Predictive Modelling of Electronic Materials: A Review of Deep Learning Techniques in Computer Engineering

This review evaluates the application of deep learning (DL) for the predictive modeling of electronic materials in computer engineering. We analyzed peer-reviewed literature across four major databases, focusing exclusively on advanced architectures like Graph Neural Networks (GNNs) and Generative models. Results indicate these models accurately predict critical properties, such as band gaps and thermal conductivity, for next-generation semiconductors, 2D materials, and memristors. These high accuracies are achieved because architectures like GNNs effectively capture complex 3D spatial interactions without requiring manual feature engineering. However, practical fabrication remains hindered by data scarcity, algorithmic opacity, and a profound "Sim-to-Real Gap". While DL accelerates predictive design, sustaining Moore's Law ultimately requires developing autonomous "Self-Driving Labs" and Large Material Models to bridge digital predictions with physical synthesis.

Hali tarjima qilinmagan