Speaker
Description
Cross-correlations between surveys which retain line-of-sight (LOS) structure - such as line-intensity-mapping (LIM) or spectroscopic surveys - and projected fields - such as CMB lensing - represent a powerful avenue for extracting cosmological information from next-generation experiments. Common approaches include tomographic binning of the radially-resolved field into 2D slices, projecting the field into a single 2D map using an assumed kernel, or correlating individual line-of-sight Fourier modes with the projected field directly. Given the improving quality of large-scale structure measurements, it is useful to systematically assess the optimality of various cross-correlation approaches. To this end, we develop a framework for constructing Fisher-optimal projection kernels for the 2Dx2D case which we frame as a quadratic optimization problem solved numerically over redshift compression weights. Using 21cm LIM and CMB lensing as the prototypical toy model, we find that Fisher-optimal projection recovers more information than naive kernel choices in the presence of foreground filtering. We additionally analyze the impact of light-cone evolution on the information content of such cross-correlations.