AI Failure Prediction for Semiconductor Devices Under Stress
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting semiconductor device failures are inefficient and inaccurate, as they fail to determine the exact location, time, and cause of failure within the device, and require cumbersome simulations or physical removal of device layers.
Innovation Solution
A method using AI models trained with in-situ measurements and feature extraction techniques to predict failure parameters of semiconductor devices under stress conditions, including high temperatures and radiation, without the need for TCAD simulations or physical removal of layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If TCAD simulations are used to simulate semiconductor device behavior, then failure prediction capability is improved, but computational complexity and time consumption increase significantly
Solution Approach 1:
The patent pre-processes semiconductor device data including material parameters, structural parameters, and operating parameters before failure occurs. This preliminary data preparation enables rapid failure prediction when stress conditions are applied, avoiding the need for time-consuming simulations at the moment of failure analysis.
Solution Approach 2:
The patent creates a digital twin or virtual model of the semiconductor device that replicates its physical characteristics and behavior. This virtual model can be analyzed repeatedly without affecting the actual device, enabling rapid failure prediction through data comparison rather than repeated physical simulations.
2Measurement precision
If in-situ measurements are performed to determine failure parameters, then measurement precision is improved, but device complexity and measurement system complexity increase
Solution Approach 1:
The patent employs a measurement system that can perform multiple functions: capturing optical emissions, measuring electrical parameters, and characterizing thermal conditions simultaneously. This multi-functional approach reduces the need for separate specialized measurement systems for each parameter type.
Solution Approach 2:
The patent uses optical emissions as an intermediary signal that provides information about internal device conditions without requiring direct physical access to failure points. The optical signals act as mediators that translate internal stress and damage states into measurable external parameters.
3Measurement precision
If ex-situ electroluminescence imaging techniques are used, then failure location identification is improved, but device structure must be physically altered by removing layers
Solution Approach 1:
The patent performs optical emission measurements in-situ during device operation or at failure points, capturing failure location information before any physical sectioning or layer removal is required. This preliminary measurement preserves the complete device structure for further analysis or operational use.
Solution Approach 2:
The patent replaces mechanical sectioning and physical layer removal with optical measurement techniques. By using optical emissions to visualize failure locations, the need for mechanical disruption of the device structure is eliminated entirely.
4Measurement precision
If multiple iterations of AI model training are performed with different data sets, then prediction accuracy is improved, but computational resources and training time increase
Solution Approach 1:
The patent implements continuous or incremental training of the AI model where each new data set builds upon previous training results. This continuous improvement approach allows the model to progressively enhance accuracy without completely retraining from scratch, reducing redundant computational energy consumption.
Data Source
AI summary
A method for predicting failure parameters of semiconductor devices can include receiving a set of data that includes (i) characteristics of a sample semiconductor device, and (ii) parameters characterizing a stress condition. The method further includes extracting a plurality of feature values from the set of data and inputting the plurality of feature values into a trained model executing on the one or more processors, wherein the trained model is configured according to an artificial intelligence (AI) algorithm based on a previous plurality of feature values, and wherein the trained model is operable to output a failure prediction based on the plurality of feature values. Further, the method includes generating, via the trained model, a predicted failure parameter of the sample semiconductor device due to the stress condition.