Abstract:The purpose of this work is to study an optimal control problem for a semilinear elliptic partial differential equation with a linear combination of Dirac measures as a forcing term; the control variable corresponds to the amplitude of such singular sources. We analyze the existence of optimal solutions and derive first and, necessary and sufficient, second order optimality conditions. We develop a solution technique that discretizes the state and adjoint equations with continuous piecewise linear finite elements; the control variable is already discrete. We analyze the convergence properties of discretizations and obtain, in two dimensions, an a priori error estimate for the underlying approximation of an optimal control variable.
What problem does this paper attempt to address?
This paper attempts to solve an optimal control problem for a semilinear elliptic partial differential equation with Dirac measure as the forcing term. Specifically, the main research objectives include:
1. **Existence analysis**: Prove the existence of the optimal solution.
2. **Optimality conditions**: Derive the first - order necessary optimality conditions as well as the second - order necessary and sufficient optimality conditions.
3. **Discretization method**: Develop a method for discretizing the state and adjoint equations using continuous piecewise - linear finite elements, where the control variable is already discrete.
4. **Convergence analysis**: Analyze the convergence properties of the discretization and obtain a priori error estimates for the optimal control variable in the two - dimensional case.
### Problem background
The paper considers an optimal control problem for a semilinear elliptic partial differential equation (PDE) with Dirac measure as the forcing term. The specific form is as follows:
\[
-\Delta y + a(\cdot, y)=\sum_{z \in D} u_{z} \delta_{z} \quad \text{in } \Omega, \quad y = 0 \quad \text{on } \partial \Omega,
\]
where:
- \(\Omega\subset\mathbb{R}^{d}\) is an open, bounded convex polyhedral region.
- \(D\) is a finite ordered subset in \(\Omega\), denoted as \(D = \{z_{1}, z_{2}, \ldots, z_{\ell}\}\).
- \(\delta_{z}\) represents the Dirac delta function at point \(z\in D\).
- \(u=\{u_{z}\}_{z \in D}\in\mathbb{R}^{\ell}\) is the control variable, satisfying \(a_{z}\leq u_{z}\leq b_{z}\) for each \(z\in D\).
### Objective
The objective of the paper is to solve the following optimization problem:
\[
\min J(y, u)=\frac{1}{2}\|y - y_{d}\|_{L^{2}(\Omega)}^{2}+\frac{\alpha}{2}\|u\|_{\mathbb{R}^{\ell}}^{2},
\]
subject to the above - mentioned semilinear elliptic PDE.
### Main contributions
1. **Error estimates**: For the basic finite - element discretization, an \(L^{2}\) error estimate is obtained, which is optimal in terms of regularity. In addition, an approximately optimal \(L^{1}\) error estimate is also derived.
2. **Existence and optimality conditions**: It is proved that there exists at least one global solution to the optimal control problem, and the first - order necessary and second - order necessary and sufficient optimality conditions are analyzed.
3. **Convergence of discretization**: It is proved that a subsequence of the global solutions of the appropriately discretized problem converges to the global solution of the continuous optimal control problem. It is also proved that the continuous strictly local solution can be approximated by the local minimum of the above - mentioned discretized problem.
4. **Error estimates**: In the two - dimensional case, an error estimate for the approximately optimal optimal control variable is obtained.
### Conclusion
Through rigorous mathematical analysis and numerical methods, the paper solves the optimal control problem of semilinear elliptic partial differential equations with Dirac measure, providing theoretical guarantees and practical application guidance.