Abstract:An innovative methodology that leverages artificial intelligence (AI) and graph representation for semiconductor device encoding in TCAD device simulation is proposed. A graph-based universal encoding scheme is presented that not only considers material-level and device-level embeddings, but also introduces a novel spatial relationship embedding inspired by interpolation operations typically used in finite element meshing. Universal physical laws from device simulations are leveraged for comprehensive data-driven modeling, which encompasses surrogate Poisson emulation and current-voltage (IV) prediction based on drift-diffusion model. Both are achieved using a novel graph attention network, referred to as RelGAT. Comprehensive technical details based on the device simulator Sentaurus TCAD are presented, empowering researchers to adopt the proposed AI-driven Electronic Design Automation (EDA) solution at the device level.

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve This paper aims to revolutionize TCAD (Technology Computer-Aided Design) simulation of semiconductor devices by introducing a universal encoding method based on graph representation. Specifically, the paper proposes the following innovations: 1. **Universal Encoding Method**: A universal encoding method based on graph structure is proposed. This method not only considers material-level and device-level embeddings but also introduces a new spatial relationship embedding method inspired by finite element mesh interpolation operations. 2. **Integrated API**: An integrated API has been developed to efficiently extract complex data structures from proprietary commercial software and convert them into forms compatible with machine learning frameworks such as PyTorch. This allows researchers to easily exchange data between device simulators and machine learning platforms. 3. **Relative Position Information**: A novel method is proposed that uses relative position information as edge features in the graph attention network model (RelGAT). This method dynamically assigns the importance of nodes based on the physical device structure. 4. **Case Studies**: The application of the proposed method is demonstrated through two case studies, including Poisson equation simulation based on node regression and current-voltage prediction under the drift-diffusion model based on graph regression. Through these innovations, the paper aims to address the challenges present in current semiconductor device design and simulation, particularly in handling complex design elements, material diversity, and process variations. Additionally, the method emphasizes openness and inclusivity, not relying on proprietary software, thereby promoting broader research collaboration and development.

Revolutionizing TCAD Simulations with Universal Device Encoding and Graph Attention Networks

A Physics-Informed Neural Networks Algorithm for Simulating Semiconductor Devices

TCAD Device Simulation With Graph Neural Network

Bridging TCAD and AI: Its Application to Semiconductor Design

TCAD Simulation of Novel Semiconductor Devices

AI for EDA/Physical Design: Driving the AI Revolution: The Crucial Role of 3D-IC

Framework for TCAD Augmented Machine Learning on Multi-I-v Characteristics Using Convolutional Neural Network and Multiprocessing

A Survey for Electronic Design Automation Based on Graph Neural Network

Transistor Compact Model Based on Multigradient Neural Network and Its Application in SPICE Circuit Simulations for Gate-All-Around Si Cold Source FETs

Simulation and Modeling of Novel Electronic Device Architectures with NESS (Nano-Electronic Simulation Software): A Modular Nano TCAD Simulation Framework

A Neural Network-based Framework for Accelerated Device-Circuit Electrothermal Co-Simulations in GAAFETs

Device Modeling Based on Cost-Sensitive Densely Connected Deep Neural Networks

Restructuring TCAD System: Teaching Traditional TCAD New Tricks

Graph Attention Network for Lane-Wise and Topology-Invariant Intersection Traffic Simulation

Late Breaking Results: Fast System Technology Co-Optimization Framework for Emerging Technology Based on Graph Neural Networks

Versatile Multi-stage Graph Neural Network for Circuit Representation

A Survey of Graph Neural Networks for Electronic Design Automation

New-Generation Design-Technology Co-Optimization (DTCO): Machine-Learning Assisted Modeling Framework

Comprehensive studies on deep learning applicable to TCAD

DeepGate3: Towards Scalable Circuit Representation Learning