Yayi Wei

Yayi Wei


Research Focus & Metrics

Academic Career Length

21Years (2006-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2024 in Opto-Electronic Advances
Verified
Journal Article
Closed access
PDF file
ZLZiqi Li
Ziqi Li
LDLisong Dong
Lisong Dong
XMXu Ma
Xu Ma
YWYayi Wei
Yayi Wei
Extreme ultraviolet (EUV) lithography with high numerical aperture (NA) is a future technology to manufacture the integrated circuit in sub-nanometer dimension. Meanwhile, source mask co-optimization (SMO) is an extensively used approach for advanced lithography process beyond 28 nm technology node. This work proposes a novel SMO method to improve the image fidelity of high-NA EUV lithography system. A fast high-NA EUV lithography imaging model is established first, which includes the effects of mask three-dimensional structure and anamorphic magnification. Then, this paper develops an efficient SMO method that combines the gradient-based mask optimization algorithm and the compressive-sensing-based source optimization algorithm. A mask rule check (MRC) process is further proposed to simplify the optimized mask pattern. Results illustrate that the proposed SMO method can significantly reduce the lithography patterning error, and maintain high computational efficiency.

23 April 2019 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Verified
Journal Article
Closed access
YCYing Chen
Ying Chen
YLYibo Lin
Yibo Lin
TGTianyang Gai
Tianyang Gai
YSYajuan Su
Yajuan Su
YWYayi Wei
Yayi Wei
DPDavid Z. Pan
David Z. Pan
Lithography simulation is computationally expensive for hotspot detection. Machine learning based hotspot detection is a promising technique to reduce the simulation overhead. However, most learning approaches rely on a large amount of training data to achieve good accuracy and generality. At the early stage of developing a new technology node, the amount of data with labeled hotspots or non-hotspots is very limited. In this paper, we propose a semi-supervised hotspot detection with self-paced multi-task learning paradigm, leveraging both data samples w./w.o. labels to improve model accuracy and generality. Experimental results demonstrate that our approach can achieve 4.6%-6.5% better accuracy at the same false alarm levels than the state-of-the-art work using 10%-50% of training data.

4 October 2021 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Verified
Journal Article
Closed access
YLYibo Lin
Yibo Lin
TQTong Qu
Tong Qu
ZLZongqing Lu
Zongqing Lu
YSYajuan Su
Yajuan Su
YWYayi Wei
Yayi Wei
The net orders in detailed routing are crucial to routing closure, especially in most modern routers following the sequential routing manner with the rip-up and reroute scheme. In advanced technology nodes, detailed routing has to deal with complicated design rules and large problem sizes, making its performance more sensitive to the order of nets to be routed. In the literature, the net orders are mostly determined by simple heuristic rules tuned for specific benchmarks. In this work, we propose an asynchronous reinforcement learning (RL) framework to automatically search for optimal ordering strategies and a transfer learning (TL) algorithm to improve performance. By asynchronous querying, the router, pretraining the RL agents, and finetuning with the TL algorithm, we can generate high-performance routing sequences to achieve a 26% reduction in the DRC violations and a 1.2% reduction in the total costs compared with the state-of-the-art detailed router.

15 December 2021 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Verified
Journal Article
Closed access
TGTianyang Gai
Tianyang Gai
TQTong Qu
Tong Qu
SWShuhan Wang
Shuhan Wang
XSXiaojing Su
Xiaojing Su
RXRenren Xu
Renren Xu
TYTianchun Ye
Tianchun Ye
Layout hotspot detection is one of the most important issues for the reliability enhancement of integrated circuits. Machine learning-based hotspot detectors have shown their advantages of efficiency and generalization compared with computationally intensive lithography process simulation. However, most machine learning-based hotspot detectors only accept layout clips of fixed size as input with the potential defect whose location is restricted at the center of each clip. Therefore, they cannot be used directly for multiple hotspots detection in a large area, which occurs frequently in real design cases. In this article, we build a new end-to-end hotspot detector based on a fully convolutional network, which has the flexibility of detecting a various number of hotspots in a layout of any size at one time. Moreover, we also develop a transfer learning scheme matching our proposed detector network, which can reduce the requirement of sample number when setting up a new model for a more advanced technology node. The experimental results demonstrate our proposed hotspot detector outstanding among state-of-the-art works and the transfer learning scheme is effective.

30 March 2023 in Microsystems & Nanoengineering
Verified
Journal Article
Open access
PDF file
DHDandan Han
Dandan Han
SDSen Deng
Sen Deng
TYTianchun Ye
Tianchun Ye
YWYayi Wei
Yayi Wei
Plasmonic lithography, which uses the evanescent electromagnetic (EM) fields to generate image beyond the diffraction limit, has been successfully demonstrated as an alternative lithographic technology for creating sub-10 nm patterns. However, the obtained photoresist pattern contour in general exhibits a very poor fidelity due to the near-field optical proximity effect (OPE), which is far below the minimum requirement for nanofabrication. Understanding the near-field OPE formation mechanism is important to minimize its impact on nanodevice fabrication and improve its lithographic performance. In this work, a point-spread function (PSF) generated by a plasmonic bowtie-shaped nanoaperture (BNA) is employed to quantify the photon-beam deposited energy in the near-field patterning process. The achievable resolution of plasmonic lithography has successfully been enhanced to approximately 4 nm with numerical simulations. A field enhancement factor (F) as a function of gap size is defined to quantitatively evaluate the strong near-field enhancement effect excited by a plasmonic BNA, which also reveals that the high enhancement of the evanescent field is due to the strong resonant coupling between the plasmonic waveguide and the surface plasmon waves (SPWs). However, based on an investigation of the physical origin of the near-field OPE, and the theoretical calculations and simulation results indicate that the evanescent-field-induced rapid loss of high-k information is one of the main optical contributors to the near-field OPE. Furthermore, an analytic formula is introduced to quantitatively analyze the effect of the rapidly decaying feature of the evanescent field on the final exposure pattern profile. Notably, a fast and effective optimization method based on the compensation principle of the exposure dose is proposed to reduce the pattern distortion by modulating the exposure map with dose leveling. The proposed pattern quality enhancement method can open new possibilities in the manufacture of nanostructures with ultrahigh pattern quality via plasmonic lithography, which would find potentially promising applications in high density optical storage, biosensors, and plasmonic nanofocusing.

24 July 2019 in Optics Express
Verified
Journal Article
Closed access
PDF file
JLJiaxin Lin
Jiaxin Lin
LDLisong Dong
Lisong Dong
TFTaian Fan
Taian Fan
Xu MaXu Ma
Xu Ma
YWYayi Wei
Yayi Wei
TYTianchun Ye
Tianchun Ye
Extreme ultraviolet (EUV) lithography emerges as a promising technique to fabricate next-generation integrated circuits. In order to improve the lithography imaging fidelity, source optimization (SO) technique is widely used to compensate for the imaging distortion. This paper develops an efficient learning-based SO approach for EUV lithography under the compressive sensing (CS) framework. The dimensionality of EUV-SO problem is significantly reduced by sparsely sampling the layout pattern. Then, the EUV-SO is formulated as an l1-norm inverse reconstruction problem based on the sparse prior of source patterns. The cost function is established based on a rigorous imaging model to take into account the characteristic effects in EUV lithography systems. In addition, a learning-based method is proposed to jointly optimize the source dictionary and projection matrix according to the sparsity and incoherence conditions in CS theory. The optimal source dictionary and projection matrix can be learned from a set of training samples collected from typical layout features in advance. Then, the optimized dictionary and projection matrix can be repetitively used in the following SO algorithms. Based on a set of simulations, the proposed SO method is proved to achieve good performance in both imaging fidelity and computational efficiency.