An enhanced finite element method for a class of variational problems exhibiting the Lavrentiev gap phenomenon

Xiaobing Feng, Stefan Schnake

Authors: Xiaobing Feng and Stefan Schnake

In the realm of the calculus of variations, certain problems present a unique numerical challenge known as the Lavrentiev gap phenomenon. This occurs when the minimum value of an energy functional $\mathcal{J}(v)$ differs significantly depending on whether the minimization is performed over the space of smooth (Lipschitz continuous, $W^{1,\infty}$) functions or a broader space allowing for singularities ($W^{1,1}$). Standard Finite Element Methods (FEM) typically seek solutions in discrete subspaces of $W^{1,\infty}$, causing them to "get stuck" on the wrong side of the gap and fail to converge to the true singular minimizer.

This paper introduces an Enhanced Finite Element Method designed to bridge this gap using a simple yet effective gradient cut-off procedure.

The Lavrentiev Gap and the Failure of Standard FEM

Many physical models in nonlinear elasticity and image processing involve energy functionals where the true minimizer is singular. For such problems, standard FEM often results in an "energy explosion" or convergence to an incorrect, non-singular value. Because the discrete energy functional over-penalizes the high gradients necessary to approximate a singularity, the numerical optimizer cannot reach the global minimum in the $W^{1,1}$ space.

The Enhanced FEM Approach

The core innovation of this work is the introduction of a modified discrete energy functional $\mathcal{J}_h^\alpha$ that incorporates a cut-off procedure.

  1. Gradient Capping: The authors define a function $\chi_h^\alpha$ that truncates the magnitude of the gradient if it exceeds a threshold defined by the mesh size $h$ (specifically $O(h^{-\alpha})$ for a positive constant $\alpha$).
  2. Discrete Energy: By minimizing a functional where the gradient input is passed through this cut-off, the method prevents the artificial energy penalty associated with singularities in discrete spaces.
  3. Parameter Selection: A significant contribution is a sufficient condition for choosing the parameter $\alpha$ based on the structure of the energy density function $f$, removing the need for a priori knowledge of the solution's exact regularity.

Numerical Results

The method was evaluated using classic benchmarks, most notably Maniá’s 1-D Problem, where the minimizer is the singular function $u(x) = x^{1/3}$.

Enhanced FEM solutions converging to the singular minimizer $u(x)=x^{1/3}$ for Maniá's problem.

As shown in the figure above, while standard FEM would converge to a simple linear function $u(x)=x$, the enhanced FEM successfully recovers the singular profile. The dashed lines represent the discrete solutions for various mesh sizes, clearly approaching the solid blue curve of the true minimizer.

Close-up of the discrete solution near the singularity at $x=0$.

A detailed view near the origin demonstrates how the piecewise linear elements in the enhanced FEM are able to resolve the steep gradients of the $x^{1/3}$ singularity without triggering the typical energy instability that plagues standard methods.

Key Insights

  • Restored Convergence: The method successfully reached the singular minimum in both 1-D (Maniá) and 2-D (Foss) problems. In the 2-D case, the enhanced method reached the true energy minimum ($\approx 0$) while the standard method remained trapped at the Lipschitz minimum ($\approx 1$).
  • Efficiency: The implementation is remarkably simple, requiring only minor modifications to standard optimization loops.
  • Convergence Rates: For Maniá’s problem, a convergence rate of approximately $O(h^{1.5})$ was observed for the energy functional, limited by the nature of the singularity rather than the polynomial degree of the elements.

Conclusion

The proposed enhanced FEM provides a robust and "black-box" friendly solution for variational problems exhibiting the Lavrentiev gap. By dampening the numerical penalty of singularities through a mesh-dependent cut-off, the method allows standard optimizers to find the true $W^{1,1}$ minimizers. Future work aims to integrate this approach with adaptive mesh refinement, using the cut-off trigger as a natural indicator for where higher resolution is required.

Summary updated

Abstract

This paper develops an enhanced finite element method for approximating a class of variational problems which exhibit the \textit{Lavrentiev gap phenomenon} in the sense that the minimum values of the energy functional have a nontrivial gap when the functional is minimized on spaces $W^{1,1}$ and $W^{1,\infty}$. To remedy the standard finite element method, which fails to converge for such variational problems, a simple and effective cut-off procedure is utilized to design the (enhanced finite element) discrete energy functional. In essence the proposed discrete energy functional curbs the gap phenomenon by capping the derivatives of its input on a scale of $O(h^{-α})$ (where $h$ denotes the mesh size) for some positive constant $α$. A sufficient condition is proposed for determining the problem-dependent parameter $\a$. Extensive 1-D and 2-D numerical experiment results are provided to show the convergence behavior and the performance of the proposed enhanced finite element method.
BibTeX
@inproceedings{feng2018enhanced,
  author = {Xiaobing Feng and Stefan Schnake},
  title = {{An enhanced finite element method for a class of variational problems exhibiting the Lavrentiev gap phenomenon}},
  booktitle = {Communications},
  year = {2018},
  doi = {10.4208/cicp.oa-2017-0046},
}

Loading related content...

Figure