Response-Surface Mobility Model for Semiconductor Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional drift-diffusion simulations are computationally expensive and impractical for modeling new transistor geometries, substrate orientations, strain conditions, and channel materials, especially for advanced CMOS transistors, as they fail to capture non-equilibrium, quasi-ballistic transport phenomena.
Innovation Solution
The method involves running ensemble Monte Carlo simulations across a Design of Experiment space to extract mobility parameters, constructing a response-surface mobility model, and using this model to run drift-diffusion simulations for different configurations, reducing the need for extensive ensemble Monte Carlo simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ensemble Monte Carlo simulations are used to model non-equilibrium quasi-ballistic transport, then accuracy of transistor performance prediction is improved, but computational cost and simulation time increase significantly
Solution Approach 1:
The patent performs preliminary ensemble Monte Carlo simulations to extract mobility parameters and construct response-surface models before actual device design optimization. This preliminary action captures the complex non-equilibrium transport physics once, and the resulting models are then reused for rapid exploration of design spaces, avoiding repeated expensive EMC simulations while maintaining accuracy.
Solution Approach 2:
The patent creates simplified drift-diffusion models with mobility parameters copied and calibrated from ensemble Monte Carlo simulations. These copied models replicate the complex non-equilibrium transport behavior of EMC simulations but execute much faster, enabling efficient design exploration without sacrificing predictive accuracy for new transistor geometries and materials.
2Measurement precision
If comprehensive mobility models are developed for all possible transistor configurations, then modeling accuracy is improved, but development time and computational expense increase
Solution Approach 1:
The patent develops universal response-surface mobility models that function across multiple transistor configurations, channel materials, and geometries. Instead of creating separate models for each configuration, the universal models are constructed from ensemble Monte Carlo data spanning the design space and can predict performance for new configurations without re-calibration, significantly reducing development time while maintaining accuracy.
Solution Approach 2:
The patent uses parameter extraction and response-surface methodology to create compact mobility models with a small set of fitting parameters that capture the essential physics across different conditions. By changing parameters based on material type and geometry rather than developing entirely new models, the approach achieves comprehensive coverage with minimal development effort.
3Productivity
If drift-diffusion simulations are used for screening new technology options, then computational efficiency is improved, but ability to capture non-equilibrium transport phenomena deteriorates
Solution Approach 1:
The patent introduces response-surface mobility models as an intermediary between ensemble Monte Carlo simulations and drift-diffusion device simulations. These intermediary models embed the complex non-equilibrium transport physics from EMC simulations into a computationally efficient drift-diffusion framework, allowing rapid screening of new technologies while preserving accuracy for non-equilibrium effects.
Data Source
AI summary
A method for simulating semiconductor devices includes running ensemble Monte Carlo (EMC) simulations of a plurality of semiconductor devices having a first plurality of configurations in a Design of Experiment (DoE) space to produce EMC results. Mobility parameters are extracted across the DoE space from the EMC results. A response-surface mobility model is constructed using the extracted mobility parameters. The response-surface mobility model is used to run a drift-diffusion simulation of a semiconductor device with a different configuration from the first plurality of configurations.