Collaborative Research: Elements: DiffAI: Diffraction Database and Intelligent Analysis

Modern technologies depend on understanding how atoms are arranged inside materials. This atomic arrangement determines how materials conduct electricity, withstand extreme environments, store energy, and perform in applications ranging from microelectronics to national defense systems. X-ray diffraction (XRD) and related scattering techniques are powerful tools for revealing atomic structure, yet analyzing XRD data is a complex task, especially since it is usually large in volume. Although thousands of diffraction patterns are generated every year in laboratories and national facilities, the results remain scattered across publications or stored locally without standardized formats, limiting its reuse and slowing scientific progress. There is no public database of experimental powder diffraction data. This project addresses this need. This project will develop DiffAI, an open, community driven platform that will host public experimental powder diffraction data, associated metadata, and provide artificial intelligence (AI) tools for automated analysis. These will enable more accurate structure determination of materials with complicated atomic arrangements, such as quantum materials that may underlie future quantum information technology. By making high-quality diffraction data findable, accessible, and reusable, DiffAI will accelerate and lower barriers for materials discovery. By democratizing access to experimental data and machine learning models, DiffAI will enable efficient analysis of diffraction data and foster collaboration within the global research community. Through open-access tools, student training, and community workshops, DiffAI aims to establish a global standard for sharing and analyzing diffraction data, ultimately driving progress in materials characterization and discovery.

This project advances the foundations of scientific cyberinfrastructure in three key ways: 1) A novel, extensible data architecture for experimental diffraction that will combine metadata schemas, JavaScript Object Notation (JSON) based data records, and a public data repository for experimental powder diffraction patterns that supports scalable, persistent storage of heterogeneous diffraction datasets. Persistent Digital Object Identifiers (DOIs), curated releases, and open APIs will facilitate reproducible workflows and long-term sustainability. 2) Automated agentic large language model (LLM) workflows for large-scale data extraction and digitization that identify relevant literature, detect and classify XRD figures, extract labels, and digitize plots into machine-readable formats. The team also plan to develop software tools for more automated data and metadata capture from laboratory instruments and synchrotron X-ray and neutron diffractometers at national laboratories, thereby creating a generalizable blueprint for automated experimental data recovery, an emerging need across multiple scientific domains. 3) Building on prior NSF work, DiffAI will implement domain-adapted AI models integrated into cyberinfrastructure that bridge synthetic training sets with real experimental data for automated XRD data and metadata validation tasks. These will enable more accurate structure determination for complex martials, such as quantum materials. Collectively, these advances will provide a scalable, community-driven cyberinfrastructure element that enables modern, AI-ready diffraction workflows.


This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Materials Research Section within the Directorate for Mathematical and Physical Sciences.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Award Number: 2513798
Principal Investigator: Simon Billinge
Funds Obligated: $270,000
State: NY