3:30–5:00 pm ERC 161
Title: Building the Next-Generation low-redshift type Ia Supernova Sample with YSE and a dash of AI
Abstract: The Dark Energy Survey's 'Dovekie' reanalysis suggests the Universe may not be compatible with a cosmological constant (3.2σ tension in Flat w₀wₐCDM) - tantalizingly close to rewriting cosmology, but not yet definitive. With the imminent start of Rubin Observatory's Legacy Survey of Space and Time (LSST), we will begin to follow thousands of type Ia supernovae (SNe Ia). And soon, these SNe Ia will be complemented by higher redshift IR observations from NASA's Nancy Grace Roman Space Telescope and Schmidt Sciences' Lazuli Space Telescope. Together, these surveys will answer this cosmological question, provided we can tackle the dominant source of systematic error with SNe Ia - their astrophysics. Improving SNe Ia as standard candles requires understanding their progenitors and host environments and controlling survey selection effects - all tasks where next-generation AI can help. I will describe the upcoming Young Supernova Experiment (YSE) Data Release 2, and how we are using Pan-STARRS and DECam to build a low-redshift sample of ~1000 photometrically classified SNe Ia, to act as a low-z anchor for Rubin, Roman and Lazuli. I will describe recent work at the NSF-Simons SkAI Institute developing SELDON (Supernova Explosions Learned by Deep ODE Networks) - a foundation model that can predict what a supernova will do next, even from early observations, and reconstruct what we missed, even with sparse observations. Together, YSE's low-redshift anchor and AI-driven analysis tools will help us determine whether dark energy is truly constant, or if we're glimpsing new physics.
Zoom: https://uchicago.zoom.us/j/95272936732?pwd=HXVM29KzFU5yU8WvLcXKsxa9HlSTLE.1
