Related Experiment Video
Updated: Jun 4, 2026
07:20
Discovery and Synthesis Optimization of Isoreticular Al(III) Phosphonate-Based Metal-Organic Framework Compounds Using High-Throughput Methods
Published on: October 6, 2023
The Open Materials 2024 (OMat24) inorganic materials dataset and models
Luis Barros-Luque1, Muhammed Shuaibi2, Xiang Fu2
1FAIR, Meta, Menlo Park, CA, USA. lbluque@meta.com.
Nature Computational Science
|June 2, 2026
Summary
The Open Materials 2024 dataset accelerates inorganic material discovery using artificial intelligence. This large, diverse dataset improves machine learning model accuracy for predicting material properties.
Area of Science:
- Materials Science
- Computational Chemistry
- Artificial Intelligence
Background:
- Inorganic material discovery is crucial for applications like climate change mitigation and semiconductor manufacturing.
- Artificial intelligence (AI) offers significant potential to accelerate materials simulation, discovery, and design.
- A gap exists in openly available, reproducible datasets and models compared to proprietary solutions.
Purpose of the Study:
- To introduce the Open Materials 2024 (OMat24) dataset, a comprehensive resource for inorganic materials.
- To address the limitations of existing datasets and models in terms of openness and reproducibility.
- To enable advancements in AI-driven materials science.
Main Methods:
- Compilation of over 110 million density functional theory (DFT) calculations.
- Inclusion of diverse chemistries, materials, and configurations within the dataset.
- Training and evaluation of machine learning interatomic potential (MLIP) models on the OMat24 dataset.
Main Results:
- MLIP models trained on OMat24 achieved top performance on the Matbench-Discovery leaderboard.
- Achieved F1 scores >0.9 for stability and ~20 meV/atom accuracy for formation energy.
- Demonstrated high accuracy in thermal conductivity and phonon prediction benchmarks, correcting prior models' softening bias.
Conclusions:
- The OMat24 dataset represents a significant step-change in inorganic material property prediction accuracy.
- The dataset's diversity resolves systematic underprediction issues found in models trained on less diverse data.
- OMat24 empowers the research community to develop superior AI models for materials discovery.
Related Concept Videos
02:00Molecular Models
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
01:23Properties of Organometallic Compounds
Organometallic compounds are compounds that contain a carbon–metal bond. Carbon belongs to an organyl group like alkyl, aryl, allyl, or benzyl groups. The metal can be from Group I or Group II of the periodic table, a transition metal, or a semimetal.
02:18Network Covalent Solids
Network covalent solids contain a three-dimensional network of covalently bonded atoms as found in the crystal structures of nonmetals like diamond, graphite, silicon, and some covalent compounds, such as silicon dioxide (sand) and silicon carbide (carborundum, the abrasive on sandpaper). Many minerals have networks of covalent bonds.
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
02:37Metallic Solids
Metallic solids such as crystals of copper, aluminum, and iron are formed by metal atoms. The structure of metallic crystals is often described as a uniform distribution of atomic nuclei within a “sea” of delocalized electrons. The atoms within such a metallic solid are held together by a unique force known as metallic bonding that gives rise to many useful and varied bulk properties.
All metallic solids exhibit high thermal and electrical conductivity, metallic luster, and malleability. Many...
All metallic solids exhibit high thermal and electrical conductivity, metallic luster, and malleability. Many...
01:19Fluid Mosaic Model
Scientists identified the plasma membrane in the 1890s and its principal chemical components (lipids and proteins) by 1915. The model for plasma membrane structure, proposed in 1935 by Hugh Davson and James Danielli, was the first model to be widely accepted in the scientific community. The model was based on the plasma membrane's "railroad track" appearance in early electron micrographs. Davson and Danielli theorized that the plasma membrane's structure resembled a sandwich with the analogy of...
02:54Molecular and Ionic Solids
Crystalline solids are divided into four types: molecular, ionic, metallic, and covalent network based on the type of constituent units and their interparticle interactions.
Molecular Solids
Molecular crystalline solids, such as ice, sucrose (table sugar), and iodine, are solids that are composed of neutral molecules as their constituent units. These molecules are held together by weak intermolecular forces such as London dispersion forces, dipole-dipole interactions, or hydrogen bonds, which...
Molecular Solids
Molecular crystalline solids, such as ice, sucrose (table sugar), and iodine, are solids that are composed of neutral molecules as their constituent units. These molecules are held together by weak intermolecular forces such as London dispersion forces, dipole-dipole interactions, or hydrogen bonds, which...