Modulated Stokes-Dirac Structure

• Modulated Stokes-Dirac Structure is a framework where state-dependent skew operators enable dynamic energy modulation in systems like the 2D Navier–Stokes equations.
• It distinguishes between state-dependent modulation, topological augmentation using harmonic forms, and geometric lifting via contact structures.
• Recent work demonstrates structure-preserving discretization techniques that accurately maintain energy and enstrophy balances in nonlinear fluid simulations.

A modulated Stokes-Dirac structure, in the strict sense supported by the recent port-Hamiltonian literature, is a Stokes-Dirac interconnection in which the skew operator is state-dependent, so that the distributed power-conserving structure itself varies with the evolving field variables. In the cited literature, this usage appears most directly in the port-Hamiltonian formulation of the nonlinear 2D incompressible Navier–Stokes equations in vorticity–stream function variables, where the interconnection takes the form $J=J(\psi,\omega)$ (Bendimerad-Hohl et al., 9 Jul 2025). Two closely related lines of work are often relevant to the same topic but are not modulation theories in this strict sense: a topological geometric extension of the classical Stokes-Dirac framework by harmonic forms on manifolds with boundary and non-trivial topology (Nishida et al., 2018), and a contact-geometric lifting of distributed-parameter port-Hamiltonian systems with respect to Stokes-Dirac structures to contact bundles (Goto, 2017).

1. Terminological scope and conceptual boundaries

The expression “modulated Stokes-Dirac structure” is not used uniformly across the literature. In the strict control-theoretic sense represented by the 2025 formulation, modulation means that the interconnection operator is state-dependent while remaining skew-symmetric in the appropriate weak sense; the nonlinear transport term is placed in the interconnection itself rather than in the Hamiltonian or in dissipation (Bendimerad-Hohl et al., 9 Jul 2025). This yields a genuine modulation of the Stokes-Dirac operator.

By contrast, the 2018 work on global energy flows states explicitly that its contribution is not a “modulated Stokes-Dirac structure” in the usual state-dependent or time-varying sense. Its extension is topological: the classical Stokes-Dirac structure is augmented by harmonic terms arising from the Hodge-Morrey-Friedrichs decomposition on manifolds with boundary and non-trivial topology (Nishida et al., 2018). Likewise, the 2017 contact-geometric treatment does not introduce a separate theory of modulation, time-varying structure matrices, or state-dependent Dirac structure deformation; instead, it presents the standard Stokes-Dirac structure and rewrites a quadratic subclass of distributed-parameter port-Hamiltonian systems as contact Hamiltonian vertical vector fields on contact bundles (Goto, 2017).

A common misconception is therefore to group all three developments under a single notion of “modulation.” The literature supports a sharper distinction: state-dependent modulation in the Navier–Stokes port-Hamiltonian formulation, topology-induced augmentation in the harmonic extension, and geometric lifting in the contact-bundle formulation.

2. Classical Stokes-Dirac framework

The classical Stokes-Dirac structure is the distributed analogue of the skew-symmetric Dirac interconnection familiar from finite-dimensional port-Hamiltonian systems. On an $n$-dimensional connected manifold $Z$, with integers $p,q$ satisfying

$0\le p,q\le n,\qquad p+q=n+1,$

the standard formulation uses flow and effort variables related by

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$

The associated bilinear power pairing defines orthogonality, and the Stokes-Dirac structure $D_Z$ satisfies

$D_Z=D_Z^\perp,$

which is the defining isotropy property in the paper’s formulation (Goto, 2017).

An equivalent notation, recalled in the topological extension paper on a compact oriented subdomain $M\subset N$ with boundary $\partial M$, writes

$n$0

with boundary port variables

$n$1

In both formulations, integration by parts and Stokes’ theorem generate the boundary power variables and ensure power continuity (Nishida et al., 2018).

The basic power balance is

$n$2

or, with the corresponding notation of the contact-geometric paper,

$n$3

The boundary pairing is derived through Green’s formula,

$n$4

which links exterior differentiation, codifferentiation, and boundary traces in the weak formulation (Nishida et al., 2018).

3. State-dependent modulation in nonlinear fluid port-Hamiltonian systems

The strict notion of a modulated Stokes-Dirac structure appears in the 2025 treatment of the nonlinear 2D incompressible Navier–Stokes equations. The starting point is the standard finite-dimensional port-Hamiltonian representation

$n$5

with

$n$6

In the infinite-dimensional setting, the distributed analogue uses operators $n$7 satisfying

$n$8

This is the standard Stokes-Dirac mechanism by which the interior skew relation produces power balance and boundary port variables (Bendimerad-Hohl et al., 9 Jul 2025).

For the homogeneous incompressible Navier–Stokes equations,

$n$9

the 2D vorticity–stream function variables are

$Z$0

The paper emphasizes two equivalent Hamiltonian viewpoints. The kinetic-energy viewpoint gives

$Z$1

while the enstrophy viewpoint gives

$Z$2

Combining them yields a coupled port-Hamiltonian system with Hamiltonian

$Z$3

where

$Z$4

The defining feature is that the interconnection is modulated: $Z$5 In the displayed system of the paper, this appears as

$Z$6

The modulation therefore means that the nonlinear transport term is encoded as a state-dependent redistribution of energy inside the skew interconnection rather than as artificial dissipation or ad hoc forcing. The paper remarks that this is “an original feature”: the Hamiltonian is separable, while coupling occurs only through modulation of $Z$7 (Bendimerad-Hohl et al., 9 Jul 2025).

4. Stokes-Lagrange structure and boundary energy ports

The modulated Stokes-Dirac formulation in the fluid example is coupled to a second geometric object, the Stokes-Lagrange structure, introduced to treat differential or implicit constitutive relations that do not fit the usual explicit relation $Z$8. In the general distributed setting, the operators

$Z$9

together with energy-port operators

$p,q$0

satisfy the symmetry relation

$p,q$1

This is the distributed analogue of $p,q$2, now with energy-control terms built in (Bendimerad-Hohl et al., 9 Jul 2025).

In the Navier–Stokes case, the relation

$p,q$3

is treated as an implicit constitutive law within the Stokes-Lagrange part of the formulation. The weak form produces three types of ports: power ports $p,q$4, associated with boundary power exchange; energy ports $p,q$5, associated with the implicit constitutive relation for the stream function; and resistive ports from viscous dissipation. The explicit energy-port contribution in the weak formulation is

$p,q$6

The boundary control variables extracted from the weak form include

$p,q$7

$p,q$8

$p,q$9

Under no-slip conditions, the formulation keeps

$0\le p,q\le n,\qquad p+q=n+1,$0

as constraints, so that the problem is treated as a DAE-like system in which boundary vorticity acts as a Lagrange multiplier. In this setting,

$0\le p,q\le n,\qquad p+q=n+1,$1

plays the role of vorticity generation at the boundary, and the enstrophy balance becomes

$0\le p,q\le n,\qquad p+q=n+1,$2

while the kinetic energy satisfies

$0\le p,q\le n,\qquad p+q=n+1,$3

The modulated formulation is used precisely to make this separation between interior dissipation and boundary enstrophy generation explicit (Bendimerad-Hohl et al., 9 Jul 2025).

5. Topological augmentation by harmonic forms

The 2018 topological extension shows that the classical Stokes-Dirac structure is incomplete on manifolds with non-trivial topology, because exact and coexact components alone do not capture all global energy flows. On a compact oriented complete manifold with boundary, the Hodge-Morrey-Friedrichs decomposition is written as

$0\le p,q\le n,\qquad p+q=n+1,$4

where $0\le p,q\le n,\qquad p+q=n+1,$5 is exact, $0\le p,q\le n,\qquad p+q=n+1,$6 is coexact, $0\le p,q\le n,\qquad p+q=n+1,$7 is coexact harmonic, and $0\le p,q\le n,\qquad p+q=n+1,$8 is tangential harmonic. The extension is therefore not a state-dependent or time-dependent modulation; it is the classical Stokes-Dirac structure augmented by harmonic terms reflecting topology (Nishida et al., 2018).

The analytic setting uses the standard operators

$0\le p,q\le n,\qquad p+q=n+1,$9

with $$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$0-pairing

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$1

Harmonic fields are

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$2

with tangential and normal harmonic subspaces

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$3

The Hodge-Morrey decomposition

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$4

and Friedrichs decompositions

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$5

supply the harmonic sector required for global bookkeeping. The topological backbone is the Hodge isomorphism

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$6

which links cohomology classes directly to harmonic subspaces.

The classical boundary power term is then extended to

$$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$7

The first term is the usual boundary energy flow associated with exact fields. The second term is the harmonic boundary energy flow arising from the topology of $$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$8. The corresponding integrability condition for solving $$f_p = (-1)^r\, d e_q,\qquad f_q = d e_p,\qquad f_a = -\left.e_q\right|_{\partial Z},\qquad e_a = (-1)^{n-q}\left.e_p\right|_{\partial Z}, \qquad r=pq+1.$$9 with boundary data is

$D_Z$0

and the explicit harmonic boundary energy relations are

$D_Z$1

$D_Z$2

These are explicitly identified as harmonic boundary energy flows (Nishida et al., 2018).

The sphere–torus comparison clarifies the topological content. For $D_Z$3, $D_Z$4, so there are no nontrivial loop or harmonic circulation modes. For $D_Z$5, $D_Z$6, so there are two independent non-contractible loops and nontrivial harmonic modes. The additional terms in the boundary energy expressions on $D_Z$7 represent global circulation modes and topological obstructions to representing everything as exact flows.

6. Contact-geometric reinterpretation

The contact-geometric description treats distributed-parameter port-Hamiltonian systems with respect to Stokes-Dirac structures on a contact bundle

$D_Z$8

where $D_Z$9 is a Riemannian base manifold and each fiber $D_Z=D_Z^\perp,$0 is a contact manifold. The fiber contact form is in Darboux form

$D_Z=D_Z^\perp,$1

with Reeb vertical vector field

$D_Z=D_Z^\perp,$2

The contact Hamiltonian vertical vector field $D_Z=D_Z^\perp,$3 is defined by

$D_Z=D_Z^\perp,$4

and in local coordinates this becomes

$D_Z=D_Z^\perp,$5

The distributed-parameter port-Hamiltonian system appears after restricting the contact dynamics to a Legendre submanifold defined by adapted functions encoding the constitutive constraints (Goto, 2017).

For the quadratic subclass, the energy functional is

$D_Z=D_Z^\perp,$6

with total Legendre transform

$D_Z=D_Z^\perp,$7

The constitutive relations are

$D_Z=D_Z^\perp,$8

For $D_Z=D_Z^\perp,$9, the paper proves that distributed-parameter port-Hamiltonian systems with respect to Stokes-Dirac structures can be written as contact Hamiltonian vertical vector fields on the bundle, with fiber dimensions $M\subset N$0, $M\subset N$1, $M\subset N$2, or $M\subset N$3 depending on the choice of $M\subset N$4 and the spatial dimension.

This framework does not introduce a modulated Stokes-Dirac structure in the strict sense. Its contribution is a fiberwise, metric-dependent, Legendre-constrained representation of the standard Stokes-Dirac dynamics. The Riemannian metric and Hodge star act as geometric parameters in the constitutive relations, and for the quadratic case the induced metric

$M\subset N$5

gives a dually flat space in the sense of information geometry (Goto, 2017).

7. Discretization, preserved balances, and limitations

The 2025 formulation develops a structure-preserving space discretization through a partitioned finite element method. The principle is to integrate by parts only where needed to expose the port variables and to choose compatible finite element spaces so that the discrete operators inherit the same symmetry and skew-symmetry relations as the continuous system. The required algebraic targets are: discrete $M\subset N$6 skew-symmetric, discrete $M\subset N$7, and discrete resistive matrices symmetric positive semidefinite (Bendimerad-Hohl et al., 9 Jul 2025).

For the 2D Navier–Stokes case, $M\subset N$8 is approximated in $M\subset N$9 with an Argyris element, $\partial M$0 in $\partial M$1 with a $\partial M$2 Lagrange space, and boundary controls and observations with a $\partial M$3 boundary element family. The state approximations are

$\partial M$4

The semi-discrete system is

$\partial M$5

The nonlinear matrices satisfy

$\partial M$6

which is the discrete analogue of modulated skew-adjointness.

At the semi-discrete level, the discrete kinetic energy and enstrophy are

$\partial M$7

with total Hamiltonian

$\partial M$8

The paper proves

$\partial M$9

$n$00

Thus the semi-discrete method preserves the exact algebraic power structure: nonlinear transport remains energy-neutral due to skew-symmetry, while dissipation is represented by symmetric positive terms.

For time discretization, the paper uses a staggered scheme inspired by earlier work. The modulated Dirac operator is frozen over half steps, and the linearized subsystem is discretized with an implicit midpoint rule. The reported result is that both kinetic energy and enstrophy are preserved very accurately, and that the enstrophy balance error is near machine precision in the numerical experiments.

The same paper states several limitations. The Navier–Stokes example is restricted to a simply connected 2D domain. The no-slip condition is handled as a constraint/Lagrange-multiplier condition rather than eliminated into a nonlocal boundary condition. A full convergence and stability theory for all such modulated systems is left as future work. These limitations delineate the current scope of the strict modulated Stokes-Dirac framework, while the earlier topological and contact-geometric papers indicate two complementary generalizations: one cohomological and one geometric, neither of which is a modulation theory in the strict state-dependent sense (Bendimerad-Hohl et al., 9 Jul 2025).