Options
Spectrum Diagnosis and Parallel Optical Simulation
Date Issued
2010
Date
2010
Author(s)
Huang, Kuan-Lu
Abstract
To ensure the quality of the nanoprint fabricated optical gratings, optical scatterometry (OS) is an efficient and effective mean to diagnose the actual fabricated geometry. To facilitate the diagnosis process, efficient pattern matching algorithms over a huge database are of great importance.
In this thesis, we propose an efficient algorithm using minimum error square approach used to matching in a huge simulated spectrum database in order to obtain the original geometric configuration inversely.We use Singular Value Decomposition to do compression on large database and the use of hierarchical moment to perform matching algorithm; our searching and diagnosis algorithm is extremely fast and accurate. It is over 3000x faster than a exhausted searching algorithm within 0.1% accuracy.
The second part is to introduce the use of parallel computing in the imaging of microlithography for acceleration. As the VLSI technology feature sizes quickly shrink smaller than the wavelength of exposure light sources, the diffraction effects have made the exposed patterns significantly deviated from the original intended mask pattern. Therefore, the quality of microlithography simulation is an important part of the VLSI manufacturing process. However, it takes considerable time to produce image. In the thesis, we use CUDA, which is a general purpose parallel computing architecture that leverages the parallel compute engine in NVIDIA graphics processing units (GPUs) to speed up the image generation in Microlithography simulation.
In this thesis, we propose an efficient algorithm using minimum error square approach used to matching in a huge simulated spectrum database in order to obtain the original geometric configuration inversely.We use Singular Value Decomposition to do compression on large database and the use of hierarchical moment to perform matching algorithm; our searching and diagnosis algorithm is extremely fast and accurate. It is over 3000x faster than a exhausted searching algorithm within 0.1% accuracy.
The second part is to introduce the use of parallel computing in the imaging of microlithography for acceleration. As the VLSI technology feature sizes quickly shrink smaller than the wavelength of exposure light sources, the diffraction effects have made the exposed patterns significantly deviated from the original intended mask pattern. Therefore, the quality of microlithography simulation is an important part of the VLSI manufacturing process. However, it takes considerable time to produce image. In the thesis, we use CUDA, which is a general purpose parallel computing architecture that leverages the parallel compute engine in NVIDIA graphics processing units (GPUs) to speed up the image generation in Microlithography simulation.
Subjects
Optical Scatterometry, Singular Value Decomposition
Moment Matching
Abbe’s method
CUDA
File(s)
No Thumbnail Available
Name
ntu-99-R96943021-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):4aa35487a9cd770d04b2852a3d98fe74