Physics Briefing for VBS VW semi-leptonic

Using the LHC as a vector boson collider

After the Higgs boson discovery in the year 2012, the standard model of particle physics offers a complete and consistent description of elementary particle interactions that, despite the many attempts, to date has not been falsified by experimental evidence. Nevertheless,  several hints indicate that a more comprehensive theory that includes and extends it  may exist, justifying the indirect evidence of dark matter, naturally giving non-null mass to neutrinos, explaining the apparent excess of matter over anti-matter in the universe. As it already happened in the past for the classical mechanics with respect to relativity and quantum physics, and for quantum physics with respect to the quantum field theory, also in this case differences between the standard model and a new theory are expected to arise at large energies, where our experimental apparatuses have not been able to arrive yet. The vector boson scattering, in short VBS, is one ideal place to search for a direct evidence that something unexpected may be happening at the Large Hadron Collider (LHC). This process happens at a hadron collider when the particles from the incoming proton beams emit vector bosons, which in turn interact among themselves. This peculiar configuration is very rare, but it also produces specific features that particle physicists exploit to select the events of interest in the data analysis. Without the so-called Higgs mechanism, which is the mathematical grounding for the existence of the Higgs boson, the probability of the vector bosons interaction couldn't even be calculated; therefore the VBS is one ideal playground to confirm the standard model or, more interestingly, to search for anomalies in one of the less experimentally known sectors of the theory. To actually observe it physicists search, among the collision events recorded by the Compact Muon Solenoid (CMS), those that contain two vector bosons resulting from the scattering and the remnants of the incoming protons, which show up as collimated showers of particles usually called "jets". Because they directly come from the LHC proton beams, such jets feature very high energy, characterising the events. Vector bosons, instead, have an extremely short life and decay into fermion pairs almost immediately after the interaction: it's these decay products that are then looked for in the CMS recordings. This analysis describes the case when one of the two vector bosons decays hadronically into two quarks, while the other one, a W boson, disintegrates into a neutrino and an electron or a muon. This specific decay pattern is particularly favourable in terms of expected number of events thanks to the large probability of the hadronic vector boson decay, while retaining the identification capabilities connected to the presence of an electron or muon in the event. Figure [eventDisplay] shows a display of a simulated VBS event happening in the CMS detector, where ...  are visible (this to be fixed when the image is done) .... Since the identification of electrons and muons are very different, as they are the determination of their properties from the raw outputs of the CMS detector, the data have been split into corresponding categories analysed separately. A further splitting is due to the fact that the vector boson hadronic decay  may generate two separate jets or a single one, depending on the angular distance between the quarks it produces. Unfortunately, in all the categories this analysis suffers from a very large background: of the many possible interaction types that may happen at the LHC, those that could mimic a VBS-like event are a lot, as a matter of fact about 1,600 times more, and their subtraction is the major problem of the study. To achieve this goal, the analysers designed a deep learning algorithm to discard uninteresting events, exploiting a combination of the most sensitive features of the interaction kinematics. To control the impact of each of them in the algorithm outcome, explainable artificial intelligence techniques have been employed: as a consequence, the final result gains in robustness, thanks to the detailed understanding  of its behaviour. Figure [DNN] shows the distribution of the deep learning algorithm for one of the analysis categories (DNN in the image), where the black dots represent the observed counts in the data, while the coloured areas correspond to the various sources of backgrounds, stacked on top of each other. The white area above them is the VBS signal, shown also as a red line at the bottom of the histogram.

It is clear that no region exists where the background is absent; therefore, one of the largest feats of this study has been a reliable estimate of them, in order to achieve a trustworthy measurement of the VBS signal. To maximise the sensitivity in the VBS determination, then, the analysis team decided to perform a multi-parametric fit of the DNN distribution, which in one single chi-squared minimisation determined the backgrounds and allowed to establish the existence of the VBS signal with the significance of  4.4 standard deviations, which means that the backgrounds should over-fluctuate 4.4 times their uncertainty to mimic the signal. The number of observed events allows also to determine the probability of the process to happen,  which is measured to be 1.9 ± 0.5 picobarns, compatible within its uncertainty with the value calculated with the theory of 2.2 picobarns. The uncertainty in this measurement is largely affected by the statistical fluctuations in the number of recorded events, which means that  the analysis of more data collected in the future will become more and more precise, until the error will be dominated by its systematic component, which is the intrinsic limit due to the experimental apparatus at our disposal. With increased precision we will be able to test with unprecedented accuracy whether the standard model actually describes the physics observed at the LHC, or another new theory will be necessary to explain the observations.

-- PietroGovoni - 2021-07-12

• res_mu_3.png: