The Higgs boson, discovered at the Large Hadron Collider (LHC) in 2012, plays a central role in the Standard Model of particle physics, endowing elementary particles such as quarks with mass through its interactions. The Higgs boson’s interaction with the heaviest “third-generation” quarks – top and bottom quarks – has been observed and found to be in line with the Standard Model. But probing its interactions with lighter “second-generation” quarks, such as the charm quark, and the lightest “first-generation” quarks – the up and down quarks that make up the building blocks of atomic nuclei – remains a formidable challenge, leaving unanswered the question of whether or not the Higgs boson is responsible for generating the masses of the quarks that make up ordinary matter.

Researchers study the Higgs boson's interactions by looking at how the particle decays into – or is produced with – other particles in high-energy proton–proton collisions at the LHC. At a seminar held at CERN last week, the CMS experiment collaboration reported the results of the first search for a Higgs boson decaying into a pair of charm quarks in collision events where the Higgs boson is produced alongside two top quarks. Exploiting cutting-edge AI techniques, this novel search has been used to set the most stringent limits to date on the interaction between the Higgs boson and the charm quark.

The production of a Higgs boson in association with a top-quark pair, with the Higgs boson decaying into pairs of quarks, is not only a rare process at the LHC but one that is particularly challenging to distinguish from similar-looking background collision events. That’s because quarks immediately produce collimated sprays (or “jets”) of hadrons that travel only a small distance before decaying, making it especially difficult to identify jets originating from charm quarks that are created in the decay of a Higgs boson from jets originating from other types of quark. Traditional identification methods, referred to as “tagging”, struggle to efficiently recognise charm jets, necessitating the development of more advanced discrimination techniques.

“This search required a paradigm shift in analysis techniques,” explains Sebastian Wuchterl, a research fellow at CERN. “Because charm quarks are harder to tag than bottom quarks, we relied on cutting-edge machine-learning techniques to separate the signal from backgrounds.”

The CMS researchers tackled two major hurdles using machine-learning models. The first was the identification of charm jets, which was performed by employing a type of algorithm called a graph neural network. The second was to distinguish Higgs boson signals from background processes, which was addressed with a transformer network – the type of machine learning that is behind ChatGPT but trained to classify events instead of generating dialogues. The charm-tagging algorithm was trained on hundreds of millions of simulated jets to allow it to recognise charm jets with higher accuracy.

Using data collected from 2016 to 2018, combined with the results from previous searches for the decay of the Higgs boson into charm quarks via other processes, the CMS team set the most stringent limits yet on the interaction between the Higgs boson and the charm quark, reporting an improvement of around 35% compared to previous constraints. This places significant bounds on potential deviations from the Standard Model prediction.

“Our findings mark a major step,” says Jan van der Linden, a postdoctoral researcher at Ghent University. “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.”
As the LHC continues to collect data, refinements in charm tagging and Higgs boson event classification could eventually allow CMS, and its companion experiment ATLAS, to confirm the Higgs boson’s decay into charm quarks. This would be a major step towards a complete understanding of the Higgs boson’s role in the generation of mass for all quarks and provide a crucial test of the 50-year-old Standard Model.