Curvature penalization of strongly anisotropic interfaces models and their phase-field approximation

Basic Information

• Paper ID: 2510.13275
• Title: Curvature penalization of strongly anisotropic interfaces models and their phase-field approximation
• Authors: J.-F. Babadjian, B. Buet, M. Goldman
• Classification: math.AP (Mathematical Analysis)
• Publication Date: October 15, 2025 (arXiv preprint)
• Paper Link: https://arxiv.org/abs/2510.13275

Abstract

This paper investigates the effects of anisotropy on both sharp interface and diffuse interface models. When surface tension is a convex function of the interface normal vector, the anisotropy is termed weak anisotropy, which typically ensures lower semicontinuity of the associated energy. However, when surface tension depends on the normal vector in a non-convex manner, this strong anisotropy can lead to instabilities related to the loss of functional lower semicontinuity. The paper studies the regularization effects of adding Willmore-type higher-order terms to the energy. Two classes of problems are considered: the first is a non-convex anisotropic generalization of perimeter, and the second is a non-convex anisotropic Mumford-Shah functional. In both cases, lower semicontinuity of the energy with respect to natural convergence modes is established, as well as Γ-convergence-type results via phase-field approximation.

Research Background and Motivation

Problem Background

1. Anisotropic interface problems: In materials science, phase transition theory, and fracture mechanics, interface energy often exhibits anisotropic characteristics, meaning surface tension depends on the interface orientation.
2. Challenges of strong anisotropy: When the surface tension function φ(ν) is non-convex with respect to the normal vector ν, it leads to:
   • Lack of lower semicontinuity of the energy functional
   • Minimizing sequences may exhibit rapid oscillations
   • Non-existence of minima in the classical sense
3. Practical application needs:
   • Faceting phenomena in crystal growth
   • Crack propagation in brittle materials
   • Segmentation problems in image reconstruction

Research Motivation

Existing approaches to handling strong anisotropy problems fall into two main categories:

1. Relaxation method: Replace the original function φ with its convex envelope φ**, but this alters the physical essence
2. Curvature regularization: Add Willmore-type curvature terms to suppress oscillations

This paper adopts the second approach, conducting an in-depth investigation of the regularization effects of curvature penalization on strongly anisotropic problems.

Core Contributions

1. Established lower semicontinuity for strongly anisotropic perimeter functionals: Proved the lower semicontinuity property of the anisotropic perimeter functional F(E) with curvature penalization (Theorem 1.1)
2. Constructed Γ-convergence theory for phase-field approximations: For the perimeter problem, established Γ-convergence of the phase-field energy Fε to the sharp interface energy F (Theorem 1.3)
3. Extended to Mumford-Shah type problems:
   • Proved lower semicontinuity of the Mumford-Shah functional G(γ) with curvature and point energy (Theorem 4.4)
   • Established corresponding phase-field approximation theory (Theorem 1.4)
4. Handled singular structures: Capable of treating crack tips and multiply-connected points and other singular geometric structures
5. Generalized Gauss-Bonnet formula: Established a generalized Gauss-Bonnet formula in the appendix for 1-rectifiable varifolds with square-integrable first variation (Theorem 5.1)

Methodology Details

Problem Formulation

Perimeter Problem

Consider the energy functional: $F(E) = \int_{\partial^*E} \phi(\nu_E) \, d\mathcal{H}^{d-1} + \int_{\partial^*E} |H_E|^2 \, d\mathcal{H}^{d-1}$

where:

• E is a set of finite perimeter
• φ: S^{d-1} → ℝ₊ is a continuous but possibly non-convex anisotropy function
• H_E is the mean curvature vector

Mumford-Shah Problem

Consider the energy functional: $G^{(\gamma)}(u) = \int_{\Omega\setminus J_u} |\nabla u|^2 dx + \int_{J_u} (\phi(\nu_u) + |H_{J_u}|^2) d\mathcal{H}^1 + \gamma \mathcal{H}^0(P_{J_u})$

where:

• u ∈ SBV²(Ω) is a special function of bounded variation
• J_u is the jump set
• P_ is the set of endpoints of the jump set

Phase-Field Approximation Models

Phase-Field Approximation for Perimeter Problem

$F_ε(v) = \frac{1}{c_0}\int_Ω \phi_ε(\nabla v)\left(\frac{ε}{2}|\nabla v|^2 + \frac{1}{ε}W(v)\right)dx + \frac{1}{c_0}\int_Ω \frac{1}{ε}\left(-ε\Delta v + \frac{1}{ε}W'(v)\right)^2 dx$

Phase-Field Approximation for Mumford-Shah Problem

Using three phase-field variables (u,v,w): $G_ε^{(\gamma)}(u,v,w) = \frac{1}{4}\int_Ω (1+v)^2|\nabla u|^2 dx + \text{interface terms} + \text{curvature terms} + \text{point energy terms}$

Technical Innovations

1. Application of Varifold Theory

• Exploit weak* convergence properties of varifolds
• Establish connections between varifold convergence and rectifiability
• Prove convergence of strongly anisotropic terms in the limit

2. Use of Monotonicity Formulas

For a 1-rectifiable varifold V: $\Theta^1(\mu_V, x_0) \leq \frac{\mu_V(B_r(x_0))}{2r} + \frac{1}{2}\int_{B_r(x_0)} |H_V| d\mu_V$

3. Treatment of Point Energy

Through blow-up analysis and the Gauss-Bonnet formula, it is proved that: $\liminf_{ε→0} G_{ε,β_ε}(w_ε, B_r(x_0)) ≥ 4π$

Experimental Setup

This is a purely theoretical work, with main results verified through mathematical proofs rather than numerical experiments. The primary "verification" is embodied in:

Theoretical Verification Framework

1. Lower semicontinuity proofs: Via varifold convergence theory
2. Γ-convergence proofs: Separate proofs of liminf and limsup inequalities
3. Regularity requirements: C² rectifiability assumptions on limiting sets

Technical Assumptions

• The anisotropy function φ satisfies: $\frac{1}{C} ≤ φ(z) ≤ C$
• Parameter scaling relations: $\frac{ε|\log ε|}{β_ε} → 0$, $\frac{β_ε}{η_ε} → 0$

Experimental Results

Main Theoretical Results

Theorem 1.1 (Lower Semicontinuity for Perimeter Problem)

If {E_n} converges to E in L¹, and V_E has bounded first variation in L²μ_E, then: $\liminf_{n→∞} F(E_n) ≥ F(E)$

Theorem 1.3 (Γ-Convergence for Perimeter Problem)

For sets E with C² boundary:

1. Lower bound: $F(E) ≤ \liminf_{ε→0} F_ε(v_ε)$
2. Upper bound: There exists a recovery sequence such that $F(E) = \lim_{ε→0} F_ε(\bar{v}_ε)$

Theorem 1.4 (Γ-Convergence for Mumford-Shah Problem)

For u ∈ A(Ω), there exists γ₀ > 0 such that when γ ∈ (0,γ₀):

1. Lower bound: $G^{(\gamma)}(u) ≤ \liminf_{ε→0} G_ε^{(\gamma)}(u_ε,v_ε,w_ε)$
2. Upper bound: There exists a recovery sequence achieving equality

Key Technical Results

Lemma 4.2 (Density Improvement)

For Γ ∈ C(Ω) and a varifold V satisfying μ_V ≥ ℋ¹⌊Γ, at each point x₀ ∈ P_Γ: $Θ¹(μ_V, x_0) ≥ Θ¹(Γ, x_0) + \frac{1}{2}$

Theorem 5.1 (Generalized Gauss-Bonnet Formula)

For a 1-rectifiable varifold V in ℝ²: $\int_{ℝ²} |H_V| dμ_V ≥ 2π$

Related Work

Research on Anisotropic Problems

• Weak anisotropy: Phase-field approximations by Lussardi (2015) and others
• Strong anisotropy: Phase-field models by Torabi et al. (2009)
• Relaxation theory: Elastica functionals by Bellettini-Mugnai (2004)

Research on Willmore Energy

• Classical results: De Giorgi conjecture by Röger-Schätzle (2006)
• Phase-field approximations: Singular perturbation problems by Nagase-Tonegawa (2007)

Mumford-Shah Theory

• Original model: Mumford-Shah (1989)
• Curvature regularization: Coscia (2000), Braides-March (2006)

Conclusions and Discussion

Main Conclusions

1. Regularization effects of curvature penalization: The Willmore term effectively suppresses oscillations induced by strong anisotropy
2. Effectiveness of phase-field approximations: Established rigorous mathematical connections from diffuse interfaces to sharp interfaces
3. Treatment of singular structures: First systematic treatment of singularities such as crack tips in free discontinuity problems

Limitations

1. Regularity assumptions: Limiting sets must possess C² regularity
2. Parameter restrictions: γ must be sufficiently small in the Mumford-Shah problem
3. Dimensional restrictions: Main results focus on 2D and 3D cases

Future Directions

1. Numerical methods: Development of efficient numerical algorithms
2. Higher-dimensional extensions: Extension to higher-dimensional cases
3. Applied research: Specific applications in materials science and image processing

In-Depth Evaluation

Strengths

1. Theoretical rigor: Complete mathematical proofs with refined technical treatment
2. Methodological innovation: Clever combination of varifold theory and phase-field methods
3. Problem importance: Addresses the important theoretical problem of strong anisotropy
4. Technical contributions: Generalization of the Gauss-Bonnet formula to the varifold setting

Weaknesses

1. Lack of numerical verification: Purely theoretical work without computational examples
2. Limited applications: Insufficient discussion of practical application scenarios
3. Strong assumptions: Stringent requirements on regularity and parameters

Impact

1. Theoretical value: Provides a new mathematical framework for strongly anisotropic problems
2. Methodological contribution: Application of varifold methods in phase-field theory
3. Foundation for future research: Provides important theoretical basis for related fields

Applicable Scenarios

1. Materials science: Crystal growth and phase transition processes
2. Fracture mechanics: Mathematical modeling of crack propagation
3. Image processing: Segmentation problems with curvature constraints
4. Geometric analysis: Further development of varifold theory

References

The paper cites 47 important references covering multiple fields including varifold theory, phase-field methods, and Mumford-Shah theory, providing a solid theoretical foundation for this research.

Overall Assessment: This is a high-quality theoretical mathematics paper that makes important contributions to the study of strongly anisotropic interface problems. Although it lacks numerical experiments, its rigorous mathematical analysis and innovative technical methods give it significant academic value in the related field.