Denoising Diffusion Probabilistic Model for Realistic and Fast Generated Euclid-like Data for Weak Lensing Analysis

Understanding and mitigating measurement systematics in weak lensing (WL) analysis requires large data sets of realistic galaxies with diverse morphologies and colors. Missions like Euclid, the Nancy Roman Space Telescope, and Vera C. Rubin Observatory’s Legacy Survey of Space and Time will provide unprecedented statistical power and control over systematic uncertainties. Achieving the stringent shear measurement requirement of ∣m∣ < 10−3 demands analyzing 109 galaxies. Accurately modeling galaxy morphology is crucial, as it is shaped by complex astrophysical processes that are not yet fully understood. Subtle deviations in shape and structural parameters can introduce biases in shear calibration. The interplay between bulges, disks, star formation, and mergers contributes to morphological diversity, requiring simulations that faithfully reproduce these features to avoid systematics in shear measurements. Generating such a large and realistic data set efficiently is feasible using advanced generative models like denoising diffusion probabilistic models. In this work, we extend Hubble Space Telescope (HST) data across Euclid’s broad optical band using the Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey and develop a generative AI tool to produce realistic Euclid-like galaxies while preserving morphological details. We validate our tool through visual inspection and quantitative analysis of galaxy parameters, demonstrating its capability to simulate realistic Euclid galaxy images, which will address WL challenges and enhance calibration for current and future cosmological missions.