Turbulence — The Last Unsolved Problem in Classical Physics

When Werner Heisenberg was asked what he would ask God if given the chance, he reportedly said: “Why relativity, and why turbulence? I really think He may have an answer to the first question.” Richard Feynman called turbulence “the most important unsolved problem of classical physics.”

Turbulence is everywhere — in the air around a plane wing, in blood flowing through an artery, in the plasma inside a fusion reactor, in the convective cells of stars. We can describe it statistically (Kolmogorov, 1941), simulate it approximately (with enormous computational cost), and predict its gross effects — but we cannot solve it mathematically. The Navier-Stokes equations that govern fluid flow remain one of the seven Millennium Prize Problems, with a $1 million prize unclaimed since 2000.

Confidence: established (Kolmogorov scaling laws, statistical descriptions); emerging (AI/ML approaches to modeling); open (mathematical proof of Navier-Stokes existence/smoothness)

Key Facts

• Navier-Stokes equations: formulated 1822 (Navier) and 1845 (Stokes); governing equations of all fluid motion
• Millennium Prize: the Clay Mathematics Institute offers $1M for proof (or disproof) that smooth solutions always exist for the 3D Navier-Stokes equations. As of 2026, unclaimed.
• Turbulence involves 5+ orders of magnitude in length scale simultaneously — from large energy-containing eddies to tiny dissipation scales (Kolmogorov scale, ~0.1–1 mm in air), all interacting
• The Reynolds number (Re = ρvL/μ) is the key parameter: below ~2300, flow is laminar; above ~4000, turbulence appears. Commercial jets fly at Re ~ 10⁷.
• Turbulence is irreversible despite time-symmetric governing equations — a deep connection to the concept-arrow-of-time problem
• Clear-air turbulence (invisible to radar) increased 55% since the 1970s and is accelerating with climate change

Kolmogorov’s 1941 Theory — A Triumph That Doesn’t Explain Itself

Andrey Kolmogorov’s 1941 papers are among the most influential in 20th-century physics. His key insight — the energy cascade hypothesis:

• Large-scale stirring (wind shear, pipe flow, etc.) injects energy at large scales
• Energy cascades downward through an inertial subrange — large eddies break into smaller eddies, which break into smaller eddies…
• At the smallest scales (Kolmogorov scale), energy dissipates into heat via viscosity
• The energy spectrum in the inertial subrange follows a universal k⁻⁵/³ power law — verified experimentally across an extraordinary range of conditions

What’s remarkable: Kolmogorov’s law works, confirmed in jet streams, laboratory pipe flows, ocean currents, and stellar interiors. But the mechanism by which the cascade develops remains incompletely understood:

• What exactly causes large vortices to break into smaller ones?
• Why does the cascade produce the specific k⁻⁵/³ spectrum rather than something else?
• Why do velocity gradients amplify in a way that is independent of viscosity at high Reynolds numbers? (This is the “anomalous dissipation” problem — viscosity shouldn’t matter, but energy still dissipates. How?)

The 2024 work on Kolmogorov–Zakharov spectra and dissipation-range mode coupling explores these questions but has not resolved them.

Why Turbulence Is Mathematically Hard

The Navier-Stokes equations are nonlinear partial differential equations. The nonlinearity creates:

1. Sensitive dependence on initial conditions (chaos): tiny perturbations grow exponentially, making long-term prediction impossible in principle. Weather is a paradigm case.
2. Scale coupling: motions at all scales interact simultaneously. You can’t separate the large-scale flow from the small-scale turbulence — they are mutually coupled.
3. Blow-up question: mathematicians cannot prove that solutions remain smooth for all time. It’s theoretically possible that the equations develop a singularity (infinite velocity gradient at a point) in finite time. This is the Millennium Prize question.

Claude Shannon’s information theory provides a useful frame: a turbulent flow is a system generating information (new structure) at every scale, faster than any finite observer can track. Turbulence is, in some sense, maximum-entropy fluid motion — the most disordered state a fluid can achieve while still conserving energy and momentum.

The Navier-Stokes Millennium Prize: Current Status (2026)

Still unsolved. The Clay Mathematics Institute’s $1M prize remains unclaimed. Only the Poincaré conjecture has been solved among the original seven Millennium Problems — Navier-Stokes is one of six remaining.

The closest recent progress:

DeepMind + Gómez-Serrano: Unstable Singularities (September 2025, arXiv 2509.14185) — The most significant mathematical advance in years. Javier Gómez-Serrano (Brown University) partnered with Google DeepMind to discover new families of unstable singularities in the Euler equations (the inviscid limit of Navier-Stokes). They embedded mathematical constraints directly into PINNs pushed to near-machine precision, enabling computer-assisted proofs. Significance: mathematicians believe no stable singularities exist for the 3D Euler/Navier-Stokes equations — so understanding any singularity type is a step toward resolving whether blow-up is possible. Manifold Markets prediction markets updated upward (slightly) after this preprint.

Convex integration (Eyink et al., 2025) — Constructs “wild” weak solutions to Euler equations, relevant to Onsager’s conjecture about energy dissipation in inviscid flows. These are mathematically valid but physically unphysical (they violate energy conservation).

Sreenivasan & Schumacher, Annual Reviews (2025): “What Is the Turbulence Problem, and When May We Regard It as Solved?” — A landmark synthesis paper that establishes a taxonomy of what “solving” turbulence would mean (from engineering metrics to full mathematical regularity). With a special Physics of Fluids issue honoring Sreenivasan’s 75th birthday (October 2025).

The Budden claim (2025, not validated): A startup founder claimed an AI-assisted Navier-Stokes solution. The Clay Institute had not validated it as of early 2026; prediction markets moved rapidly against it.

AI/ML Approaches (2024–2026)

The failure of analytical methods has driven increasingly sophisticated ML approaches:

Physics-Informed Neural Networks (PINNs)

• Neural networks trained to satisfy Navier-Stokes as a constraint, not just fit data
• Now pushed to near-machine precision for rigorous computer-assisted mathematical proofs (Gómez-Serrano / DeepMind 2025)
• Can reconstruct 3D turbulent fields from sparse sensor data using underdetermined RANS equations

Explainable AI reveals vortices are overrated (University of Michigan, Phys.org, November 2025): Using XAI to identify which structures in turbulent flow are most physically influential — key finding: Reynolds stresses near and far from the wall and streaks at moderate distances are most important; vortices are less important than previously assumed. This overturns a long-standing intuition in wall turbulence modeling.

Generative AI for subgrid-scale modeling (arXiv, October 2025): Diffusion models and GANs can now represent the full distribution of subgrid-scale stresses conditioned on resolved velocity fields, capturing backscatter of energy from small to large scales — something classical LES models fundamentally cannot do.

Fourier Neural Operators (FNOs): Learn maps between function spaces; can run turbulence simulations 1000× faster than DNS after training. Now mature.

Mamba Neural Operator (NeurIPS 2024): State-space model applied to PDEs; captures long-range spatiotemporal dependencies. Adaptive state-space matrices with alias-free design.

LEX v1.6.0 (February 2026): JAX-based, GPU-accelerated large-eddy simulation model with automatic differentiation, targeting the “gray zone” (~1 km scales) in climate/weather models.

Key limitation: AI models learn to emulate turbulence statistics but do not solve the equations. They generalize poorly to conditions far from their training distribution. Engineering tools, not mathematical proofs.

Real-World Stakes: Why Turbulence Matters

Aviation: Clear-Air Turbulence and Climate Change

Clear-air turbulence (CAT) — the invisible, radar-transparent turbulence encountered at cruise altitude — is the leading cause of in-flight injuries. Critical 2024 findings:

• University of Reading analysis (2024, Geophysical Research Letters): moderate-to-severe CAT increased 60–155% over key routes (North Atlantic, North Pacific, East Asia) between 1980–2021
• 2025 projections: over some busy routes, turbulence expected to “double, treble, or quadruple” in coming decades due to climate change strengthening upper-level wind shear in the jet stream
• Mechanism: warming troposphere + stable stratosphere = greater wind shear at the jet stream core = more Kelvin-Helmholtz instability = more CAT
• Cost: ~$500M/year in US aviation from turbulence damage and injuries; expected to rise significantly

Blood Flow and Cardiovascular Disease

Turbulent blood flow in arteries is not just a physics curiosity — it is a cause of cardiovascular disease:

• Normally, blood flows laminarly through straight vessels; it transitions to turbulence at branching points, curves (aortic arch), and stenoses (narrowed arteries)
• Turbulent flow creates oscillatory shear stress on vessel walls that damages endothelial cells, promotes inflammation, and drives atherosclerotic plaque formation
• This is why atherosclerosis clusters at branching points and curves — where turbulence is highest
• Understanding turbulence in blood flow is an active medical research area with direct implications for cardiovascular treatment

Nuclear Fusion (ITER/Plasma Turbulence)

Plasma turbulence is the primary obstacle to achieving net energy gain in tokamak reactors:

• Hot plasma in a tokamak is supposed to be confined by magnetic fields, but micro-turbulence causes heat to leak out across field lines
• Gyrokinetic turbulence simulation is among the most computationally demanding calculations in science

2024–2025 plasma turbulence breakthroughs:

• Multi-scale coupling surprise (October 2025): When large-scale plasma turbulence decreases, small-scale turbulence intensifies and becomes less spatially deformed — unexpected cross-scale energy coupling that overturns assumptions about confinement optimization
• Rotation doesn’t universally help (ORNL, Frontier exascale supercomputer): Plasma rotation was thought to universally improve confinement; ORNL simulations show this is wrong — the subtle heavy-ion/light-electron interaction must be captured, and rotation can make things worse in some regimes
• Princeton PPPL (2024): Identified a plasma escape mechanism where turbulence disrupts the last confinement surface, widening the heat load on ITER’s divertor plates — this changes ITER engineering assumptions
• Reinforcement learning real-time control: RL is now controlling tearing instabilities in the DIII-D tokamak at 25ms intervals — the first real-time ML-driven turbulence suppression in a fusion device

Turbulence and Deep Physics

Turbulence connects to some of the deepest unresolved questions in physics:

concept-arrow-of-time: The Navier-Stokes equations are time-reversible, but turbulence is irreversible — a turbulent flow never spontaneously “un-mixes.” This is the same asymmetry as thermodynamic entropy, but expressed in fluid dynamics. Turbulence is irreversibility in action.

Quantum turbulence — a universal law discovered (June 2025, FSU): Superfluid helium-4 confines all rotation to quantized vortex filaments (tubes of precisely fixed circulation, set by ℏ). These form quantum turbulence — tangles of quantized vortices that statistically mimic classical turbulence at large scales. The June 2025 FAMU-FSU result discovered a universal law of quantum vortex dynamics: when vortices reconnect, they separate faster than they approach, creating energy bursts. Multiple simultaneous reconnections trigger distinctive quantum turbulence states not seen in classical fluids — potentially explaining the intermittency (rare violent bursts) that plagues classical turbulence models. The Kolmogorov k⁻⁵/³ spectrum appears in quantum turbulence above the inter-vortex spacing scale, suggesting the cascade is a statistical attractor independent of the microscopic physics (classical or quantum). This is one of the most profound universality claims in physics.

Chaos and coherent structures: Turbulence is chaos, but structured chaos. A 2025 paper (arXiv 2506.13417) explores how coherent structures — recognizable swirling patterns within turbulence — coexist with underlying chaos, representing partial ordering of the chaotic attractor. The Ruelle-Takens-Newhouse route to turbulence (1971) proposed chaotic attractors appear after only a few bifurcations — confirmed experimentally. The strange attractor of turbulence is a fractal object in infinite-dimensional phase space; Kolmogorov’s multifractal corrections to K41 are a direct manifestation.

Emergence: turbulence is a paradigm case of concept-emergence — simple rules (Navier-Stokes) producing irreducible complexity. The complexity is not in the rules; it is in how the rules interact across scales.

The Most Unexpected Connection: Brain Turbulence

The most surprising 2024 cross-domain result — turbulence physics applied to psychiatry:

Whole-brain turbulent dynamics predict antidepressant response (Molecular Psychiatry, 2024): Patients with lower brain turbulence amplitude were non-responders to antidepressants; higher amplitude predicted positive response 8 weeks before treatment began. This is clinically actionable — a turbulence measurement of brain dynamics outperforms other baseline predictors.

The framework (Turbulence as a framework for brain dynamics, PubMed 39716558, 2024): The brain operates near criticality (the edge between ordered and chaotic dynamics). “Brain turbulence” is defined as rich spatiotemporal variability in local synchronization of coupled neural signals — formally analogous to fluid turbulence’s spatiotemporal variability in velocity gradients. This measure discriminates between brain states (conscious vs. unconscious, healthy vs. depressed, acute vs. recovered TBI).

Why does this make sense? The mixing properties of turbulence — efficient energy/information transfer across scales — may be the mechanism by which the brain integrates information across regions. The same cascade structure that moves kinetic energy from large to small eddies may move neural information from global networks to local columns. The same mathematical tools (energy spectral analysis, intermittency measures) that describe jet stream turbulence describe cortical dynamics.

This connects turbulence directly to the concept-hard-problem-of-consciousness and information integration theories — Tononi’s Integrated Information Theory predicts that consciousness correlates with the richness of information integration, which turbulent dynamics would maximize near criticality.

Confidence: emerging — quantitatively testable, but the mechanistic interpretation is speculative.

Cross-Realm Connections

• event-bronze-age-collapse: societal collapse as turbulence — cascade instabilities in over-coupled networks, where small perturbations cascade across scales. Both Bronze Age collapse and fluid turbulence are essentially the same mathematical phenomenon (nonlinear cascade dynamics) in different substrates.
• concept-mycelium-networks: mycelium transport networks face turbulence-analog challenges in nutrient flow — flow stability in branched networks is governed by similar mathematics
• concept-distributed-cognition: turbulence is an emergent phenomenon with no “central controller” — same architecture as distributed cognition. Intelligence and fluid chaos may be related varieties of the same mathematical structure.
• concept-fermi-paradox: plasma turbulence has limited fusion energy for 70+ years; if fusion had worked in the 1970s, human civilization would look very different. Turbulence may be a “Great Filter” for technological civilizations — limiting nuclear fusion access.
• Climate / dest-trappist-1: understanding atmospheric turbulence is essential for modeling the habitability of exoplanet atmospheres, including TRAPPIST-1 worlds

The Millennium Prize Question

The Clay Mathematics Institute’s formulation: Do smooth solutions always exist for the 3D Navier-Stokes equations? If yes, prove it. If no, provide a counter-example (a blow-up solution).

As of 2026, no proof exists in either direction. The best current result (Tao, 2016) shows that a modified version of Navier-Stokes can blow up — but not the real physical equations. The gap between what mathematicians can prove and what physics demands has been a productive frontier for decades.

If a proof of existence were found, it would likely require fundamentally new mathematical tools — and those tools might illuminate why turbulence behaves the way it does, potentially breaking open the physics as well.

See Also

• event-bronze-age-collapse — cascade dynamics in complex systems
• concept-emergence — simple rules, complex behavior
• concept-mycelium-networks — distributed networks and flow
• concept-fermi-paradox — turbulence as a potential filter on technological civilizations
• concept-distributed-cognition — decentralized complexity