🔢 mathematics

Global weak solutions to the Cahn-Hilliard equation with degenerate mobility and singular diffusion

This paper establishes the existence of global weak solutions to the three-dimensional Cahn-Hilliard equation with degenerate mobility and singular diffusion in convex domains by deriving novel $L^4(0,T; H^2(\Omega))$ estimates for the solution and its arcsine transformation without relying on the classical entropy-based approach.

Original authors: Monica Conti, Andrea Giorgini, Greta Ricchi

Published 2026-08-07

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Monica Conti, Andrea Giorgini, Greta Ricchi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where materials don't just sit still but constantly rearrange themselves, like a crowd of people at a party slowly sorting into groups based on who they like. This is the fascinating realm of phase separation, a process where a mixed substance spontaneously splits into distinct regions. Think of oil and water separating in a bottle, or the colorful swirls in a marble cake. Scientists study this using mathematical models called Cahn-Hilliard equations. These equations act like a rulebook for how the "concentration" of different ingredients changes over time.

Two main characters drive this story: mobility and diffusion. Mobility is like how easily the ingredients can move around; in some materials, they get stuck or "frozen" when they reach a pure state (like 100% oil or 100% water). Diffusion is the force that smooths things out, trying to mix everything evenly. The tricky part of this specific paper is that the material behaves strangely at the extremes: the ability to move vanishes exactly when the mixture becomes pure, and the "smoothing" force becomes infinitely strong. It's like trying to run through a crowd that suddenly turns into a solid wall the moment you try to reach the edge. Understanding how these materials evolve is crucial for making better plastics, polymers, and advanced materials, but the math gets incredibly messy when things get stuck or blow up at the edges.

This paper, written by Monica Conti, Andrea Giorgini, and Greta Ricchi, tackles the messy math of these "stuck" materials. They prove that even when the rules get weird—where movement stops at the pure phases and the smoothing force goes wild—there is still a valid, global solution to the problem. In simpler terms, they showed that the material's evolution doesn't break the universe or disappear into a mathematical black hole; it follows a predictable path all the way through time.

The authors' main achievement is finding a new way to measure the "roughness" of the solution. Usually, mathematicians use a tool called "entropy" to keep track of how things change, but that tool failed here because of the weird behavior at the edges. Instead, the team invented a clever trick: they changed their point of view. They introduced a new character, a function called $\phi(u) = \arcsin(u)$ , which acts like a special lens. When they looked at the problem through this lens, the impossible math suddenly became manageable. They proved that both the material's concentration and this new "lens" function stay well-behaved in a specific mathematical sense (specifically, they belong to a space called $L^4(0, T; H^2(\Omega))$ ). This means the solution is smooth enough to be real and reliable, even in three-dimensional space.

To get there, the researchers didn't just stare at the hard problem; they built a bridge. They created a series of "approximate" problems where the rules were slightly tweaked to be easier to handle (making the mobility slightly positive instead of zero). They solved these easier versions and then carefully squeezed the "tweak" parameter down to zero. Along the way, they had to prove that the solutions to these easy versions didn't go crazy as the tweaks disappeared. They used a special mathematical inequality (a tool that relates the shape of a domain to the behavior of functions on it) and the fact that their container was convex (shaped like a ball or a box, with no inward dents) to keep everything under control.

The paper explicitly rules out the idea that these solutions might be chaotic or non-existent under these specific conditions. They don't just suggest it might work; they provide a rigorous mathematical proof that a global weak solution exists for any amount of time $T$ , provided the starting material has finite energy. They also show that their solution fits into a broader framework used by other scientists (specifically the work of Cancès and Matthes), confirming that their new lens approach is consistent with existing theories.

So, what did they find? They found that even in the most stubborn, "stuck" scenarios of phase separation, nature follows a consistent mathematical script. The material will separate, the domains will grow and merge, and the process will continue indefinitely without the math breaking down. By proving this, the authors have given scientists a solid foundation to simulate and understand these complex materials, ensuring that when engineers design new polymers, the math behind the scenes is as stable as the materials themselves.

Technical Summary: Global Weak Solutions to the Cahn-Hilliard Equation with Degenerate Mobility and Singular Diffusion

Problem Statement
The paper addresses the initial-boundary value problem for the Cahn-Hilliard system modeling phase separation in polymer blends, governed by the Flory-Huggins-de Gennes free energy. The system is defined on a bounded, smooth, convex domain $\Omega \subset \mathbb{R}^3$ over a time interval $(0, T)$ . The governing equations are:
$\partial_t u = \text{div}(b(u)\nabla\mu), \quad \mu = -a(u)\Delta u - \frac{a'(u)}{2}|\nabla u|^2 + f(u) - \theta_0 u,$
subject to homogeneous Neumann boundary conditions and an initial datum $u_0$ .

The specific physical regime under investigation involves:

Degenerate Mobility: $b(u) = 1 - u^2$ , which vanishes at the pure phases $u = \pm 1$ .
Singular Diffusion: $a(u) = \frac{2}{1-u^2}$ , which becomes singular at the pure phases.
Singular Potential: The free energy density involves the Flory-Huggins entropy $F(u)$ , which is singular at $u = \pm 1$ .

The primary mathematical challenge lies in the simultaneous presence of degenerate mobility and singular diffusion. While existence results exist for cases where either the mobility is non-degenerate or the diffusion is constant, the combination of both degeneracy and singularity in three dimensions, particularly with the specific Flory-Huggins-de Gennes structure, presents significant analytical difficulties. Standard entropy-based approaches, which typically yield $L^2(0, T; H^2(\Omega))$ estimates, are insufficient to handle the singular terms without relying on the chemical potential $\mu$ being well-defined in standard Lebesgue spaces, which is not guaranteed here.

Methodology
The authors employ a regularization approach combined with a novel change of variables and a new elliptic estimate.

Approximation Scheme: The degenerate mobility $b(u)$ and singular coefficient $a(u)$ are approximated by strictly positive sequences $b_\varepsilon(u) = 1 - u^2 + \varepsilon$ and $a_\varepsilon(u) = 2/b_\varepsilon(u)$ . Crucially, the approximation preserves the fundamental algebraic identities linking the coefficients: $a_\varepsilon(u)b_\varepsilon(u) = 2$ and $b_\varepsilon(u)f'(u) = \theta$ . The potential $\Psi$ is kept in its original singular form.
Existence of Approximating Solutions: Using existing results for non-degenerate mobility and regular diffusion (specifically from Giorgini et al. [13]), the authors establish the existence of global weak solutions $u_\varepsilon$ for the regularized system. These solutions satisfy $|u_\varepsilon| < 1$ almost everywhere.
Change of Variable: To handle the singular diffusion, the authors introduce the transformation $\phi(u) = \arcsin(u)$ . This choice is motivated by the identity $\frac{a(u)}{2}|\nabla u|^2 = |\nabla \phi(u)|^2$ . The regularized version is $\phi_\varepsilon(u) = \arcsin(u/\sqrt{1+\varepsilon})$ .
Key Elliptic Estimate: The core of the analysis is the derivation of uniform bounds for $\phi_\varepsilon(u_\varepsilon)$ $ϕ_{ε} (u_{ε})$ and $u_\varepsilon$ $u_{ε}$ in $L^4(0, T; H^2(\Omega))$ $L^{4} (0, T; H^{2} (Ω))$ . Unlike previous works relying on entropy estimates (which yield $L^2$ $L^{2}$ in time), the authors derive a stronger $L^4$ $L^{4}$ estimate. This is achieved by:
- Testing the equation with $-\Delta u_\varepsilon$ .
- Utilizing an integration by parts formula (Dal Passo, Garcke, and Grun) involving an auxiliary function $h_\varepsilon$ tailored to the specific structure of $\phi_\varepsilon$ .
- Exploiting the convexity of the domain $\Omega$ to ensure the non-negativity of boundary terms.
- Combining these with the energy inequality to control the singular terms.
Passage to the Limit: Compactness arguments (Aubin-Lions lemma) are used to extract a limit function $u$ . A critical step involves proving that the term $\varepsilon \int f(u_\varepsilon) \text{div}\eta$ vanishes as $\varepsilon \to 0$ . This is achieved by establishing a specific bound on $\|f(u_\varepsilon)\|_{L^2(0,T; L^1(\Omega))}$ that grows at most as $\varepsilon^{-1/2}$ , which is compensated by the prefactor $\varepsilon$ .

Key Contributions and Results

Theorem 1.1 (Existence of Global Weak Solutions): The paper proves the existence of global weak solutions to the Cahn-Hilliard system with degenerate mobility and singular diffusion in 3D bounded, smooth, convex domains, assuming finite initial energy.
- Regularity: The solution $u$ satisfies $-1 \le u \le 1$ a.e. and belongs to $L^\infty(0, T; H^1(\Omega)) \cap L^4(0, T; H^2(\Omega)) \cap C([0, T]; L^2(\Omega))$ .
- Enhanced Regularity for Transformed Variable: The transformed variable $\phi(u) = \arcsin(u)$ also belongs to $L^4(0, T; H^2(\Omega))$ . This $L^4$ -in-time regularity is a significant improvement over the standard $L^2$ -in-time regularity obtained via entropy methods.
- Weak Formulation: The solution satisfies a specific integral identity involving a flux vector $J \in L^2(0, T; L^2(\Omega))$ , defined distributionally to handle the singularity where $b(u)=0$ . The formulation avoids explicit dependence on the chemical potential $\mu$ in the flux definition.
- Energy Inequality: The solution satisfies the energy inequality $E(u(t)) + \int_0^t \int_\Omega |\tilde{J}|^2 dx ds \le E(u_0)$ , where $\tilde{J}$ is related to the flux.
Theorem 1.3 (Equivalent Formulation): The authors demonstrate that their notion of weak solution is equivalent to the formulation introduced by Cancès and Matthes [6] for nonlocal systems. They define a weighted chemical potential $q = -2\Delta \phi(u) + \sqrt{b(u)}\Psi'(u)$ and show that the solution satisfies an integral identity involving $q$ . This connects the local singular model to the gradient-flow structure in weighted Wasserstein spaces.

Significance and Claims
The paper claims that the derivation of $L^4(0, T; H^2(\Omega))$ estimates for both the solution $u$ and the transformed variable $\phi(u)$ without relying on the classical entropy-based approach is a key novelty. This stronger integrability is obtained by fully exploiting the energy bounds and the specific algebraic relations between the mobility and diffusion coefficients.

The authors note that the convexity of the domain is a crucial assumption, consistent with previous analytical studies of singular diffusion models, and extending these results to non-convex domains remains an open problem. Furthermore, while uniqueness and long-time dynamics characterization remain open challenges even for constant diffusion coefficients, this work establishes the foundational existence theory for the physically relevant case of degenerate mobility and singular diffusion in three dimensions.

The techniques developed, particularly the new elliptic estimate, are suggested to be adaptable to a broader class of mobility functions (e.g., $b(u) = (1-u^2)^m$ with $m > 1$ ) and potentially to coupled Navier-Stokes/Cahn-Hilliard systems, though these extensions are reserved for future work.

More like this