Using Artificial Intelligence to uncover the initial conditions of the Local Universe

• Applicant: Dr. Noam Isaac Libeskind
• Subject Area: Astrophysics and Astronomy
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 569604537

While our understanding of the Universe’s large-scale structure has advanced, a significant challenge remains in deciphering the nature of dark matter and gravity on quasilinear transitional scales. These scales mark the transition from highly non-linear, virialized structures, such as galaxy groups and clusters, to the more linear large-scale structure of the universe. Understanding the dynamics and gravitational effects in this transitional regime is crucial for a comprehensive cosmological model. Modeling quasilinear scales entails grappling with the intertwined effects of gravity, peculiar velocities, and redshift distortions. Traditional analysis methods struggle to accurately capture these complexities, which poses challenges in elucidating the true nature of dark matter and gravity in this regime. Linking dark matter and gravity to observations on intermediate scales demands meticulous modeling of the local cosmic web. This involves generating simulations that accurately mimic the local Universe by creating constrained initial conditions to be evolved forward using simulations. Traditionally, these initial conditions are generated using approximate dynamics, such as the Zeldovich approximation, second-order perturbation theory, and Peebles’ action principle, albeit with varying degrees of success. This proposal seeks to address this challenge by developing a comprehensive AI/ML framework for generating the relevant initial conditions. Leveraging generative models like GANs or VAEs offers distinct advantages, as they can be trained on complex mock catalogs without relying on specific dynamics or statistical assumptions, thereby enabling more robust exploration of various scenarios. With robust mock galaxy catalogs extracted from the simulations, various methods for probing signatures of dark matter and dark energy on intermediate scales will be employed. Simulations with varying mass and spatial resolutions will be conducted to extract cosmological information from the large-scale structure down to galaxy scales, including insights into the origins of present day local group galaxies.

• DFG Programme: Research Grants
• International Connection: Israel
• Cooperation Partner: Professor Adi Nusser, Ph.D.