Electron-scattering-aware mixed-signal IC placement with reinforcement learning for EBL technologies

Hajijafari, Mohammad (2022) Electron-scattering-aware mixed-signal IC placement with reinforcement learning for EBL technologies. Masters thesis, Memorial University of Newfoundland.

Full text not available from this repository.


As the technology node shrinks to the nanometer regime, the demand for new lithography methods with high resolution and low cost is increasing. Electron beam lithography (EBL) is one of the promising next-generation lithography (NGL) technologies that can tackle both challenges compared to the traditional lithography methods. Electron scattering, which causes pattern distortion in layout, is one of the main challenges for industry to widely adopt EBL in technologies below 22nm. Two major effects generated by electron scattering in EBL process are proximity effect and fogging effect. This thesis proposes a reinforcement-learning (RL) placement method that trains a neural network as an agent to effectively control the fogging and proximity effects in the EBL technologies. To speed up our method compared to other popular placement approaches (e.g., absolute coordinate based analytical placement, simulated annealing (SA) based placement, advantage actor critic (A2C) based placement), we benefit from the following innovations: using topological floorplan representation for manipulating our layouts during placement, and deploying an RL trained agent that can intelligently take actions. To more effectively tackle mixed-signal ICs, our method focuses on the sensitive analog devices, which are better protected from potential variations due to the fogging and proximity effects of other digital/analog portions. The experimental results show that our proposed placer is able to efficiently decrease the variation of the fogging and proximity effects among sensitive III devices in the analog portion up to 89.26% and 95.22% respectively, while it is 15, 4.3, and 5 times faster than the analytical RL-based placement, SA-based and A2C-based placement methods, respectively. In summary, our proposed approach has three main contributions: 1) to the best of our knowledge, our work is the first study that considers the fogging and proximity effects in analog portion of mixed-signal ICs, 2) we apply deep Q-network (DQN) based placement to handle the fogging and proximity effects that improves the quality and speed of placement by intelligently choosing actions, and 3) we introduce a new RL placer in this study, which is based on a topological representation scheme resulting in much smaller configuration space and in turn faster placement operation.

Item Type: Thesis (Masters)
URI: http://research.library.mun.ca/id/eprint/15743
Item ID: 15743
Additional Information: Includes bibliographical references (pages 92-99) -- Restricted until May 30, 2024
Keywords: mixed-signal IC placement, electron beam lithography, fogging effect, proximity effect, reinforcement learning
Department(s): Engineering and Applied Science, Faculty of
Date: May 2022
Date Type: Submission
Digital Object Identifier (DOI): https://doi.org/10.48336/CA33-1Q83
Library of Congress Subject Heading: Lithography, Electron beam; Reinforcement learning

Actions (login required)

View Item View Item