Abstract

Turbulence in fluid simulation is typically treated as an emergent and largely uncontrollable phenomenon, where visual outcomes are difficult to predict and even harder to shape. This limits its use as a controllable element in graphics and visual effects.
We introduce a mechanism-aware extension of Stable Fluids that turns turbulence into a designable feature. By exposing the underlying operator geometry as a controllable primitive, our method enables artists and simulation designers to directly control when turbulence emerges, which scales dominate, and how visible breakup evolves, all from the same initial condition.
Unlike conventional approaches that rely on global parameter tuning, our framework provides a scale-resolved and time-local control mechanism via a shell-time field, allowing predictable and repeatable manipulation of turbulence onset and structure.
This transforms turbulence from an uncontrollable byproduct into a controllable visual effect, opening a new direction for physics-based simulation in computer graphics.

Bio.

Min-Jhe Lu is an Assistant Professor at the Institute of Computational and Modeling Science at National Tsing Hua University. His research focuses on multiscale dynamics, geometric structures in nonlinear systems, and mechanism-aware modeling of complex physical and biological processes.
His recent work explores how operator geometry and spectral structure can be leveraged to design and control emergent phenomena, with applications ranging from active matter and tumor growth to turbulence control in fluid simulation.

All faculty and students are welcome to join.