Close Menu
    Facebook X (Twitter) Instagram
    SciTechDaily
    • Biology
    • Chemistry
    • Earth
    • Health
    • Physics
    • Science
    • Space
    • Technology
    Facebook X (Twitter) Pinterest YouTube RSS
    SciTechDaily
    Home»Technology»This AI Can Find the Missing Atoms That X-Rays Cannot See
    Technology

    This AI Can Find the Missing Atoms That X-Rays Cannot See

    By Oliver Morsch, Paul Scherrer InstituteAugust 9, 2026No Comments5 Mins Read
    Facebook Twitter Pinterest Telegram LinkedIn WhatsApp Email Reddit
    Share
    Facebook Twitter LinkedIn Pinterest Telegram Email Reddit
    Artistic Illustration of AI Adding Hydrogen Atoms
    In the inpainting method, the artificial intelligence system is trained to preserve a known crystal structure (blue, black, and red spheres) and only insert the missing hydrogen atoms (blurred on the left, white spheres on the right). Credit: Paul Scherrer Institute PSI/Giovanni Pizzi

    An AI model adapted from image inpainting can reconstruct missing atoms in crystal structures with a reported 97 percent success rate.

    A crystal structure can appear nearly complete yet still be unusable for computer simulations because some of its atoms are missing. Hydrogen is a frequent source of these gaps, and researchers have now adapted an artificial intelligence technique used to repair images to predict where those hidden atoms belong.

    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, in which AI recognizes and interprets visual information, to incomplete crystal structures.

    Specialized AI systems already assist materials research. Microsoft’s MatterGen, for example, can build complex crystal structures from basic information about which elements should be included and their proportions. Researchers can then use the generated structures in computer simulations to investigate potential materials.

    The new method instead focuses on filling missing positions within structures that are already known experimentally. Completing those records 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.

    Missing hydrogen sidelines promising materials

    “For our simulations of material properties, we rely on information in databases telling us where each atom is located in a crystal structure,” says Timo Reents, a doctoral candidate in Giovanni Pizzi’s group.

    Hydrogen creates a persistent difficulty. Although it is often part of a crystal lattice, standard X-ray diffraction methods used to determine atomic arrangements do not detect it easily. As a result, crystal visualizations may place hydrogen atoms inaccurately or leave them out 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 researchers turned to AI to recover that missing information.

    Timo Reents and Giovanni Pizzi
    Timo Reents (left) and Giovanni Pizzi have taught an artificial intelligence system to find missing positions of hydrogen atoms in crystal structures. Credit: Paul Scherrer Institute PSI/Mahir Dzambegovic

    An image tool finds missing atoms

    Their solution borrows from diffusion models used in computer vision. When these systems restore an absent portion of an image, the process is known as inpainting. A model might, for instance, reconstruct a dog’s paw that was hidden in the original photograph.

    Older methods would cover the entire image with random information known as noise, then rebuild the full scene. Modern approaches concentrate stronger noise in the unknown area while disturbing the visible parts much less.

    This selective process gives the model a reliable starting point. Rather than recreating everything, it can use the surrounding information to infer what belongs in the gap.

    XtalPaint rebuilds only the gaps

    Although selective inpainting is well established in computer vision, it had not previously been available for reconstructing missing atomic positions. Pizzi’s group addressed that problem by developing XtalPaint, an open-source model adapted from Microsoft’s MatterGen.

    “This combines the advantages of modern computer vision and crystal reconstruction: Noise is added only to the unknown positions within the crystal – the known positions remain largely unchanged during the process,” Timo Reents explains.

    Because XtalPaint retains most of the existing structure, it can use the known atoms as a guide throughout reconstruction instead of beginning from a heavily disrupted crystal.

    “With step-by-step reconstruction, XtalPaint can orient itself to the existing crystal from the very beginning,” Reents says. “This increases the success rate and also saves computing power.”

    Tests reach a 97 percent success rate

    The researchers tested XtalPaint by taking crystal structures with known hydrogen positions, removing those coordinates, and asking the model to restore them.

    In 87 percent of the tests, XtalPaint recovered the previously known arrangement. In another ten percent, it generated configurations that were more energetically stable than the recorded structures.

    “Overall, this means a success rate of 97 percent for XtalPaint,” Reents says.

    The model could now be used to restore missing hydrogen coordinates in materials databases. “We can now use our method, for example, to complete structures in databases with the missing hydrogen positions,” says Pizzi.

    Pizzi and his colleagues have also used the method to identify database errors that can occur when information is transferred from original scientific papers. Its potential use extends beyond hydrogen. The same approach can reconstruct positions for lithium and sodium, two elements that are important in the search for new battery materials.

    Reference: “Score-based diffusion models for accurate crystal-structure inpainting and reconstruction of hydrogen positions” by Timo Reents, Arianna Cantarella, Marnik Bercx, Pietro Bonfà and Giovanni Pizzi, 11 June 2026, npj Computational Materials.
    DOI: 10.1038/s41524-026-02090-1

    This research was supported by the NCCR MARVEL, a National Centre of Competence in Research, funded by the Swiss National Science Foundation (grant number 205602). This work was supported by a MARVEL INSPIRE Potentials Master’s Fellowship.

    Never miss a breakthrough: Join the SciTechDaily newsletter.
    Follow us on Google and Google News.

    Artificial Intelligence Crystals Machine Learning Materials Science Paul Scherrer Institute
    Share. Facebook Twitter Pinterest LinkedIn Email Reddit

    Related Articles

    Google Scientists Discovered 380,000 New Materials Using Artificial Intelligence

    MIT’s AI System Reveals Internal Structure of Materials From Surface Observations

    Merging Artificial Intelligence and Physics Simulations To Design Innovative Materials

    AI Used To Predict Synthesis of Complex Novel Materials – “Materials No Chemist Could Predict”

    MIT Uses AI To Accelerate the Discovery of New Materials for 3D Printing

    AI Accurately Predicts Material Properties To Break Down a Previously Insurmountable Wall

    Novel Machine Learning Technique To Identify Structural Similarities and Trends in Materials

    Isaac Newton May Have Met His Match: New AI Tool Calculates Materials’ Stress and Strain Based on Photos

    Machine Learning Boosts the Search for New “Superhard” Materials

    Leave A Reply Cancel Reply

    • Facebook
    • Twitter
    • Pinterest
    • YouTube

    Don't Miss a Discovery

    Subscribe for the Latest in Science & Tech!

    Trending News

    Scientists Find Two New Ways To Break Down “Forever Chemicals” in Water

    Pesticide Exposure May Carry a Serious Neurological Risk Years Later

    Ozempic Could Fight Addiction Too – Scientists Finally Know How

    Two Mysterious ‘Ghost’ Ancestors Are Hiding in Our DNA

    This Widely Used Plastic Could Be Harming Your Liver

    Don’t Miss: August Brings Two Eclipses and the Perseid Meteor Shower

    Brain Waves Once Dismissed As Noise May Help Build Our Reality

    Perseverance Finds Organic Carbon Preserved Just Beneath the Martian Surface

    Follow SciTechDaily
    • Facebook
    • Twitter
    • YouTube
    • Pinterest
    • Newsletter
    • RSS
    SciTech News
    • Biology News
    • Chemistry News
    • Earth News
    • Health News
    • Physics News
    • Science News
    • Space News
    • Technology News
    Recent Posts
    • This AI Can Find the Missing Atoms That X-Rays Cannot See
    • Scientists Catch a Hidden Electronic State Forming Almost Instantly
    • Starlink Satellites Reveal a Hidden Atmosphere 300 Miles Above Earth
    • Cancer Cells Exploit a Hidden Immune “Off Switch”
    • Rare Human Brain Study Reveals How an Alzheimer’s Drug May Protect the Brain
    Copyright © 1998 - 2026 SciTechDaily. All Rights Reserved.
    • Science News
    • About
    • Contact
    • Editorial Board
    • Privacy Policy
    • Terms of Use

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