Disordered Optics for Multidimensional Information Processing
Photonic platforms with multiplexing capabilities are of profound importance for high-dimensional information processing. With the rapid expansion of data volume, growing pressure on the sensor and post-processing hardware calls the development of physical pre-processing interfaces. Complex photonic...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/140093 |
| _version_ | 1826194633790586880 |
|---|---|
| author | Li, Xinhao |
| author2 | Fang, Nicholas X. |
| author_facet | Fang, Nicholas X. Li, Xinhao |
| author_sort | Li, Xinhao |
| collection | MIT |
| description | Photonic platforms with multiplexing capabilities are of profound importance for high-dimensional information processing. With the rapid expansion of data volume, growing pressure on the sensor and post-processing hardware calls the development of physical pre-processing interfaces. Complex photonic systems aim to address the challenges as the interfaces between raw signals and sensors, which improve the efficiency of information reconstruction. In this thesis, we explore disordered photonic devices for multidimensional information processing and compatible fabrication techniques.
The first part of the thesis is about diffractive optical elements (DOE) for spectral imaging. We designed a spatially modulated DOE filter that can efficiently sample in the Fourier transformed domain and facilitate spectral image reconstruction. The DOE layer distinguishes the main Fourier spectral components, and the second spatial modulation layer mediates spectral aliasing. Unlike conventional snapshot spectral imagers, our design does not require sub-super-pixel-level sensing, which enables the efficient usage of sensor resolution. We further demonstrated a grayscale stencil lithography technique for efficient and customizable manufacturing of DOE or multilayer optics with spatial thickness variation.
The second part of the thesis is about scattering reservoir computers (RC). Complex optical medium shows great potential for large-scale optical RC thanks to the intrinsic parallelism and scalability. We identify the trade-off between the fading memory and non-normality of scattering RC, which determines their memory capacity and resistance to noise. Further, we proposed a transient amplification method to fully harness the high noise resistance and high dimensionality of non-normal scattering RC. We further developed a dynamic hydrogel scatter, and projection lithography method for 2D patterning of gain/loss materials, which are promising for applications in dynamic light modulation and re-configurable optical computing. |
| first_indexed | 2024-09-23T09:59:23Z |
| format | Thesis |
| id | mit-1721.1/140093 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T09:59:23Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1400932022-02-08T03:53:36Z Disordered Optics for Multidimensional Information Processing Li, Xinhao Fang, Nicholas X. Massachusetts Institute of Technology. Department of Mechanical Engineering Photonic platforms with multiplexing capabilities are of profound importance for high-dimensional information processing. With the rapid expansion of data volume, growing pressure on the sensor and post-processing hardware calls the development of physical pre-processing interfaces. Complex photonic systems aim to address the challenges as the interfaces between raw signals and sensors, which improve the efficiency of information reconstruction. In this thesis, we explore disordered photonic devices for multidimensional information processing and compatible fabrication techniques. The first part of the thesis is about diffractive optical elements (DOE) for spectral imaging. We designed a spatially modulated DOE filter that can efficiently sample in the Fourier transformed domain and facilitate spectral image reconstruction. The DOE layer distinguishes the main Fourier spectral components, and the second spatial modulation layer mediates spectral aliasing. Unlike conventional snapshot spectral imagers, our design does not require sub-super-pixel-level sensing, which enables the efficient usage of sensor resolution. We further demonstrated a grayscale stencil lithography technique for efficient and customizable manufacturing of DOE or multilayer optics with spatial thickness variation. The second part of the thesis is about scattering reservoir computers (RC). Complex optical medium shows great potential for large-scale optical RC thanks to the intrinsic parallelism and scalability. We identify the trade-off between the fading memory and non-normality of scattering RC, which determines their memory capacity and resistance to noise. Further, we proposed a transient amplification method to fully harness the high noise resistance and high dimensionality of non-normal scattering RC. We further developed a dynamic hydrogel scatter, and projection lithography method for 2D patterning of gain/loss materials, which are promising for applications in dynamic light modulation and re-configurable optical computing. Ph.D. 2022-02-07T15:23:44Z 2022-02-07T15:23:44Z 2021-09 2021-09-30T17:29:08.185Z Thesis https://hdl.handle.net/1721.1/140093 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Li, Xinhao Disordered Optics for Multidimensional Information Processing |
| title | Disordered Optics for Multidimensional Information Processing |
| title_full | Disordered Optics for Multidimensional Information Processing |
| title_fullStr | Disordered Optics for Multidimensional Information Processing |
| title_full_unstemmed | Disordered Optics for Multidimensional Information Processing |
| title_short | Disordered Optics for Multidimensional Information Processing |
| title_sort | disordered optics for multidimensional information processing |
| url | https://hdl.handle.net/1721.1/140093 |
| work_keys_str_mv | AT lixinhao disorderedopticsformultidimensionalinformationprocessing |