---
title: 'Lions'' Exponent: Critical Navier–Stokes Thresholds'
url: https://www.emergentmind.com/topics/lions-exponent
type: topic
---

# Lions' Exponent: Critical Navier–Stokes Thresholds

Lions’ exponent denotes a threshold associated with J.-L. Lions that appears in more than one analytical context. In the theory of fractional or hyperviscous Navier–Stokes equations, the three-dimensional Lions exponent is the dissipation power $\alpha_L=5/4$, equivalently $m_L=5/2$ when the diffusion is written as $\Lambda^m=( -\Delta)^{m/2}$; it is the $L^2$-energy-critical boundary and the first exponent at which uniqueness in the energy class holds by classical methods. In fractional Sobolev theory, the expression also occurs in the critical Sobolev-Lions exponent $p_s^*=\frac{Np}{N-sp}$ and in the variable-exponent analogue $p_s^*(x)=\frac{N\,p(x)}{N-s\,p(x)}$ [2006.11861] [2508.16906] [2005.00617].

## 1. Definitions, notation, and scope

In the fluid-mechanical literature represented here, the generalized incompressible Navier–Stokes system on $\mathbb{R}^3$, $\mathbb{T}^3$, or $\mathbb{R}^d$ is obtained by replacing the Laplacian by a fractional power. One standard form is
$$
\partial_t u + (u\cdot\nabla)u + \nabla p + (-\Delta)^\alpha u = f,\qquad \nabla\cdot u=0,
$$
with $\widehat{(-\Delta)^\alpha u}(k)=|k|^{2\alpha}\hat u(k)$; in this notation, larger $\alpha$ gives stronger dissipation, $\alpha=1$ is the classical Navier–Stokes case, and $\alpha=0$ formally corresponds to Euler [2006.11861].

A notational variant writes the dissipative term as $\Lambda^m u=( -\Delta)^{m/2}u$. In that convention Lions’ threshold is $m_L=1+d/2$, whereas in the $( -\Delta)^\alpha$ convention the same threshold becomes $\alpha_L=m_L/2=(1+d/2)/2$. For $d=3$, this yields $m_L=5/2$ and $\alpha_L=5/4$; one paper states explicitly that “because his notation differs by a half-power one finds in our notation $\alpha_L=5/4$” [2508.16906] [2006.11861].

The same name also occurs in concentration-compactness and embedding theory. For $0<s<1$ and $1<p<\frac{N}{s}$, the classical fractional Sobolev space $W^{s,p}(\mathbb{R}^N)$ carries the critical Sobolev-Lions exponent
$$
p_s^*:=\frac{Np}{N-sp},
$$
while in the variable-exponent setting the corresponding local critical exponent is
$$
p_s^*(x):=\frac{N\,p(x)}{N-s\,p(x)}.
$$
These exponents govern Lions-type vanishing lemmas and compactness statements rather than fluid-dynamical well-posedness [2005.00617].

## 2. Scaling and the energy-critical threshold

For the hyperdissipative Navier–Stokes equation on $\mathbb{T}^3$,
$$
\partial_t u + \mathbb{P}\nabla\cdot(u\otimes u) + \nu(-\Delta)^\alpha u = 0,
$$
the natural parabolic scaling is
$$
u_\lambda(t,x)=\lambda^{2\alpha-1}u(\lambda^{2\alpha}t,\lambda x).
$$
Under this scaling, $C_tL_x^2$ is critical precisely at $\alpha=5/4$; for $\alpha>5/4$ it is subcritical, while for $\alpha<5/4$ it is supercritical [2205.10260].

The same conclusion appears in the $\Lambda^m$ notation. If
$$
\partial_t u + (u\cdot\nabla)u + \nabla\pi + \nu\Lambda^m u = 0,
$$
then under $x\to\lambda x$, $t\to\lambda^{2m}t$ one sets
$$
u_\lambda(t,x)=\lambda^{2m-1}u(\lambda^{2m}t,\lambda x),
$$
and the $L^2$-energy scales as $\|u_\lambda\|_{L_x^2}^2=\lambda^{4m-2-d}\|u\|_{L^2}^2$. Energy invariance therefore requires $4m-2-d=0$, hence $m=1+d/2$ and, in dimension three, $\alpha=m/2=5/4$ [2508.16906].

A complementary heuristic in dimension three balances “gain $2\alpha$ derivatives from dissipation” against “lose $5/2$ derivatives from nonlinearity,” yielding the condition $2\alpha\ge 5/2$, i.e. $\alpha\ge 5/4$ [2006.11861]. In the deterministic hyper-viscous setting, this is also the point at which the nonlinear term can be absorbed by the dissipative term through suitable Sobolev embeddings; one summary states that at $\nu=5/4$ “this balance is exactly sharp” [2411.06133].

## 3. Lions’ theorem and the Leray–Hopf regime

A Leray–Hopf solution of the forced fractional Navier–Stokes system is a divergence-free field
$$
u\in L^\infty(0,T;L^2)\cap L^2(0,T;H^\alpha)
$$
that satisfies the weak form of the equation, attains the initial data in the strong $L^2$ sense, and obeys the energy inequality
$$
\|u(t)\|_{L^2}^2
+
2\int_s^t\|(-\Delta)^{\alpha/2}u(\tau)\|_{L^2}^2\,d\tau
\le
\|u(s)\|_{L^2}^2
+
2\int_s^t\!\!\int u\cdot f\,dx\,d\tau
$$
for almost every $0\le s<t\le T$ [2306.06358].

Lions’ classical result asserts that for the forced system on $\mathbb{R}^3$ or $\mathbb{T}^3$, if $\alpha\ge 5/4$, then for any $u_0\in L^2$ and sufficiently regular forcing $f$, there exists a unique global-in-time Leray–Hopf solution. In the torus setting, one formulation gives, for divergence-free $v_0\in L^2(\mathbb{T}^3)$ with zero mean and $\theta\ge 5/4$, a unique global weak solution
$$
v\in L^\infty(0,T;L^2(\mathbb{T}^3))\cap L^2(0,T;H^\theta(\mathbb{T}^3))
$$
satisfying the distributional equation and the energy equality
$$
\frac12\|v(t)\|_{L^2}^2+\nu\int_0^t\|(-\Delta)^{\theta/2}v(s)\|_{L^2}^2\,ds
=
\frac12\|v_0\|_{L^2}^2
$$
[1808.07595] [2306.06358].

Later work refines this energy-class picture through generalized Ladyzhenskaya–Prodi–Serrin criteria. For the hyper-viscous scaling
$$
u(t,x)\mapsto \lambda^{2\nu-1}u(\lambda^{2\nu}t,\lambda x),
$$
the mixed space $L_t^\gamma W_x^{s,p}$ is scaling-invariant when
$$
\frac{2\nu}{\gamma}+s+\frac{3}{p}=2\nu-1.
$$
One summary states that if a solution lies in such a scaling-invariant regime with $2\le\gamma,p\le\infty$ and $s>0$, then it must coincide with the unique Leray–Hopf solution; taking $(\gamma,p,s)=(2,6,1)$ recovers Lions’ well-posedness for $\nu>5/4$ [2411.06133].

## 4. Sharpness below $5/4$

The modern sharpness theory shows that the threshold $5/4$ is not merely an artifact of Lions’ argument. For the periodic hyperviscous Navier–Stokes equation, Luo–Titi proved that whenever $0\le\theta<5/4$ there exist infinitely many weak solutions
$$
v\in C_t(L_x^2)\cap C_t(W_x^{2\theta-1,1})
$$
to
$$
\partial_t v+\nabla\cdot(v\otimes v)+\nabla p+\nu(-\Delta)^\theta v=0,\qquad \nabla\cdot v=0,
$$
which can realize a prescribed compactly supported energy profile and, in particular, yield infinitely many weak solutions with zero initial data [1808.07595].

A stronger result in the Leray–Hopf class was obtained for the forced fractional Navier–Stokes equation. For every $\frac12<\alpha<\frac54$, there exists a forcing $f\in C_c^\infty((0,T)\times\mathbb{R}^3)$ and two distinct Leray–Hopf solutions
$$
u_1,u_2\in L^\infty(0,T;L^2)\cap L^2(0,T;H^\alpha)
$$
with $u|_{t=0}=0$, so uniqueness fails for every $\alpha\in(\frac12,\frac54)$ in the forced energy class [2306.06358].

The stochastic theory reaches the same threshold from a different direction. For the three-dimensional stochastic generalized Navier–Stokes equations with additive noise or linear multiplicative noise, Yamazaki proved non-uniqueness in law on finite time intervals for $\alpha\in(1/3,5/4)$, extending convex-integration methods together with a probabilistic gluing or stopping-time argument [2006.11861]. Across these results, the common conclusion is that the Lions exponent is sharp for uniqueness questions tied to the classical energy framework.

## 5. Beyond the Lions exponent: supercritical spaces and non-Leray–Hopf solutions

The statement “$\alpha\ge 5/4$ implies uniqueness” is correct only in the Leray–Hopf or, more generally, the LPS-critical framework. Above the Lions exponent, recent work shows that uniqueness can fail in supercritical spaces even though the classical Leray–Hopf solution remains unique.

For $\alpha\in[5/4,2)$ on $\mathbb{T}^3$, Li–Qu–Zeng–Zhang proved strong non-uniqueness in mixed spaces $L_t^\gamma W_x^{s,p}$ whenever the generalized Ladyženskaja–Prodi–Serrin condition is breached:
$$
\frac{2\alpha}{\gamma}+\frac{3}{p}+s>2\alpha-1.
$$
They identified two sharp endpoints,
$$
(s,\gamma,p)=\left(\frac{3}{p}+1-2\alpha,\infty,p\right),\qquad
(s,\gamma,p)=\left(\frac{2\alpha}{\gamma}+1-2\alpha,\gamma,\infty\right),
$$
and constructed infinitely many weak solutions that coincide with the unique Leray–Hopf solution near the initial time, remain arbitrarily close in the supercritical norm, and are smooth outside a fractal set of singular times with zero Hausdorff $\mathcal{H}^{\eta_*}$ measure [2205.10260].

The stochastic hyper-viscous theory exhibits an analogous phenomenon. For any deterministic divergence-free initial datum $u_0\in L^2$ and any $\nu\in[5/4,2)$, Cao–Zeng–Zhang constructed infinitely many probabilistically strong and analytically weak solutions in
$$
L_\Omega^rL_t^\gamma W_x^{s,p}
$$
for $(s,\gamma,p)$ lying in two supercritical regimes $S_1,S_2$ strictly beyond the LPS critical line. The same summary states that outside the LPS-critical spaces one sees non-uniqueness even though $\nu>5/4$, whereas within the scaling-invariant LPS regimes one retains classical uniqueness relative to the Leray–Hopf solution [2411.06133].

The stochastic endpoint $\alpha=5/4$ also supports a distinct positive theory. For the three-dimensional Navier–Stokes equations forced by space-time white noise and diffused by $\Lambda^{5/2}$, Yamazaki proved a unique global-in-time mild solution for almost every noise realization and, separately, uniqueness of a high-low weak solution formulation for $L^2$ initial data [2508.16906]. This indicates that the endpoint behavior depends decisively on the solution class and on whether one works inside or outside the energy/LPS-critical framework.

## 6. Extensions to magnetohydrodynamics and to Sobolev compactness theory

The Lions threshold extends to generalized magnetohydrodynamics. In the deterministic MHD system
$$
\partial_t u + (u\cdot\nabla)u + \nabla p + \nu_1(-\Delta)^{m_1}u=(b\cdot\nabla)b,\qquad \nabla\cdot u=0,
$$
$$
\partial_t b + (u\cdot\nabla)b - (b\cdot\nabla)u + \nu_2(-\Delta)^{m_2}b=0,\qquad \nabla\cdot b=0,
$$
Wu extended Lions’ result by showing uniqueness whenever both $m_1,m_2\ge (d+2)/4$; in $d=3$, this means $m_1,m_2\ge 5/4$ [2410.02196]. In the stochastic additive-noise setting, Yamazaki then used convex integration with temporal intermittency to prove non-uniqueness from prescribed initial data for $m_1,m_2\in(1,2)$, constructing infinitely many adapted weak-analytic, strong-probabilistic solutions [2410.02196].

In a separate analytical lineage, the “Lions exponent” refers not to fluid dissipation but to an embedding threshold. For $0<s<1$ and $1<p<\frac Ns$, the critical Sobolev-Lions exponent
$$
p_s^*=\frac{Np}{N-sp}
$$
controls the classical fractional Lions vanishing lemma: if $(u_n)\subset W^{s,p}(\mathbb{R}^N)$ is bounded and
$$
\lim_{n\to\infty}\sup_{y\in\mathbb{R}^N}\int_{B(y,r)}|u_n|^p\,dx=0
$$
for some $r>0$, then $u_n\to0$ in $L^q(\mathbb{R}^N)$ for every $p<q<p_s^*$ [2005.00617].

Bahrouni–Ounaïes generalized this structure to fractional Sobolev spaces with variable exponents. For
$$
W^{s,p(\cdot),\tilde p(\cdot,\cdot)}(\Omega),
$$
with symmetric $\tilde p(x,y)$, the local critical exponent becomes
$$
p_s^*(x)=\frac{N\,p(x)}{N-s\,p(x)}.
$$
Their Lions-type lemma states that if $(u_n)$ is bounded in $W^{s,p(\cdot),\tilde p(\cdot,\cdot)}(\mathbb{R}^N)$ and its local $L^{q(\cdot)}$ mass vanishes for some $q$ satisfying $p(x)<q(x)<p_s^*(x)$, then $u_n\to0$ in $L^{a(\cdot)}(\mathbb{R}^N)$ for every $a$ with $p(x)<a(x)<p_s^*(x)$. Under radiality assumptions, they also obtain the compact embedding
$$
W^{s,p(\cdot),\tilde p(\cdot,\cdot)}_{\rm rad}(\mathbb{R}^N)\hookrightarrow L^{a(\cdot)}(\mathbb{R}^N)
$$
for every such $a$ [2005.00617].

Taken together, these usages show that Lions’ exponent functions as a threshold notion: in fluid equations it separates regimes of energy-based uniqueness, sharp non-uniqueness, and supercritical ill-posedness phenomena; in Sobolev analysis it marks the critical boundary for vanishing and compactness mechanisms.

Source: https://www.emergentmind.com/topics/lions-exponent