Duchon–Robert Framework

• The Duchon–Robert framework is a geometric-analytic method that defines the defect in local energy conservation using velocity increments in weak Euler flows.
• It employs coarse-graining and increment-based formulations to decompose the energy defect into flux, cascade, and commutator terms, linking dissipation to Hölder regularity.
• The framework extends to negative Besov spaces and geometric analyses via Hausdorff measures, informing intermittency constraints in compressible and MHD models.

The Duchon–Robert framework is a geometric-analytic formulation of anomalous dissipation for weak solutions of the incompressible Euler equations. Its central object is the Duchon–Robert distribution, which appears in the local energy balance as the defect of local energy conservation and therefore provides a distributional encoding of inviscid dissipation generated by lack of smoothness. In recent work, the framework has been sharpened into a regularity theory in negative Besov spaces, a geometric theory of dissipative sets via Hausdorff measures, an intermittency constraint for structure-function exponents, an increment-based diagnostic for experimental flows, a template for local exact Yaglom-type laws in several hydrodynamic models, and a compressible generalization with additional baropycnal and viscous-density couplings (Rosa et al., 14 Feb 2025, Kuzzay et al., 2016, Wang et al., 2023, Zinchenko et al., 5 Aug 2025).

1. Foundational formulation

For a weak Euler solution $u$ with pressure $q$, Duchon and Robert showed that one has the local energy balance

$$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$

where $D$ is the Duchon–Robert distribution. It measures the defect in local energy conservation; when $D\neq 0$, the flow dissipates energy anomalously even though the Euler equations are inviscid (Rosa et al., 14 Feb 2025).

An equivalent increment-based form, emphasized in experimental and coarse-grained treatments, defines the DR dissipation by

$$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$

with $$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$ and $G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$. This makes the framework explicitly local in scale and based on velocity increments rather than pointwise derivatives (Kuzzay et al., 2016).

The scaling statement

$$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$

immediately links DR to Hölder regularity. If $\delta u(\ell)\sim \ell^h$, then

$q$0

Hence $q$1 implies $q$2, $q$3 is the critical Onsager/Kolmogorov case, and $q$4 allows divergence or nonvanishing anomalous dissipation (Kuzzay et al., 2016).

2. Coarse-graining and the modified energy identity

A central structural device is space mollification at scale $q$5. The framework introduces

$q$6

$q$7

$q$8

$q$9

These quantities separate the defect into an energy defect term, a flux/pressure term, and a commutator or cascade term (Rosa et al., 14 Feb 2025).

The key identity is the modified energy identity

$$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$0

Within this decomposition, $$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$1 is an energy defect term, small in negative norms; $$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$2 is a flux/pressure-related term, also small in negative norms; and $$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$3 is the main cascade or commutator term, the one that carries the leading dissipation scaling (Rosa et al., 14 Feb 2025).

This organization is analytically decisive because the dissipation regularity is obtained by balancing the small terms against the singular scaling of $$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$4. A plausible implication is that the framework does not treat anomalous dissipation as a purely qualitative defect, but as an object whose regularity can be read off from the scale-by-scale asymptotics of coarse-grained commutators.

3. Besov regularity, Hausdorff geometry, and intermittency

Under spatial Besov regularity assumptions,

$$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$5

the Duchon–Robert distribution satisfies

$$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$6

This is the central regularity statement: spatial Besov regularity of the velocity implies improved negative Besov regularity of the dissipation distribution (Rosa et al., 14 Feb 2025).

The exponent

$$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$7

acts as a dissipation regularity threshold. If $$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$8, then the classical Onsager conclusion is recovered: $$\partial_t \frac{|u|^2}{2}+\operatorname{div}\!\left(\left(\frac{|u|^2}{2}+q\right)u\right)=-D \qquad \text{in }\mathcal D'_{x,t},$$9 For $D$0, $D$1 may be nonzero, but it must lie in the specified negative Besov class. The same decomposition also yields further Onsager-type consequences: if

$D$2

then $D$3; and if

$D$4

then the kinetic energy

$D$5

is Hölder continuous in time with exponent

$D$6

A corresponding Minkowski intermittency statement excludes nontrivial dissipation when the dissipative set has sufficiently small upper Minkowski dimension under the same Besov hypothesis (Rosa et al., 14 Feb 2025).

When $D$7 is a real-valued Radon measure, the framework becomes geometric. If $D$8, then

$D$9

for any $D\neq 0$0 satisfying

$D\neq 0$1

Thus the dissipation cannot concentrate on sets that are too small in Hausdorff dimension. If in addition $D\neq 0$2, then for every compact $D\neq 0$3 there is $D\neq 0$4 such that

$D\neq 0$5

This is a quantitative upper bound on the local density of dissipation (Rosa et al., 14 Feb 2025).

The intermittency corollary states that if $D\neq 0$6 is concentrated on a set $D\neq 0$7 with

$D\neq 0$8

and if $D\neq 0$9 with nontrivial $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$0 on $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$1, then necessarily

$$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$2

If $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$3, then for any $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$4, the bound forces

$$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$5

This provides a rigorous PDE version of the statement that lower-dimensional dissipative sets are incompatible with naive Kolmogorov $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$6 scaling for higher moments, and the paper correspondingly derives a restriction on structure-function exponents $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$7 relative to the naive law $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$8 (Rosa et al., 14 Feb 2025).

4. Increment diagnostics, circulation production, and comparison with BKM

The DR framework has an operational experimental form because it depends only on increments and a smoothing kernel. The direct DR criterion uses $$\mathscr{D}(\vec u) \overset{def}{=} \lim_{\ell\rightarrow 0} \mathscr{D}_\ell (\vec u) = \lim_{\ell\rightarrow 0} \ \frac{1}{4\ell} \int_\mathcal{V} d\vec r \ (\vec\nabla G_\ell)(\vec r) \cdot \delta\vec u(\vec r) \ |\delta\vec u (\vec r)|^2,$$9 as a local detector for singularities with

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$0

by searching for regions where $$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$1 does not vanish as $$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$2 decreases (Kuzzay et al., 2016).

Eyink’s circulation-production criterion enlarges the detectable range. It is based on

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$3

with turbulent vortex-force

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$4

Its scaling

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$5

shows detectability for singularities with

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$6

The hierarchy

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$7

means that circulation production can detect rough structures that need not generate DR anomalous energy dissipation (Kuzzay et al., 2016).

Planar stereoscopic PIV can be sufficient. The planar analogue

$$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$8

agrees asymptotically with the three-dimensional object when the field is regular in the missing direction: $$\delta \vec u(\vec r) = \vec u(\vec x+\vec r)-\vec u(\vec x)$$9 Thus, if $G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$0, it indicates a genuine singular structure in the full flow, although SPIV cannot detect singularities living only along the missing direction (Kuzzay et al., 2016).

Criterion	Diagnostic quantity	Detectable threshold
Duchon–Robert	$G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$1	$G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$2
Eyink circulation	$G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$3	$G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$4
Beale–Kato–Majda	$G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$5	vorticity blow-up criterion

The BKM criterion,

$G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$6

is a vorticity blow-up condition, whereas DR is an anomalous-energy-dissipation criterion formulated in increments. In the boundary-layer wind-tunnel experiment, strong-DR regions and strong-vorticity regions were substantially correlated: the paper reports $G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$7 between $G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$8 and $G_\ell(\vec r)=\ell^{-3}G(\vec r/\ell)$9, $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$0 between planar DR and $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$1, $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$2 between strong-transfer regions in 2D and 3D DR maps, and $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$3 between circulation maps and DR events (Kuzzay et al., 2016).

5. DR-type defect measures and local exact laws in related models

A major extension of the Duchon–Robert perspective is the identification of a defect distribution for each quadratic invariant whose conservation can fail for weak or rough solutions. The general pattern is: mollify the equations, derive a local balance for the mollified quantities, isolate the nonlinear commutator, and pass to the limit $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$4. The commutator limit is the DR-type dissipation term, and it matches a local third-order Yaglom law (Wang et al., 2023).

For the temperature equation

$$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$5

the paper defines

$$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$6

and obtains

$$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$7

For inviscid MHD in Elsässer variables, analogous defects $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$8 and $$\mathscr{D}_\ell(\vec u)=O\!\left(\frac{\delta u(\ell)^3}{\ell}\right), \qquad \delta u(\ell)=\sup_{|\vec r|<\ell}|\delta\vec u(\vec r)|$$9 yield local balances for $\delta u(\ell)\sim \ell^h$0 and $\delta u(\ell)\sim \ell^h$1, together with

$\delta u(\ell)\sim \ell^h$2

In primitive MHD variables, the same method produces DR-type local defects for total energy and cross-helicity, denoted $\delta u(\ell)\sim \ell^h$3 and $\delta u(\ell)\sim \ell^h$4, and exact identities relating them to six third-order quantities $\delta u(\ell)\sim \ell^h$5 (Wang et al., 2023).

For helicity in the ideal Euler equations, the framework yields a defect $\delta u(\ell)\sim \ell^h$6 in the local helicity balance

$\delta u(\ell)\sim \ell^h$7

together with a local helicity $\delta u(\ell)\sim \ell^h$8 law

$\delta u(\ell)\sim \ell^h$9

The same logic is extended to the Oldroyd-B model and to six new $q$00 laws for subgrid-scale $q$01-models, including a cross-helicity law for Leray-$q$02 MHD (Wang et al., 2023).

System	Conserved quantity	DR-type statement
Temperature advection	$q$03	$q$04
Inviscid MHD	Elsässer energies	$q$05, $q$06
Primitive MHD	Energy, cross-helicity	defects $q$07, $q$08
Euler	Helicity	$q$09
Oldroyd-B	Energy	$q$10

In this literature, the DR and Eyink formulations are treated as essentially equivalent local exact-law frameworks: the former emphasizes the dissipation distribution arising from roughness, and the latter the third-order increment law, but the mollification limit identifies the two (Wang et al., 2023).

6. Compressible generalization and distinct uses of the Duchon–Robert name

The compressible extension works with the density-weighted velocity

$q$11

and the auxiliary fields

$q$12

The exact local kinetic-energy balance becomes

$q$13

with

$q$14

$q$15

$q$16

Here $q$17 is the compressible analogue of the classical DR term, $q$18 is a new pressure–density–velocity coupling, and $q$19 is a viscous stress–density–velocity coupling (Zinchenko et al., 5 Aug 2025).

The comparison with Aluie’s coarse-graining framework identifies the correspondences

$q$20

$q$21

$q$22

This suggests that the incompressible DR picture survives in compressible flow as one part of a broader increment-based dissipation theory in which baropycnal transfer and density-dependent viscous effects appear as additional channels (Zinchenko et al., 5 Aug 2025).

In one-dimensional compressible shock flows, the framework detects strong local maxima of $q$23 at shock fronts, and these maxima sharpen as $q$24 decreases. For the singular profile $q$25, the local dissipation behaves as

$q$26

Consequently, $q$27 gives divergence as $q$28, $q$29 gives finite nonzero dissipation $q$30, and $q$31 gives vanishing dissipation. For the compressible baropycnal term in the symmetric case, finite local dissipation occurs at the different critical exponent $q$32 (Zinchenko et al., 5 Aug 2025).

The literature also uses the Duchon–Robert name in a distinct sense for an analytic fixed-point method in Fourier-measure spaces for vortex sheets and related interface problems. In that setting, the framework consists of analytic-in-time and analytic-in-space solutions, diagonalization of the linearized interface system in Fourier space, Duhamel operators integrating from $q$33 to $q$34, and nonlinear estimates in $q$35-type spaces; it was extended to Rayleigh–Taylor vortex sheets and to the Muskat problem with small initial data (Beck et al., 2012). This is a separate usage from the anomalous-dissipation framework, even though both originate in work associated with Duchon and Robert.