← Latest papers
⚛️ lattice

Total Angular Momentum Assignments in the Hadron Spectrum

This paper presents a data-driven XGBoost model trained on PDG hadron data to predict the total angular momentum JJ of hadronic resonances with unknown spins, conditional on parity hypotheses, thereby offering new insights that both align with and challenge existing theoretical assignments.

Original author: The HADLENS Collaboration, S. Rostami, Omid Zamani Jamshidi, B. Dehghanzad, K. Azizi

Published 2026-11-19
📖 6 min read🧠 Deep dive

Original author: The HADLENS Collaboration, S. Rostami, Omid Zamani Jamshidi, B. Dehghanzad, K. Azizi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the author. For technical accuracy, refer to the original paper. Read full disclaimer

The universe is built from a small set of fundamental particles that stick together to form everything we see, from the atoms in our bodies to the stars in the sky. Most of the matter around us is made of protons and neutrons, which are themselves composed of even smaller pieces called quarks. These quarks are held together by a powerful force that acts like a cosmic glue, but unlike the gravity that holds planets in orbit, this force behaves in ways that are difficult to calculate with standard math. Because of this, physicists cannot simply solve an equation to predict exactly what kinds of particles can exist or how they will behave. Instead, they rely on a mix of high-energy experiments to spot new particles and theoretical models to guess their properties. One of the most important properties to identify is a particle's total spin, which describes how it rotates and how its internal parts are arranged. Knowing this spin, along with other characteristics like mass and parity, allows scientists to place a particle into the correct slot in the grand family tree of matter, known as the hadron spectrum. Without these labels, a newly discovered particle is just a nameless bump in the data, and its true nature remains a mystery.

For decades, scientists have been filling in this family tree, but many spots remain blank or uncertain. The Particle Data Group, the international organization that compiles the known properties of all subatomic particles, lists hundreds of resonances—short-lived particles that appear and disappear quickly—where the mass is known but the spin is not. Some of these particles are well-understood combinations of quarks, while others might be exotic new forms of matter, such as particles made of four or five quarks instead of the usual two or three. Theorists have proposed many different ideas about what these missing particles might be, but without experimental proof, these ideas remain just hypotheses. To bridge the gap between theory and the unknown, a team of researchers has turned to machine learning, a branch of artificial intelligence that allows computers to find patterns in large datasets. By teaching a computer to recognize the relationship between the known properties of particles and their spins, the team has created a tool that can predict the spin of particles that have never been measured.

The researchers, working under the name HADLENSa, built their model using a vast collection of data from the Particle Data Group. They fed the computer information about hundreds of known mesons and baryons, which are the two main families of particles made of quarks. The computer learned to look at specific features of each particle, such as which types of quarks it contains, its mass, its electric charge, and its parity, a property that describes how the particle behaves when viewed in a mirror. The goal was to teach the machine to predict the total spin of a particle based on these features. However, there was a catch: the model could not predict the spin and the mirror-behavior at the same time. Instead, the researchers had to tell the computer to assume a specific mirror-behavior for each particle, and then the computer would predict the spin that best fit that assumption. This approach allowed them to generate a list of likely spins for particles where the experimental data was missing or unclear.

To ensure their predictions were reliable, the team used a sophisticated method of training that involved testing the model against itself in many different ways. They split their data into groups, training the computer on some parts and checking its accuracy on others, repeating this process until the model was as sharp as possible. They also used a genetic algorithm, a technique inspired by natural selection, to help the computer decide which pieces of information were most important for making a correct guess. Once the model was ready, the researchers applied it to a long list of particles whose spins were unknown. They compared the computer's predictions with the various theories that physicists had already proposed for these same particles. The results showed a strong agreement in many cases. For example, for a particle called K(1630), the model predicted a spin of one with a specific mirror-behavior, which matched the leading theory that described it as a specific type of excited quark combination. Similarly, for a group of particles known as the hidden-charm pentaquarks, the model's predictions aligned well with the most popular theoretical ideas about their internal structure.

However, the study also highlighted where the current understanding of the universe might be incomplete. In several instances, the computer's prediction did not match the most common theoretical guess. For the particle known as Σ(3000), the model suggested a spin of one-half, while a recent theoretical study had argued for a spin of five-halves. The researchers noted that this difference does not necessarily mean the computer is wrong or that the theory is wrong; rather, it points to a need for new experiments to settle the debate. The same was true for a particle called Ω(2250), where the model predicted a spin of three-halves, conflicting with a theoretical assignment of five-halves. These disagreements are valuable because they act as signposts, telling experimentalists exactly which particles to measure next and which theoretical ideas to re-examine. The team also found that for some particles, like the Ξ(2500), there were no reliable theoretical predictions to compare against at all, leaving the computer's output as the only guide for future investigation.

The final output of this work is not just a list of numbers, but a systematic database of hadronic states with their most likely spin assignments. This resource provides a clear starting point for future research, offering a way to test whether different theoretical models are pointing in the same direction or if they are fundamentally at odds. By providing these predictions, the researchers have given the scientific community a new tool to organize the chaotic landscape of subatomic particles. The study confirms that machine learning can successfully uncover hidden patterns in the rules that govern matter, even in a regime where traditional math struggles. While the computer cannot replace the need for physical experiments, it can tell scientists where to look and what to expect, turning a sea of unknowns into a map with clear destinations. As new data comes in from particle colliders, these predictions can be checked and refined, slowly filling in the missing pieces of the cosmic puzzle.

Drowning in papers in your field?

Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.

Try Digest →