ATLAS one step closer in the search for rare Higgs boson decays to muons

23 July 2020 | By

Higgs Candidates,Proton Collisions,Event Displays,Physics,ATLAS
Figure 1: A Run 2 ATLAS event containing two muons (red) with mass compatible with that of the Higgs boson, and two forward jets (yellow cones). (Image: ATLAS Collaboration/CERN)

"Who ordered that?" commented physicist Isidor Isaac Rabi when the muon was discovered in 1936. In the 80 years since, scientists have learnt a lot about the muon’s role in our Universe and have studied its properties with extreme precision. Muons have even been used via decays of intermediate weak bosons in the detection of new particles, such as the Higgs boson – now the centerpiece of its own extremely rich field of research.

In the Standard Model, elementary particles acquire mass through interaction with the Higgs field: the stronger the interaction, the larger the mass of the particle. So far, physicists have collected conclusive evidence of the Higgs boson interacting with bosons and the heaviest elementary fermions belonging to the third fermion generation (tau-lepton as well as top and bottom quarks). Yet to date, there is no indication whether the Higgs boson interacts with the next lighter fermions, muon or charm quark, belonging to the second fermion generation.

The ATLAS Collaboration has released a new paper on the search for the Higgs-boson decay to a pair of muons. The new study uses the entire dataset collected by the ATLAS experiment during Run 2 of the LHC (2015–2018) to give a first hint of this elusive process.

The ATLAS Collaboration studied the entire dataset collected during Run 2 of the LHC (2015–2018) to give a first hint of the elusive Higgs-boson decay to muon pairs.

The wheat from the chaff

Despite a simple experimental signature, spotting this rare decay continues to be a challenge. This is due to the low probability of the Higgs boson decaying to muons (predicted to be just 0.02%) and the large number of events from similar Standard Model background processes that can dominate the search. Only 0.2% of selected muon-pair events with masses between 120 and 130 GeV from proton–proton collisions are expected to come from a Higgs-boson decay.

Fortunately, a signal can be distinguished from background processes by looking at the shape of the mass distribution of the precisely measured muon pairs. Higgs-boson events will cluster around the Higgs-boson mass of 125 GeV, producing a narrow peak that can be distinguished from the smoothly-falling distribution of background events. By fitting the invariant-mass spectrum, ATLAS physicists are able to directly constrain the background and extract a possible signal.

Plots or Distributions,Physics,ATLAS
Figure 2: The invariant-mass spectrum of the reconstructed muon-pairs in ATLAS data. Events are weighted according to the expected signal-to-background ratio of their category. In the top panel, the signal-plus-background fit is visible in blue, while in the lower panel the fitted signal (in red) is compared to the difference between the data and the background model. (Image: ATLAS Collaboration/CERN)

Divide et impera

To further increase the sensitivity of their analysis, ATLAS physicists divided their events into 20 mutually-exclusive “categories”. These categories focussed on the features of an event – such as the number and properties of its additional jets or leptons – to target specific production modes of the Higgs boson, including the scattering of two gluons or two weak bosons, and the associated production with a weak W or Z boson or a top-quark pair. Inside these categories, events were further split using dedicated multivariate discriminants (Boosted Decision Trees). As a result of this complex division, ATLAS physicists could separate out the few Higgs-boson-like events from the more common, but less Higgs-boson-like, events.

In addition, ATLAS physicists developed a robust (and ambitious) background-modelling strategy using a variety of simulation techniques to create more than 10 billion simulated events. Detailed ATLAS detector simulations (totalling about five times the Run-2 dataset) were complemented by dedicated fast simulation samples (more than 100 times the dataset). The fast simulation samples were crucial to ensure that the overwhelming backgrounds could not mimic a false signal, while maximising the analysis sensitivity to a real signal.

Let the die be cast

The new ATLAS result gives a first hint of the Higgs boson decaying to a muon pair; the significance of the observed signal amounts to 2.0 standard deviations and the ratio of the observed signal yield to the one expected in the Standard Model is 1.2 ± 0.6. The data, together with the signal-plus-background fit, are shown in Figure 2, where data events are weighted to reflect the signal-to-background ratio of their respective categories. More data to be collected in Run 3 (2022–2024) and during the operation of the High-Luminosity LHC will help close in on this first hint.