ATLAS reports the first single-experiment observation of s-channel single top-quark production at 13 TeV, measuring a 10.8 pb cross-section
Synopsis
Using 13 TeV proton–proton collision data corresponding to 140 fb⁻¹ and selecting events with exactly one electron or muon, missing transverse momentum, and exactly two b-quark-initiated jets, ATLAS separates the signal from top-quark pair, W+jets, and t-channel single top-quark backgrounds with a Graph Neural Network; a likelihood fit to the network output yields an observed (expected) signal significance of 5.1 (5.2) standard deviations and a measured cross-section of 10.8 ± 1.5 (stat.) ± 1.6 (syst.) pb, consistent with the Standard Model prediction.
cross-section is heavily suppressed because the leading-order (LO) process, illustrated in Figure 1, requires a sea antiquark in the initial state, although higher-order corrections involving quark-gluon initial states provide substantial enhancements [9, 10]. The ATLAS Collaboration previously reported evidence for this process at center-of-mass energies of 8 TeV [11] and 13 TeV [12], achieving observed (expected) significances of 3.2 (3.9) and 3.3 (3.9) standard deviations, respectively. Additionally, searches by the CMS Collaboration at 7 TeV and 8 TeV [13] yielded an observed significance of 2.5 standard deviations.
arXiv · Page 2Interpretation
The first single-experiment observation of s-channel single top-quark production, with an observed (expected) signal significance of 5.1 (5.2) standard deviations, above the threshold for direct detection. The process had previously evaded direct detection at the Large Hadron Collider because of its highly suppressed cross-section; this work fills that gap with a single-experiment observation. Based on 13 TeV proton–proton collision data with 140 fb⁻¹ integrated luminosity, an event selection requiring exactly one electron or muon, missing transverse momentum, and exactly two b-quark-initiated jets, and a likelihood fit to the network output for the significance.
The measured s-channel single top-quark production cross-section is σ_{t b̄ + t̄ b} = 10.8 ± 1.5 (stat.) ± 1.6 (syst.) pb, consistent with the Standard Model prediction. It provides a quantitative measurement of the process rather than significance evidence alone, adding a direct constraint to the characterization of the top quark's electroweak interactions. Statistical and systematic uncertainties are reported separately, and the measured value is consistent with the Standard Model prediction.
The analysis uses a Graph Neural Network to separate the signal from top-quark pair, W+jets, and t-channel single top-quark backgrounds. Applying a Graph Neural Network to this multi-background separation task for a highly suppressed signal is a key methodological component of the analysis. The method is described in the summary, and the network output is subsequently used in a likelihood fit to extract the signal.
Perspective
The result applies to s-channel single top-quark production as defined by the event selection described in the summary (exactly one electron or muon, missing transverse momentum, exactly two b-quark-initiated jets) in 13 TeV proton–proton collisions with 140 fb⁻¹ integrated luminosity. It moves the process from a direct-detection gap into the set of experimentally measured processes, serving as a benchmark for top-quark physics, electroweak-interaction studies, and new-physics searches.
The summary does not give the Graph Neural Network architecture, training, or validation scheme, nor does it break down the sources of systematic uncertainty; the specific thresholds in the event selection, the background-estimation methods, and the fit parameterization are likewise not described in the summary. These details remain important for judging the robustness of the measurement at a finer level and are open questions that require the full text.