---
title: Incompressible DiPerna-Lions Flows
url: https://www.emergentmind.com/topics/incompressible-diperna-lions-flows
type: topic
---

# Incompressible DiPerna-Lions Flows

Incompressible DiPerna–Lions flows are generalized Lagrangian flows associated with divergence-free vector fields whose spatial regularity is too weak for classical ODE theory but still sufficient for transport and continuity equations to be well posed in a renormalized sense. In this setting, the flow is typically defined only almost everywhere, or as a regular Lagrangian flow with bounded compression, and incompressibility upgrades bounded compression to measure preservation. The theory is fundamentally Eulerian—built from transport identities, renormalization, and weak stability—but it also recovers geometric and dynamical information about the flow map itself, including preservation of Lebesgue measure, transport of co-moving volumes, commutativity criteria, and asymptotic mixing properties [1402.4788] [2603.27491] [2510.02921].

## 1. Foundational framework

The basic objects are divergence-free vector fields \(b\) or \(v\) with Sobolev regularity, typically \(L^1_tW^{1,p}_x\) or \(BV\), and the associated transport equation
\[
\partial_t \rho + b\cdot \nabla \rho = 0,
\]
equivalently, in the incompressible case,
\[
\partial_t \rho + \nabla\cdot(\rho b)=0.
\]
In the metric-measure formulation of Ambrosio–Trevisan, the vector field is encoded by a derivation \(\boldsymbol b\), the continuity equation is written as
\[
\partial_t u_t+\operatorname{div}(u_t\boldsymbol b_t)=0,
\]
and the associated regular flow \(X\) is required to solve the ODE in the sense that, for every test function \(f\), \(f(X(\cdot,x))\) is absolutely continuous and
\[
\frac{d}{dt}f(X(t,x))=df(\boldsymbol b_t)(X(t,x))
\]
for \(\mathfrak m\)-a.e. \(x\), together with bounded compressibility
\[
X(t,\cdot)_\#\mathfrak m \le C\,\mathfrak m.
\]
When \(\operatorname{div}\boldsymbol b=0\), the incompressible case corresponds to the measure-preserving regime [1402.4788].

On bounded domains, the same philosophy persists but with additional measurable-set subtleties. For vector fields \(v\) with weak derivatives in \(L^1_{\mathrm{loc}}\) and \(\nabla\cdot v\in L^1_{\mathrm{loc}}(\mathbb R;L^\infty(\Omega))\), the generalized flow map \(X(s,t,x)\) satisfies the DiPerna–Lions compressibility estimate
\[
\frac{1}{c^{s,t}}\mathcal L^3(A)\le \mathcal L^3\big(X(s,t,\cdot)^{-1}(A)\big)\le c^{s,t}\mathcal L^3(A),
\qquad
c^{s,t}=\exp\Big(\|\nabla\cdot v\|_{L^1(I^{s,t};L^\infty(\Omega))}\Big).
\]
Thus incompressibility, \(\nabla\cdot v=0\), gives \(c^{s,t}=1\), hence exact preservation of Lebesgue measure for inverse images [2603.27491].

This framework is not restricted to Euclidean smooth settings. Ambrosio–Trevisan develop analogues of DiPerna–Lions theory on Dirichlet-form and metric-measure spaces, including \({\sf RCD}(K,\infty)\) spaces, replacing classical vector fields by derivations, convolution by the heat semigroup, and Euclidean commutators by semigroup commutators [1402.4788]. A plausible implication is that incompressible DiPerna–Lions flows are best understood as a structural class of measure-preserving nonsmooth dynamics, rather than as a narrowly Euclidean construction.

## 2. Renormalization and Eulerian well-posedness

The defining mechanism of the theory is renormalization. For a weak solution \(\rho\) of
\[
\partial_t \rho + u\cdot \nabla \rho = 0
\]
with \(\nabla\cdot u=0\), the renormalized formulation asserts that for every admissible nonlinear function \(\beta\),
\[
\partial_t \beta(\rho)+u\cdot\nabla\beta(\rho)=0.
\]
In the survey “Conserved quantities and regularity in fluid dynamics” [2003.07807], the DiPerna–Lions theorem is stated in the incompressible form: if
\[
u\in L^1(0,T;W^{1,1}(\mathbb T^d)),
\]
then every bounded weak solution is renormalized. The proof proceeds by spatial mollification, yielding a commutator
\[
R^\varepsilon = \operatorname{div}(pu)^\varepsilon-\operatorname{div}(p^\varepsilon u),
\]
and uses the fact that in the limit the defect becomes \(p\,\nabla\cdot u\), which vanishes in the divergence-free case [2003.07807]. This is the canonical Eulerian replacement for the classical chain rule along smooth trajectories.

The bounded-domain theory retains the same structure. For \(\Omega\subset\mathbb R^N\) smooth and bounded, if
\[
u\in L^1(0,T;W_0^{1,q}(\Omega)),\qquad \nabla\cdot u=0,
\]
then the transport equation on \((0,T)\times\Omega\) admits weak solutions in \(L^\infty_tL^p_x\), every weak solution is renormalized, uniqueness holds under the additional assumption \(u\in L^1_tC(\overline\Omega)\), and one has \(L^p\)-norm conservation together with
\[
\rho\in C([0,T];L^p(\Omega))
\]
for \(1<p<\infty\). The same paper proves a bounded-domain stability theorem of DiPerna–Lions type: if \(u^n\to u\) in \(L^1_tL^1_x\) and the corresponding renormalized solutions are uniformly bounded in \(L^\infty_tL^p_x\), then \(\rho^n\to \rho\) in \(C([0,T];L^p(\Omega))\) [2103.09695].

A complementary uniqueness route is duality. The paper “A uniqueness lemma with applications to regularization and incompressible fluid mechanics” proves that distributional solutions of
\[
\partial_t a + \operatorname{div}(a v)-\nu\Delta a =0,\qquad a(0)=0,
\]
with
\[
a\in L^p_tL^q_x,\qquad v\in L^{p'}_tW^{1,q'}_x,\qquad \operatorname{div}v=0,
\]
must vanish identically. The method is DiPerna–Lions-like in its commutator structure, but it uses a backward adjoint equation and a maximum principle rather than a full renormalization theory [1612.04138]. This suggests that, within incompressible transport, renormalization is not the only route to uniqueness, although it remains the organizing principle of the theory.

## 3. Geometry of the flow map

Although the theory is Eulerian at its core, several recent works recover genuinely geometric statements about nonsmooth incompressible flow maps. The most direct example is the treatment of co-moving volumes in “Co-moving volumes and Reynolds transport theorem in DiPerna-Lions theory” [2603.27491]. There, the inverse image \(X(s,t,\cdot)^{-1}(A)\) of a Borel set is always measurable, but the forward image \(X(s,t,A)\) need not be measurable. The paper resolves this by trimming \(A\) by a suitable null set \(\dot N_A^{s,t}\), defining measurable regular co-moving volumes
\[
X(s,t,A\setminus \dot N_A^{s,t}),
\]
and proving that these trimmed images have the expected measure. In the incompressible case,
\[
\mathcal L^3\big(X(s,t,\cdot)^{-1}(A)\big)=\mathcal L^3(A),
\qquad
\mathcal L^3\big(X(s,t,A\setminus \dot N_A^{s,t})\big)=\mathcal L^3(A).
\]
The same work establishes a Reynolds transport theorem on these co-moving volumes, which in the divergence-free case reduces to
\[
\frac{d}{ds}\int_{X(s,t,A\setminus \dot N_A^{s,t})} g(s,x)\,dx
=
\int_{X(s,t,A\setminus \dot N_A^{s,t})} (\partial_s g+v\cdot\nabla g)(s,x)\,dx
\]
for a.e. \(s\) [2603.27491]. A common misconception is therefore incorrect: nonsmooth incompressible DiPerna–Lions flows do not destroy the moving-domain calculus, but they require measurable trimming of forward images.

Another structural question is commutativity. For autonomous vector fields
\[
X,Y\in W^{1,p}_{\mathrm loc}\cap L^\infty(\mathbb R^n,\mathbb R^n),\qquad \div X,\div Y\in L^\infty,
\]
the paper “On the commutativity of flows of rough vector fields” proves that commutativity of the regular Lagrangian flows,
\[
\Phi_t^X\circ\Phi_s^Y=\Phi_s^Y\circ\Phi_t^X,
\]
is equivalent not merely to vanishing Lie bracket
\[
[X,Y]=DY\,X-DX\,Y=0,
\]
but to vanishing bracket together with weak differentiability of \(\Phi_t^X\) in the direction of \(Y\) [2011.08133]. In the incompressible case, \(\div X=\div Y=0\), the compressibility density is identically \(1\), so the flows are measure preserving and the criterion becomes a weak Frobenius theorem for rough measure-preserving dynamics. This clarifies that, at low regularity, algebraic involutivity alone does not recover the smooth commutativity theory; a hidden directional regularity of the flow map is also required.

## 4. Role in incompressible fluid mechanics

The fluid-mechanical significance of incompressible DiPerna–Lions flows is especially transparent in two-dimensional vorticity dynamics. In “On the vanishing viscosity limit for 2D incompressible flows with unbounded vorticity”, the incompressible Navier–Stokes vorticity equation on \(\mathbb T^2\),
\[
\partial_t \omega^\nu + u^\nu\cdot \nabla \omega^\nu = \nu \Delta \omega^\nu + g^\nu,
\qquad
u^\nu=K*\omega^\nu,
\]
is shown to converge, along vanishing viscosity, to the Euler transport equation
\[
\partial_t \omega + u\cdot \nabla \omega = g,\qquad u=K*\omega,
\]
with strong convergence
\[
\omega^{\nu_k}\to \omega \qquad\text{in } C([0,T];L^p(\mathbb T^2))
\]
for every \(p>1\), assuming strong convergence of initial vorticities and forcings in the natural \(L^p\) spaces [2007.01091]. The paper’s central point is that DiPerna–Lions renormalization is precisely what makes the low-regularity Euler limit meaningful. For \(p<4/3\), the product \(u\omega\) is not comfortably interpretable in raw distributional form, whereas the renormalized identity
\[
\partial_t \beta(\omega)+u\cdot \nabla\beta(\omega)=\beta'(\omega)\,g
\]
remains well defined because \(\beta(\omega)\) is bounded [2007.01091].

Incompressibility enters in three essential ways in that argument: the velocity is divergence free, \(L^p\) norms are preserved modulo forcing in the inviscid problem, and the Biot–Savart law converts scalar vorticity into a Sobolev velocity field,
\[
\omega\in L^p(\mathbb T^2)\Longrightarrow u\in W^{1,p}(\mathbb T^2).
\]
This is exactly the regularity threshold that permits direct use of DiPerna–Lions theory for \(p>1\); the endpoint \(p=1\) is excluded because the standard Sobolev regularity required by the theory fails there [2007.01091].

The duality paper [1612.04138] provides a second fluid-mechanical application. It proves uniqueness for transport-diffusion and frozen vorticity equations at the Leray scale, gives a new proof of Serrin regularity for Navier–Stokes, and, in 2D Euler, shows that a weak solution with zero initial data and vorticity
\[
\omega\in L^\infty(\mathbb R_+;L^p(\mathbb T^2)),\qquad p\ge 2,
\]
must vanish identically [1612.04138]. Together with the vanishing-viscosity result, this indicates that the DiPerna–Lions framework is not merely a well-posedness device for passive scalars; it governs nonlinear hydrodynamic limits and low-regularity uniqueness mechanisms inside incompressible fluid dynamics.

## 5. Extensions beyond the deterministic Euclidean setting

Several later developments embed incompressible DiPerna–Lions flows into broader theories. One direction is compressibility. The paper “Flows for non-smooth vector fields with subexponentially integrable divergence” studies Sobolev vector fields with
\[
\operatorname{div}b\in L^1_t\operatorname{Exp}(L/\log L),
\]
constructs forward and backward flows, and obtains explicit Orlicz bounds on the Radon–Nikodym densities of pushforward measures. In this framework, the incompressible case is the rigid specialization in which the Lebesgue-density is exactly \(1\), so the flow preserves measure [1507.04016]. This suggests that incompressible DiPerna–Lions flows are structurally the zero-compressibility limit of a much subtler density-evolution theory.

A second direction is stochastic. On compact Riemannian manifolds, “Quasi Invariant Stochastic Flows of SDEs with Non-smooth Drifts on Riemannian Manifolds” proves existence and uniqueness of a \(\nu\)-a.e. stochastic invertible flow for Stratonovich SDEs with Sobolev drift, together with the explicit density formula
\[
\frac{d(\nu\circ x_t^{-1})}{d\nu}(x)
=
\exp\!\left\{
\int_0^t \operatorname{div}X_0(x_s(x))\,ds
+
\int_0^t \operatorname{div}X_k(x_s(x))\circ dW_s^k
\right\}.
\]
If all vector fields \(X_0,X_1,\dots,X_m\) are divergence free, then this density is \(1\), so the stochastic flow is volume preserving; if only the drift is divergence free, a nontrivial random density remains [1007.1486]. The deterministic incompressible criterion therefore bifurcates under noise: drift incompressibility is no longer sufficient by itself.

For nondegenerate Itô diffusions with rough coefficients, “Stochastic Lagrangian Flows for SDEs with rough coefficients” formulates stochastic Lagrangian flows and almost-everywhere stochastic flows as probabilistic analogues of regular Lagrangian flows. In the constant-diffusion, divergence-free case, the Fokker–Planck equation preserves the constant density \(1\), so Lebesgue measure is preserved at the level of one-time marginals. The paper applies this to every Leray–Hopf solution of 3D Navier–Stokes, obtaining a unique stochastic flow in the DiPerna–Lions sense [1911.08133]. The related paper “Regular Flows for Diffusions with Rough Drifts” shows that additive Brownian noise can produce a weakly differentiable stochastic flow under only mixed \(L^{r,q}\) integrability of the drift, together with a Constantin–Iyer type circulation formula for suitable weak Navier–Stokes solutions [1405.5856].

A third direction is nonsmooth geometry. Ambrosio–Trevisan generalize the continuity-equation, superposition, and regular-flow theory to metric measure spaces and \({\sf RCD}(K,\infty)\) spaces, replacing classical vector fields by derivations and classical gradients by Carré du Champ structures [1402.4788]. This indicates that incompressible DiPerna–Lions flows are not tied to smooth ambient manifolds, but belong to a larger class of measure-preserving low-regularity dynamics.

## 6. Ergodic structure, asymptotic mixing, and conceptual boundaries

A recent development pushes the theory from well-posedness toward smooth-ergodic-type asymptotics. The paper “Lyapunov exponents, entropy and mixing for DiPerna-Lions flows” studies \(1\)-periodic divergence-free vector fields on \(\mathbb T^d\), interprets the time-one map \(X_1\) as a measure-preserving system, and develops an Oseledets-type theory for incompressible DiPerna–Lions flows [2510.02921]. For \(BV\), \(1\)-periodic incompressible fields, it proves existence of Lyapunov exponents \(\lambda_i(x)\) and invariant splittings \(E_x^i\); for Sobolev fields with \(p>1\), it establishes a Ruelle inequality
\[
h_{\mathcal L^d}(X_1)\le \int_{\mathbb T^d}\sum_i \lambda_i^+(x)m_i(x)\,dx.
\]
The same paper derives asymptotic bounds on regularity propagation in a logarithmic Sobolev scale and asymptotic lower bounds on mixing in both \(\dot H^{-1}\) and geometric mixing scales, yielding an asymptotic form of Bressan’s mixing conjecture [2510.02921]. This shows that incompressible DiPerna–Lions flows support a nontrivial ergodic theory despite the lack of classical differentiability.

At the same time, several endpoint and structural limitations remain explicit in the literature. The vanishing-viscosity theorem in 2D excludes \(p=1\) because the standard Sobolev regularity of the velocity required by DiPerna–Lions fails there [2007.01091]. The ergodic paper leaves open finiteness of entropy and validity of Ruelle’s inequality at the \(BV\) or \(W^{1,1}\) endpoint, and does not prove the full non-asymptotic form of Bressan’s conjecture [2510.02921]. This suggests that the incompressible theory is mature for \(p>1\), but the endpoint \(p=1\) remains a critical boundary for several central questions.

A further conceptual boundary concerns terminology. The DiPerna–Lions theory of incompressible flows concerns nonsmooth ODEs, transport equations, regular Lagrangian flows, and renormalization. It is distinct from the DiPerna–Majda theory of generalized Young measures and measure-valued Euler solutions. The latter studies oscillation and concentration in weak limits of velocity fields and does not address regular Lagrangian flow maps [1101.3499]. Confusing these two “DiPerna” frameworks obscures a real divide: incompressible DiPerna–Lions flows are a theory of transport and trajectories, whereas DiPerna–Majda solutions are a theory of weak-limit statistics.

Overall, incompressible DiPerna–Lions flows now constitute a broad research area rather than a single theorem. They provide the Eulerian-Lagrangian dictionary for divergence-free Sobolev dynamics, organize low-regularity incompressible transport, support sharp 2D hydrodynamic limits, admit geometric refinements such as Reynolds transport and commutativity criteria, and increasingly interface with ergodic theory, stochastic analysis, and nonsmooth geometry [2603.27491] [2007.01091] [2510.02921].

Source: https://www.emergentmind.com/topics/incompressible-diperna-lions-flows