Abstract:A posteriori error estimates are an important tool to bound discretization errors in terms of computable quantities avoiding regularity conditions that are often difficult to establish. For non-linear and non-differentiable problems, problems involving jumping coefficients, and finite element methods using anisotropic triangulations, such estimates often involve large factors, leading to sub-optimal error estimates. By making use of convex duality arguments, exact and explicit error representations are derived that avoid such effects.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: to achieve accurate a posteriori error control for variational problems through convex duality and explicit flux reconstruction methods. Specifically, the paper aims to solve the following problems:
1. **A posteriori error estimation for nonlinear and non - differentiable problems**: For nonlinear and non - differentiable problems, problems with jump coefficients, and finite element methods using anisotropic triangulations, traditional a posteriori error estimations usually involve large factors, resulting in sub - optimal error estimations. The paper hopes to derive accurate and explicit error representations that avoid these problems by using convex duality theory.
2. **Improving the accuracy of error estimations**: Traditional error estimation methods often rely on regularity conditions that are difficult to establish, which limits their scope of application. The method proposed in the paper can provide more accurate a posteriori error estimations without these stringent regularity conditions.
3. **Developing highly adaptable error estimators**: The paper hopes to develop an a posteriori error estimator applicable to a wide range of variational problems (including non - quadratic, non - differentiable, constrained, and degenerate problems), which are difficult to handle with classical methods.
### Specific Objectives
- **Deriving accurate a posteriori error estimations**: Through convex duality relations, derive accurate and explicit error representation formulas, thereby avoiding the large - factor problems common in traditional methods.
- **Wide range of application**: Ensure that the proposed method is applicable not only to linear problems, but also to nonlinear, non - differentiable, etc. problems, and is applicable to multiple numerical methods (such as mixed methods and non - conforming methods).
- **High computational efficiency**: Construct approximately optimal discrete dual vector fields through local post - processing procedures, thereby avoiding the high computational cost of solving dual problems.
### Main Contributions
- **Introducing the generalized Prager - Synge identity**: Generalize the classical Prager - Synge identity to a wider range of convex minimization problems, providing a new perspective on error representation.
- **Discrete reconstruction formulas**: Propose discrete reconstruction formulas for Poisson problems and other model problems, making it possible to approximate the original problem and the dual problem simultaneously without directly solving the dual problem.
- **Global equivalence**: Prove the global equivalence between the primal - dual gap estimator and the residual - type estimator, providing a theoretical basis for the convergence of adaptive algorithms.
Through these improvements, the paper provides a new theoretical basis and practical tools for a posteriori error estimation of variational problems, which helps to improve the accuracy and reliability of numerical simulations.