22-05-2025
AI helps CERN physicists charm Higgs boson into revealing rare decay
Researchers at CERN have used artificial intelligence (AI) to explore one of the Higgs boson's most elusive behaviors, shedding light on its subtle interaction with charm quarks and bringing science closer to understanding how the so-called God particle gives matter its mass.
Discovered in 2012 at the Large Hadron Collider (LHC), the Higgs boson, the fundamental force-carrying particle of the Higgs field, plays a crucial role in the Standard Model of particle physics by interacting with particles like quarks to bestow mass.
While earlier experiments confirmed the Higgs boson's interactions with heavier third-generation quarks, both top and bottom, it remains a major challenge to study its connection with second-generation quarks like the charm quark, and the lighter first-generation up and down quarks that form the building blocks of atomic nuclei.
Now, as part of the Compact Muon Solenoid (CMS) experiment, one of two major general-purpose particle detectors at the LHC, the research team announced that they've set the most precise limits yet on the Higgs boson's decay into charm quarks, marking a major step toward understanding how it gives mass to matter.
Producing a Higgs boson alongside a pair of top quarks, followed by its decay into quark pairs, is a rare event at the LHC and especially difficult to tell apart from background collisions that appear nearly identical.
This is because quarks don't appear as distinct particles but instead form dense, collimated sprays of hadrons, called jets, that travel only a short distance before decaying. This makes it especially difficult for methods such as tagging to distinguish charm quark jets from those produced by other quarks.
"This search required a paradigm shift in analysis techniques," Sebastian Wuchterl, PhD, a CERN research fellow explains. "Because charm quarks are harder to tag than bottom quarks, we relied on cutting-edge machine-learning techniques to separate the signal from backgrounds."
To overcome the challenge, the CMS team turned to advanced AI, leveraging two machine-learning models tailored for the challenge. First, they used a type of algorithm called a graph neural network to enhance charm jet identification. These algorithms treat each jet as a network of particles, learning how to spot subtle structural patterns unique to charm quark decays.
The team then tackled the second hurdle, distinguishing Higgs boson signals from background processes, with a transformer network, best known for powering AI language models like ChatGPT.
This network was repurposed to classify entire collision events, distinguishing those likely to feature a Higgs boson decaying into charm quarks. The charm-tagging algorithm was trained on hundreds of millions of simulated jets, enabling it to identify charm jets with significantly greater accuracy.
The analysis focused on data collected from 2016 to 2018, specifically targeting collision events in which the Higgs boson is produced together with a pair of top quarks. By combining this dataset with results from previous studies, CMS achieved a roughly 35 percent improvement in the sensitivity of Higgs–charm interaction measurements.
"Our findings mark a major step," Jan van der Linden, PhD, a postdoctoral researcher at Ghent University, said in a press release. 'With more data from upcoming LHC runs and improved analysis techniques, we may gain direct insight into the Higgs boson's interaction with charm quarks at the LHC - a task that was thought impossible a few years ago."
The researchers hope that as the LHC collects more data, improved charm tagging and event classification could enable CMS and its companion experiment ATLAS to confirm the Higgs boson's decay into charm quarks. This would mark a massive step toward fully understanding how the Higgs gives mass to all quarks and would provide a crucial test of the 50-year-old Standard Model.
