Artificial Intelligence Accelerates Nuclear Event Investigations
Scientists have applied artificial intelligence (AI) and advanced computing to speed up the identification process following nuclear events such as explosions, accidents, or industrial emissions. Traditionally, uncovering key details about these events involves painstaking laboratory work to analyze radioactive debris and trace the origins of materials used. The chemical reactions that follow a nuclear event produce hundreds of isotopes and compounds, many of which decay rapidly, making the analysis complex and time-consuming.
Researchers at the Department of Energy’s Pacific Northwest National Laboratory (PNNL) have demonstrated how generative AI and machine learning, combined with cloud computing from Microsoft, can accelerate this analysis. Their work shows that AI can address complicated chemistry questions that arise when interpreting the mixture of radioactive debris from a nuclear explosion.
Speeding Up Laboratory Analysis
One of the ongoing challenges is to shorten the time required to identify crucial information about nuclear detonations. This research advances that goal by prioritizing and targeting specific chemical processes, ultimately reducing laboratory time. The study’s results were published in the journal Physical Chemistry Chemical Physics and presented at the Methods and Applications of Radioanalytical Chemistry conference.
Nic Uhnak, a radiochemist at PNNL who led the study, explains, “There’s a tremendous amount of radiochemistry that needs to be done to determine the fingerprints of a nuclear explosion.” Scientists must work quickly within a highly complex chemical environment, characterized by intense radiation and multiple simultaneous reactions.
Uhnak likens the post-detonation analysis to figuring out the details of a baked cake: identifying the source and quantity of ingredients and the baking conditions. The number of questions after a nuclear explosion is orders of magnitude more complex.
Collaborating to Enhance Nuclear Forensics
PNNL is part of a coalition of national laboratories and law enforcement agencies responsible for maintaining the U.S. nuclear forensics capability. The lab provides critical data that helps inform decisions and attributions. In this project, the team identified chemical forms likely present in nuclear debris and asked which reactions would occur, what lab experiments were necessary, and the optimal order to conduct these tests.
The debris from a nuclear explosion contains many elements, including uranium, strontium, iron, and cerium, in various chemical forms. Typically, analysis involves dissolving these materials in an aqueous solution such as nitric acid, followed by chemical separations to isolate components.
AI-Enabled Computational Chemistry
The team applied AI-driven high-performance computing to simulate complex chemical interactions and calculate properties like stability constants. These constants indicate how strongly ions or molecules bond, how molecular complexes form or break apart, and how energy moves through the system. This information guides scientists in conducting efficient chemical separations to determine what happened and trace the materials’ origins.
Generative AI can explore and calculate properties for a vast number of molecular combinations far beyond what is feasible in the laboratory. “Generative AI calculates in many dimensions at once, in a way that is difficult for a person,” said Hadi Dinpajooh, a computational chemist involved in the study. This capability significantly reduces the time needed to evaluate all possibilities.
PNNL scientists believe AI-driven chemical separation modeling could also benefit other nuclear science areas, such as producing medical isotopes like molybdenum-99, which is vital for cancer diagnosis and requires similar chemical separations.
Partnership with Industry and Use of Advanced Computing
The mathematical challenges in this work are substantial. To manage these, PNNL partnered with Microsoft to use Azure Quantum Elements, a cloud computing platform. The system leveraged NVIDIA’s powerful GPUs, including 230 NVIDIA H100 units, and a total of 55 terabytes of RAM to perform the complex calculations. This analysis represents only one step in the overall process following a nuclear detonation.
Paul Rigor, a computer scientist managing the PNNL-Microsoft collaboration, helped align the research needs with Microsoft’s computing capabilities. According to Uhnak, “This first paper along these lines is a baby step, but it’s an important step. Anything we can do to speed the process of analysis is a win.”
Ongoing Efforts in Nuclear Forensics
Research on nuclear devices and debris is part of a broader and ongoing program at PNNL focused on nuclear forensics. The laboratory plays a critical role in the nation’s ability to analyze nuclear and radioactive materials and events, including the intricate science involved in nuclear detonations.
- PNNL’s Generative AI for Science, Energy, and Security Science and Technology investment supports this work.
- The lab is conducting 45 projects spanning autonomous experimentation, predictive phenomics, cybersecurity, nuclear security, and grid modernization.
- Funding also comes from PNNL’s Center for AI and Center for Continuum Computing.
This collaboration between AI, chemistry, and high-performance computing marks a promising approach to improving nuclear event investigations and related scientific challenges.
Your membership also unlocks:
