Comparison of Cross-Correlation Methods for Line Intensity Mapping

Abstract

Line intensity mapping (LIM) is a technique for producing 3D maps of the Universe by scanning the sky with a spectrometer sensitive to a range of wavelengths corresponding to the redshifted spectral lines of atoms or molecules, such as hydrogen or carbon, commonly found in galaxies and the diffuse media around them. While LIM experiments have successfully detected the 21 cm line of neutral hydrogen, other lines that reveal large-scale structure or astrophysical processes remain undetected. Many LIM experiments are in development or are underway to fill this gap, but will likely suffer from contamination from systematics, like Galactic foregrounds, or noise. Cross-correlation techniques offer the smoothest route for making detections and constraining astrophysical processes in this regime. In this work, we apply three cross-correlation techniques (stacking, the conditional voxel intensity distribution (CVID), and the cross power spectrum) to simulated LIM maps produced using [CII] luminosity models for a pathfinder LIM experiment (EXCLAIM). We find that these cross-correlation techniques allow for mean detection of the target signal line ([CII]) at redshifts 2.5-3.5 at the 4.5$σ$, 3.9$σ$, and 8.4$σ$ level, respectively, and offer moderate constraints on the line emission model. Under a futuristic scenario with reduced noise, the techniques improve substantially, with detections at the 44.0$σ$, 24.6$σ$, and 34.3$σ$ levels and percent-level constraints. Each technique offers unique information, with the strongest constraints achieved by using the three techniques in combination.

Figures (16)

• Figure 1: A patch of simulated LIM maps generated in Limlam Mocker, as described in Sec. \ref{['sec:simcomps']}. (a), (b), (c), and (e) are individual map components. (d) is the sum of (a), (b) and (c) convolved by the angular and spectral resolution of the EXCLAIM instrument. (f) is the sum of (d) and (e). Note that (a), (b), (c), and (d) are in log$_{10}$(Jy/sr) units and (e) and (f) are in MJy/sr.
• Figure 2: The cross power spectrum between the mock QSO catalogs and simulated LIM maps. The solid purple curve is the cross power for the LIM map with signal and all contaminants included averaged over 1,000 realizations, with the purple shaded region around it being the standard deviation of the realizations. The blue dashed curve is for a LIM map with just [CII] signal and the red dashed curve is for a LIM map with just CO interlopers and Galactic foregrounds. The instrument resolution convolves all maps used. (a) exhibits significant contamination at low $k$ due to foregrounds. (b) depicts the same spectra but with the lowest $k_x$, $k_y$, and $k_z$ modes removed.
• Figure 3: Example stacks resulting from the stacking process. These depict the frequency slice of the stack in which the QSO (or random location) is located. Coordinates are relative to the center. All stacks are scaled according to the minimum/maximum values of the stack in (b). The black square outlines the central 9 voxels of the stack slice that will be included in the summed intensity observable for the stack.
• Figure 4: Example stacks resulting from the stacking process. These depict the central voxels of the stack across frequency slices. Coordinates are relative to the center. All stacks are scaled according to the minimum/maximum values of the stack in (b). The black dashed lines outline the central 3 voxels of the stack that will be included in the summed intensity observable for the stack.
• Figure 5: An example of a stacking bootstrap test. The red dashed line indicates the weighted sum of the central 27 voxels of the stack depicted in Figs. \ref{['fig:ang_stacks']}(b) and \ref{['fig:freq_stacks']}(b). The violet histogram is the distribution of the weighted sums for 250 stacks performed on random locations on the LIM map.