Navier–Stokes Existence and Smoothness
Table of Contents
The motion of a viscous incompressible fluid is described by the Navier–Stokes equations, first written down by Claude-Louis Navier in 1822 and given their modern form by George Gabriel Stokes. Whether smooth solutions to these equations can always be continued for all time (or whether they can spontaneously develop a singularity at some finite time) is one of the deepest open problems in mathematics, and one of the seven Clay Millennium Prize Problems, carrying a 1,000,000$ prize for a solution.
Let $u_0 : \mathbb{R}^3 \to \mathbb{R}^3$ be a smooth divergence-free vector field. Does there exist a smooth solution $u(x,t)$, $p(x,t)$ to the 3D incompressible Navier–Stokes equations $$\partial_t u + (u \cdot \nabla)u - \nu\Delta u + \nabla p = 0, \qquad \nabla \cdot u = 0, \qquad u(\cdot,0) = u_0$$ defined for all $t > 0$ and satisfying $\int_{\mathbb{R}^3}|u(x,t)|^2,dx < C$ for all $t \geq 0$? A solution or a counterexample (a smooth $u_0$ for which no such smooth solution exists) both qualify for the prize.
The Equations and Their Scaling #
Compared to the Euler equations (which describe inviscid flow), the Navier–Stokes equations add the viscous term $\nu\Delta u$, where $\nu > 0$ is the kinematic viscosity. This term dissipates energy and regularises the flow locally. The central tension is that the nonlinear term $(u\cdot\nabla)u$ can concentrate energy at small spatial scales faster than viscosity can diffuse it away.
Scaling symmetry. The Navier–Stokes equations are invariant under the rescaling $$u(x,t) \mapsto \lambda u(\lambda x,, \lambda^2 t), \qquad p(x,t) \mapsto \lambda^2 p(\lambda x,, \lambda^2 t).$$ A norm is critical (or scale-invariant) if it is preserved by this rescaling. The critical norm in $L^p(\mathbb{R}^3)$ is $L^3$, since $|\lambda u(\lambda\cdot)| _{L^3} = |u| _{L^3}$. The energy norm $|u| _{L^2}$ is subcritical: it scales as $\lambda^{1/2}|u| _{L^2}$, which shrinks under the rescaling $\lambda \to \infty$ (i.e., zoom into small scales). This mismatch is the core of the difficulty: global energy control does not prevent concentration at arbitrarily small scales.
2D global regularity. In two dimensions the scaling is different: the enstrophy $|\nabla u|_{L^2}^2$ is scale-invariant and is controlled by the energy. Global regularity in 2D follows from this enstrophy estimate, a fact known since the 1960s. In 3D no analogous critical quantity is controlled globally, and the problem is open.
The Hierarchy of Known Results #
Leray–Hopf Weak Solutions (1934) #
For any $u_0 \in L^2(\mathbb{R}^3)$ divergence-free, there exists a global weak solution $u \in L^\infty(0,\infty;, L^2) \cap L^2(0,\infty;, H^1)$ satisfying the energy inequality $$|u(t)| _{L^2}^2 + 2\nu\int _0^t |\nabla u| _{L^2}^2, ds \leq |u_0| _{L^2}^2.$$
Leray’s construction, via a compactness argument on regularised equations, produces a solution that is globally defined but potentially not smooth, and the term “weak” refers to the fact that the equations are satisfied only in an integral (distributional) sense, not pointwise. The energy inequality is the only bound available globally. Whether Leray–Hopf solutions are unique, or whether they are the same as smooth solutions when the initial data is smooth, is unknown.
Partial Regularity: The CKN Theorem #
The best known result limiting the size of potential singularities is the following.
For any suitable weak solution to the 3D Navier–Stokes equations, the set of space-time singular points has parabolic Hausdorff dimension at most 1. In particular, at any given time the spatial singular set has Hausdorff dimension at most $\dfrac{1}{2}$.
A “suitable weak solution” is a weak solution satisfying a local energy inequality. The CKN theorem proves that singularities, if they exist, cannot fill a curve or surface: they can occupy at most a set of dimension one in space-time. This is the most quantitative partial regularity result available and was simplified by Lin (1998). Scheffer (1977) had earlier shown singular times have Hausdorff dimension at most $\dfrac{1}{2}$.
Conditional Regularity: Ladyzhenskaya–Prodi–Serrin #
If a weak solution additionally satisfies $u \in L^r(0,T;, L^s(\mathbb{R}^3))$ with $\dfrac{2}{r} + \dfrac{3}{s} = 1$ and $3 < s \leq \infty$, then $u$ is smooth on $(0,T]$.
The condition $\dfrac{2}{r} + \dfrac{3}{s} = 1$ is precisely the scale-invariant line in the $(r,s)$ plane: membership in any of these spaces implies regularity. The family ranges from $(r,s)=(\infty, 3)$ (critical $L^3$ control in space, uniform in time) to $(r,s)=(2,\infty)$ (square-integrable $L^\infty$ control in time). These are conditional results: they do not prove that a weak solution lies in such a space, only that if it does, it must be smooth.
The Critical Endpoint: Escauriaza–Seregin–Šverák #
If $u$ is a Leray–Hopf weak solution with $\sup _{t \in [0,T^*)} |u(\cdot,t)| _{L^3(\mathbb{R}^3)} < \infty$, then $u$ can be extended as a smooth solution past $T^*$.
The endpoint case $s=3$ of the LPS family is the critical one: $L^3(\mathbb{R}^3)$ is exactly the scale-invariant norm for Navier–Stokes. The ESS proof is substantially harder than the subcritical cases; it uses a compactness argument to reduce to a smooth, backwards self-similar solution and then invokes a backwards uniqueness theorem for parabolic equations to rule it out.
Tao’s Quantitative Criterion #
If a smooth finite-energy solution first becomes singular at time $T^*$, then $$\limsup_{t \uparrow T^*} \dfrac{|u(\cdot,t)| _{L^3(\mathbb{R}^3)}}{\bigl(\log\log\log\tfrac{1}{T^*-t}\bigr)^c} = \infty$$ for some absolute constant $c>0$. In particular, the critical $L^3$ norm must blow up at least as fast as a triple-logarithm in $(T^*-t)^{-1}$.
Tao’s result is the first supercritical regularity criterion for Navier–Stokes: it gives quantitative information about the blowup rate that goes (by a triple logarithm) beyond what scaling alone can detect. The proof quantifies the compactness arguments in the ESS proof, replacing each use of a compactness method by an explicit Carleman inequality, and propagates lower bounds for the vorticity across dyadic annuli. The triple-exponential dependence in Tao’s bound has since been localised and sharpened by Barker–Prange (2021) and others.
The Supercriticality Problem #
The fundamental analytical obstruction is that Navier–Stokes is supercritical with respect to the only globally controlled norm ($L^2$): the energy.
Define the critical regularity index as the Sobolev exponent $s$ such that $\dot{H}^s(\mathbb{R}^3)$ is scale-invariant. For Navier–Stokes, $s = 1/2$. The energy controls $\dot{H}^0 = L^2$ (subcritical), and regularity theory requires control at $\dot{H}^1$ (critical viscous norm) or $L^3$ (critical Lebesgue norm). There is a regularity gap between what is globally available ($L^2$) and what is needed ($L^3$ or $\dot{H}^1$). Every known approach to closing this gap runs into the same obstruction: the nonlinearity can create structure at arbitrarily small scales that the subcritical $L^2$ bound cannot see.
Tao (2016) made this gap precise by constructing an averaged Navier–Stokes system, where the bilinear nonlinearity $(u\cdot\nabla)u$ is replaced by a carefully designed convex average of related nonlinearities, for which finite-time blowup can be rigorously proved. This construction does not produce a counterexample to the true Navier–Stokes equations, but it demonstrates that the specific algebraic structure of the nonlinearity is load-bearing: any proof of global regularity must use something specific about $(u\cdot\nabla)u$ that is not shared by its averages.
Research Directions #
1. Improving the Quantitative Blowup Rate #
Tao’s triple-logarithmic rate is the sharpest known lower bound on blowup of the critical $L^3$ norm. Scaling considerations suggest that the true rate, if blowup occurs, should be much faster; conjecturally $|u|_{L^3} \sim (T^*-t)^{-\delta}$ for some $\delta > 0$, analogous to Type I blowup in nonlinear heat equations. The gap between the triple-logarithmic lower bound and the conjectured power-law rate represents the frontier of quantitative regularity theory. Closing even part of this gap, for instance establishing a single-logarithmic or power-of-log lower bound, would require new ideas beyond Carleman estimates.
2. Type I vs. Type II Blowup #
A blowup is called Type I if the scale-invariant norm $|u(\cdot,t)|_{L^3}$ grows no faster than $O((T^-t)^{-1/2})$ near $T^$. It is Type II otherwise. For the Navier–Stokes equations, ruling out Type I blowup would be a significant advance: all self-similar singularities (where $u(x,t) = (T^*-t)^{-1/2}U(x/(T^*-t)^{1/2})$) are of Type I, and several results (including work of Ružička and Seregin) already rule them out under mild additional assumptions. Whether all Type I blowup can be excluded, leaving only the less structured Type II, is open.
3. Uniqueness of Weak Solutions #
Leray–Hopf weak solutions exist globally, but they may not be unique. This is a separate, equally deep question: even if all smooth solutions extend globally, one must also ask whether weak solutions coincide with smooth ones when started from smooth data. Recent work of Buckmaster and Vicol (2019) showed that weak solutions below the Ladyzhenskaya–Prodi–Serrin threshold are indeed non-unique, using convex integration techniques developed for the Euler equations (De Lellis–Székelyhidi). Whether Leray–Hopf solutions with the energy inequality are unique is still open and is perhaps the central problem in the weak solution theory.
4. Self-Similar and Discretely Self-Similar Solutions #
Self-similar solutions of the form $u(x,t) = (T^*-t)^{-1/2} U(x/(T^*-t)^{1/2})$ satisfy a nonlinear elliptic system for the profile $U$. Several non-existence theorems show that backward self-similar solutions with certain integrability must be trivial (Nečas–Ružička–Šverák, 1996). The case of discretely self-similar solutions, where $u(x,t) = \lambda u(\lambda x, \lambda^2 t)$ for a fixed $\lambda \neq 1$, is less understood and was recently revisited. Whether the set of self-similar profiles that could appear as blowup limits is empty is not known.
5. Computer-Assisted Proofs via Rigorous Numerics #
The Chen–Hou approach to Euler singularities (2025) used a computer-assisted proof framework: construct a numerical approximate profile, then verify its stability rigorously using interval arithmetic. For Navier–Stokes the presence of viscosity complicates such an approach (the profile is dissipated rather than transported), but the same framework (dynamical rescaling plus nonlinear stability verification) might in principle detect or rule out singularities in specific axi-symmetric geometries. Applying and adapting the Hou group’s methods to the viscous problem is an active direction.
6. The Zero-Viscosity Limit and Euler–Navier–Stokes Connection #
As $\nu \to 0$, Navier–Stokes formally converges to Euler. The precise relationship is subtle: in the presence of boundaries (Prandtl layers) or after a potential Euler singularity, the zero-viscosity limit can fail to hold in strong norms. If Euler develops a finite-time singularity at time $T^*_E$ from smooth data (as Chen–Hou suggest for bounded domains), then for small $\nu$ the Navier–Stokes solution must either also develop a near-singularity or be regularised by viscosity before $T^*_E$. Whether viscosity is always sufficient to regularise an Euler singularity, or whether a Navier–Stokes singularity can arise from a nearby Euler one, is entirely open.
References #
- Fefferman, C. L. (2000). Existence and smoothness of the Navier–Stokes equation. Clay Mathematics Institute Millennium Prize Problems. https://www.claymath.org/wp-content/uploads/2022/06/navierstokes.pdf
- Leray, J. (1934). Sur le mouvement d’un liquide visqueux emplissant l’espace. Acta Mathematica, 63, 193–248.
- Hopf, E. (1951). Über die Anfangswertaufgabe für die hydrodynamischen Grundgleichungen. Mathematische Nachrichten, 4(1–6), 213–231.
- Caffarelli, L., Kohn, R., & Nirenberg, L. (1982). Partial regularity of suitable weak solutions of the Navier–Stokes equations. Communications on Pure and Applied Mathematics, 35(6), 771–831.
- Ladyzhenskaya, O. A. (1967). On uniqueness and smoothness of generalized solutions to the Navier–Stokes equations. Zapiski Nauchnykh Seminarov LOMI, 5, 169–185.
- Escauriaza, L., Seregin, G. A., & Šverák, V. (2003). $L_{3,\infty}$-solutions of the Navier–Stokes equations and backward uniqueness. Russian Mathematical Surveys, 58(2), 211–250.
- Tao, T. (2019). Quantitative bounds for critically bounded solutions to the Navier–Stokes equations. arXiv:1908.04958. Published in Nine Mathematical Challenges, AMS, 2021, pp. 149–193.
- Tao, T. (2016). Finite time blowup for an averaged three-dimensional Navier–Stokes equation. Journal of the American Mathematical Society, 29(3), 601–674.
- Buckmaster, T. & Vicol, V. (2019). Nonuniqueness of weak solutions to the Navier–Stokes equation. Annals of Mathematics, 189(1), 101–144.
- Barker, T. & Prange, C. (2021). Localized quantitative estimates and potential blow-up rates for the Navier–Stokes equations. Communications in Mathematical Physics, 385, 717–792.