By CMS Collaboration

In the thrilling world of particle physics, our mission at the Large Hadron Collider (LHC) is to explore the mysteries of the universe by conducting experiments and pushing the boundaries of the Standard Model (SM). The SM is a wonderful theoretical framework that describes the fundamental particles and their interactions, but we know for certain that it is incomplete. Some things are surely missing… maybe some particles that we haven’t seen yet?

You might think that all we need is to produce those new particles at the LHC, much like we successfully did for the Higgs boson. But it is not that easy. We can only produce so much energy in a single collision at the LHC, and if these new particles are extremely heavy, they will not show up. Luckily, we have a way to test their existence even if we cannot see them directly: we can “feel” their presence. Indeed, new heavy particles can affect some interactions and make them deviate from the SM prediction. So, we focus on identifying the LHC processes most sensitive to these deviations and search for them in the data collected by our experiments.

One fascinating process that we investigate is the production of a top quark together with a W boson and a Z boson, the tWZ process. Some brilliant theory colleagues of ours have studied this process in detail and realized that it is particularly sensitive to the presence of new particles in a high-energy regime. The tWZ production has a very low cross section, meaning that it is very unlikely to happen in proton-proton collisions at the LHC. More specifically, the tWZ cross section is about 0.14 pb, while the total inelastic proton-proton cross section is about 100’000’000’000 pb. We decided to take on the challenge and look for it! Because of its rarity, we search the tWZ process in a final state with three or four leptons (Fig. 1). This yields a cleaner signature from an experimental standpoint or, in other words, it is easier to identify in the detector, as can be seen in Fig. 2.

sketch of signal

Figure 1: Illustration of tWZ production in proton-proton collisions in a final state with three charged leptons.

In an experimental search, the process that we look for (the signal) can be mimicked by other processes (the backgrounds), which make the search more difficult; as when you are looking for a friend in a crowd. Among the background processes, the most overwhelming for our search is the production of two top quarks in association with a Z boson, ttZ. One of the first analysis steps is to simulate both the signal and all the backgrounds. The LHC smashes bunches of protons every 25 ns and many physics processes can occur each time. By simulating these collisions with a computer, we know how many events per process to expect, which we can then compare to the collected data to establish the sensitivity of our analysis, a measure of how well we should detect new particles or observe a particular process.

Figure 2: A data event recorded by the CMS detector at 13 TeV, compatible the tWZ signal: the Z boson decays to a pair of electrons (green lines) whereas the W boson decays into quarks that produce jets of particles (yellow cones). The top quark decays to a b quark, appeared as a b jet (orange cone) and a W boson decaying to a muon (red line) and neutrino. You can rotate and zoom in this separate page.

Here comes one of the first challenges: simulating the signal. Indeed, tWZ is a unique process because of some quantum-level interference with other physics processes, such as ttZ. The complexity lies in the decay of the top quark, which is almost always to a W boson and a bottom quark, t→Wb. So, if by chance a bottom quark is produced along with tWZ, we will face a tWbZ event that is indistinguishable from ttZ. To avoid any confusion, we use methods that, in practice, count the number of "Wb" resonances: a ttZ event has two of them, while a tWZ event has only one!

Our signal simulation is now well defined. We then proceed to maximize the separation between tWZ and all other backgrounds; you may guess that it is vital to discriminate tWZ against ttZ. For this purpose, we employ a Deep Neural Network (DNN) that receives a number of features distinguishing between signal and background, and classifies events as originating from tWZ, ttZ, or other processes.

Last but not least, we looked at the most favorable region to find deviations from the SM: the high-energy tails of our kinematic distributions. Here, the jets from the top quark receive an energy boost that makes them very close in space, and appear as one fat jet in the detector, which we can reconstruct. Instead, in the case where the W boson from the top quark decays to leptons, we again employ a DNN multivariate discrimination technique that gives us the probability that a given charged lepton originates from an energetic top quark, to better identify the signal in this region.

Now we have all the ingredients needed for our statistical analysis! We fit the data in several categories (the collisions with high energy and the ones with different numbers of detected leptons and jets) as a superposition of the signal and background components, to evaluate the strength of the signal.

We measure a cross section of 0.37 ± 0.11 pb. We have computed that the observed result has a probability of occurring by chance of less than 0.3%, and thus we have evidence that the process exists! The post-fit distribution of the tWZ DNN output is shown in Fig. 3.

tWZ output node score

Figure 2: Distribution of the DNN output score for the tWZ node, after the fit to data. The hatched band represents the total uncertainty.

With this analysis, we looked where no one had ever looked before and established a groundwork for future searches. What’s next? With the larger event samples collected in 2022 and beyond, we can further refine our measurement and dive deeper into the high-energy region to uncover physics beyond the SM. Exciting times lie ahead!

Read more about these results:

  • Do you like these briefings and want to get an email notification when there is a new one? Subscribe here
Date of publication