Abstract:Manufacturing of semiconductor chips involves hundreds of process steps. For high-throughput semiconductor manufacturers, it is not feasible to get all the quality measurements of wafers at each step because of expensive costs. Based on the limited measurements of sampled wafers, virtual metrology (VM) allows predicting the relevant quality variables without increasing the number of physical measurements. Traditional data-driven VM methods are constructed using a predefined model along with feature extractions in data pre-processing. Often the identified VM model is not reliable if the constructed model and the extracted features are improper. In this paper, the convolutional neural network (CNN) based Gaussian process regression (GPR) VM mode (CNN-GPR) calibrated by maximization of the Bayesian posterior density distribution is systematically developed. CNN is used to extract information from massive wafer data through layers of the cascade-connected convolving filters. With the extracted features, the GPR model associates the stochastic nature of the operating steps with the quality variables to obtain the predicted means as well as the quantitated uncertainty levels. Unlike the traditional VM models where the extracted features and the prediction models are generated separately, the CNN-GPR model is sensitive to the variations of the quality variables and the features are determined simultaneously through the covariance matrices with information from both of the process and the quality variables. The proposed CNN-GPR model performs better as it represents the complex process of the whole set of measured variables. Through the studies on a real semiconductor process, the effectiveness of the proposed algorithm is demonstrated and compared against the regular regression algorithms.

Low-cost High-Accuracy Variation Characterization for Nanoscale IC Technologies Via Novel Learning-Based Techniques.

Accelerating Chip Design with Machine Learning: from Pre-Silicon to Post-Silicon

Just-In-Time Modeling with Variable Shrinkage Based on Gaussian Processes for Semiconductor Manufacturing

Variation-aware gate sizing and clustering for post-silicon optimized circuits.

Development of Convolutional Neural Network Based Gaussian Process Regression to Construct a Novel Probabilistic Virtual Metrology in Multi-Stage Semiconductor Processes

Machine Learning-Assisted Device Modeling With Process Variations for Advanced Technology

Machine Learning Based Variation Modeling and Optimization for 3D ICs.

Improving Semiconductor Device Modeling for Electronic Design Automation by Machine Learning Techniques

Extraction of Process Variation Parameters in FinFET Technology Based on Compact Modeling and Characterization

Variation-Resilient FeFET-Based In-Memory Computing Leveraging Probabilistic Deep Learning

Improvement of TCAD Augmented Machine Learning Using Autoencoder for Semiconductor Variation Identification and Inverse Design

Statistically Modeling I–V Characteristics of CNT-FET with LASSO

Data-Efficient Prediction of Minimum Operating Voltage via Inter- and Intra-Wafer Variation Alignment

Wafer-level Variation Modeling for Multi-site RF IC Testing via Hierarchical Gaussian Process

Efficient Variability- And Reliability-Aware Device-Circuit Co-Design: From Trap Behaviors To Circuit Performance

A Novel Process Variation Model for Test Cost Reduction in Wafers with Addressable Monitoring Structures

An IC Manufacturing Yield Model Considering Intra-Die Variations.

A Big-Data Approach to Handle Many Process Variations: Tensor Recovery and Applications

Device Modeling Based on Cost-Sensitive Densely Connected Deep Neural Networks

Predictive Compact Modeling of Random Variations in FinFET Technology for 16/14nm Node and Beyond

Experimental investigation and design optimization guidelines of characteristic variability in silicon nanowire CMOS technology