An AI model adapted from image inpainting can reconstruct missing atoms in crystal structures with a reported 97 percent success rate. The approach was developed by a team led by Giovanni Pizzi of the PSI Center for Scientific Computing, Theory and Data, working with researchers from the universities of Parma and Modena in Italy. Described in npj Computational Materials, the method applies computer vision to incomplete crystal structures. Hydrogen is a frequent source of gaps, as standard X-ray diffraction methods do not detect it easily, leaving hydrogen atoms inaccurately placed or omitted entirely. Simulations require precise atomic coordinates to predict characteristics such as electrical and thermal conductivity. “If the information about the hydrogen atoms is missing, that’s a problem,” says Giovanni Pizzi. “Often, we can’t use several thousand potentially interesting materials for our simulations precisely for this reason.” The new method focuses on filling missing positions within structures already known experimentally, which could make previously inaccessible materials available for simulation or substantially improve existing models, supporting searches for properties useful in areas such as hydrogen storage and potentially new superconductors. Timo Reents, a doctoral candidate in Pizzi’s group, noted that simulations rely on database information about atomic locations. The AI system is trained to preserve a known crystal structure and only insert the missing hydrogen atoms.