MAP-MAKING ALGORITHMS FOR CMB POLARIZATION EXPERIMENTS

  • Jack Bewley

Student thesis: Master of Science by Research

Abstract

The EM radiation that we observe as the CMB provides an incredible wealth of information regarding our universe. Particularly, the elusive primordial B-mode polarization signal is a topic of paramount interest in modern CMB cosmology as potential evidence for cosmic inflation. This B-mode signal has yet to be de- tected due to its faint nature and the many complications when observing at the largest distances possible. The CMB pipeline, of which map-making algorithms comprise just one part, is therefore a process which demands constant review and development. In this thesis, I compare the performance of two map-makers: ex- tended binning (operating in real space) and Fourier binning (utilising transforms to Fourier space) in several simulated scenarios. Each of the two algorithms are tested as both single-detector and di↵erenced-pair implementations. Near-perfect recovery of the input signal is reported using each algorithm in the control case of signal-only idealistic simulations. An artificially simulated di↵erential pointing error is then introduced to the simulations with near-perfect map recovery once again seen in all of these scenarios. The limitations of this pointing systematic implementation are also explored. Another systematic e↵ect, di↵erential gain error, is then simulated for only the di↵erenced-pair implementations of the two map-makers. While they are unable to remove the gain errors without a priori information regarding the calibration of at least one of the detectors, we see that the gain errors propagate through the algorithms exactly as expected which at least provides a solid grounding for which to explore other methods for di↵eren- tial gain mitigation. Both algorithms are then tested on simulations containing each of three noise models: Gaussian noise; 1/f noise generated independently for each detector; and 1/f noise with detector-detector correlations. A surpris- ing parity is observed between the two algorithms when tested as single-detector methods. There are, however, di↵erences between the methods when applied to a di↵erenced pair. In the case of Gaussian noise, we see a noise penalty su↵ered by Fourier binning on account of its solving for unnecessary quantities. Inclusion of a 1/f noise component yields, surprisingly, the opposite observation. The high frequency modulation of our sky polarization, caused by our fast-rotating HWPs, is postulated as an explanation for this. The results presented in this thesis give promising indication that either of the algorithms may prove a valuable asset to future CMB cosmology and encourage further, more rigorous testing.
Date of Award1 Aug 2021
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorDaniel Thomas (Supervisor) & Michael Brown (Supervisor)

Cite this

'