Evolution equations of second order with nonconvex potential and linear damping: existence via convergence of a full discretization

Etienne Emmrich, David ŠiškaJournal of Differential Equations 2013 · DOI

TL;DR

This paper tackles the challenge of proving the existence of solutions for a class of second-order evolution equations with linear damping and a nonconvex potential. The authors achieve this by developing and analyzing a novel "full discretization" scheme, combining implicit time stepping with a Galerkin method, which also offers a path to numerical approximation.

What Problem Does This Paper Address?

Many physical phenomena, like nonlinear viscoelasticity and elastodynamics, are described by second-order evolution equations of the form $u'' + Au' + Bu = f$. A key difficulty in analyzing these equations arises when the term $B$, which often represents nonlinear material behavior, is derived from a nonconvex potential. Nonconvexity means that standard mathematical tools that rely on convexity (like monotonicity of the operator $B$ itself) don't directly apply, making it hard to prove the existence of solutions. Existing methods often require the potential to be convex or have limited growth, which restricts their applicability to real-world problems where potentials can be more complex and grow faster. This paper aims to overcome these limitations by providing a general framework for proving the existence of solutions even with nonconvex potentials, and by showing that a specific numerical discretization scheme converges to these solutions.

Methods

The core of the paper's contribution lies in its full discretization approach. This method combines two key techniques:

  • Implicit Time Stepping: Instead of using explicit time-stepping methods (which can be unstable for certain problems), the authors employ an implicit Euler scheme. This means that the solution at the current time step depends on the solution at the same time step, leading to a system of equations that needs to be solved at each step.
  • Galerkin Scheme: For the spatial discretization, a Galerkin method is used. This involves approximating the solution within finite-dimensional subspaces of the function spaces. The paper assumes the existence of a suitable Galerkin scheme for the relevant function spaces.

The authors then prove the existence of a unique solution to this fully discrete problem and derive crucial a priori estimates. These estimates are essential for showing that as the discretization parameters (time step and spatial mesh size) go to zero, the approximate solutions converge to a weak solution of the original continuous problem. A key technical element is the use of a generalized Andrews-Ball-type condition on the operator $(B + \lambda A)$ (where $\lambda \ge 0$), which is a weaker condition than requiring $B$ itself to be monotone, and is crucial for handling the nonconvexity.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and the development of a numerical method, rather than empirical testing on specific datasets.

Key Insights and Ideas

  • Handling Nonconvexity: The paper demonstrates that by combining implicit time stepping with a Galerkin method and a generalized Andrews-Ball condition, one can prove the existence of solutions for evolution equations with nonconvex potentials, a significant step beyond previous work.
  • Convergence of Full Discretization: The work establishes that a specific full discretization scheme converges to a weak solution. This is important because it not only proves existence but also provides a theoretically sound basis for numerical approximations of these complex problems.
  • Generalization of Existing Conditions: The paper shows how the Andrews-Ball condition, previously used in specific contexts, can be generalized and applied in a more abstract setting, broadening its applicability.
  • Bridging Theory and Computation: The paper effectively connects theoretical existence results with a practical numerical method, offering a pathway for researchers to solve problems that were previously intractable due to analytical difficulties.

Key Results

  • Existence of Weak Solutions: The paper proves the existence of weak solutions for a class of second-order evolution equations with linear damping and nonconvex potentials, under specific assumptions on the operators and function spaces.
  • Convergence of Approximate Solutions: A subsequence of the fully discrete approximate solutions is shown to converge strongly in $L^2(0, T; V_A)$ and weakly* in $L^\infty(0, T; V)$ to a weak solution of the original problem.
  • Uniqueness for Discrete Problem: The paper establishes the uniqueness of the solution to the fully discrete problem under certain conditions on the time step.
  • A Priori Estimates: Crucial a priori estimates for the discrete solutions are derived, which are fundamental for proving convergence.

Limitations and Future Work

  • Assumptions on Spaces and Operators: The existence proof relies on specific assumptions about the function spaces (reflexive, separable Banach spaces, Gelfand triples) and the operators (boundedness, strong positivity of $A$, demicontinuity and weak coercivity of $B$, and the generalized Andrews-Ball condition). Relaxing some of these assumptions could extend the applicability of the results.
  • Initial Conditions: For the nonconvex case, the proof requires the initial data $u_0$ to belong to $V_A$, which is a stronger condition than belonging to $V_B$.
  • Constant Operators: The paper assumes that the operators are constant in time. The authors suggest that extending the results to time-dependent operators, provided the assumptions hold uniformly in time, would be a natural next step.
  • Nonmonotone Perturbations: The paper focuses on linear damping and a potential operator $B$. Incorporating nonmonotone (strongly continuous) perturbations is mentioned as a direction for future work.
  • Open Questions: The paper notes that the stability of Galerkin schemes (Assumption P) is an open question in certain general settings, although it is satisfied in many practical applications.

Summary updated

BibTeX
@article{emmrich2013evolution,
  author = {Etienne Emmrich and David Šiška},
  title = {{Evolution equations of second order with nonconvex potential and linear damping: existence via convergence of a full discretization}},
  journal = {Journal of Differential Equations},
  year = {2013},
  doi = {10.1016/j.jde.2013.07.065},
}
Figure