EPIC: Efficient prediction of IC manufacturing hotspots with a unified meta-classification formulation

Abstract

In this paper we present EPIC, an efficient and effective predictor for IC manufacturing hotspots in deep sub-wavelength lithography. EPIC proposes a unified framework to combine different hotspot detection methods together, such as machine learning and pattern matching, using mathematical programming/optimization. EPIC algorithm has been tested on a number of industry benchmarks under advanced manufacturing conditions. It demonstrates so far the best capability in selectively combining the desirable features of various hotspot detection methods (3.5-8.2% accuracy improvement) as well as significant suppression of the detection noise (e.g., 80% false-alarm reduction). These characteristics make EPIC very suitable for conducting high performance physical verification and guiding efficient manufacturability friendly physical design.

Sequential Sampling Methods for Finding Classification Boundaries in Engineering Applications

“…Classification tasks can be found in numerous engineering applications, including safety, 1−3 process control, 1,2,4 bioprocesses, 5−8 product formulation and design, 7,9,10 material science, 9,10 additive manufacturing, 11 and others. These applications include real-time classification of processes (reactive), 2,11 and classification boundary identification in high parameter spaces to attain desired conditions (proactive).…”

“…The second approach, referred to as iterative sampling or active learning (AL), is characterized by making informed predictions based on the available samples to guide the selection of subsequent sampling regions. 9,18,19 AL is typically more efficient and requires fewer samples to obtain the same amount of information about a system. It is best used in the exploration of large parameter spaces, where evaluating new conditions may be costly.…”

“…It is best used in the exploration of large parameter spaces, where evaluating new conditions may be costly. 9,12,19 Since AL requires new samples during each iteration, the sampling techniques in the first approach can be employed here but on a subset of the entire input space. Though grid and random searches are easier to implement and understand, AL or sequential sampling has increased in popularity due to software packages becoming more accessible to scientists and engineers.…”