SFA III: Planet formation and atmospheres

"Zonal flows" are narrow, fast bands of fluid all moving in one direction – Earth's jet stream, the colorful stripes of Jupiter, and ocean currents are familiar examples. Remarkably, the same phenomenon appears in the disks of gas and dust around young stars, where planets are born. These bands emerge from tiny turbulent fluctuations that global simulations cannot resolve. This project uses highly resolved local simulations to train a physics-informed neural network that captures the effect of the small-scale turbulence, then plugs it into global models. One exciting application: testing whether the mysterious dark rings seen in planet-forming disks by the ALMA telescope are caused by zonal flows rather than hidden planets. It is multi-scale because large ordered bands emerge from unresolvably small turbulence, and interdisciplinary because the identical physics connects protoplanetary disks, planetary atmospheres, and Earth's climate and ocean currents.

Telescopes can now analyze the atmospheres of planets orbiting distant stars – but the planet's faint signal is buried under light from its host star and interference from Earth's own atmosphere, like a whisper in a crowded room. Today's detection methods rely on comparing data with pre-computed model templates, which can miss anything the models did not anticipate. This project uses physics-informed neural networks – AI constrained by the laws of physics – to extract planetary signals directly from the data. It combines simple one-dimensional chemistry models with full three-dimensional climate models of these alien worlds, plus descriptions of processes too small to simulate directly, such as clouds and turbulence. It is multi-scale because tiny, local processes like cloud formation shape the global appearance of a planet's atmosphere, and interdisciplinary because it merges astronomy, atmospheric and climate science, and machine-learning research.