Closest Point Methods for Eigenvalue Problems from Inhomogeneous Structures

本项目的目标是开发基于最近点法的正向特征值求解器，并针对非均匀结构的形状优化问题开发高效的算法，包括：（1）一般约束和边界条件，（2）四阶方程（例如BiLaplace算子）和积分微分方程，以及（3）一般曲面。最近点法是一种基于标准笛卡尔网格离散的曲面偏微分方程数值求解新方法。它将被扩展到解决一般的特征值问题。极值特征值的方法是基于瑞利公式和一个有效的重排算法，以实现最佳配置。将研究两种类型的重排方法：完全重排和部分重排。完全重排方法在每次迭代中寻找最优重排，而部分重排方法则进行适度的改变以获得满意的结果。基于形状导数和拓扑导数的方法是部分重排的例子。为了证明数值方法的能力和效率，它将被应用到非均匀材料和人口动力学的问题。 这项工作的广泛影响来自其广泛的应用。PI将应用数值方法解决以下问题：（1）识别具有频率控制的复合材料弦和膜，（2）找到具有最佳导电性的复合材料，（3）设计具有所需极值频率的复合材料板，以及（4）研究群体生物学中的特征值优化和来自不同模态的图像中的形状识别，包括磁共振图像和光学图像。相干图像此外，该技术将打开一扇新的大门，计算光谱信息的一般表面上没有网格的表面上，并提供了一个更好的理解一般表面上的形状优化。作为这项工作的一部分开发的软件将被纳入研究生学习的数字课程，并将免费向公众提供。在未来三年，PI将在即将到来的SIAM和国际会议上组织关于最近点方法和形状优化的小型研讨会，以吸引更多科学家的兴趣，并邀请更多代表性不足的群体发言，以拓宽该领域。

The goal of this project is to develop forward eigenvalue solver based on closest point method and efficient algorithms for shape optimization for inhomogeneous structure with (1) general constrains and boundary conditions, (2) fourth order equations (e.g. BiLaplace Operator) and integro-differential equations, and (3) general surfaces. The closest point method is a new numerical technique to solve PDEs on a surface based on standard Cartesian grid discretization via the closest point extension. It will be extended to solve general eigenvalue problems. The approach for extremum eigenvalue is based on Rayleigh formulation and an efficient rearrangement algorithm to achieve optimal configuration. Two types of rearrangement approaches will be investigated: full rearrangement and partial rearrangement. The full rearrangement approach looks for the optimal rearrangement at each iteration while the partial rearrangement approach takes moderate changes to have the satisfactory result. The approaches based on shape derivatives and topological derivatives are examples of partial rearrangement. To demonstrate the capability and efficiency of the numerical approach, it will be applied to problems from inhomogeneous materials and population dynamics. The broader impact of the work arises from its wide ranges of applications. The PI will apply the numerical approaches to problems including (1) identifying of composite strings and membranes with frequency control, (2) finding composite materials with optimal conductivity, (3) designing composite plates with desired extremum frequency, and (4) investigating eigenvalue optimization in population biology and shape identification in images from different modalities including magnetic resonance images and optical coherence images. Moreover, the techniques will open a new door to compute spectral information on general surfaces without meshes on surfaces and provide an improved understanding of shape optimization on general surfaces. Software developed as part of this work will be incorporated into numerical courses in graduate study and will be freely available to the public. In the coming three years, the PI will organize mini symposiums on closest point method and shape optimization in the coming SIAM and international conferences to interest more scientists and invite more speakers in the underrepresented groups to broaden the field.