Crystal– tag –
-
Interpretable 2D Material Design: Dynamic Collision Fingerprint Outperforms Traditional Descriptors
ACS Omega USA Overview A recent comparative study published in ACS Omega investigates various structural representations for 2D materials, highlighting the promising potential of the Dynamic Collision Fingerprint (DCF). The research addr... -
LLMs Drive Autonomous Research to Boost Material Bandgap Predictions
arXiv USA Overview A recent preprint introduces an autonomous research loop powered by large language models (LLMs) that significantly enhances the optimization of crystal graph networks for electronic bandgap prediction. This self-consi... -
AI Workflow Unites Foundation Models and Bayesian Optimization for Cost-Efficient Materials Discovery
arXiv USA Overview A new preprint introduces a novel Bayesian materials design workflow that integrates surrogate-gated generation with foundation model embeddings. This approach, leveraging pre-trained ORB embeddings and Gaussian proces... -
AI-Powered Latent Genetic Algorithm Accelerates Crystal Structure Prediction
arXiv USA Overview A new 'Latent Genetic Algorithm (LGA),' detailed in an arXiv preprint, revolutionizes crystal structure prediction (CSP) by leveraging latent representations learned from pre-trained generalized interatomic potentials ... -
Nested Graph Neural Network Revolutionizes Property Prediction for Complex Solid Solutions
Edinburgh Research Explorer UK Overview Researchers at the University of Edinburgh have developed the Solid Solution Nested Graph Neural Network (SSNGNN), a novel AI framework that significantly boosts the predictive accuracy for chemica... -
Google DeepMind and Meta Spearhead Curated AI Resources for Scientific Breakthroughs
GitHub USA Overview A new 'awesome-ai-for-science' GitHub repository, a meticulously curated list of AI tools and datasets, has been released to accelerate scientific discovery across physics, chemistry, biology, and materials science. S... -
Physics-Informed Foundation Model “CLOUD,” Pre-trained on Over 6 Million Crystals, Revolutionizes Materials Property Prediction
springermedizin.de Germany Overview The "CLOUD" model, a scalable, physics-informed, Transformer-based foundation model for crystal representation learning, has been introduced. Pre-trained on over 6 million crystals, CLOUD encodes cryst... -
DeepMind’s GNoME and Microsoft’s MatterGen Drastically Accelerate AI-Driven Materials Discovery, Rapidly Screening Millions of Inorganic Crystals
AI CERTs News USA Overview Advanced AI pipelines like DeepMind's GNoME and Microsoft's MatterGen are leveraging graphene neural networks and machine learning potentials to screen millions of inorganic crystals at unprecedented speeds. Th... -
arXiv Paper Evaluates Universal MLIPs, Finds DFT Fine-Tuning Essential for Accuracy in Reactive Processes
arXiv USA Overview A new arXiv study evaluated five universal machine-learning interatomic potentials (MLIPs) for quantitative materials modeling involving reactive events. It revealed that while universal MLIPs are becoming general-purp... -
arXiv Presents Structural Novelty Analysis for AI-Generated Crystals, Reveals Bias Towards Known Prototypes
arXiv USA Overview A new workflow for assessing the structural novelty of inorganic crystals generated by AI models has been published on arXiv. The study evaluates whether AI-generated crystals are duplicates, reproducible via elemental...
