The ATLAS and CMS experiments have successfully detected the production of a quartet of top quarks during high-energy proton collisions inside the Large Hadron Collider. Four-top production is 4,000 times less common than even the production of Higgs bosons.
“It’s just incredible that we’re able to observe this process,” says Nedaa Alexandra Asbah, a postdoc at Harvard.
Top quarks are the most massive fundamental particles, weighing in at the same mass as a caffeine molecule. Scientists hope that by studying these chart-topping particles, they can learn more about the Higgs field, which gives quarks and other fundamental particles their masses.
“Studying the four-top-quark production is a great way to look for new physics,” says Melissa Quinnan, a postdoc at the University of California, San Diego.
The Standard Model of particle physics, the best model scientists have to describe the behavior of subatomic particles, predicts how often four-top production should occur. Scientists look for and study rare processes like this to test whether experimental measurements match what the Standard Model tells them to expect.
“Undiscovered heavy particles could influence the rate at which we see this process,” says Meng-ju Tsai, a graduate student at the University of Michigan. “For instance, the LHC could be producing two top quarks and a heavy cousin of the Higgs boson that then decays into an additional two top quarks. This would increase the rate at which we see four top quarks beyond the value predicted by the Standard Model.”
The LHC uses Albert Einstein’s equation E=mc2 to transform the energy stored inside protons into massive fundamental particles. Protons themselves are not fundamental but made up of point-like particles called quarks which are held together by other fundamental particles called gluons. As the protons accelerate, their gluons gain energy. When two gluons collide, their energy can transform into a host of other particles, including top quarks.
According to Tsai, the production of four top quarks is one of the rarest and heaviest processes ever observed at the LHC. If the Standard Model is correct, then only around one in 4 trillion LHC collisions should generate four top quarks at once.
If there had been no improvements to the LHC research program since its first run, it would have taken physicists around 1,000 years of operation to see this process. Luckily, there’s been significant progress, thanks to upgrades to both the accelerator and detectors. Since 2012, the ATLAS and CMS experiments have increased their experimental data sets sixfold.
Physicists have also improved their analysis techniques. Top quarks are so short-lived that scientists will never see them directly. Instead, scientists look for the particles top quarks produce as they decay.
The problem is that the decay patterns produced by top quarks are almost identical to the patterns left by many other subatomic processes. “Everything we do is to try to disentangle the signal from background,” Asbah says.
Scientists use theoretical models and computer simulations to predict what their signal will look like: the kinds of particles their detectors should see and their trajectories and momenta. Because these signatures are so similar to background events, scientists often use machine-learning algorithms to help them sort signal from background.
Quinnan compares this work to isolating a person in a photograph. “With machine learning, applications like Photoshop can map the background of a digital photograph in a much more complex way,” she says. “Our analysis is similar.”
But in this case, the signal and background can look so similar that it’s less like removing an unsightly parking lot from a selfie, and more like isolating a specific person in a grainy picture taken during a massive concert. “We had such a large and complicated background that we couldn’t simulate it; we had to use the data,” Quinnan says. “This was the first time CMS used machine learning to estimate a data-driven background.”
While a photographer can visually check an image to make sure the background has been cleanly removed, physicists don’t have the same luxury. Until they are ready to analyze their data, they won’t even check the signal region, as a safeguard against biasing themselves.
For these kinds of physics analyses, scientists identify all the relevant features of their signal and develop a sketch of what the signal event should look like in the experimental data. They also create sketches of similar-looking background events, in which only one or two of the features are tweaked. Physicists then feed these imitation signal events and imitation backgrounds into their machine-learning algorithms and see how well they perform.
“We had long discussions and validations to cross-check that the modeling for the machine-learning output is decent,” Tsai says.
Only after the algorithms have been verified on imitation data do the scientists apply the same techniques to the signal region of their real data.
The new results from ATLAS and CMS are consistent with the Standard Model, given the systematic and statistical uncertainties. ATLAS’s measured statistical significance is 6.1 sigma, which means there is only a one-in-a-billion chance that the findings are the result of signal-like background variations and not of a previously undiscovered process. Likewise, CMS’s statistical significance is 5.5 sigma, which is also beyond the 5-sigma threshold physicists need to claim the discovery of a previously unseen particle or process.
Even though the results are still consistent with the Standard Model, both experiments are seeing more signal events than they would normally expect based on the Standard Model’s predictions. “It can be either data fluctuation or actually new physics playing a role here,” Tsai says. “Although we cannot conclude that we have observed new physics beyond the Standard Model, this analysis provides some hints and direction that we can continue looking into.”
If the excess grows over time, it could be evidence of a heavy fundamental particle producing four top quarks and thus boosting the rate. “By the end of Run 3, we will have tripled our data set,” Asbah says. “We will be able to pin this process down.”
Even if the analysis doesn’t lead to new physics, for Tsai, getting to work on this kind of research has been a dream come true.
“When I was a high school student in Taiwan, I was attracted by the discovery of the Higgs boson,” Tsai says. “Even though I didn’t know what exactly a Higgs boson was, I was thrilled and wished that I would be able to come to Geneva and join the collaboration in CERN one day, working with physicists there to discover new physics. And now, everything has come true.”