Shape optimization of optical microscale inclusions

Manaswinee Bezbaruah,Matthias Maier,Winnifried Wollner

2024-01-05

Abstract:This paper describes a class of shape optimization problems for optical metamaterials comprised of periodic microscale inclusions composed of a dielectric, low-dimensional material suspended in a non-magnetic bulk dielectric. The shape optimization approach is based on a homogenization theory for time-harmonic Maxwell's equations that describes effective material parameters for the propagation of electromagnetic waves through the metamaterial. The control parameter of the optimization is a deformation field representing the deviation of the microscale geometry from a reference configuration of the cell problem. This allows for describing the homogenized effective permittivity tensor as a function of the deformation field. We show that the underlying deformed cell problem is well-posed and regular. This, in turn, proves that the shape optimization problem is well-posed. In addition, a numerical scheme is formulated that utilizes an adjoint formulation with either gradient descent or BFGS as optimization algorithms. The developed algorithm is tested numerically on a number of prototypical shape optimization problems with a prescribed effective permittivity tensor as the target.

Numerical Analysis,Optimization and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to design optical micro - structures with specific macroscopic dielectric tensors through shape optimization methods, especially optical metamaterials of periodic micro - scale inclusions composed of low - dimensional materials. Specifically, the author aims to expand the two - scale homogenization theory by introducing a deformation field, and based on this, establish a shape optimization problem to minimize the difference between the calculated effective dielectric tensor and the target dielectric tensor.

### Main problem description

1. **Control of the effective dielectric tensor**:
The author hopes to make the final macroscopic effective dielectric tensor reach the predetermined target value by changing the geometric shape of the micro - structure (i.e., the deformation field). This goal can be expressed by the following formula:
\[
C(\hat{\chi}; \hat{q}) := \frac{1}{2} \| \varepsilon_{\text{eff}}[\hat{\chi}, \hat{q}] - \varepsilon_{\text{target}} \|_F^2 + \frac{\alpha}{2} \| \nabla \hat{q} \|_w^2 + \beta \int_{\hat{Y}} P(\hat{y}; \hat{q}) d\hat{y}
\]
where $\varepsilon_{\text{eff}}$ is the effective dielectric tensor, $\varepsilon_{\text{target}}$ is the target dielectric tensor, $\hat{q}$ is the deformation field, $\alpha$ and $\beta$ are regularization parameters, and $P(\hat{y}; \hat{q})$ is a penalty term to ensure the good quality of the deformed mesh.

2. **Extension of the homogenization theory**:
The author introduces a deformation field $\hat{q}$, and through the homogenization theory, transforms the influence of the micro - scale structure into an equivalent macroscopic dielectric tensor. The deformed cell problem is proved to be well - posed, and the solution is regular. This provides a theoretical basis for the well - posedness of the shape optimization problem.

3. **Numerical solution and optimization algorithm**:
In order to achieve the above - mentioned optimization goals, the author proposes the gradient descent method based on the adjoint method and the improved BFGS algorithm. These algorithms perform iterative optimization by calculating the gradient of the objective function with respect to the deformation field, thereby gradually approaching the optimal solution.

### Key points of the solution

- **Introduction of the deformation field**: By introducing the deformation field $\hat{q}$ to describe the geometric changes of the micro - structure, thereby affecting the effective dielectric tensor.
- **Application of the homogenization theory**: Use the homogenization theory to simplify the complex micro - scale structure into an equivalent macroscopic dielectric tensor.
- **Design of the optimization algorithm**: Adopt the adjoint method and the gradient descent / BFGS algorithm for numerical solution to ensure the efficiency and accuracy of the optimization process.
- **Setting of the penalty term**: Introduce an appropriate penalty term to ensure the quality of the deformed mesh and the rationality of the physical meaning.

In summary, the main purpose of this paper is to design micro - structures with specific optical properties through shape optimization methods, especially to manipulate and optimize the effective dielectric tensor of photonic crystals at the nanoscale, thereby achieving special optical phenomena such as near - zero refractive index.