Abstract
Although modern cosmology has been able to reveal a wealth of information about the Universe, there are still many questions to be answered, such as the details of how the first stars ionized the Universe in the Epoch of Reionization (EoR). The CO Mapping Array Pathfinder (COMAP) is a line intensity mapping experiment aiming to answer this question. Currently in the early stages, much of the present effort of COMAP goes into understanding and mitigating systematic effects. To do this we process the raw telescope data through a series of filters, cleaning it of correlated noise and systematics, before making sky maps. In this thesis, we contribute to this effort by presenting three improvements to the COMAP data analysis pipeline.
The first of these is to extend the COMAP pipeline by a simulation pipeline, with which we can simulate the COMAP telescope picking up additional signal. By estimating the factor by which the simulated signal has been attenuated by filtering the raw data, we estimate the pipeline transfer function. This additional step of the COMAP pipeline is critical as it gives us the possibility to adjust for any underestimated signal power and its error bars. In particular, when computing the ensemble-averaged transfer function from estimates of three different signal realizations as well as ∼ 63h of observational data, the peak transfer efficiency of the filters is found to be 85 − 90% at scales kk,k⊥ ≥ 0.1Mpc−1, while resulting in an almost complete attenuation at scales kk,k⊥ ≤ 0.05Mpc−1. Thus, on these scales, the signal estimates of the COMAP pipeline and its errors should be adjusted upwards by a factor inverse proportional to this transfer function estimate.
Secondly, we implement a baseline destriper in the COMAP mapmaker, which by fitting residual long-timescale modes such as ground pickup, standing waves or unknown systematics, can better resolve the large-scale modes of the astrophysical signal perpendicular to the line-of-sight. Of the tested baseline lengths, the 10s baseline destriper is found to maximize the transfer function, yielding a peak transfer efficiency of ∼ 95% and outperforming the currently implemented noise weighted binning of the highpass filtered time stream by up to 20 − 25% at the lower k⊥ region.
Finally, a principal component analysis of the dataset of feed-feed pseudo-cross spectra, computed from different data splits, is shown to aid in identifying spectra showing signs of systematic effects. Subtracting leading principal components of the data that could correspond to systematic effects is found to result in a modest signal loss of 20 − 50% in a few of the k-bins, introducing a potential new tool in the COMAP pipeline to clean otherwise discarded data of systematic effects.
Veiledere: Professor Hans K. K. Eriksen og Ingunn K. Wehus, Institutt for teoretisk astrofysikk, UiO
Intern sensor: Professor Øystein Elgarøy, Institutt for teoretisk astrofysikk, UiO
Ekstern sensor: Researcher Elina Keihänen, University of Helsinki