High-energy collisions investigate the internal structure of subatomic particles depicted as webs like neural networks of quantum connections. This graphic highlights how physicists use AI/ML to map Quark-Gluon structures within particles and search for new physics beyond standard models. Credit: Brandon Kriesten/Argonne National Laboratory.
In the world of particle physics, where scientists unravel the mysteries of the universe, artificial intelligence (AI) and machine learning (ML) are creating waves in ways that enhance understanding of the most fundamental particles. The central parton distribution function (PDF) is at the center of this search. These complex mathematical models are important for predicting the results of high energy physics experiments testing standard models of particle physics.
PDF is a mathematical model that helps scientists understand the internal function of protons, particles found in the nuclei of atoms. Protons are made up of even smaller particles known as quarks and gluons, all known as partons. PDFS explains how these partons are distributed within protons, providing a map of where essentially these small particles are likely to be found, and how much momentum they carry.
This information helps scientists predict the outcome of high-energy physics experiments. For example, those done in a large hadron crider, protons are crushed together to explore basic forces and particles.
Modeling these features is difficult due to their complexity and limited availability of experimental data. However, AI and ML provide new ways to analyze and understand these complex features by processing the large amount of data collected by collider facilities.
At the U.S. Department of Energy (DOE) Argonne National Laboratory, theoretical physicists Tim Hobbs and Brandon Kriesten will pioneer the use of AI/ML to tackle the challenges of modeling PDFs, improving both the accuracy of PDFs derived from data and interpretation of ML models. This means that scientists can more easily identify patterns, relationships, and underlying principles within PDFs, and the techniques used to extract them can lead to more informed and reliable conclusions.
“Particle physics deals with basic or basic particles,” explained Hobbs. “The current focus is to find the cracks in the standard model that was completed in the 1970s. Despite its strength, we find them incomplete due to hints like cosmology's dark matter.”
PDFDECODER: Bridging theory and data
Recent research, publications by Hobbs and Kriesten Physical Review dwe have introduced the “PDFDECODER” framework. Uses the encoder decoder model, a type of neural network architecture. These models simplify complex data into a more manageable format and reconstruct the original data from this simplified version.
Reconstructing PDFs is important as scientists can predict particle behavior in high energy physics experiments. The key properties of PDFs are captured through “Mellin Moments,” a mathematical formula that summarises the distribution of these particles.
“This model uses generated AI to fill in gaps and reproduces initial conditions,” Kriesten said. In this context, “initial condition” refers to the starting parameters or configuration required to accurately model the distribution of quarks and glues within protons.
“We looked at how we could decipher PDFs from Mellin's moment and looked into a variety of possible solutions,” he added.
This approach increases the accuracy of particle physics predictions by ensuring that the reconstructed PDFs are intimately consistent with the actual data. In particular, lattice gauge calculations allow for more accurate PDF models. This is a computational technique that delves into the complexity of quantum chromodynamics, and is a theory that explains powerful force-coupled quarks and gluons.
By incorporating Mellin Moment data, the PDFDecoder framework provides a new way to integrate lattice information into PDF research, enhancing the relationship between theoretical models and experimental findings.
High-energy physics experiments are attempting to reveal unknown particles or indications of interactions beyond the standard model (BSM). However, the persistent uncertainty of theoretical predictions, experimental data, and parton distribution functions can obscure these findings. AI/ML methods provide solutions to solve these complexities in particle physics. Credit: Tim Hobbs/Argonne National Laboratory.
Understand AI decision-making in theoretical models
Another study has been published in Journal of High Energy PhysicsHobbs and Kriesten have announced another framework called “Xai4PDF”. This framework uses explanatory AI technology, a method designed to make the decision-making process of AI models more transparent and easier to understand.
The XAI4PDF framework uses the ResNet architecture. This is a type of neural network that relies on shortcuts to improve training efficiency. These shortcuts allow the network to bypass specific layers, allowing you to easily train deep networks without losing important information.
This framework classifies PDFs based on underlying theoretical assumptions. It not only determines which theoretical models best fit a particular PDF, but also tracks how certain assumptions affect the behavior of these PDFs. This provides valuable insight into the factors leading to AI decisions and helps researchers understand the impact of various theoretical parameters.
By adapting techniques originally developed for image recognition, researchers have created powerful tools for analyzing complex theoretical models in particle physics.
“We reused the tools from computer vision,” explained Kriesten. “This helps us understand how different theoretical assumptions change the characteristics of PDFs.”
Together, these frameworks represent important advances in the application of AI/ML in particle theory.
Change the future of high energy physics
“Our work is focusing on using AI/ML to unravel complex problems in particle physics,” says Hobbs. “By improving understanding and accuracy of PDFs, we are paving the way for more accurate predictions in high-energy physics experiments.”
As AI continues to move forward, its role in particle physics is expected to deepen, potentially revealing more secrets in the universe. Hobbs and Kriesten are optimistic about the potential for AI/ML transformation in theoretical physics. They plan to expand the framework to cover a wider range of particle interactions and explore basic models to fully capture the complexity of particle physics.
By pushing the boundaries of AI/ML applications, they not only advance high energy physics, but also set the stage for future discoveries that can redefine understanding of the universe.
Other contributors to this work include Jonathon Gomprecht of the University of Arizona.
detail:
Brandon Kriesten et al., learn PDFs through interpretable latent representations of Mellin spaces; Physical Review d (2025). doi:10.1103/physrevd.111.014028
Brandon Kriesten et al., Explanatory AI classification of Parton density theory; Journal of High Energy Physics (2024). doi:10.1007/jhep11 (2024)007
Provided by Argonne National Laboratory
Quote: Decoding high energy physics with AI and machine learning (June 13, 2025) Retrieved from https://phys.org/news/2025-06-06-06-06-decoding-high-energy-physics-ai.html
This document is subject to copyright. Apart from fair transactions for private research or research purposes, there is no part that is reproduced without written permission. Content is provided with information only.
