Search for higgsinos in compressed mass spectra using low-momentum tracks in $pp$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

TL;DR

This paper tackles the challenge of probing higgsinos in compressed mass spectra by deploying two complementary ATLAS searches that target distinct mass-splitting regimes. It leverages ISR-boosted events and novel low-$p_T$ track techniques, including displaced-track and soft-lepton tagging powered by neural discriminants, to identify higgsino decays to soft pions or leptons. No excess over the Standard Model is observed, leading to 95% CL limits that exclude chargino masses up to about 126 GeV for $\Delta m$ in 0.3–2 GeV, effectively superseding LEP bounds. The combination of displaced-track and 1$\ell$1T strategies provides broad coverage of the compressed higgsino parameter space and demonstrates ATLAS’s ability to probe challenging signatures with soft final-state particles.

Abstract

This paper presents two searches for the electroweak production of higgsinos with compressed mass spectra using 140 fb$^{-1}$ of $\sqrt{s}=13$ TeV proton-proton collision data collected by the ATLAS experiment at the Large Hadron Collider. Events are required to feature an energetic jet, large missing transverse momentum, and at least one low-momentum charged particle that serves as a candidate higgsino decay product. In the first search, targeting higgsino mass splittings in the range of 0.3-1 GeV, the higgsinos are expected to predominantly decay into pions that are identified as low-momentum charged particles with large transverse impact parameters due to the long higgsino lifetime ($cτ\approx\mathcal{O}$(0.1-1 mm)). The second search targets larger mass splittings in the range of 1-3 GeV, where the higgsinos are expected to decay promptly into low-momentum leptons, one of which is identified by dedicated low-momentum electron or muon taggers based on neural networks utilising tracking and calorimeter information. No significant excess above the Standard Model prediction is observed in either search and the results are used to set lower limits on the masses of the higgsino-like charginos and neutralinos within a simplified model. Together, these searches exclude chargino masses below 126 GeV at 95% confidence level for mass splittings between the chargino and lightest neutralino in the range of 0.3-2 GeV, representing the first ATLAS constraints in this parameter space and surpassing the limits previously set by the LEP experiments.

Figures (15)

• Figure 1: Example diagrams for the targeted signal processes featuring a jet ($j$) from initial-state radiation for (a) the displaced track search and (b) the 1$\ell$1T search. While only the $\hbox{$\tilde{\chi}^0_1$}\xspace\hbox{$\tilde{\chi}^\pm_1$}\xspace$ ($\hbox{$\tilde{\chi}^0_2$}\xspace\hbox{$\tilde{\chi}^\pm_1$}\xspace$) diagram is shown for the displaced track (1$\ell$1T) search, the additional production processes described in Section \ref{['sec:samples']} are considered as well.
• Figure 2: Simulated true-track efficiency and fake-track rate as a function of track for (a) the low- electron-track ID and (b) the low- muon-track ID. The scale factors obtained from dedicated calibration methods are applied with corresponding uncertainties. The total uncertainty including the statistical and systematic uncertainties are shown, where the former uncertainty is found to be negligible. The true track is defined as a track matched to a lepton originating from $\hbox{$\tilde{\chi}^0_2$}\xspace\to\hbox{$\tilde{\chi}^0_1$}\xspace\ell^+\ell^-$ decay in signal events satisfying the event selections defined later in Section \ref{['sec:selection']}. The fake track is defined as a track in $W^\pm+$jets events that satisfies the same selection, excluding the track associated with the lepton from the $W^\pm$ decay.
• Figure 3: Overall structure of the pNN model used for event selection. The $\Delta m(\hbox{$\tilde{\chi}^0_2$}\xspace,\hbox{$\tilde{\chi}^0_1$}\xspace)$ and 16 kinematic variables are used as inputs, and there are six hidden layers with 64 nodes each. The nodes are fully connected. L1 and L2 regularisations are applied to each layer for increased signal sensitivity. The output pNN score ranges from 0 to 1, with 1 meaning that the input event is signal event-like.
• Figure 4: ROC curves of pNN of each pNN-$\theta$ for (a) the electron channel and (b) the muon channel. The performance is better for smaller $\Delta m(\hbox{$\tilde{\chi}^0_2$}\xspace,\hbox{$\tilde{\chi}^0_1$}\xspace)$ values for both channels. The steps in the curves are due to the limited statistical precision of the MC samples used.
• Figure 5: Schematic of the four validation region categories defined in the event- vs. track-level NN score plane. Each VR is further divided into three or four slices depending on the event and track NN scores. VR-DT-Lep regions have the same score requirements as the SR-DT regions but with the 1L requirement and are thus denoted in parentheses.