Close Menu
Versa AI hub
  • AI Ethics
  • AI Legislation
  • Business
  • Cybersecurity
  • Media and Entertainment
  • Content Creation
  • Art Generation
  • Research
  • Tools

Subscribe to Updates

Subscribe to our newsletter and stay updated with the latest news and exclusive offers.

What's Hot

Fast set-fit reasoning with optimal Intel on Xeon

June 22, 2025

Hugging Wiz Research and facial partners to improve AI security

June 21, 2025

Piclumen Art V1: Transform AI Art Generation with Advanced Visual Models for 2025 | AI News Details

June 21, 2025
Facebook X (Twitter) Instagram
Versa AI hubVersa AI hub
Sunday, June 22
Facebook X (Twitter) Instagram
Login
  • AI Ethics
  • AI Legislation
  • Business
  • Cybersecurity
  • Media and Entertainment
  • Content Creation
  • Art Generation
  • Research
  • Tools
Versa AI hub
Home»Tools»Millions of new materials discovered with deep learning
Tools

Millions of new materials discovered with deep learning

By January 5, 2025No Comments7 Mins Read
Share Facebook Twitter Pinterest LinkedIn Tumblr Reddit Telegram Email
Share
Facebook Twitter LinkedIn Pinterest Email

the study

Published November 29, 2023 Author

Amir Merchant and Ekin Dogs Cubek

AI tool GNoME discovers 2.2 million new crystals containing 380,000 stable materials that could power future technologies

Modern technology, from computer chips and batteries to solar panels, relies on inorganic crystals. To enable new technologies, crystals must be stable or they may break down. Obtaining new stable crystals can require months of painstaking experimentation.

Today, in a paper published in Nature, we share the discovery of 2.2 million new crystals. This is almost 800 years of knowledge. We introduce Graph Networks for Materials Exploration (GNoME), a new deep learning tool that dramatically increases the speed and efficiency of discovery by predicting the stability of new materials.

Using GNoME, we have doubled the number of technically viable materials known to humanity. Of the 2.2 million predictions, 380,000 are the most stable, making them promising candidates for experimental synthesis. These candidates include materials that have the potential to develop future innovative technologies, from superconductors to powering supercomputers to next-generation batteries that make electric cars more efficient. It is.

GNoME shows the potential to use AI to discover and develop new materials at scale. External researchers in labs around the world independently created 736 of these new structures experimentally in parallel work. A team of researchers at Lawrence Berkeley National Laboratory, in partnership with Google DeepMind, also published a second paper in Nature showing how AI predictions can be leveraged for autonomous materials synthesis.

We have made GNoME predictions available to the research community. We plan to deliver 380,000 materials to materials projects that are expected to be stable. The Materials Project is currently processing the compounds and adding them to the online database. We hope these resources will advance research in inorganic crystals and unlock the potential of machine learning tools to guide experiments.

Accelerate material discovery with AI

Approximately 20,000 of the experimentally identified crystals in the ICSD database are computationally stable. Computational approaches from the Materials Project, the Open Quantum Materials Database, and the WBM database have increased this number to 48,000 stable crystals. GNoME expands the number of stable materials known to humanity to 421,000.

Previously, scientists searched for new crystal structures by tweaking known crystals or experimenting with new combinations of elements. This was an expensive trial-and-error process that took months with limited results. Over the past decade, computational approaches led by the Materials Project and other groups have contributed to the discovery of 28,000 new materials. But so far, new AI-powered approaches have reached fundamental limits in their ability to accurately predict experimentally viable materials. GNoME’s 2.2 million material discoveries represent approximately 800 years of knowledge and represent an unprecedented scale and level of predictive accuracy.

For example, 52,000 new layered compounds similar to graphene have the potential to revolutionize electronics through the development of superconductors. Previously, about 1,000 such substances had been identified. They also discovered 528 potential lithium-ion conductors, 25 times more than previous research, that could be used to improve the performance of rechargeable batteries.

We publish predicted structures for 380,000 materials that are most likely to be produced in the lab and used in viable applications. For a material to be considered stable, it must not decompose into similar compositions at lower energies. For example, carbon in structures like graphene is stable compared to carbon in diamond. Mathematically, these materials lie on a convex hull. The project discovered 2.2 million new crystals that are stable by current scientific standards and sit below the convex hull of previous discoveries. Of these, 380,000 are considered the most stable and lie on the “final” convex hull, a new criterion we have set for material stability.

GNoME: Using graph networks for materials exploration

GNoME uses two pipelines to discover low-energy (stable) materials. While the structural pipeline creates candidates with structures similar to known crystals, the compositional pipeline follows a more randomized approach based on chemical formulas. The outputs of both pipelines are evaluated using established density functional theory calculations, and their results are added to the GNoME database to inform the next round of active learning.

GNoME is a state-of-the-art graph neural network (GNN) model. GNNs are particularly suited for discovering new crystalline materials because their input data takes the form of graphs that can be likened to connections between atoms.

GNoME was originally trained using publicly available data on crystal structures and their stability through the Materials Project. We used GNoME to generate new candidate crystals and predict their stability. Density functional theory is used in physics, chemistry, and materials science to understand the structure of atoms, which is important for assessing the stability and for evaluating the predictive power of a model during progressive training cycles. We repeatedly checked its performance using an established computational technique known as (DFT). of crystals.

We used a training process called “active learning” that dramatically improved GNoME’s performance. GNoME generates structural predictions for novel stable crystals and tests them using DFT. The resulting high-quality training data was fed back into the model training.

Our work improves the discovery rate for material stability predictions from about 50% to 80% based on MatBench Discovery, an external benchmark set by previous state-of-the-art models. We were also able to scale up the efficiency of our model by improving the detection rate from less than 10% to more than 80%. Such efficiency gains can have a significant impact on the amount of computing required per detection.

“Recipes” for new materials using AI

The GNoME project aims to reduce the cost of discovering new materials. External researchers independently created 736 new GNoME materials in the lab and demonstrated that the model’s predictions about stable crystals accurately reflect reality. We have opened our database of newly discovered crystals to the research community. By providing scientists with a complete catalog of promising “recipes” for new candidate materials, we hope they can test and potentially create the best materials.

After completing the latest discovery effort, we searched the scientific literature and found that 736 of the computational discoveries were independently realized by external teams around the world. Above are six examples ranging from a first-of-its-kind alkaline earth diamond-like optical material (Li4MgGe2S7) to a potential superconductor (Mo5GeB2).

The ability to rapidly develop new technologies based on these crystals depends on the ability to manufacture them. In a paper led by Berkeley Lab collaborators, researchers showed that robotics labs can rapidly produce new materials using automated synthesis techniques. Using materials from the Materials Project and stability insights from GNoME, this autonomous lab has created new recipes for crystalline structures, successfully synthesized over 41 new materials, and pioneered new approaches to AI-driven materials synthesis. It opened up possibilities.

A-Lab is a Berkeley Lab facility where artificial intelligence teaches robots to make new materials. Photo credit: Marilyn Sargent/Berkeley Lab

New materials for new technology

Building a more sustainable future requires new materials. GNoME discovers 380,000 stable crystals with the potential to develop greener technologies, from better batteries for electric cars to superconductors for more efficient computing I did.

Our work, and that of our collaborators at Berkeley Lab, Google Research, and teams around the world, shows the potential for using AI to guide materials discovery, experimentation, and synthesis. We hope that GNoME, in conjunction with other AI tools, will revolutionize materials discovery today and help shape the future of the field.

Read our paper in Nature

Acknowledgment

This work would not have been possible without our wonderful co-authors: Simon Batzner, Sam Schoenholz, Muratahan Aykol, and Gowoon Cheon. We would also like to thank Doug Eck, Jascha Sohl-dickstein, Jeff Dean, Joëlle Barral, Jon Shlens, Pushmeet Kohli, and Zoubin Ghahramani for sponsoring the project. Lizzie Dorfman for product management support. Andrew Pierson for program management support. We thank Ousmane Loum for computing resources. Luke Metz for his assistance with infrastructure development. Ernesto Ocampo for assisting with initial work on the AIRSS pipeline. Insightful discussions by Austin Sendek, Bilge Yildiz, Chi Chen, Chris Bartel, Gerbrand Ceder, Joy Sun, JP Holt, Kristin Persson, Lusann Yang, Matt Horton, and Michael Brenner. Google DeepMind team for ongoing support.

author avatar
See Full Bio
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
Previous ArticleThe artificial intelligence (AI) company’s value hit $2 trillion last year, and Wall Street thinks it could rise even more in 2025
Next Article UAE AI regulation ‘prioritizes innovation but also risks gaps’

Related Posts

Tools

Fast set-fit reasoning with optimal Intel on Xeon

June 22, 2025
Tools

Hugging Wiz Research and facial partners to improve AI security

June 21, 2025
Tools

Text2SQL using Face Dataset Viewer API and MotherDuckDuckDB-NSQL-7B

June 21, 2025
Add A Comment
Leave A Reply Cancel Reply

Top Posts

Piclumen Art V1: Next Generation AI Image Generation Model Launches for Digital Creators | Flash News Details

June 5, 20253 Views

New Star: Discover why 보니 is the future of AI art

February 26, 20252 Views

How to build an MCP server with Gradio

April 30, 20251 Views
Stay In Touch
  • YouTube
  • TikTok
  • Twitter
  • Instagram
  • Threads
Latest Reviews

Subscribe to Updates

Subscribe to our newsletter and stay updated with the latest news and exclusive offers.

Most Popular

Piclumen Art V1: Next Generation AI Image Generation Model Launches for Digital Creators | Flash News Details

June 5, 20253 Views

New Star: Discover why 보니 is the future of AI art

February 26, 20252 Views

How to build an MCP server with Gradio

April 30, 20251 Views
Don't Miss

Fast set-fit reasoning with optimal Intel on Xeon

June 22, 2025

Hugging Wiz Research and facial partners to improve AI security

June 21, 2025

Piclumen Art V1: Transform AI Art Generation with Advanced Visual Models for 2025 | AI News Details

June 21, 2025
Service Area
X (Twitter) Instagram YouTube TikTok Threads RSS
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms and Conditions
  • Disclaimer
© 2025 Versa AI Hub. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.

Sign In or Register

Welcome Back!

Login to your account below.

Lost password?