Yale is joining a new National Science Foundation-funded center focused on understanding and controlling turbulence, a problem that affects systems from oceans and atmospheres to fusion reactors and spacecraft. The $30 million TEMPEST collaboration, based at Michigan State University, will combine theory, computation, artificial intelligence and experiments to develop predictive models for consequential applications.
Key facts
- The National Science Foundation announced a $30 million grant for TEMPEST on Aug. 26.
- TEMPEST stands for Transformative Explorations in Multi-Physics and Engineering of Scientific Turbulence.
- The center will be based at Michigan State University and includes Yale researchers Anna Gilbert and Rajit Manohar.
- The collaboration aims to build predictive models of real-world turbulence and the computing hardware needed to run them.
What this means for universities and researchers
TEMPEST creates a cross-institution research effort linking physics, mathematics, engineering and AI. Yale’s role includes work on physics-informed AI and machine learning, experimental-data decisions, and computing architectures, while researchers across the center will connect theory, experiments and computation around turbulence problems.
The center brings together researchers from Yale, Michigan State University, Baylor University, Auburn University, the University of Rochester, San Jose State University, and Texas A&M University-Corpus Christi. Los Alamos National Laboratory, Sandia National Laboratories, Lawrence Livermore National Laboratory, Pacific Fusion, and General Atomics are listed as unfunded partners.
What will the TEMPEST collaboration do?
TEMPEST will seek to understand and control turbulence under real-world conditions, where multiple physical effects can interact across scales. Its researchers will integrate theory, computation, AI techniques and experimentation to create predictive models for applications in which reliable treatment of turbulence is important.
Turbulence has long been an unsolved problem in physics. Standard textbook cases can isolate an ordinary fluid, a simple geometry and one physical effect, with a separation between large energy-containing scales and smaller scales where energy dissipates. Under those conditions, turbulence can show universal statistics and simulations can use generic rules for unresolved scales.
Many practical and extreme settings do not follow those simplified conditions. Magnetic fields, radiation, chemistry and particle-scale effects can operate together in phenomena such as fusion-reactor turbulence, flow around a reentering spacecraft, and turbulence in a dying star. The center’s work is intended to address this more complex landscape.
How will Yale contribute to the center?
Anna Gilbert, the John C. Malone Professor of Statistics and Data Science in Yale’s Faculty of Arts and Sciences, will lead work described as the center’s AI and machine-learning connector. Her group will work with mathematicians to translate provable features of physical systems into constraints for models, and use data to identify results that may merit further proof.
Gilbert’s group will also work with experimentalists on what data to collect and how much to measure so that measurements are meaningful. Across applications, the work will involve analysis, benchmarking, validation and verification of computational methods and mathematical approximations, alongside their integration with empirical results.
Rajit Manohar, the John C. Malone Professor of Electrical & Computer Engineering and Computer Science in Yale Engineering, will focus on machines capable of running the models. His group will explore computing architectures, including clockless chips designed alongside related algorithms to use a fraction of the power required by current hardware.
Why does better turbulence modelling matter?
Scientists say greater understanding of turbulence could support challenges across science and engineering. Turbulence plays a role in energy movement, the spread of pollutants and heat through oceans and the atmosphere, fluid motion in chemical processing plants, and plasma behavior in extreme environments.
The stated goal is not only to develop models, but to develop models and hardware that can be trusted because they are built on the relevant physics. That focus connects the center’s mathematical, experimental, AI and hardware work.
Frequently asked questions
What is the TEMPEST collaboration?
TEMPEST, short for Transformative Explorations in Multi-Physics and Engineering of Scientific Turbulence, is an NSF-funded science and technology center based at Michigan State University. It will bring together researchers in physics, mathematics, engineering and AI to understand and control turbulence.
How much funding did TEMPEST receive?
The National Science Foundation announced a $30 million grant for TEMPEST on Aug. 26. The funding supports the center’s work on predictive models of real-world turbulence and related computing approaches.
