High-level data analysis in the COMAP intensity mapping experiment

The CO Mapping Array Pathfinder (COMAP) is an intensity mapping experiment targeting the emission from rotational lines of carbon monoxide, redshifted to the frequency range of 26-34 GHz. In this first phase of a planned longer-term program, the primary science goal is the detection of CO(1-0) signal, aiming to constrain its power spectrum at the Epoch of Galaxy Assembly (z = 2.4-3.4). It is, therefore, crucial to devise an efficient method of extracting this signal from the obtained three-dimensional line intensity maps. We also need to make sure that any experimental systematic, giving rise to undesirable signs of excess power, is sufficiently suppressed. In this thesis, we develop an effective approach addressing both of these demands and apply it to analyze the first year of gathered data. We present how the low-level analysis pipeline converts raw data provided by the telescope into filtered and calibrated temperature sky maps. Then, we design the Feed-feed Pseudo CROSS Spectrum (FPXS) method, the foundation of the high-level data analysis in COMAP. FPXS is used as a principal tool for inferring the signal spectrum from the maps and assessing the quality of the current data set. Moreover, by employing accurate signal simulations, we study and quantify the impact of different effects on the measured signal. The resulting transfer functions are applied to derived spectra in order to correct the bias present in the maps. Specifically, we consider the instrumental beam smoothing the signal on small angular scales, the choice of the spectral resolution of the maps, affecting small scales in the line-of-sight dimension, as well as the attenuation of the signal on large scales, introduced by various procedures performed during the low-level data analysis. In the considered range of cosmological scales, the results associated with constant elevation scans (CES) indicate that we are effectively suppressing the sources of systematics below the current level of white noise and that the measurement uncertainties integrate down as expected from the radiometer equation. Already at this point (before accumulating enough data to reach a suitable detection sensitivity), the upper limit derived from CES results is excluding the brightest models for the CO signal at 95\% confidence. Nevertheless, the data obtained by scanning the sky according to the Lissajous pattern exhibit clear signs of excess power caused by residual systematics. We suspect that these arise due to effects correlated with the pointing of the telescope, requiring a more complicated treatment prior to using Lissajous data in the scientific analysis