An imbalance aware lithography hotspot detection method based on HDAM and pre-trained GoogLeNet

Abstract

Due to the continuous shrinkage of transistor size and the ever-increasing complexity of integrated circuit design layout, great challenges arise in optical lithography-any defect on the mask will be transferred to the silicon wafer, which may lead to severe defects such as open circuit and short circuit. These defects on masks are called hotspots. Before transferring the circuit layout on the mask to the silicon wafer, the entire mask must be inspected to accurately find out the hotspots before optical lithography. Although traditional lithography hotspot detection approaches, such as pattern matching and machine learning, have gained satisfactory results the performance of the model degrades when encountering problems such as complex layout and data imbalance. In this paper, a hotspot detection method based on hybrid data enhancement, data compression and pre-trained GoogLeNet is proposed to solve the aforementioned problems. Our study shows that the average recall rate can be up to 98.3%. Meanwhile, the false alarm is reduced and the F1-score is 63.5%. Experimental results show that the proposed method achieves better performance on the ICCAD 2012 contest benchmark compared to hotspot detection methods based on deep or representative machine learning.

Lithography Hotspot Detection Method Based on Transfer Learning Using Pre-Trained Deep Convolutional Neural Network

“…The proposed method takes the strategy of calculating the local density to down-sample the original pattern, thereby achieving data compression. The proposed method is different from the other methods [19,26]. On the one hand, the resolution of the down-sampled patterns is higher, which is 240 × 240 pixels.…”

“…In recent years, transfer learning has developed rapidly and has been widely used [22][23][24]. A transfer-learning-based hotspot-detection method has begun to emerge [25,26]. Accuracy, recall, precision, and F1 score are commonly used as evaluation indicators for machine learning [27].…”

“…A good hotspot-detection model should perform well in F1 score, which means it has a high hotspot-detection rate and low false-alarm rate. Although the available machinelearning-based hotspot-detection methods perform well in recall, they still have insufficient precision and F1 score [16,18,19,25,26]. A high false-alarm rate will increase the postprocessing steps and increase the turn-around time of IC manufacturing.…”

“…ICCAD 2012 benchmark suite [30] is used for model training and model verification. Comparisons with Samsung's hotspot-detection method based on deep CNNs [16] and the hotspot-detection methods based on transfer learning in the past two years [25,26] were carried out. The results show that the proposed method performs well in accuracy, recall, precision, and F1 score.…”