From diffusion to transmission via EDP-convergence: a paradigmatic multiscale limit

Technical Summary: From Diffusion to Transmission via EDP-Convergence

Problem Statement
The paper investigates the rigorous derivation of effective gradient structures for nonlinear diffusion equations on a one-dimensional domain $D = ]-1, 1[$ containing a thin, highly resistive membrane layer of width $\varepsilon$ near $x=0$. The diffusion coefficient (mobility) $A_\varepsilon$ is scaled to be of order $\varepsilon$ within this layer, while remaining $O(1)$ in the bulk regions. While the convergence of the solutions $v_\varepsilon$ to a limiting system with transmission conditions (a membrane model) is well-understood, the authors address a more subtle question: Does the associated family of gradient structures (defined by a free energy $E$ and a dissipation potential $R_\varepsilon$) converge to a limiting gradient structure?

Specifically, the paper studies the convergence in the sense of the Energy-Dissipation Principle (EDP). Unlike standard solution convergence, EDP-convergence seeks to identify the unique limiting dissipation potential $R_{\text{eff}}$ that governs the macroscopic kinetics. A central challenge is that while the microscopic dissipation potentials $R_\varepsilon$ are quadratic (leading to linear Onsager relations), the effective macroscopic dissipation potential $R_{\text{eff}}$ may exhibit non-quadratic, even exponential, behavior due to the singular limit of the membrane layer.

Methodology
The authors employ the theory of evolutionary $\Gamma$-convergence for gradient systems. The core methodology involves:

1. Rescaling and Variable Transformation: The thin membrane layer $[0, \varepsilon]$ is stretched to a fixed interval $I_0 = [0, 1]$ via the transformation $y = x/\varepsilon$. This reveals a separation of time scales: the density adjusts instantaneously within the membrane to a Non-Equilibrium Steady State (NESS) profile determined by the boundary values, while the bulk evolves on a slower diffusive time scale.
2. $\Gamma$-Convergence of Dissipation Functionals: The authors analyze the $\Gamma$-convergence of the total dissipation functional $D_\varepsilon$ (defined via the energy-dissipation balance) to a limit $D_0$. This requires proving both a $\Gamma$-liminf and a $\Gamma$-limsup estimate.
   • Leminf Estimate: Established using weak convergence methods, convexification techniques (via a transformation $\psi$), and the characterization of the limiting flux in the membrane part. A key difficulty is the lack of temporal compactness in the membrane, which is overcome by showing the limiting flux is spatially constant and absolutely continuous.
   • Limsup Estimate: Constructed via a recovery sequence involving temporal smoothing and flux corrections. This part relies on stronger concavity assumptions on the mobility and energy density to utilize Jensen's inequality.
3. NESS and BER Functions: The derivation of the effective membrane potential relies on the theory of Non-Equilibrium Steady States (NESS) characterized as saddle points of "BER functions" (B-functionals for Energy and Dissipation). This abstract theory allows the reduction of the microscopic cell problem to an effective kinetic relation.
4. Cell Problem Formulation: The effective membrane dissipation is derived from a cell problem minimizing a functional $M$ over profiles $u(y)$ in the membrane, subject to fixed boundary values and a fixed flux $\kappa$.

Key Contributions and Results

• Derivation of Effective Kinetic Relations: The paper rigorously derives the effective dual dissipation potential $R^*_{\text{eff}}$ for the membrane. It is shown that $R^*_{\text{eff}}$ consists of bulk terms (quadratic) and a membrane term $R^*_{\text{mb}}$ that depends on the jump in chemical potential $[\xi]_0$ and the traces of the density $v_\pm$.
• Dependence on Microscopic Parameters: A crucial finding is that the effective kinetic relation depends on the specific choices of the free energy $E$ and mobility $m$, not just on the combined exponent $\alpha$ of the resulting PDE.
  • For power-law structures $m(v) = v^\beta$ and $E = E_p$ (where $E''(v) \sim v^{p-2}$), the effective potential $R^*_{\text{mb}}$ depends on both $\beta$ and $p$ independently.
  • Exponential Kinetics: In the case of the linear Fokker-Planck equation ($\alpha=1$) with the Otto gradient structure ($\beta=1, p=1$, corresponding to Boltzmann entropy), the limiting membrane potential is exponential. Specifically, $R^*_{\text{mb}} \sim \cosh(\beta [\xi]_0)$, recovering the classical Marcelin-De Donder kinetics.
  • Power-Law Kinetics: For $p > 1$, the kinetic relation exhibits power-law behavior $R^*_{\text{mb}} \sim |[\xi]_0|^\gamma$ with $\gamma = 1 + \alpha/(p-1)$.
• Rigorous Proof of EDP-Convergence: The paper provides the first rigorous proof of EDP-convergence for this class of degenerate diffusion problems, establishing the convergence of solutions to the effective gradient flow under appropriate initial conditions.
• Explicit Formulas: The authors provide explicit formulas for the membrane potential and its dual for various power-law regimes, including the critical case of Boltzmann entropy.

Significance and Claims
The paper claims that EDP-convergence is a flexible framework capable of handling degenerate limits where the structure of the dissipation potential fundamentally changes (e.g., from quadratic to exponential or general power laws).

• Separation of Energetics and Kinetics: The authors address the apparent paradox that the effective kinetic relation depends on the microscopic energy functional $E$. They argue that while Otto's philosophy of separating energetics and kinetics holds on a single scale, singular limits mix these scales. The "cell problem" (minimizing dissipation in the membrane) couples the microscopic energetics (via $E$) and kinetics (via $m$) to produce the macroscopic kinetic relation.
• Physical Relevance: The work justifies the use of non-quadratic, exponential kinetic relations (like Marcelin-De Donder) for membrane transmission as the rigorous macroscopic limit of linear diffusion with specific microscopic gradient structures, rather than as ad-hoc phenomenological models.
• Generality: The results generalize previous formal calculations (e.g., in [LM*17]) to a broad class of mobilities and energy densities, covering a wide range of exponents $\beta$ and $p$.

The authors remain modest regarding generalizations, noting that while the theory suggests applicability to higher dimensions and reaction-diffusion systems, a rigorous analysis of these cases remains a challenge for future research. The paper focuses strictly on the one-dimensional membrane problem and the rigorous derivation of the effective gradient structure.