Leray-Hopf Weak Solution

• A Leray–Hopf weak solution is a distributional divergence-free solution to the incompressible Navier–Stokes equations that satisfies the energy inequality, initial conditions, and appropriate boundary constraints, covering various domains and equations with fractional models, manifolds, and periodic domains
• Leray's theorem guarantees the global existence of at least one such solution for $L^2$ initial data in three dimensions, but uniqueness remains an open problem in the classical setting.
• Numerical approximation and uniqueness criteria involve approximation schemes and additional regularity assumptions, including spatial and temporal regularity in different geometries and viscosities

A Leray–Hopf weak solution is an energy-class weak solution of the incompressible Navier–Stokes equations that satisfies the distributional equations, incompressibility, appropriate initial-data attainment, and a global energy inequality. For the three-dimensional system, it is the principal global existence concept: Leray’s theorem provides at least one such solution for every divergence-free $L^2$ initial datum, while uniqueness and global regularity for arbitrary data remain unresolved in the classical setting. In Euclidean domains, manifolds, bounded domains, periodic domains, fractional models, and coupled systems, the precise definition is adapted to the dissipation, geometry, forcing, and boundary conditions.

1. Definition and analytical framework

For the incompressible Navier–Stokes equations on $\mathbb R^3$ with viscosity normalized to one,

$$\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$$

a Leray–Hopf weak solution with $u_0\in L^2_\sigma(\mathbb R^3)$ satisfies

$$u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$$

for every $T&gt;0, together with weak temporal continuity</p> <p>$u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$$</p> <p>The initial datum is attained strongly in$$L^2$:$\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$$The weak formulation is imposed against divergence-free test fields</p> <p>$$\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$$</p> <p>The pressure does not appear because</p> <p>$$\int_{\mathbb R^3}\nabla p\cdot\varphi\,dx=0.$$</p> <p>The defining admissibility condition is the energy inequality</p> <p>$$\mathbb R^3$0

or, with viscosity $\mathbb R^3$1,

$\mathbb R^3$2

The two components of the energy estimate have distinct meanings: $\mathbb R^3$3 is kinetic energy, while the time integral of $\mathbb R^3$4 is viscous dissipation. The cancellation

$\mathbb R^3$5

for divergence-free fields is the fundamental mechanism underlying the formal energy balance.

On bounded Lipschitz domains $\mathbb R^3$6 with no-slip boundary condition, the standard spaces are

$\mathbb R^3$7

and

$\mathbb R^3$8

A Leray–Hopf solution belongs to

$\mathbb R^3$9

satisfies the weak equation in the Gelfand framework

$\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$0

and obeys

$\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$1

The same structure appears in bounded-domain existence theory and in convergence proofs for numerical schemes (Kuroki et al., 2018, Weber, 15 Sep 2025).

2. Energy inequalities, representatives, and temporal regularity

The ordinary global energy inequality begins at the initial time. A stronger formulation is the strong energy inequality, which restarts the estimate at almost every positive time: $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$2 for almost every $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$3 and every $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$4.

The distinction is essential because weak solutions are temporal equivalence classes and their $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$5 representatives need not have canonical values at every time. In the periodic setting, every weak solution satisfies an almost-everywhere differential identity

$\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$6

This identity alone does not imply the integrated strong energy inequality, because the classical chain rule for $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$7 is not automatically available.

A representative-based characterization is given by the equivalence between the Leray–Hopf property and monotone-decreasing kinetic energy. Specifically, a weak solution on $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$8 is Leray–Hopf precisely when it possesses a representative $\partial_tu+(u\cdot\nabla)u-\Delta u+\nabla p=0, \qquad \nabla\cdot u=0, \qquad u(0)=u_0,$9 such that

$u_0\in L^2_\sigma(\mathbb R^3)$0

is monotone decreasing and

$u_0\in L^2_\sigma(\mathbb R^3)$1

Monotonicity permits recovery of the integrated strong energy inequality, although possible singular parts of the monotone energy function are not excluded (Lee, 2022).

The energy inequality yields strong attainment of the initial datum in the standard Leray representative: $u_0\in L^2_\sigma(\mathbb R^3)$2 Weak temporal continuity is generally available throughout the interval, while strong continuity at every positive time is more delicate in three dimensions. Additional regularity assumptions can improve this behavior. For example, critical initial-data conditions in $u_0\in L^2_\sigma(\mathbb R^3)$3 and suitable Besov subspaces yield quantitative convergence toward the heat flow,

$u_0\in L^2_\sigma(\mathbb R^3)$4

which is used to initiate a weak–strong uniqueness argument at $u_0\in L^2_\sigma(\mathbb R^3)$5 (Barker, 2016).

3. Existence, approximation, and compactness

Leray–Hopf existence is obtained by approximation, most commonly through Galerkin schemes. On a bounded domain, one selects finite-dimensional solenoidal spaces

$u_0\in L^2_\sigma(\mathbb R^3)$6

and seeks

$u_0\in L^2_\sigma(\mathbb R^3)$7

The finite-dimensional equations possess solutions on every finite interval. Testing by $u_0\in L^2_\sigma(\mathbb R^3)$8 gives the exact Galerkin energy identity

$u_0\in L^2_\sigma(\mathbb R^3)$9

Uniform estimates imply weak-* compactness in $u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$0 and weak compactness in $u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$1. Compactness arguments then permit passage to the limit in the nonlinear term and produce a global weak solution.

For time-dependent forces on bounded domains, the assumption

$u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$2

is sufficient for global existence. Uniformly local integrability allows persistent forcing without requiring finite total force energy on $u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$3. The resulting solution satisfies uniform kinetic-energy and accumulated-dissipation bounds, while long-time averages can be extracted along subsequences. Their weak limits satisfy stationary Reynolds equations containing a Reynolds stress (Berselli et al., 2018).

Numerical approximation follows the same compactness architecture. The finite-difference Chorin projection method uses a tentative velocity with a discrete gradient bound and a projected velocity satisfying discrete incompressibility. Strong $u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$4 compactness of the tentative velocities, together with weak convergence of discrete gradients, yields a Leray–Hopf limit under

$u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$5

A second-order BDF2 projection method with conforming finite elements uses discrete energy inequalities, refined time-translation estimates, and Simon’s compactness theorem. Under

$u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$6

the intermediate, projected, and extrapolated velocities converge, up to a subsequence, strongly in

$u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$7

to a Leray–Hopf weak solution (Kuroki et al., 2018, Weber, 15 Sep 2025).

4. Uniqueness and regularity properties

The existence theory does not imply uniqueness in three dimensions. Arbitrary energy-class Leray–Hopf solutions are known to exist globally, but uniqueness and global regularity remain open for the classical three-dimensional equations.

Uniqueness is available under additional regularity. A typical weak–strong uniqueness argument considers two solutions $u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$8 and $u\in L^\infty(0,T;L^2_\sigma(\mathbb R^3)) \cap L^2(0,T;\dot H^1(\mathbb R^3))$9, sets $u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$0, and derives

$u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$1

If the strong solution is sufficiently small in a critical path norm near the initial time, Grönwall’s inequality implies $u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$2. Critical $u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$3 data and closure subspaces of critical Besov spaces provide short-time uniqueness among all weak Leray–Hopf solutions (Barker, 2016).

Global uniqueness follows under Prodi–Serrin-type conditions. If one solution satisfies

$u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$4

then the difference estimate closes by Hölder, Sobolev interpolation, and Grönwall. Lorentz-space extensions include

$u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$5

with $u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$6, and the weak endpoint

$u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$7

under a sufficiently small critical norm (Barker, 2016).

A related result shows that certain very weak solutions automatically become Leray–Hopf solutions. On $u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$8, if a distributional solution has finite-energy initial data, converges weakly to the initial datum, and satisfies a critical Prodi–Serrin condition on every interval $u\in C_w([0,T];L^2_\sigma(\mathbb R^3)).$9, then it acquires

$L^2$0

regularity, the energy inequality, weak continuity, and strong attainment of the initial value. The proof uses an auxiliary energy solution and a backward adjoint problem (Galdi, 2018).

In the hyperdissipative system

$L^2$1

Leray–Hopf solutions in the range

$L^2$2

possess partial regularity. For initial data in $L^2$3, the spatial singular set at blow-up times satisfies

$L^2$4

and finite unions of the singular sets satisfy

$L^2$5

The proof uses Littlewood–Paley localization, good and bad cubes, energy barriers, and frequency-space dissipation rather than a local energy inequality or pressure estimates (Ożański, 2020).

5. Nonuniqueness and limitations of the solution concept

The Leray–Hopf conditions are not universally sufficient for uniqueness in all related systems.

For the geometric Navier–Stokes equation on the hyperbolic plane $L^2$6, there exist non-unique Leray–Hopf solutions with identical initial data. Let $L^2$7 be a bounded nonconstant harmonic function and set

$L^2$8

Because

$L^2$9

the nonlinear convection term is a pressure gradient. Distinct functions $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$0 satisfying the same energy constraint and the same initial value produce distinct finite-energy solutions. Thus negative curvature changes the admissible energy modes sufficiently to destroy uniqueness within the Leray–Hopf class (Chan et al., 2010).

For fractional Navier–Stokes on $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$1,

$\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$2

convex integration yields infinitely many Leray–Hopf solutions for

$\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$3

They can have smooth identical initial data, belong to $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$4 for every $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$5, and possess distinct prescribed kinetic-energy profiles. Their total energy-dissipation functional can be strictly decreasing on a short interval. The construction treats the fractional dissipation as an error in a Navier–Stokes–Reynolds iteration and relies on the condition $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$6 (Rosa, 2018).

Transport noise does not automatically restore uniqueness. For fractional Navier–Stokes equations on $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$7 with Stratonovich transport noise, there exist infinitely many analytically weak, probabilistically strong solutions on a fixed stochastic basis, with the same deterministic initial datum and the same Brownian motion. Their pathwise Leray–Hopf energy inequality holds on a positive random interval for sufficiently small fractional exponent $\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$8 (Lange et al., 2024).

Other stochastic fractional systems exhibit both local nonuniqueness and local uniqueness depending on the force. Linear multiplicative noise can be removed through

$\lim_{t\to0^+}\|u(t)-u_0\|_{L^2}=0.$9

reducing the problem to a random-coefficient PDE. At critical force scaling, distinct local Leray–Hopf-type solutions can be constructed, whereas perturbations with strictly better temporal decay recover local uniqueness. Specially constructed stochastic forces can also produce two distinct global solutions (Chen et al., 23 Mar 2025).

These results do not imply that every Navier–Stokes equation or every forcing is non-unique. They show that the Leray–Hopf energy inequality, although sufficient for global existence, does not by itself guarantee uniqueness in several geometric, fractional, or stochastic regimes.

6. Long-time behavior, inviscid limits, and applications

For forced three-dimensional flows on bounded domains, Leray–Hopf energy estimates support subsequential long-time averaging. If

$\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$0

then along suitable sequences $\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$1,

$\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$2

The Reynolds stress is defined by

$\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$3

The limiting mean flow satisfies a stationary Reynolds system with $\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$4 as an additional stress. The turbulent dissipation obeys

$\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$5

If the forcing is asymptotically stationary, the external turbulent flux vanishes and the Reynolds stress is dissipative on the mean flow (Berselli et al., 2018).

Leray–Hopf solutions also form a central framework for vanishing-viscosity limits. For bounded domains in dimensions two and three, uniformly positive local second-order structure-function exponents imply that weak $\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$6 limits of Navier–Stokes Leray–Hopf solutions solve the forced Euler equations. The result applies to both no-slip and Navier-friction boundaries. It does not require strong spacetime convergence, and it does not guarantee energy conservation: viscous dissipation may remain nonzero in the limit as anomalous dissipation (Drivas et al., 2018).

Low-frequency information can refine long-time behavior. For Hall–MHD, every two-dimensional Leray–Hopf solution becomes smooth after finite time, while in three dimensions only specially constructed solutions are shown to gain eventual magneto-vorticity regularity. Under low-frequency pseudomeasure assumptions, Fourier splitting yields algebraic decay estimates. The two-dimensional result applies to every Leray–Hopf solution because eventual smoothness is available; the corresponding three-dimensional decoupled estimates require global strong solutions or specially constructed weak solutions (Jung et al., 2 Jun 2026).

The Leray–Hopf concept also extends to coupled systems. In the three-dimensional Beris–Edwards system, a Leray–Hopf type solution has

$\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$7

satisfies the distributional equations, and obeys both the physical free-energy inequality and an expanded inequality suitable for weak–strong uniqueness. Stable quartic bulk energy, $\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$8, supplies coercivity and controls the $\varphi\in C_c^\infty([0,T)\times\mathbb R^3;\mathbb R^3), \qquad \nabla\cdot\varphi=0.$9-tensor contributions (Zhang et al., 16 May 2026).

In two-dimensional inhomogeneous Navier–Stokes without vacuum, the Leray–Hopf class is substantially better behaved. For bounded positive density,

$\int_{\mathbb R^3}\nabla p\cdot\varphi\,dx=0.$0

global existence, uniqueness, stability, immediate positive-time regularization, and energy equality can be obtained. The proof combines approximation by immediately strong solutions, renormalized transport theory, negative-Sobolev estimates for the density difference, and a relative-energy argument (Škondrić, 23 Apr 2025, Crin-Barat et al., 2024).

The breadth of these applications illustrates the role of the Leray–Hopf conditions: they provide a robust energy-admissible compactness class, but their consequences depend decisively on dimension, dissipation, geometry, forcing, coupling, and additional regularity.