Lions's Energy Conservation Criteria

• Lions's Energy Conservation Criteria are sufficient space-time conditions that guarantee energy equality for weak solutions by imposing u ∈ L⁴(0,T;L⁴(Ω)).
• They extend the classical incompressible Navier–Stokes framework by incorporating Prodi–Lions–Shinbrot ideas through mixed velocity–gradient formulations and anisotropic generalizations.
• The criteria distinguish energy conservation from full regularity, offering practical insights for bounded domains, compressible flow, and adaptations via commutator estimates.

Lions's energy conservation criteria are sufficient space-time conditions under which weak solutions satisfy an exact energy balance rather than only an energy inequality. In the literature surveyed here, the classical criterion attributed to Lions for the homogeneous incompressible Navier–Stokes equations is

$u\in L^4(0,T;L^4(\Omega)),$

for which Leray–Hopf weak solutions satisfy energy equality. Subsequent work places this condition within a broader Prodi–Lions–Shinbrot tradition, and extends it to mixed velocity–gradient formulations, gradient-based criteria in bounded domains, anisotropic componentwise conditions, and compressible Navier–Stokes analogues (Wang et al., 2021, Berselli et al., 2024, Wang et al., 10 Sep 2025, Chen et al., 14 Feb 2026).

1. Classical incompressible formulation

For 3D incompressible Navier–Stokes, the natural weak-solution class is the Leray–Hopf class

$$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$

together with the distributional equations and the global energy inequality

$$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$

Within this framework, Lions's classical condition is the additional assumption

$u\in L^4(0,T;L^4(\Omega)),$

which upgrades the inequality to the exact energy equality

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$

This is the formulation explicitly identified as the “famous Lions criterion” in both Wang–Ye and the anisotropic paper (Wang et al., 2021, Wang et al., 10 Sep 2025).

The criterion belongs to a class of conditional energy-equality statements that are weaker in ambition than full regularity theorems. The literature discussed here repeatedly distinguishes such conditions from Serrin-type regularity criteria. In particular, Serrin’s assumption

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$

is presented as stronger, since under it a weak solution “will immediately become a classical one” in the incompressible setting (Wang et al., 2021). Lions's criterion is therefore best understood as an energy-balance threshold rather than a regularity threshold.

Historically, the papers place Lions alongside Prodi as an early source for determining “minimal space-time assumptions necessary for energy conservation, particularly in the viscous case.” That historical framing is explicit in the bounded-domain Euler/Navier–Stokes paper, which treats its own gradient criteria as part of the same Prodi–Lions lineage (Berselli et al., 2024).

2. Lions within the velocity-based criteria hierarchy

The modern interpretation of Lions's criterion is inseparable from its relation to Shinbrot- and Serrin-type mixed $L^p_tL^q_x$ classes. The papers considered here present the classical Shinbrot criterion as

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$

and the $3

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{1}{p}+\frac{3}{q}=1,\qquad 3<q<4.$$

These conditions contain the Lions point $$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$0, since

$$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$1

The anisotropic paper also cites the $$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$2 criterion in this form and treats it as complementary to Shinbrot’s $$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$3 branch (Wang et al., 10 Sep 2025).

Wang–Ye give a unifying perspective. In their incompressible specialization, the mixed criterion

$$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$4

implies energy equality. They state explicitly that the special case $$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$5 reduces to Lions’s criterion because

$$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$6

while $$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$7 is already part of the Leray–Hopf class. They further show that interpolation with the natural energy bounds recovers velocity-only criteria: $$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$8 or

$$u\in L^\infty(0,T;L^2(\Omega))\cap L^2(0,T;H^1(\Omega)),$$9

This suggests that Lions’s condition is not an isolated endpoint but a symmetric point inside a larger admissible family (Wang et al., 2021).

The 2026 compressible paper adopts exactly this Lions–Shinbrot interpretation. It states that prior compressible arguments reached only a Serrin-type or sub-Shinbrot regime, whereas its objective is to bridge the gap toward the classical Lions/Shinbrot criteria by using the same two branches,

$$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$0

and

$$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$1

now adapted to compressible flow (Chen et al., 14 Feb 2026).

3. Mixed velocity–gradient and gradient formulations

A major development after the classical $$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$2 formulation is the replacement of a pure velocity condition by a mixed condition on $$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$3 and $$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$4. Wang–Ye derive such a criterion for incompressible and general compressible Navier–Stokes, and in the incompressible case state that it contains Lions, Shinbrot, the Beirão da Veiga–Yang extension, and gradient criteria of Berselli–Chiodaroli and Zhang (Wang et al., 2021).

In 3D incompressible flow, the gradient-only criteria recovered from their mixed framework are

$$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$5

or

$$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$6

These criteria are not presented as replacements for Lions’s criterion in a literal sense; rather, the paper treats them as consequences of a broader commutator-based framework whose symmetric point reproduces Lions exactly.

The bounded-domain paper pushes the gradient viewpoint further. For 3D incompressible Euler in a bounded domain with slip boundary condition $$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$7, it proves that weak solutions conserve energy provided

$$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$8

The paper emphasizes that this is “exactly the same” criterion already known for Navier–Stokes weak solutions in the $$\|u(t)\|_{L^2(\Omega)}^2+2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds\le \|u_0\|_{L^2(\Omega)}^2.$$9 branch: $u\in L^4(0,T;L^4(\Omega)),$0 It also recalls the lower Navier–Stokes branch

$u\in L^4(0,T;L^4(\Omega)),$1

but does not prove the Euler analogue in that range (Berselli et al., 2024).

The technical significance of the bounded-domain result is that it avoids pressure estimates and ordinary translation-based convolution, both of which are obstructed by boundaries. Instead, it uses a divergence-preserving approximation $u\in L^4(0,T;L^4(\Omega)),$2 constructed from a transversal deformation of the domain. A plausible implication is that Lions-type energy criteria can be transported to settings where the classical whole-space mollification argument is unavailable, provided one replaces translation invariance by a boundary-compatible regularization scheme.

4. Compressible Lions–Shinbrot criteria

The compressible extension of Lions's energy-conservation philosophy is formulated most explicitly in "Shinbrot Type Criteria for Energy Conservation of the Compressible Navier-Stokes Equations" (Chen et al., 14 Feb 2026). The underlying system is the barotropic compressible Navier–Stokes equation

$u\in L^4(0,T;L^4(\Omega)),$3

with $u\in L^4(0,T;L^4(\Omega)),$4, $u\in L^4(0,T;L^4(\Omega)),$5, posed on $u\in L^4(0,T;L^4(\Omega)),$6 or on a bounded $u\in L^4(0,T;L^4(\Omega)),$7 domain in $u\in L^4(0,T;L^4(\Omega)),$8, $u\in L^4(0,T;L^4(\Omega)),$9. The target conclusion is the full energy identity

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$0

For constant viscosity, the theorem assumes

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$1

together with

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$2

Under these hypotheses, weak solutions satisfy energy equality for any $$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$3. The result includes the Lions point $$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$4, and the paper stresses that this is conceptually important because earlier compressible criteria required $$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$5, far from the Lions threshold.

The paper also treats a degenerate-viscosity model,

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$6

with energy balance

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$7

Here the theorem assumes

$$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$8

and the same two-branch Shinbrot-type condition, now allowing the endpoint $$\|u(t)\|_{L^2(\Omega)}^2 +2\int_0^t \|\nabla u(s)\|_{L^2(\Omega)}^2\,ds = \|u_0\|_{L^2(\Omega)}^2.$$9, $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$0. The constant-viscosity result allows vacuum but excludes $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$1; the degenerate-viscosity result requires density bounded away from zero and includes that endpoint.

The paper presents this as a genuine Lions–Shinbrot advance for compressible flow. The main obstructions are the time derivative of the nonlinear momentum $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$2, the continuity equation, vacuum, and pressure work. A central point is that previous compressible energy-equality results only reached a Serrin-type or sub-Shinbrot regime, whereas the new theorem descends to the true Lions/Shinbrot range by replacing the standard strong commutator estimate with a weaker distributional one.

5. Anisotropic extensions of Lions's criterion

The anisotropic paper generalizes Lions’s energy-equality condition by allowing the horizontal and vertical components of velocity to have different integrability (Wang et al., 10 Sep 2025). Writing

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$3

its main viscous theorem states that energy equality for Leray–Hopf weak solutions of 3D Navier–Stokes on $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$4 holds if

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$5

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$6

with

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$7

The paper remarks explicitly that the special choice

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$8

reduces to the famous Lions energy-balance class.

This componentwise generalization is accompanied by three further sufficient conditions. One uses only the horizontal velocity,

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{d}{q}\le 1,\quad q\ge d,$$9

including the endpoint $L^p_tL^q_x$0. A second uses the horizontal gradient,

$L^p_tL^q_x$1

A third uses anisotropic Besov regularity,

$L^p_tL^q_x$2

with

$L^p_tL^q_x$3

The structural mechanism is the divergence-free condition

$L^p_tL^q_x$4

which allows one to rewrite

$L^p_tL^q_x$5

The paper stresses that this ensures every nonlinear term can be arranged so that at least one horizontal component appears. This same mechanism underlies the inviscid anisotropic theorem, where one component may lie in the largest borderline Besov space: $L^p_tL^q_x$6 with

$L^p_tL^q_x$7

At $L^p_tL^q_x$8, this permits

$L^p_tL^q_x$9

while the horizontal part remains in the refined $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$0 class. The paper presents this as a genuinely anisotropic phenomenon, since isotropic conservation in the full largest space $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$1 is not expected.

6. Proof mechanisms, scope, and limitations

Across the papers considered here, Lions-type criteria are proved by controlling the nonlinear energy flux through mollification and commutator estimates rather than by proving full regularity. In Wang–Ye, the mixed velocity–gradient condition is tailored to the defect

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$2

The derivative falls on one factor through $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$3, and the other factor is balanced by

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$4

This commutator structure is the reason the parameter family contains Lions’s $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$5 condition as the symmetric point $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$6 (Wang et al., 2021).

In the bounded-domain Euler paper, the corresponding obstruction is geometric rather than compressible: ordinary convolution near $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$7 breaks tangency and incompressibility. The proof therefore uses the divergence-preserving approximation $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$8, time mollification, and a test function of the form

$$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{2}{p}+\frac{2}{q}=1,\qquad q\ge 4,$$9

The paper emphasizes that the argument requires no pressure estimates and that the troublesome boundary contributions are neutralized by the approximation itself (Berselli et al., 2024).

In the compressible setting, the main new obstruction is the temporal commutator created by $3

$3

and proves a weak-type commutator statement: $3Chen et al., 14 Feb 2026).

The literature also states several limitations. Wang–Ye formulate the incompressible theorem on the periodic domain $3Wang et al., 2021). The bounded-domain Euler paper proves only the $3Berselli et al., 2024). The compressible Shinbrot-type theorem with constant viscosity does not obtain the endpoint $3Chen et al., 14 Feb 2026). The anisotropic paper explicitly identifies $$u\in L^p(0,T;L^q(\Omega)),\qquad \frac{1}{p}+\frac{3}{q}=1,\qquad 3<q<4.$$0 as an interesting open problem (Wang et al., 10 Sep 2025).

A final interpretive point common to all four papers is that energy equality should not be conflated with regularity. The criteria are sufficient and are not claimed to be sharp in an Onsager sense. In particular, the bounded-domain Beltrami analysis stresses that conditions leading to energy conservation are significantly distinct from those implying regularity, and the compressible work similarly frames its theorem as part of the Lions/Shinbrot energy-equality tradition rather than an exact criticality theory (Berselli et al., 2024, Chen et al., 14 Feb 2026).