Papex

Open problem · Physics · Open

Turbulence

Can a theory predict the statistics and internal structures of turbulent flow?

Summary

Most flows in nature and engineering are turbulent, with interacting eddies over a wide range of sizes; despite statistical theories since Kolmogorov, a complete description is lacking, and Feynman called turbulence the most important unsolved problem of classical physics.

Why it is open

Turbulent incompressible flows are modelled by the Navier–Stokes equations [10], but Andrey Kolmogorov's influential 1941 statistical theory rests on similarity hypotheses, including local isotropy, rather than a derivation from those equations [1]. Measured departures from its scaling predictions, known as intermittency, remain unresolved [15]. Pressure couples fluid-particle motion nonlocally, making its statistics difficult to derive [3].

There are several distinct questions. The “zeroth law” asks whether energy dissipation remains finite as viscosity vanishes. In 2023, Elia Bruè and Camillo De Lellis described strong experimental support but an unresolved rigorous result with uniformly regular forcing and initial data [9]. Whether smooth solutions break down is the separate Navier–Stokes existence and smoothness problem. The onset of turbulence also depends on the flow geometry and disturbances [8].

Outstanding goals include analytical predictions of anomalous scaling exponents [15], a proof of the zeroth law under uniform bounds [9], and direct measurement of transition critical exponents [8]. These address different parts of turbulence, rather than one interchangeable test of success.

What is known

Kolmogorov's theory (1941–1962). Kolmogorov pictured large eddies passing energy to ever smaller ones until viscosity turns it into heat [1]. He assumed that at high Reynolds numbers (R = LU/ν, for a typical length L, velocity U and viscosity ν) the small-scale statistics away from boundaries depend only on ν and the mean dissipation rate per unit mass, ε, and at intermediate scales on ε alone; the mean square velocity difference across a distance r then grows as (εr) to the power 2/3 [1]. The predicted structure function, the moment of order p of velocity differences, grows as r to the power p/3; the four-fifths law relates its third moment to ε and r [2, 10]. In 1962 Kolmogorov incorporated fluctuating dissipation using a lognormal hypothesis, modifying the exponents [2].

Measured statistics (2003–2026). Simulations by Yukio Kaneda and colleagues on up to 4096³ grid points gave an energy spectrum close to Kolmogorov's, with an exponent steeper by about 0.1 [4]. From their group's simulations, Kartik Iyer, Katepalli Sreenivasan and P. K. Yeung reported in 2020 that the four-fifths law holds over an intermediate range of scales, that the second-order exponent approaches 0.72 rather than 2/3, and that the exponents of transverse velocity differences saturate at about 2 at high orders, which they read as a sign of vortex sheets [7]. A May 2026 preprint by Dipendra Gupta and Gregory Bewley reports longitudinal-exponent saturation in shear layers near 2.2 ± 0.1 at high orders; this experimental setting differs from isotropic simulations [6].

Anomalous dissipation. Grid-turbulence experiments and the near-constant drag of bluff bodies [10], and Kaneda's simulations, in which the normalised dissipation rate tends to a constant as the viscosity decreases [4], support the zeroth law. Gregory Eyink notes that "an extremely slow decay" can never be ruled out, that in pipes and channels there is evidence for the anomaly only along hydraulically rough walls, not smooth ones, and that Kaneda's simulation starts from data with a built-in "quasi-singularity" [10].

Onsager's conjecture (1949–2023). In 1949 Lars Onsager proposed that dissipation could occur without viscosity in "ideal" turbulence, whose velocity could then not be Hölder continuous (velocity differences bounded by a power of distance) with exponent above 1/3 [10]. Energy conservation above 1/3 was proved by Eyink and by Peter Constantin, Weinan E and Edriss Titi; in 2018 Philip Isett constructed weak solutions of the Euler equations (solutions in an averaged sense) with any exponent below 1/3 that do not conserve energy, solving the conjecture [5]. In 2023 Bruè and De Lellis gave smooth forced Navier–Stokes solutions whose dissipation stays bounded away from zero as the viscosity tends to zero [9]. Eyink distinguishes these existence proofs from evidence that the constructed solutions describe physical turbulence: convex integration also permits non-unique solutions, and boundary effects require further treatment [10].

The transition in pipes (1883–2024). Pipe-flow transition depends on disturbance size and type; reducing inlet disturbances can delay its onset [8]. Near onset, turbulence forms localised puffs that, after memoryless waiting times, decay or split; their mean lifetimes grow faster than exponentially with Re but stay finite, and turbulence first becomes sustained where decay and splitting times cross, at Re ≈ 2,040 ± 10 [8]. In 2023 there was no direct confirmation that the transition belongs to the directed-percolation class, in which activity spreads or dies out and cannot restart once extinct everywhere [8]. In 2024 Grégoire Lemoult and colleagues fed measured puff interactions into renormalisation-group calculations and model simulations and concluded that "the universality class of the transition is directed percolation" [11].

Recent claims. In 2024 Alexander Migdal proposed a first-principles account of decaying turbulence in Physics of Fluids, using an Euler ensemble associated with a loop formulation of the Navier–Stokes equations [12, 13]. In long simulations, Akash Rodhiya and Sreenivasan (2026) found the agreement excellent for Birkhoff–Saffman initial spectra, except at large wavenumbers, but not for Loitsianskii–Kolmogorov–Batchelor spectra; LKB energy decay departed from the prediction, although internal spectral structure and length-scale relations still agreed [14]. Bruè and De Lellis (2026) proved that, under minimal assumptions, Migdal's "momentum" variables cannot be a nonnegative measure on any reasonable space of generalised functions unless the underlying distribution is trivial, while also analysing a finite-dimensional discretisation of the formulation [13].

Main approaches

Cascade and multifractal models. Refine Kolmogorov's picture so that the exponents depart from p/3, as in the lognormal model [2] or the multifractal model of Giorgio Parisi and Uriel Frisch, in which local Hölder exponents vary over sets of different fractal dimension [10]. Strongest evidence: many cascade models agree with experiments and simulations [15], and measured multifractal spectra support the bounds that follow from Onsager's argument [10]. Main difficulty: none is yet fully predictive [15], and at high orders the saturating transverse exponents depart from two such models [7].

Solvable models. Compute anomalous exponents in simplified problems. Strongest evidence: for a passive scalar (a quantity such as temperature carried by the flow) in a random velocity field that is Gaussian and white in time, the Kraichnan model, the exponents were derived analytically, "the first analytical description of anomalous scaling laws in turbulence" [3]; in 2026 Mailybaev and Thalabard computed exponents of all orders perturbatively in a shell model, which keeps one variable per band of scales [15]. Main difficulty: the exact methods are hard to extend to nonlinear settings, and the 2026 framework is shown only for a shell model [15].

Ideal turbulence. Treat high-Reynolds-number flow as a dissipative weak solution of the Euler equations, as Onsager proposed [10]. Strongest evidence: the proved 1/3 threshold [5], the forced Navier–Stokes examples [9] and a local, deterministic form of the four-fifths law without probabilistic assumptions [10]. Main difficulty: mathematical existence alone does not establish that a constructed solution describes physical turbulence; solid walls require additional analysis, which Eyink treats separately [10].

Transition as a phase transition. Treat puffs like the active sites of a non-equilibrium phase transition [8, 11]. Strongest evidence: directed-percolation behaviour in Couette and Waleffe flows [8], and the 2024 pipe analysis, checked by experiments that periodically re-create puff patterns [11]. Main difficulty: near the critical point, puff decay and splitting take of order 10⁷ advective time units, so direct observation would need pipes "considerably exceeding 10⁷ diameters in length" [11].

Exact coherent states. Describe turbulence through exact unstable solutions of the Navier–Stokes equations, such as travelling waves [8]. Strongest evidence: flow states resembling travelling waves have been observed in turbulent pipe flow [8]. Main difficulty: the number of such states needed "may grow explosively" with Re, and the bifurcations that lead to puffs are still unknown [8].

Open sub-questions

  1. Longitudinal saturation. Does the longitudinal saturation reported in shear layers also occur in homogeneous isotropic flow, beyond the reliable moment orders in Iyer and colleagues’ simulations [6, 7]?
  2. Dissipation from smooth data. In periodic-box simulations started from smooth data rather than a developed spectrum, does the normalised dissipation rate tend to a nonzero constant as the viscosity decreases [4, 10]?
  3. Smooth and rough walls. In pipes and channels, does a dissipation rate independent of the Reynolds number appear only with hydraulically rough walls [10]?
  4. Viscosity-independent forcing. Can anomalous dissipation in forced Navier–Stokes flows be obtained with one force that does not depend on the viscosity, or on time, as Bruè and De Lellis ask [9]?
  5. Critical exponents in pipes. Can the turbulent fraction near Re ≈ 2,040 be measured directly enough to give critical exponents, and do they match directed percolation [8, 11]?
  6. Loitsianskii–Kolmogorov–Batchelor decay. At higher Reynolds numbers or later times, does this case approach the energy decay predicted by Migdal's Euler ensemble [12, 14]?

Work on this in Papex

If you have a testable answer to one of the sub-questions above, you can draft it as a private hypothesis in Papex, or prepare the draft from Claude or ChatGPT with the Papex connector. You review the exact version, sign it and decide whether to publish it; a public record that names this problem is listed on this page. A Papex record shows who signed which version and when. It is not peer review and does not mean the hypothesis is correct.

Key references

  1. Kolmogorov, A. N. (1991). The local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers. Proceedings of the Royal Society of London A 434(1890), 9–13. English translation by V. Levin of the paper first published in Doklady Akademii Nauk SSSR 30(4) (1941). 10.1098/rspa.1991.0075 · Checked 15 Oct 2026
  2. Kolmogorov, A. N. (1962). A refinement of previous hypotheses concerning the local structure of turbulence in a viscous incompressible fluid at high Reynolds number. Journal of Fluid Mechanics 13(1), 82–85. 10.1017/S0022112062000518 · Checked 15 Oct 2026
  3. Falkovich, G., Gawędzki, K. & Vergassola, M. (2001). Particles and fields in fluid turbulence. Reviews of Modern Physics 73(4), 913–975. 10.1103/RevModPhys.73.913 · Checked 15 Oct 2026
  4. Kaneda, Y., Ishihara, T., Yokokawa, M., Itakura, K. & Uno, A. (2003). Energy dissipation rate and energy spectrum in high resolution direct numerical simulations of turbulence in a periodic box. Physics of Fluids 15(2), L21–L24. 10.1063/1.1539855 · Checked 15 Oct 2026
  5. Isett, P. (2018). A proof of Onsager's conjecture. Annals of Mathematics 188(3), 871–963. 10.4007/annals.2018.188.3.4 · Checked 15 Oct 2026
  6. Gupta, D. & Bewley, G. P. (2026). Experimental Evidence for Longitudinal Scaling Exponent Saturation in Shear Turbulence. arXiv:2605.01867v1, preprint. Source · Checked 15 Oct 2026
  7. Iyer, K. P., Sreenivasan, K. R. & Yeung, P. K. (2020). Scaling exponents saturate in three-dimensional isotropic turbulence. Physical Review Fluids 5(5), 054605. 10.1103/PhysRevFluids.5.054605 · Checked 15 Oct 2026
  8. Avila, M., Barkley, D. & Hof, B. (2023). Transition to turbulence in pipe flow. Annual Review of Fluid Mechanics 55, 575–602. 10.1146/annurev-fluid-120720-025957 · Checked 15 Oct 2026
  9. Bruè, E. & De Lellis, C. (2023). Anomalous dissipation for the forced 3D Navier–Stokes equations. Communications in Mathematical Physics 400(3), 1507–1533. 10.1007/s00220-022-04626-0 · Checked 15 Oct 2026
  10. Eyink, G. (2024). Onsager's 'ideal turbulence' theory. Journal of Fluid Mechanics 988, P1. 10.1017/jfm.2024.415 · Checked 15 Oct 2026
  11. Lemoult, G., Mukund, V., Shih, H.-Y., Linga, G., Mathiesen, J., Goldenfeld, N. & Hof, B. (2024). Directed percolation and puff jamming near the transition to pipe turbulence. Nature Physics 20(8), 1339–1345. 10.1038/s41567-024-02513-0 · Checked 15 Oct 2026
  12. Migdal, A. (2024). Quantum solution of classical turbulence: decaying energy spectrum. Physics of Fluids 36(9), 095161. 10.1063/5.0228660 · Checked 15 Oct 2026
  13. Bruè, E. & De Lellis, C. (2026). Some rigorous remarks on Migdal's momentum loop equation. Journal of Mathematical Fluid Mechanics 28(3), 45. 10.1007/s00021-026-01023-8 · Checked 15 Oct 2026
  14. Rodhiya, A. & Sreenivasan, K. R. (2026). The asymptotic state of decaying turbulence. Philosophical Transactions of the Royal Society A 384(2325), 20250021. 10.1098/rsta.2025.0021 · Checked 15 Oct 2026
  15. Mailybaev, A. A. & Thalabard, S. (2026). Perturbative anomalous exponents from Kolmogorov multipliers. Physical Review Letters 137(9), 094002. 10.1103/6qrr-3646 · Checked 15 Oct 2026

Page reviewed 15 Oct 2026

Source

Open, as listed by Wikipedia · Unsolved problems in physics. Checked 08 Oct 2026.

The status is the source’s. Papex does not decide whether a problem is solved.

Records on this problem

No public record addresses this problem yet.

More in Physics