Advancing cosmic density field reconstruction with machine learning and statistical methods

The baryon acoustic oscillation (BAO) technique is one of the most powerful probes of dark energy and has been playing an important role in large galaxy surveys in the last decade. The effects of late-time nonlinear structure formation, however, wash out the acoustic peak of the correlation function, reducing the precision of BAO distance scale measurements. Reconstructing the nonlinear density field reverses these adverse effects and increases the precision of the measurements. A standard reconstruction method has been used in observations for a decade and has reduced the uncertainty on the BAO scale by about a factor of two. However, ongoing and next-generation surveys, such as the Dark Energy Spectroscopic Instrument, Nancy Grace Roman Space Telescope, and Euclid, will offer unprecedented precision measurements of cosmological parameters and thus will greatly benefit from improvements in reconstruction to more fully realize their potential. We present new methods of reconstruction that use machine learning and statistical tools. In particular, we show our investigation of reconstruction for fields of low number densities, at the level achievable by current surveys.