Embedding-Enhanced Probabilistic Modeling of Ferroelectric Field Effect Transistors (FeFETs)

2025-08-01

AI Summary

Modeling ferroelectric field-effect transistors (FeFETs) is challenging due to fabrication and operational randomness, rendering existing deterministic and machine learning models inadequate for simultaneously capturing device variability and ensuring smooth integration into circuit-level simulators. Method: This work proposes an enhanced probabilistic compact model featuring a C^∞-continuous activation function to guarantee mathematical differentiability and a physics-informed, device-specific embedding layer that explicitly encodes heterogeneous ferroelectric properties. A mixture density network (MDN) is employed to sample synthetic device instances from the learned embedding distribution, enabling variability-aware simulation. Contribution/Results: The model achieves an R² of 0.92 in FeFET current–voltage characteristic modeling—significantly outperforming baseline approaches—while delivering high accuracy, strong generalization, seamless compatibility with circuit simulators, and scalability to diverse FeFET technologies.

Technology Category

• Machine Learning: Probabilistic Circuits and Graphical Models
• Reasoning under Uncertainty: Relational Probabilistic Models
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

FeFETs hold strong potential for advancing memory and logic technologies, but their inherent randomness arising from both operational cycling and fabrication variability poses significant challenges for accurate and reliable modeling. Capturing this variability is critical, as it enables designers to predict behavior, optimize performance, and ensure reliability and robustness against variations in manufacturing and operating conditions. Existing deterministic and machine learning-based compact models often fail to capture the full extent of this variability or lack the mathematical smoothness required for stable circuit-level integration. In this work, we present an enhanced probabilistic modeling framework for FeFETs that addresses these limitations. Building upon a Mixture Density Network (MDN) foundation, our approach integrates C-infinity continuous activation functions for smooth, stable learning and a device-specific embedding layer to capture intrinsic physical variability across devices. Sampling from the learned embedding distribution enables the generation of synthetic device instances for variability-aware simulation. With an R2 of 0.92, the model demonstrates high accuracy in capturing the variability of FeFET current behavior. Altogether, this framework provides a scalable, data-driven solution for modeling the full stochastic behavior of FeFETs and offers a strong foundation for future compact model development and circuit simulation integration.

Problem

• Model inherent randomness in FeFETs for reliable performance prediction
• Address limitations of existing deterministic and machine learning models
• Capture device-specific variability for robust circuit simulation

Innovation

• Mixture Density Network for FeFET modeling
• C-infinity activation functions for smooth learning
• Device-specific embedding captures physical variability