AtlasPublic Web 
Internal TWiki
|
Measurement of the $t\bar{t}t\bar{t}$ production cross section in pp collisions at $\sqrt{s} = 13$ TeV with the ATLAS detector ATLAS-CONF-2021-013
|
||
| Content | Preview | |
|---|---|---|
| Main document (CDS record), Physics Briefing - internal | pdf from CDS | |
| Figures Tables | - | |
| Abstract | ||
| A measurement of four-top-quark production using proton-proton collision data at a centre-of-mass energy of 13 TeV collected by the ATLAS detector at the Large Hadron Collider with an integrated luminosity of 139 fb$^{-1}$ is presented. Events are selected if they contain a single lepton (electron or muon) or an opposite-sign lepton pair, in association with multiple jets. The events are categorised according to the number of jets and how likely these are to contain b-hadrons. A multivariate technique is then used to discriminate between signal and background events. The measured four-top-quark production cross section is found to be $26^{+17}_{-15}$ fb, with a corresponding observed (expected) significance of 1.9 (1.0) standard deviations over the background-only hypothesis. The result is combined with the previous measurement performed by the ATLAS Collaboration in the multilepton final state. The combined four-top-quark production cross section is measured to be $24^{+7}_{-6}$ fb, with a corresponding observed (expected) signal significance of 4.7 (2.6) standard deviations over the background-only predictions. It is consistent within 2.0 standard deviations with the Standard Model expectation of $12.0\pm2.4$ fb. | ||
| Figures | ||
|
Figure 01a: Examples of tree-level Feynman diagrams for SM $t\bar{t}t\bar{t}$ signal (left and middle) and one of the main backgrounds, $t\bar{t}$ production in association with b-jets (right). png (50kB) pdf (24kB) |
 |
|
|
Figure 01b: Examples of tree-level Feynman diagrams for SM $t\bar{t}t\bar{t}$ signal (left and middle) and one of the main backgrounds, $t\bar{t}$ production in association with b-jets (right). png (50kB) pdf (29kB) |
 |
|
|
Figure 01c: Examples of tree-level Feynman diagrams for SM $t\bar{t}t\bar{t}$ signal (left and middle) and one of the main backgrounds, $t\bar{t}$ production in association with b-jets (right). png (98kB) pdf (46kB) |
 |
|
|
Figure 02a: Schematic view of the event categorisation in the 1L channel (left) and 2LOS channel (right). The axes represent the jet multiplicity and the b-tagging requirements defined in Table 1. 3bL (3bH) refers to the b-tagging requirement that select events with lower (higher) purity of truth b-jets amongst the three jets tagged at 70% OP. 3bV refers to the b-tagging requirement used to define the validation regions. png (136kB) pdf (54kB) |
 |
|
|
Figure 02b: Schematic view of the event categorisation in the 1L channel (left) and 2LOS channel (right). The axes represent the jet multiplicity and the b-tagging requirements defined in Table 1. 3bL (3bH) refers to the b-tagging requirement that select events with lower (higher) purity of truth b-jets amongst the three jets tagged at 70% OP. 3bV refers to the b-tagging requirement used to define the validation regions. png (112kB) pdf (46kB) |
 |
|
|
Figure 03a: Relative contribution from the signal and backgrounds in all signal, control and validation regions in the 1L channel (left) and 2LOS channel (right). 3bL (3bH) refers to the b-tagging requirement that select events with lower (higher) purity of truth b-jets amongst the three jets tagged at 70% OP. 3bV refers to the b-tagging requirement used to define the validation regions. For the $t\bar{t}$+jets background, the fraction is shown for each component with the finer classification. The $t\bar{t}$$t\bar{t}$ signal is normalized to the SM cross-section prediction. png (89kB) pdf (16kB) |
 |
|
|
Figure 03b: Relative contribution from the signal and backgrounds in all signal, control and validation regions in the 1L channel (left) and 2LOS channel (right). 3bL (3bH) refers to the b-tagging requirement that select events with lower (higher) purity of truth b-jets amongst the three jets tagged at 70% OP. 3bV refers to the b-tagging requirement used to define the validation regions. For the $t\bar{t}$+jets background, the fraction is shown for each component with the finer classification. The $t\bar{t}$$t\bar{t}$ signal is normalized to the SM cross-section prediction. png (84kB) pdf (16kB) |
 |
|
|
Figure 04a: The Njets and HTall distributions in the region with ≥ 8 jets and ≥ 3 b-jets in the 1L channel before (left) and after (right) the flavour rescaling and the sequential kinematic reweighting. The band includes the total uncertainty of the MC prediction. The ratio of the data to the total MC expectation is shown in the lower panel. The last bin in all distributions includes the overflow. png (126kB) pdf (19kB) |
 |
|
|
Figure 04b: The Njets and HTall distributions in the region with ≥ 8 jets and ≥ 3 b-jets in the 1L channel before (left) and after (right) the flavour rescaling and the sequential kinematic reweighting. The band includes the total uncertainty of the MC prediction. The ratio of the data to the total MC expectation is shown in the lower panel. The last bin in all distributions includes the overflow. png (117kB) pdf (19kB) |
 |
|
|
Figure 04c: The Njets and HTall distributions in the region with ≥ 8 jets and ≥ 3 b-jets in the 1L channel before (left) and after (right) the flavour rescaling and the sequential kinematic reweighting. The band includes the total uncertainty of the MC prediction. The ratio of the data to the total MC expectation is shown in the lower panel. The last bin in all distributions includes the overflow. png (148kB) pdf (20kB) |
 |
|
|
Figure 04d: The Njets and HTall distributions in the region with ≥ 8 jets and ≥ 3 b-jets in the 1L channel before (left) and after (right) the flavour rescaling and the sequential kinematic reweighting. The band includes the total uncertainty of the MC prediction. The ratio of the data to the total MC expectation is shown in the lower panel. The last bin in all distributions includes the overflow. png (136kB) pdf (20kB) |
 |
|
|
Figure 05a: Pre-fit comparison between data and prediction for the distributions of the sum of the pseudo-continuous b-tagging score over the six jets with the highest score in the event for the 1L channel (left) and the 2LOS channel (right). The band includes the total uncertainty of the pre-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total pre-fit expectation is shown in the lower panel. The first and last bins contain underflow and overflow events, respectively. png (160kB) pdf (22kB) |
 |
|
|
Figure 05b: Pre-fit comparison between data and prediction for the distributions of the sum of the pseudo-continuous b-tagging score over the six jets with the highest score in the event for the 1L channel (left) and the 2LOS channel (right). The band includes the total uncertainty of the pre-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total pre-fit expectation is shown in the lower panel. The first and last bins contain underflow and overflow events, respectively. png (166kB) pdf (22kB) |
 |
|
|
Figure 06a: Comparison of predicted and observed event yields in each control and signal region in the 1L channel (top) and in the 2LOS channel (bottom) before the fit (left) and after the fit (right). The band includes the total pre- and post-fit uncertainties. The dashed red line shows the signal distribution normalised to the total background yield. The ratio of the data to the total pre- or post-fit prediction is shown in the lower panel. png (121kB) pdf (21kB) |
 |
|
|
Figure 06b: Comparison of predicted and observed event yields in each control and signal region in the 1L channel (top) and in the 2LOS channel (bottom) before the fit (left) and after the fit (right). The band includes the total pre- and post-fit uncertainties. The dashed red line shows the signal distribution normalised to the total background yield. The ratio of the data to the total pre- or post-fit prediction is shown in the lower panel. png (107kB) pdf (20kB) |
 |
|
|
Figure 06c: Comparison of predicted and observed event yields in each control and signal region in the 1L channel (top) and in the 2LOS channel (bottom) before the fit (left) and after the fit (right). The band includes the total pre- and post-fit uncertainties. The dashed red line shows the signal distribution normalised to the total background yield. The ratio of the data to the total pre- or post-fit prediction is shown in the lower panel. png (110kB) pdf (20kB) |
 |
|
|
Figure 06d: Comparison of predicted and observed event yields in each control and signal region in the 1L channel (top) and in the 2LOS channel (bottom) before the fit (left) and after the fit (right). The band includes the total pre- and post-fit uncertainties. The dashed red line shows the signal distribution normalised to the total background yield. The ratio of the data to the total pre- or post-fit prediction is shown in the lower panel. png (98kB) pdf (20kB) |
 |
|
|
Figure 07a: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 1L channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (148kB) pdf (21kB) |
 |
|
|
Figure 07b: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 1L channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (133kB) pdf (19kB) |
 |
|
|
Figure 07c: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 1L channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (154kB) pdf (23kB) |
 |
|
|
Figure 07d: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 1L channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (137kB) pdf (21kB) |
 |
|
|
Figure 07e: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 1L channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (139kB) pdf (21kB) |
 |
|
|
Figure 07f: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 1L channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (129kB) pdf (19kB) |
 |
|
|
Figure 08a: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 2LOS channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (138kB) pdf (21kB) |
 |
|
|
Figure 08b: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 2LOS channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (145kB) pdf (21kB) |
 |
|
|
Figure 08c: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 2LOS channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (141kB) pdf (21kB) |
 |
|
|
Figure 08d: Comparison between data and post-fit prediction for the distributions of the BDT score in each signal region in the 2LOS channel. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (135kB) pdf (20kB) |
 |
|
|
Figure 09a: Comparison between data and prediction after the fit (`Post-Fit') for the distributions of the BDT score in each validation region. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (154kB) pdf (22kB) |
 |
|
|
Figure 09b: Comparison between data and prediction after the fit (`Post-Fit') for the distributions of the BDT score in each validation region. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (143kB) pdf (21kB) |
 |
|
|
Figure 09c: Comparison between data and prediction after the fit (`Post-Fit') for the distributions of the BDT score in each validation region. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (139kB) pdf (21kB) |
 |
|
|
Figure 09d: Comparison between data and prediction after the fit (`Post-Fit') for the distributions of the BDT score in each validation region. The band includes the total uncertainty of the post-fit computation. The dashed red line shows the signal distribution normalised to the background yield. The ratio of the data to the total post-fit computation is shown in the lower panel. png (142kB) pdf (20kB) |
 |
|
|
Figure 10: Observed and expected event yields as a function of log10(S/B), where S and B are the post-fit signal and total background yields, respectively. The bins in all fitted regions are ordered and grouped in bins of log10(S/B). The signal is shown for both the best-fit signal strength, μ=2.2, and the SM prediction, μ=1.0. The lower panel shows the ratio of the data to the post-fit background prediction, compared to the signal-plus-background prediction with the best-fit signal strength and the SM prediction. The shaded band represents the total uncertainty on the background prediction. png (127kB) pdf (18kB) |
 |
|
|
Figure 11: The nuisance parameters ranked according to their post-fit impacts on the best-fit value of μ. Only the 20 nuisance parameters with the largest impacts are shown. The empty (solid) blue rectangles illustrate the pre-fit (post-fit) impacts, corresponding to the top axis. The pre-fit (post-fit) impact of each nuisance parameter, Δμ, is calculated as the difference in the fitted value of μ between the nominal fit and the fit when fixing the corresponding nuisance parameter to θ̂ ± Δθ (θ̂ ± Δθ̂), where θ̂ is the best-fit value of the nuisance parameter and Δθ (Δθ̂) is its pre-fit (post-fit) uncertainty. The black points show the best-fit values of the nuisance parameters, with the error bars representing the post-fit uncertainties. Each nuisance parameter is shown relative to its nominal value, θ0, and in units of its pre-fit uncertainty. png (98kB) pdf (18kB) |
 |
|
|
Figure 12: Fitted signal strength in the signal-plus-background fit from the dataset for the individual channels (1L/2LOS and 2LSS/3L) and the combined signal strength from all $t\bar{t}t\bar{t}$ analysis regions. png (45kB) pdf (15kB) |
 |
|
| Tables | ||
|
Table 01: Summary of the b-tagging requirements for the event categorisation. Nb60%, Nb70% and Nb85% are defined as the number of b-tagged jets using respectively the b-tagging operating points with average expected efficiencies of 60%, 70% and 85%. 3bL (3bH) refers to the requirement that select events with lower (higher) purity of truth b-jets amongst the three jets tagged at 70% OP. 3bV refers to the requirement used to define the validation regions. png (11kB) pdf (44kB) |
 |
|
|
Table 02: Summary of the sources of systematic uncertainty for the $t\bar{t}$+jets modelling. The last column of the table lists the uncorrelated components of each systematic source. All systematic uncertainty sources are treated as uncorrelated across the $t\bar{t}$+jets components. For generator and PS choices, each $t\bar{t}$+jets component is further decomposed into a shape and a migration component. The number of uncorrelated components for each physics source is shown in the brackets. png (35kB) pdf (53kB) |
 |
|
|
Table 03: The contribution from different systematic uncertainties to the measured $t\bar{t}t\bar{t}$ production cross section, σ$t\bar{t}t\bar{t}$, grouped into categories. For each uncertainty source, the fit is repeated with the corresponding group of nuisance parameters fixed to their best-fit values, θ̂. The contribution from each source, Δσ$t\bar{t}t\bar{t}$, is then evaluated by subtracting in quadrature the uncertainty on σ$t\bar{t}t\bar{t}$ obtained in this fit from that of the full fit. The contributions from individual groups are compared to the total systematic uncertainty and the statistical uncertainty. The total systematic uncertainty is different from the sum in quadrature of the different groups due to correlations among nuisance parameters in the fit. png (36kB) pdf (61kB) |
 |
|
