The production of four top quarks ($\mathrm{t\bar{t}t\bar{t}}$) is studied with LHC proton-proton collision data samples collected by the CMS experiment at a center-of-mass energy of 13 TeV, and corresponding to integrated luminosities of up to 138 fb$^{-1}$. Events that have no leptons (all-hadronic), one lepton, or two opposite-sign leptons (where lepton refers only to prompt electrons or prompt muons) are considered. This is the first $\mathrm{t\bar{t}t\bar{t}}$ measurement that includes the all-hadronic final state. The observed significance of the $\mathrm{t\bar{t}t\bar{t}}$ signal in these final states of 3.9 standard deviations (1.5 expected) provides evidence for $\mathrm{t\bar{t}t\bar{t}}$ production, with a measured cross section of 36 $^{+12}_{-11}$ fb. Combined with earlier CMS results in other final states, the signal significance is 4.0 standard deviations (3.2 expected). The combination returns an observed cross section of 17 $\pm$ 4 (stat) $\pm$ 3 (syst) fb, which is consistent with the standard model prediction.
The jet multiplicity for $N_\textrm{b} \geq 4$ in the opposite-sign dilepton channel for the combined 2017--2018 dataset with dilepton decay categories combined. Here, $\textrm{t}\bar{\textrm{t}} + \geq 1 \textrm{b}$ refers to $\textrm{t}\bar{\textrm{t}}$ events with at least one additional b jet, $\textrm{t}\bar{\textrm{t}} + 0 \textrm{b}$ includes all other $\textrm{t}\bar{\textrm{t}}$ events not produced in association with a boson, and EW refers to events that contain W and Z bosons but no top quarks. The backgrounds and $\textrm{t}\bar{\textrm{t}}\textrm{t}\bar{\textrm{t}}$ signal (derived from the fit) are shown as a stacked histogram. The hatched bands correspond to the estimated total uncertainty after the fit.
The distribution of the BDT discriminants for the 2016--2018 data set for three different categories in the combined single-electron and single-muon channels. The three categories are defined by the number of resolved t tags ($N_\textrm{RT}$), b tags ($N_\textrm{b}$), and jets ($N_\textrm{j}$), selected as representative based on their sensitivity to signal. Here, $\textrm{t}\bar{\textrm{t}} + \geq 1 \textrm{b}$ refers to $\textrm{t}\bar{\textrm{t}}$ events with at least one additional b jet, while $\textrm{t}\bar{\textrm{t}} + 0 \textrm{b}$ includes all other $\textrm{t}\bar{\textrm{t}}$ events not produced in association with a boson. The TOP grouping contains single top quark production along with the other $\textrm{t}\bar{\textrm{t}}$ processes not explicitly shown, and EW refers to events that contain W and Z bosons but no top quarks. The backgrounds and $\textrm{t}\bar{\textrm{t}}\textrm{t}\bar{\textrm{t}}$ signal (derived from the fit) are shown as a stacked histogram. The hatched bands correspond to the estimated total uncertainty after the fit. While the bins are shown to be equal width, they do not correspond to equal width in BDT value.
The distribution of the BDT discriminants for the full 2016--2018 data set in the all-hadronic channel. The sample VR category shown is defined by $N_\textrm{RT}=1$, $N_\textrm{BT} \geq 1$, $H_T > 1400$ GeV. The background from QCD multijet and $\textrm{t}\bar{\textrm{t}}$ production is derived from control regions in the data. Estimates for the $\textrm{t}\bar{\textrm{t}}\textrm{t}\bar{\textrm{t}}$ signal and other backgrounds are shown using simulated samples. The hatched bands correspond to the estimated total uncertainty.