Hamiltonian Interface Dynamics for Reduced-Order Optimization of Incompressible Mixing
Ziqian Li 1 2, Enrique Zuazua 1 3 4 · 2026
Read Hamiltonian Interface Dynamics for Reduced-Order Optimization of Incompressible Mixing by Ziqian Li 1 2, Enrique Zuazua 1 3 4 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-04688.txt.
Introduction
Ziqian Li and Enrique Zuazua develop a reduced-order method for improving mixing in flat incompressible flows. Mixing is the stirring that spreads a dye or pollutant through a fluid. Optimising the full transport equation is expensive, so they replace it with a geometric objective: maximise how much an advected material interface stretches.
That reformulation yields a finite-size Hamiltonian control problem. It is driven by stream functions with a few parameters. They derive the adjoint equations and reduced gradients. They discretise both with a midpoint rule. That rule keeps the discrete adjoint consistent with the discrete objective. They test on cellular flow and frontogenesis benchmarks. Optimised time-dependent flows stretch interfaces very fast, close to an exponential rate. Stationary flows are far slower.
Catalog record
| CREATOR | Ziqian Li 1 2, Enrique Zuazua 1 3 4 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.04688
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Hamiltonian Interface Dynamics for Reduced-Order Optimization of Incompressible Mixing Ziqian Li 1 2, Enrique Zuazua 1 3 4 Abstract . We develop a reduced-order framework for optimizing mixing in two-dimen-sional incompressible flows. Instead of optimizing the full transport PDE, the method maximizes the length of advected material interfaces, leading to a finite-dimensional Hamiltonian control problem based on parametrized stream functions. We derive the continuous adjoint equations and reduced gradients, and discretize the forward and adjoint dynamics with the implicit midpoint rule. The resulting discrete adjoint is algebraically consistent with the derivative of the fully discrete objective, up to the tolerance of the nonlinear midpoint solves. The approach applies to bounded two-dimensional domains with smooth finite-dimensional stream-function parametrizations. Numerical experiments on cellular-flow and Doswell frontogenesis benchmarks show that the optimized time-dependent Hamiltonians generate near-exponential interface stretching and substantially faster decay of the H ˙ − 1 \dot{H}^{-1} mix-norm, in contrast with the polynomial behavior observed for stationary flows. When evaluated on a common reference transport solver, the interface-based controls produce faster H ˙ − 1 \dot{H}^{-1} decay than a Eulerian Sobolev-norm optimizer under a matched setup, while substantially reducing computational cost. We also identify a limitation of the reduced model: increasing the control basis may further improve the interface-length objective without yielding proportional gains in H ˙ − 1 \dot{H}^{-1} mixing, confirming that interface length is an effective but not fully faithful proxy for mixing in geometrically complex regimes. 1. Introduction 1.1. Problem statement and motivation Understanding and controlling mixing by incompressible transport is a central issue in fluid mechanics, scientific computing, and applications ranging from microfluidics to geophysical transport. In this paper, we consider a passive scalar \theta transported by an incompressible velocity field \mathbf{v} in a bounded planar domain \Omega\subset\mathbb{R}^{2} . The governing model is the transport equation (1.1) \partial_{t}\theta+\mathbf{v}\cdot\nabla\theta=0\qquad\text{in }(0,T)\times\Omega, together with the incompressibility and no-penetration conditions (1.2) \nabla\cdot\mathbf{v}=0\qquad\text{in }(0,T)\times\Omega,\qquad\mathbf{v}\cdot\mathbf{n}_{\Omega}=0\qquad\text{on }(0,T)\times\Gamma, where \Gamma:=\partial\Omega and \mathbf{n}_{\Omega} is the outward unit normal. The optimal mixing problem is to choose the stirring field \mathbf{v} so that the transported scalar becomes mixed as efficiently as possible. A large part of the modern literature quantifies mixing by negative Sobolev norms, often called mix-norms . The homogeneous \dot{H}^{-1}(\Omega) norm (1.3) \|\theta\|_{\dot{H}^{-1}(\Omega)}:=\sup\Bigl\{\int_{\Omega}\theta\,\phi\,dx\;:\;\int_{\Omega}\phi\,dx=0,\;\int_{\Omega}|\nabla\phi|^{2}\,dx\leq 1\Bigr\} is a natural measure of mixedness, as introduced by Mathew, Mezić, and Petzold [ 14 ] and widely employed in the analysis of mixing rates [ 5 ] . A smaller value of \dot{H}^{-1}(\Omega) norm corresponds to better mixing, reflecting the transfer of scalar content from large to small spatial scales. However, negative Sobolev norms are not the only meaningful way to quantify mixing. In many two-phase transport settings, geometric quantities such as interface deformation, geometric mixing scales, or entropy-type functionals also provide natural descriptions of the mixing process; see, for example, [ 18 , 1 , 17 ] . The main purpose of this paper is to develop such a geometric viewpoint in a Hamiltonian framework. We take the final length of the material interface as the objective functional to be optimized, rather than the final \dot{H}^{-1}(\Omega) norm of the scalar. This choice is motivated by theoretical, computational, and modelling considerations, as we now explain. In two space dimensions, an incompressible velocity field admits a stream-function representation \mathbf{v}=\nabla^{\perp}\psi , under which the characteristic dynamics becomes Hamiltonian. Denoting by J=\bigl(\begin{smallmatrix}0&-1\\ 1&\phantom{-}0\end{smallmatrix}\bigr) the symplectic matrix, the flow map satisfies \dot{\Phi}_{\psi}=J\nabla_{x}\psi(t,\Phi_{\psi}) . An initial material interface \gamma_{0} is propagated by \Phi_{\psi} , and its position X(t,s)=\Phi_{\psi}(t;\gamma_{0}(s)) and tangent vector a(t,s)=\partial_{s}X(t,s) satisfy the coupled ODE system \partial_{t}X=J\nabla_{x}\psi(t,X),\qquad\partial_{t}a=J\nabla_{x}^{2}\psi(t,X)\,a. The final length of the interface \mathcal{L}(\psi;T)=\int_{0}^{1}|a(T,s)|\,ds is therefore expressed entirely in terms of the Hamiltonian flow and its differential, replacing the Eulerian full-field mixing objective by a Lagrangian geometric one, and the method yields a substantial computational reduction, quantified in Section 5.4 . The use of \mathcal{L}(\psi;T) as a surrogate for mixing has a transparent geometric rationale. For a two-phase configuration whose transition layer is concentrated near the material interface \Gamma_{t} , the area balance between the two phases suggests the heuristic scaling (1.4) \ell(t)\,\mathcal{L}(\psi;t)\sim|\Omega|, where \ell(t) denotes the typical filament width; growth of the interface length therefore naturally corresponds to decrease of a geometric mixing scale. This argument is heuristic and does not establish equivalence with negative Sobolev mix-norms; we use it solely to motivate the reduced objective, whose predictive value is assessed a posteriori by evaluating the optimized controls on the full transport equation ( 1.1 ). The reduction underlying our framework is Hamiltonian/Lagrangian (replacing the full Eulerian transport PDE by ODEs on a small set of advected markers) rather than a multigrid hierarchy in the sense of Glowinski’s two-grid philosophy [ 9 ] ; the spirit, however, is similar: optimization is carried out on a reduced model that captures the essential geometric mechanism, and the resulting controls are ultimately evaluated on the original transport equation (see Section 5.4 ). This viewpoint is also in line with the role of interfaces as primary geometric objects in level-set methods [ 16 ] and active contour models [ 3 ] . 1.2. Main results and contributions Our contributions are twofold: methodological and computational. On the methodological side, we introduce a reduced optimal-control framework for incompressible mixing based on the geometry of an advected material interface. On the computational side, the reduction yields lower optimization cost and improved \dot{H}^{-1} mix-norm decay (in the tested benchmarks) when the resulting controls are evaluated on the original transport equation. More precisely, our main results are as follows. • A geometric reduced model for mixing based on Hamiltonian interface dynamics. We show that, in two dimensions, the advection of a material interface by an incompressible flow is naturally governed by a Hamiltonian system: the interface position X and its tangent vector a=\partial_{s}X satisfy a closed system of ODEs, and the final length of the interface is determined entirely by the Hamiltonian flow and its linearization (Proposition 2.1 ). • An adjoint-based optimal-control formulation on the reduced model. We formulate a regularized control problem that maximizes the final interface length \mathcal{L}(\psi;T) under an L^{2} penalty on the control amplitude (equivalent, via the Gram matrix of the basis, to an L^{2}(0,T;L^{2}(\Omega)) penalty on the stream function). Within a finite-dimensional smooth stream-function ansatz, we prove the existence of an optimal Hamiltonian (Theorem 3.3 ), derive the continuous adjoint system, and obtain an explicit gradient formula in terms of adjoint variables transported along marker trajectories. This is done avoiding full-domain PDE solves during the optimization loop. • A fully discrete optimization framework. We discretize the forward and adjoint systems by the implicit midpoint rule, which preserves the Hamiltonian structure and is self-adjoint. As a consequence, the discrete adjoint obtained by transposing the linearized forward map coincides with the midpoint scheme run backward in time, so that the resulting discrete gradient is algebraically consistent with the derivative of the discrete cost at the scheme level, up to the tolerance of the nonlinear solves used to implement the midpoint step. The optimization is then carried out with a Polak–Ribière conjugate-gradient method and Armijo line search (Algorithm 4.1 ). • Validation on the full transport PDE. The reduced model tracks O(N_{p}) marker points instead of O(N_{\mathrm{cell}}^{2}) Eulerian degrees of freedom per time step; this Hamiltonian/Lagrangian reduction is the direct