A search for lepton-flavour violating $τ\to 3μ$ decays with the ATLAS detector

Abstract

A search for charged lepton flavour violation in $τ\to 3μ$ decays is performed in $pp$ collisions at a centre-of-mass energy of 13 TeV using ATLAS data collected between 2016 and 2018, corresponding to an integrated luminosity of 137 $\text{fb}^{-1}$. The search focuses on the electroweak $W \to τν$ production channel. Data are collected using two-muon and three-muon triggers and a multivariate analysis is used to separate the signal from the background. An unbinned likelihood fit is then performed to the resulting three-muon invariant mass spectrum and the data are found to be compatible with the background-only hypothesis. The observed (expected) limit on the branching ratio $B(τ\to3μ)$ is found to be $8.7\times 10^{-8}$ ($7.5\times 10^{-8}$) at $90\%$ CL.

Figures (4)

• Figure 1: Distributions of representative example variables used in the BDT event classification: (a) vertex quality $p\text{-value}$, (b) muon isolation $p_{\text{T}}\xspace^\mathrm{iso}/p_{\text{T}}\xspace^\mathrm{triplet}$, (c) triplet $p_{\text{T}}\xspace$, $p_{\text{T}}\xspace^{\text{triplet}}$ and (d) significance of the transverse distance between the PV and SV $L^{\text{xy}}/\sigma_{L^{\text{xy}}}$. The plots show both the EW $\tau \to 3 \mu$ signal sample (histogram) and data from the sidebands (SB Data) (markers) after the preselection cuts. All distributions are normalised to unit area.
• Figure 2: BDT score distribution for the EW $\tau \to 3 \mu$ signal sample (histogram), and background from sideband data (SB Data) (markers). All distributions are normalised to unit area, and the error bars are too small to be visible. The vertical lines mark the boundaries of the loose ($0.88<$BDT$\leq0.975$) and medium ($0.975<$BDT$\leq0.998$) BDT bins used for the event categorisation. The tight boundary ($0.998<$BDT) is not shown as it is indistinguishable from the vertical axis.
• Figure 3: The $m_{3\mu}$ in data (markers), background-only fit (solid line) and expected EW plus HF signal, for $B(\tau \to 3 \mu) = 1\times 10^{-6}$ as an illustration (dotted line) in the (a,c,e) barrel and (b,d,f) endcap categories for the (a,b) tight, (c,d) medium and (e,f) loose BDT categories.
• Figure 4: $B(\tau\to3\mu)$ limit scan showing the expected limit (dotted line), $\pm 1(2) \sigma$ error bands (green and yellow), and the observed limit using the CL$_\text{s}$ method (solid line). A $p$-value of $0.1$ ($90\%$ CL) is shown by the horizontal line.