological model using Planck in combination with current and upcoming ground-based experiments. We begin with an overview of the current Standard Model of cosmology and the CMB. We present CosMOPED, a compressed likelihood code for Planck data at l ≥ 30, which uses the Massively Optimized Parameter Estimation and Data compression technique (MOPED) method to reduce the dimensions of the data space to one number per cosmological parameter of interest. We then construct a binned likelihood for the Planck low-l temperature and E-mode polarization, called Planck-low-py. We fit the bandpowers in two temperature bins and seven polarization bins with shifted log-normal distributions, and use these bins in a differentiable Python likelihood to facilitate ease of use of the Planck 2018 large-scale data. The ΛCDM parameters recovered with CosMOPED and Planck-low-py are consistent with the uncompressed Planck likelihoods, and a 7-parameter extended model is similarly well-constrained. We also examine some of the foreground modeling choices made in the BICEP/Keck primordial gravitational wave analysis (BK18) and estimate foreground-marginalized CMB B-mode bandpower amplitudes in the BICEP3 sky region. We use these bandpowers to construct a marginalized likelihood with no nuisance parameters. The tensor-to-scalar ratio inferred from this BK18-lite likelihood matches that using the public BK18 likelihood. Additionally, we use data from the Atacama Cosmology Telescope, the South Pole Telescope, and the Planck satellite to constrain different types of initial conditions. First we explore an adiabatic model with a broken power law instead of a single power law, and then we look at a model with both adiabatic and isocurvature fluctuations, allowing the isocurvature to vary independently in five bins. Finally, we forecast the upcoming Simons Observatory’s ability to improve on our binned isocurvature constraints at small scales.