A New Look at Dark Matter in the Roman Era: Joint Inference of High-Resolution Mass Map and Cosmology

A New Look at Dark Matter in the Roman Era: Joint Inference of High-Resolution Mass Map and Cosmology

Program ID	19013
Science Category	Large Scale Structure of the Universe
Program Type	Analysis
Category	Small
Principal Investigator	Chihway Chang
PI Institution	University of Chicago
Co-Investigators	Chun-Hao To (University of Chicago) Benjamin Remy (University of Chicago)
Abstract	Weak gravitational lensing provides one of the most direct probes of dark matter distribution. However, existing studies are limited to either low-resolution reconstructions over wide areas or high-resolution reconstructions over small fields. We propose to exploit the exceptional depth and imaging quality of early Roman weak-lensing data. This will allow us to jointly reconstruct the dark matter distribution and infer cosmological constraints in the Roman COSMOS deep field. This proposal will develop a new diffusion-model-based field-level inference framework. It will simultaneously sample the true mass map and the cosmological parameters $\Omega_m$ and $S_8$ from the observed shear catalog. Unlike traditional methods, which use simplified Gaussian priors, this approach preserves the full non-Gaussian information content of the weak-lensing field. It also delivers principled posterior uncertainties on both the reconstructed map and cosmological parameters. We will train and validate the pipeline using Roman-like mock observations based on N-body simulations. These will include realistic survey geometry, noise, masking, redshift uncertainties, and intrinsic alignments. We will then apply the method to the Roman COSMOS deep field. This will produce the first high-resolution weak-lensing mass map from Roman, reaching about 2.3 arcmin resolution and probing structure to $z\sim3$ across 9.6 deg$^2$. Preliminary forecasts indicate a 58% improvement in the $S_8$-$\Omega_m$ figure of merit compared to standard two-point analyses. The resulting pipeline and reconstructed mass maps will be released publicly. This will provide both a compelling Roman science product and a pathfinder for future mass-mapping analyses over the full Roman survey.