AI model predicts hydrogen atom positions in crystal structures
AI model predicts hydrogen atom positions in crystal structures
Publish Date: 2026-06-19 04:48:00
Source Domain: www.chemistryworld.com
Here are 5 key points summarizing the article on an AI model capable of refining atomic positions in crystal structures:
-
Refinement of Atomic Structure: An artificial intelligence (AI) model has been developed to accurately fill in or correct the positions of missing or incorrectly placed atoms such as hydrogen in crystal structures of inorganic materials.
-
Challenges in Determining Small Atom Positions: Determining the exact positions of hydrogen and similar small atoms in crystalline materials is difficult due to weak scattering of x-rays, leading to inaccuracies in simulating material structures.
-
AI Methodology: The model improves atomic positions by adding noise or additional data points and refining until the lowest energy crystal structure is produced, similar to filling empty areas in an image.
-
Training and Efficacy: The model was trained on data by removing hydrogen atom positions from known structures, and has shown good predictive success, accurately predicting crystal structures in 85% of instances and identifying more stable structures in another 12%.
-
Future Research and Applications: The AI tool holds significant promise for identifying and correcting errors in existing crystallographic data, aiding in material simulations and potentially in the design of new materials such as superconductors.
