Rapid FinFET Modelling Using an Autoencoder

2026-06-23 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Semiconductor Device Engineering · Depth: Expert, quick

Summary

A machine learning framework utilizes an autoencoder (AE) for efficient FinFET modeling. The process begins by calibrating a BSIM-CMG model to generate a dataset of current-voltage (ID-VG) characteristics. This data then trains an autoencoder, which compresses full I-V curves into a low-dimensional latent space, inherently encoding device physics. A key innovation involves explicitly incorporating drain-to-source voltage (VDS) as an input feature, enhancing the model's ability to capture bias-dependent variations. The trained model accurately reconstructs full I-V curves and directly extracts critical device metrics, including threshold voltage (VTH), subthreshold slope (SS), and peak transconductance (gm). This data-driven approach achieves high accuracy with minimal training data, offering a powerful tool for rapid device characterization, modeling, and circuit-level simulation.

Key takeaway

For AI Hardware Engineers or Research Scientists focused on FinFET design and characterization, this autoencoder framework offers a significant efficiency boost. You can achieve rapid device characterization and accurate circuit-level simulations using minimal training data. Consider integrating this data-driven compact modeling approach to accelerate design iterations and extract critical device metrics like VTH, SS, and gm more quickly.

Key insights

An autoencoder framework efficiently models FinFETs by compressing I-V curves into a latent space, enabling rapid characterization with minimal data.

Principles

• Autoencoders can compress complex device characteristics.
• Explicitly incorporating bias parameters improves model accuracy.
• Data-driven compact models achieve high accuracy with less data.

Method

Calibrate a BSIM-CMG model for ID-VG data, train an autoencoder to compress I-V curves into a latent space, and include VDS as an input feature for bias dependency.

In practice

• Use AEs for rapid device characterization.
• Apply AE models in circuit-level simulation.
• Extract VTH, SS, and gm directly from AE output.

Topics

• FinFET Modeling
• Autoencoders
• Machine Learning
• Device Characterization
• Circuit Simulation
• BSIM-CMG

Best for: AI Scientist, AI Hardware Engineer, Machine Learning Engineer, Research Scientist

Related on AIssential

• Neural Network-Based Parametric Model Reduction for Predicting Turbulent Flow for Different Vehicle Geometries — A VAE-enhanced neural network model reduces computational cost for high-fidelity turbulent flow predictions around diverse vehicle geometries.
• Deep-Learning-Based Pixelated Microwave Filter Design and Characterization using Electro-Optical Electric-Field Measurements — Deep learning combined with genetic algorithms automates pixelated microwave filter synthesis, validated by novel electro-optical measurements.
• Andrej Karpathy Open-Sources ‘Autoresearch’: A 630-Line Python Tool Letting AI Agents Run Autonomous ML Experiments on Single GPUs — Autoresearch enables AI agents to autonomously optimize ML models on single GPUs through iterative, metric-driven refinement.
• Engineering Hybrid Physics-Informed Neural Networks for Next-Generation Electricity Systems: A State-of-the-Art Review — Physics-informed machine learning integrates domain physics into models, enhancing accuracy and efficiency for data-scarce electricity systems.
• On Approximating the Dynamic Response of Synchronous Generators via Operator Learning: A Step Towards Building Deep Operator-based Power Grid Simulators — Operator Learning, specifically DeepONets, can accurately model synchronous generator dynamics for power grid simulation.
• Advances in Scientific Machine Learning for Coupled Fluid Flow and Transport — SciML, combining reduced-order models and neural networks with HPC, efficiently approximates complex coupled fluid flow and transport, significantly cutting computational costs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.