Oxide Defect Characterization from Multi-Defect RTN Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods are inadequate for characterizing and analyzing individual oxide defects in deeply-scaled CMOS devices, particularly due to the complexity of Random Telegraph Noise (RTN) signals from multiple defects.
Innovation Solution
A physics-informed machine learning methodology using Bayesian machine learning algorithms, Markov Chain models, and clustering techniques to detect and characterize individual oxide defects by analyzing charge/discharge behavior and reconstructing RTN signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional statistical methods are used to analyze oxide defects, then analysis simplicity is maintained, but measurement precision and defect characterization capability deteriorate due to inability to resolve individual defect behavior in deeply-scaled devices
Solution Approach 1:
The patent segments the RTN signal into individual defect contributions by identifying discrete current levels corresponding to different defect charge states. The Bayesian machine learning algorithm decomposes the complex multi-defect signal into separate defect events, enabling precise characterization of individual defects even when multiple defects are present in the gate oxide.
Solution Approach 2:
The patent introduces a Bayesian machine learning algorithm as an intermediary between the raw RTN measurement data and the defect characterization results. This intermediary processes the complex signal by evaluating posterior probabilities of different defect configurations, transitioning probabilities, and discrete level assignments to extract meaningful defect parameters.
2Productivity
If device dimensions are downscaled to 10 nm range, then device integration density is improved, but reliability deteriorates due to increased impact of stochastic defect behavior
Solution Approach 1:
The patent replaces conventional statistical reliability modeling with a physics-informed machine learning approach that directly models the charge/discharge dynamics of individual defects. By using Bayesian inference and Markov Chain models, the system captures the stochastic nature of defect behavior while providing deterministic parameter estimates, enabling reliable prediction of device degradation in deeply-scaled technologies.
3Adaptability or versatility
If multiple defects are present in gate oxide, then device functionality is maintained, but signal complexity increases making defect analysis difficult
Solution Approach 1:
The patent creates a probabilistic model (copy) of the physical defect system that replicates the charge/discharge behavior observed in measurements. The Bayesian framework generates posterior distributions for defect parameters by simulating multiple possible defect configurations and comparing them against the measured RTN signal, effectively copying the complex multi-defect behavior in a tractable mathematical form.
Data Source
AI summary
A method, system, and non-transitory computer-readable medium for characterizing oxide defects in semiconductor devices are described. Based on time-resolved SILC data, a total number of active oxide defects and each defect's characteristics can be determined for a deeply-scaled FET device. The discrete changes in leakage current correspond to switching of a single defect. A Bayesian-inspired algorithm is utilized to extract distinct current levels and experimental data is quantized into these extracted current levels by filtering noise. The evolution of current levels corresponds to a Markov chain. The defect currents are extracted by clustering transition probabilities and absolute differences in current levels using an affinity propagation algorithm. A maximum likelihood estimator is developed to extract a base leakage current. Lastly, defect currents are used to reconstruct the experimental data and its deconvolution into activity of individual defects provides the respective time