Mathematicians Want To Make Fluid Equations Glitch Out

Episode Summary

Executive Summary: The episode explains how mathematicians are using AI-powered methods to search for “blow-ups” or singularities in the Navier-Stokes equations, which model fluid motion. It contrasts stable singularities found in simplified systems with the harder-to-find unstable ones, and highlights recent progress from neural-network-based approaches that may bring researchers closer to resolving one of mathematics’ biggest open problems.

Main Topics: Navier-Stokes equations as a math and physics cornerstone (Priority: 5/5): The hosts frame Navier-Stokes as the fluid analogue of F=ma: a foundational differential-equation system that accurately models real fluids but remains mathematically mysterious. Blow-ups, singularities, and mathematical glitches (Priority: 5/5): The conversation defines the key open question: whether solutions can become infinite/undefined in a way that lacks physical meaning, and why such behavior matters mathematically. Simulation as a tool and its limitations (Priority: 4/5): Researchers use computer simulations to search for candidate singularities, but digital approximations can miss unstable blow-ups because of finite precision and discretization artifacts. Stable vs. unstable singularities (Priority: 5/5): A central distinction is whether a blow-up persists under tiny perturbations (stable) or disappears with minuscule changes (unstable); unstable ones are harder to detect numerically. Earlier progress in simplified fluid systems (Priority: 4/5): The episode reviews a 2013 candidate singularity in an Euler-equation setup with cylindrical symmetry that was later rigorously proven, showing stable blow-ups can exist in constrained settings. AI and neural networks as a new blow-up-finding method (Priority: 5/5): Researchers are now using physically informed neural networks to search for singularities directly, rather than stepping a simulation forward in time, yielding several unstable candidate blow-ups. Open problems and the next frontier (Priority: 4/5): The field is moving toward boundary-free 3D Euler and ultimately full 3D Navier-Stokes with viscosity, but the final Clay prize problem remains unresolved.

Key Arguments: Navier-Stokes is mathematically rich even though it successfully models real-world fluids; the open question is whether all of its solutions are physically meaningful. A blow-up/singularity is a point where a quantity becomes infinite or undefined, which is acceptable in abstract math but not in a realistic fluid model. Computer simulations are useful for detecting candidate singularities, but unstable singularities may evade them because small numerical errors change the outcome. Stable singularities can survive perturbation and therefore can be supported by simulations; unstable singularities require extreme precision and are harder to capture numerically. AI-based approaches using physically informed neural networks can search for a full solution directly, reducing the sensitivity problems of time-stepped simulation. Recent work has produced multiple unstable candidate blow-ups in several fluid setups, suggesting the method may be nearing more difficult cases. The field is progressing incrementally through simplified models before attempting the full Navier-Stokes equations with 3D geometry and viscosity.

Data Points: Navier-Stokes age: About 200 years - The equations were devised by two physicists roughly two centuries ago. Clay prize: $1 million - Awarded for proving whether Navier-Stokes has smooth global solutions or blow-ups. Stable candidate blow-up found: 2013 - Thomas Ho and Go Luo found a candidate singularity in a cylindrical Euler setup. Rigorous proof timeline: About 10 years - Ho and graduate student Jia Jae Chen later proved the 2013 candidate blow-up was real in the simplified system. New AI results: 5 to 10 - The latest AI-driven work unveiled several new unstable candidate blow-ups across three fluid setups. Dimensional frontier: 2D and 3D - The research progresses through simplified fluid geometries and dimensions before the full 3D Navier-Stokes case.

Pivotal Quotes: "Do all of their solutions make sense? Are they physical at all places, at all times?" — Charlie Wood: Summarizing the core mathematical question behind Navier-Stokes singularities. "A stable singularity is one in which, as the fluid is evolving towards this blow-up, if you poke it a little bit, it does not change its fate." — Charlie Wood: Explaining the stable-versus-unstable distinction. "It’s like trying to balance a pencil on its tip." — Tristan Buckmaster (as quoted by Charlie Wood): Describing the difficulty of using neural networks to capture unstable singular behavior directly.

Implications: AI is becoming a serious new tool for discovering extreme behavior in PDEs, potentially accelerating progress toward one of mathematics’ biggest prizes. For fluid modeling, the work sharpens the gap between simulations and exact theory.