Algorithm-driven paradigms for the optimization and simulation of photonic devices
Abstract/Contents
- Abstract
- The world has witnessed transformative advancements in photonics, encompassing a broad range of applications from AR/VR glasses to LiDAR technology employed in self-driving cars. These captivating applications heavily rely on the simulation (forward problem) and optimization (inverse problem) of photonic devices. However, both the forward and inverse problems present their own set of challenges: 1) For forward problems, how to rapidly and accurately compute electromagnetic field distributions within structured media, especially when evaluating large simulation domains and large batches of devices. 2) For inverse problems, how to efficiently identify the overall optimal device in a non-convex design space while incorporating physical and fabrication constraints. This thesis discusses how algorithms based on classical optimization and deep learning are able to overcome the aforementioned challenges and are establishing a new conceptual framework for freeform photonic engineering. In the first half of the thesis, we introduce how a design space reparameterization scheme is able to simplify the inverse design problem and can naturally enforce hard fabrication constraints in adjoint-based freeform topology optimizations. The combination of device capability, feature size constraints, and ease of manufacturability enabled by this methodology facilitate our design of robust, high performance metasurfaces such as metagratings, metalenses, and metapolarizers. In the second half of the thesis, we present a high accuracy and ultra-fast surrogate electromagnetic simulator for the forward problem that is based on physics-augmented deep learning. A central feature of this neural network surrogate simulator is the utilization of data and physics to model and exploit high-dimensional relationships between geometric structure and electromagnetic response within the constraints of Maxwell's equations. Finally, by combining the high-speed forward simulators with the reparameterization inverse design scheme, we showcase hybrid ultrafast freeform optimizers that are able to effectively search for local and global optima of dielectric nanostructures with a speed-up of three to four orders of magnitude. We anticipate that these algorithmic approaches to photonic engineering will empower practitioners to move away from asking \emph{how} to realize a photonic system and focus on \emph{what} system functionality is useful for an application.
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource. |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2023; ©2023 |
| Publication date | 2023; 2023 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Chen, Mingkun |
|---|---|
| Degree supervisor | Fan, Jonathan |
| Thesis advisor | Fan, Jonathan |
| Thesis advisor | Fan, Shanhui |
| Thesis advisor | Kochenderfer, Mykel |
| Degree committee member | Fan, Shanhui |
| Degree committee member | Kochenderfer, Mykel |
| Associated with | Stanford University, School of Engineering |
| Associated with | Stanford University, Department of Electrical Engineering |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Mingkun Chen. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis Ph.D. Stanford University 2023. |
| Location | https://purl.stanford.edu/vk352dv6605 |
Access conditions
- Copyright
- © 2023 by Mingkun Chen
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...