The 95% CL limits on the cross-section $\sigma(pp \rightarrow Z' \rightarrow \chi \chi$) times branching ratio B in fb with all statistical and systematic uncertainties, for the $R_{\text{inv}}=$0.6 signal points.

Data yield in the PFN signal region.

Yield of data in the ANTELOPE signal region in the binning used for BumpHunter fitting.

Observed profile likelihood as a function of $\sigma\times\mathcal{B}_{H\rightarrow WW^{*}}$ normalised by the SM expectation for the single-channel measurements

Two-dimensional likelihood scan of the measured values of $\sigma_{ZH}\times\mathcal{B}_{H\rightarrow WW^{*}}$ vs. $\sigma_{WH}\times\mathcal{B}_{H\rightarrow WW^{*}}$.

The observed values of the background normalisation factors for the combined 2-POI fit. The uncertainties correspond to the total of all statistical and systematic sources.

Measured cross sections times the $H\rightarrow WW^{*}$ branching ratio for the $p_T^V$ scheme.

Measured cross sections times the $H\rightarrow WW^{*}$ branching ratio for the STXS scheme. Note: the uncertainties of "-0" are due to the confidence interval reaching the minimum allowed value of the POI.

Correlation matrix of the POIs and normalisation factors for the 2-POI inclusive analysis

Correlation matrix of the parameters of interest, normalisation factors and nuissance parameters for the $p_T^V$ scheme

Correlation matrix of the parameters of interest, normalisation factors and nuissance parameters for the STXS scheme

Event data from Z-dominated SR

Distributions for data and expected signal and background contributions of the MVA score for the µ + jets channel in the 3j1b category, before the maximum likelihood fit. The vertical error bars represent the statistical uncertainty in the data, and the shaded band the uncertainty in the prediction. All uncertainties considered in the analysis are included in the uncertainty band. The lower panels show the data-to-prediction ratio. The first and last bins in each distribution include underflow and overflow events, respectively.

Observed and predicted number of events in each of the eight categories of the signal region, before the maximum likelihood fit. The vertical error bars represent the statistical uncertainty in the data, and the shaded band the uncertainty in the prediction. All uncertainties considered in the analysis are included in the uncertainty band. The lower panels show the data-to-prediction ratio.

Distributions for the e + jets final state before the maximum likelihood fit, MVA score bins for the 3j1b category and ∆R med ( j, j 0 ) bins for the other categories. The vertical error bars represent the statistical uncertainty in the data, and the shaded band the uncertainty of the prediction. All uncertainties considered in the analysis are included in the uncertainty band. The lower panels show the data-to-simulation ratio, where the simulations are adjusted according to their normalization before the fit. The first and last bins in each distribution include underflow and overflow events, respectively.

Distributions for the e + jets final state after the maximum likelihood fit, MVA score bins for the 3j1b category and ∆R med ( j, j 0 ) bins for the other categories. The vertical error bars represent the statistical uncertainty in the data, and the shaded band the uncertainty of the prediction. All uncertainties considered in the analysis are included in the uncertainty band. The lower panels show the data-to-simulation ratio, where the simulations are adjusted according to their normalization after the fit. The first and last bins in each distribution include underflow and overflow events, respectively.

Distributions for the µ + jets final state before the maximum likelihood fit, MVA score bins for the 3j1b category and ∆R med ( j, j 0 ) bins for the other categories. The vertical error bars represent the statistical uncertainty in the data, and the shaded band the uncertainty of the prediction. All uncertainties considered in the analysis are included in the uncertainty band. The lower panels show the data-to-simulation ratio, where the simulations are adjusted according to their normalization before the fit. The first and last bins in each distribution include underflow and overflow events, respectively.

Distributions for the µ + jets final state after the maximum likelihood fit, MVA score bins for the 3j1b category and ∆R med ( j, j 0 ) bins for the other categories. The vertical error bars represent the statistical uncertainty in the data, and the shaded band the uncertainty of the prediction. All uncertainties considered in the analysis are included in the uncertainty band. The lower panels show the data-to-simulation ratio, where the simulations are adjusted according to their normalization after the fit. The first and last bins in each distribution include underflow and overflow events, respectively.

Covariance matrix for the lepton+jets ML fit.

Covariance matrix for the combination ML fit.

Distributions in $S_T$ in the SR for the electron channel, after a background-only fit to the data. The signal distributions are scaled to the cross section predicted by the theory. The hatched bands show the post-fit uncertainty band, combining all sources of uncertainty. The ratio of data to the background predictions is shown in the panels below the distributions.

Distributions of the $\mathrm{NN^{SvsB}}$ output for data and the expected background after the likelihood fit in the $SR_{cQu --}$ signal region. The post-fit background expectations are shown as filled histograms, the combined pre-fit background expectations are shown as dashed lines. The signal distribution using the Wilson coefficient values $c_{tu}^{(1)}=0.04$, $c_{Qu}^{(1)}=0.1$, $c_{Qu}^{(8)}=0.1$ is shown with a dotted line, normalized to the same number of events as the background.

Summary of the observed and predicted number of events in the five 2$\ell$ control regions. The background prediction is shown after the combined likelihood fit to data under the signal-plus-background hypothesis across all control regions and signal regions. The uncertainties in the total yields are smaller than the sum in quadrature of the uncertainties in the individual contributions due to anti-correlations resulting from the likelihood fit. Int Conv refers to events with photon conversions in the inner detector, while Mat Conv refers to events with photon conversions in the detector material. QMisID refers to events with charge-misidentified lepton.

Summary of the observed and predicted number of events in the four 3$\ell$ control regions. The background prediction is shown after the combined likelihood fit to data under the signal-plus-background hypothesis across all control regions and signal regions. The uncertainties in the total yields are smaller than the sum in quadrature of the uncertainties in the individual contributions due to anti-correlations resulting from the likelihood fit. Int Conv refers to events with photon conversions in the inner detector, while Mat Conv refers to events with photon conversions in the detector material. QMisID refers to events with charge-misidentified lepton.

Comparison of the observed limits on the three WCs to the limit obtained by the ATLAS Run~1 analysis of events with $b$-tagged jets and a same-sign lepton pair [JHEP10(2015)150]. The cross-section limits from that analysis are converted to the limits on the WCs by using the fitted EFT parametrization. The bounds on the WCs are shown at the 68$\%$ CL (solid) and/or 95$\%$ CL (dashed) levels. For the ATLAS Run~1 analysis only the bounds at 95$\%$ CL are available. The vertical bar represents the SM prediction. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$.

Observed lower limits at 95$\%$ confidence level on the scale of new physics $\Lambda$ for Wilson coefficient values of 0.01, 1 and $4\pi^2$. The limits are compared with the limits obtained by the ATLAS Run~1 analysis of events with $b$-tagged jets and a same-sign lepton pair.

Observed and expected limits at 95$\%$ confidence level on the cross-section of $\sigma(pp \rightarrow tt)$.

2D likelihood scans of the Wilson coefficients $c_{tu}^{(1)}$ versus $c_{Qu}^{(1)}$. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$. In the table the $2\Delta$NLL values are shown which correspond to the minimized negative log likelihood for a given Wilson coefficient pair.

2D likelihood scans of the Wilson coefficients $c_{tu}^{(1)}$ versus $c_{Qu}^{(8)}$. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$. In the table the $2\Delta$NLL values are shown which correspond to the minimized negative log likelihood for a given Wilson coefficient pair.

2D likelihood scans of the Wilson coefficients $c_{Qu}^{(1)}$ versus $c_{Qu}^{(8)}$. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$. In the table the $2\Delta$NLL values are shown which correspond to the minimized negative log likelihood for a given Wilson coefficient pair.

1D-likelihood scan for $c_{tu}^{(1)}$. The observed and expected limits at 68\% and 95\% confidence level (CL) can be read off by finding the intersection of the likelihood scan curve with the 1$\sigma$ and 2$\sigma$ lines. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$. In the table the $2\Delta$NLL values are shown which correspond to the minimized negative log likelihood for a given Wilson coefficient value.

1D-likelihood scan for $c_{Qu}^{(1)}$. The observed and expected limits at 68\% and 95\% confidence level (CL) can be read off by finding the intersection of the likelihood scan curve with the 1$\sigma$ and 2$\sigma$ lines. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$. In the table the $2\Delta$NLL values are shown which correspond to the minimized negative log likelihood for a given Wilson coefficient value.

1D-likelihood scan for $c_{Qu}^{(8)}$. The observed and expected limits at 68\% and 95\% confidence level (CL) can be read off by finding the intersection of the likelihood scan curve with the 1$\sigma$ and 2$\sigma$ lines. The new-physics scale is set to $\Lambda = 1 \, \text{TeV}$. In the table the $2\Delta$NLL values are shown which correspond to the minimized negative log likelihood for a given Wilson coefficient value.

The cutflow table for $N1$ and $N2$ signal samples for different mass points. The samples consist of both $ee$ and $\mu\mu$ final states in equal ratio. Each sample is normalized to a total cross-section of 0.1 pb. The abbreviations used in the table are, SC = Same Charge, SF = Same Flavor.

The cutflow table for $N3$ signal samples for different mass points. The samples consist of both $ee$ and $\mu\mu$ final states in equal ratio. Each sample is normalized to a total cross-section of 10 pb. The abbreviations used in the table are, SS = Same Charge, SF = Same Flavor.

Expected and observed upper limits on the strength of HNL mixing with electron neutrinos

Expected and observed upper limits on the strength of HNL mixing with muon neutrinos

Expected and observed upper limits on the strength of HNL mixing with tau neutrinos

Table with post-fit yields in the four regions used in the measurement. The correlations of the nuisance parameters, as obtained by the fit, are considered for the calculation of the uncertainties.

Ratio of the $t\bar{t}$ to the $Z$-boson cross section compared to the predictions for several sets of parton distribution functions. The total cross section for $t\bar{t}$ production and the fiducial cross section for $Z$-boson production are used. The uncertainty on the prediction represents the PDF+$\alpha_{s}$ variations.

Nuisance parameter ranking plot for the $t\bar{t}$ signal-strength. Only the 10 most important nuisance parameters are shown. The impact of each nuisance parameter on the measured cross section, is defined as the shift induced in measured cross-section as the nuisance parameter is shifted by its pre-fit and post-fit uncertainties from its nominal best-fit value and then fixed while all other nuisance parameters profiled.

Nuisance parameter ranking plot for the ratio of the $t\bar{t}$ and $Z$-boson signal-strengths. Only the 10 most important nuisance parameters are shown. The impact of each nuisance parameter on the measured ratio, is defined as the shift induced in measured ratio as the nuisance parameter is shifted by its pre-fit and post-fit uncertainties from its nominal best-fit value and then fixed while all other nuisance parameters profiled.

Nuisance parameter ranking plot for the $Z$-boson signal-strength. Only the 10 most important nuisance parameters are shown. The impact of each nuisance parameter on the measured cross section, is defined as the shift induced in measured cross-section as the nuisance parameter is shifted by its pre-fit and post-fit uncertainties from its nominal best-fit value and then fixed while all other nuisance parameters profiled.

Distribution of ${E}_{T}^{miss}$ after the fit of the multijet backgrounds, in the electron channel, in the $t\bar{t}$ 3-jets VR, without applying the cut on ${E}_{T}^{miss}$. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit. The last bin includes the overflow. The uncertainty band indicates the simulation's statistical uncertainty, the normalisation uncertainties for different processes ($40$ % for $W$+jets production, $30$ % for multijet background and $6$ % for top-quark processes) and the multijet background shape uncertainty in each bin, summed in quadrature. The lower panel of the figure shows the ratio of the data to the prediction.

Distribution of ${E}_{T}^{miss}$ after the fit of the multijet backgrounds, in the electron channel, in the $t\bar{t}$ 4-jets VR, without applying the cut on ${E}_{T}^{miss}$. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit. The last bin includes the overflow. The uncertainty band indicates the simulation's statistical uncertainty, the normalisation uncertainties for different processes ($40$ % for $W$+jets production, $30$ % for multijet background and $6$ % for top-quark processes) and the multijet background shape uncertainty in each bin, summed in quadrature. The lower panel of the figure shows the ratio of the data to the prediction.

Distribution of $m_{T}^{W}$ after the fit of the multijet backgrounds, in the muon channel, in the signal region, without applying the cut on $m_{T}^{W}$. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit. The last bin includes the overflow. The uncertainty band indicates the simulation's statistical uncertainty, the normalisation uncertainties for different processes ($40$ % for $W$+jets production, $30$ % for multijet background and $6$ % for top-quark processes) and the multijet background shape uncertainty in each bin, summed in quadrature. The lower panel of the figure shows the ratio of the data to the prediction.

Distribution of $m_{T}^{W}$ after the fit of the multijet backgrounds, in the muon channel, in the $W$+jets VR, without applying the cut on $m_{T}^{W}$. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit. The last bin includes the overflow. The uncertainty band indicates the simulation's statistical uncertainty, the normalisation uncertainties for different processes ($40$ % for $W$+jets production, $30$ % for multijet background and $6$ % for top-quark processes) and the multijet background shape uncertainty in each bin, summed in quadrature. The lower panel of the figure shows the ratio of the data to the prediction.

Distribution of $m_{T}^{W}$ after the fit of the multijet backgrounds, in the muon channel, in the $t\bar{t}$ 3-jets VR, without applying the cut on $m_{T}^{W}$. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit. The last bin includes the overflow. The uncertainty band indicates the simulation's statistical uncertainty, the normalisation uncertainties for different processes ($40$ % for $W$+jets production, $30$ % for multijet background and $6$ % for top-quark processes) and the multijet background shape uncertainty in each bin, summed in quadrature. The lower panel of the figure shows the ratio of the data to the prediction.

Distribution of $m_{T}^{W}$ after the fit of the multijet backgrounds, in the muon channel, in the $t\bar{t}$ 4-jets VR, without applying the cut on $m_{T}^{W}$. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit. The last bin includes the overflow. The uncertainty band indicates the simulation's statistical uncertainty, the normalisation uncertainties for different processes ($40$ % for $W$+jets production, $30$ % for multijet background and $6$ % for top-quark processes) and the multijet background shape uncertainty in each bin, summed in quadrature. The lower panel of the figure shows the ratio of the data to the prediction.

Expected distributions of the MEM discriminant $P(S|X)$ in the SR, for the s-channel single-top signal, and for the $t\bar{t}$ and $W$+jets backgrounds, for MEM discriminant values larger than $2.0\times10^{-4}$. Each distribution is normalised to unity. The binning is the same as the optimised binning used in the signal extraction fit, resulting in a non-linear horizontal scale.

Distribution of the MEM discriminant $P(S|X)$ in the $W$+jets VR. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit presented in Section 5 in the paper. The uncertainty band indicates the simulation's statistical uncertainty and the normalisation uncertainties for the various processes in each bin, summed in quadrature. The ratio of the observed number to the predicted number of events in each bin is shown in the lower panel of the figure, with different vertical axis ranges. The binning is the same as the optimised binning used in the signal extraction fit described in Section 8 in the paper, resulting in a non-linear horizontal scale.

Distribution of the MEM discriminant $P(S|X)$ in the $t\bar{t}$ 3-jets VR. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit presented in Section 5 in the paper. The uncertainty band indicates the simulation's statistical uncertainty and the normalisation uncertainties for the various processes in each bin, summed in quadrature. The ratio of the observed number to the predicted number of events in each bin is shown in the lower panel of the figure, with different vertical axis ranges. The binning is the same as the optimised binning used in the signal extraction fit described in Section 8 in the paper, resulting in a non-linear horizontal scale.

Distribution of the MEM discriminant $P(S|X)$ in the $t\bar{t}$ 4-jets VR. Simulated events are normalised to the expected number of events given the integrated luminosity, after applying the normalisation factors obtained in the multijet fit presented in Section 5 in the paper. The uncertainty bands indicate the simulation's statistical uncertainty and the normalisation uncertainties for the various processes in each bin, summed in quadrature. The ratio of the observed number to the predicted number of events in each bin is shown in the lower panel of the figure, with different vertical axis ranges. The binning is the same as the optimised binning used in the signal extraction fit described in Section 8 in the paper, resulting in a non-linear horizontal scale.

Distribution of the MEM discriminant $P(S|X)$ in the SR before the fit to data, for MEM discriminant values larger than $2.0\times10^{-4}$. The lower panel of the figure shows the ratio of the data to the prediction, with different vertical axis ranges. The uncertainty band indicates the total uncertainties and their correlations in each bin. The uncertainties in the $t\bar{t}$ and $W$+jets normalisation factors, as well as in the s-channel signal cross-section, are not defined pre-fit and therefore not included. The binning is the same as the optimised binning used in the fit, resulting in a non-linear horizontal scale.

Distribution of the MEM discriminant $P(S|X)$ in the SR after the fit to data, for MEM discriminant values larger than $2.0\times10^{-4}$. The lower panel of the figure shows the ratio of the data to the prediction, with different vertical axis ranges. The uncertainty band indicates the total uncertainties and their correlations in each bin. The binning is the same as the optimised binning used in the fit, resulting in a non-linear horizontal scale.

Distribution of the MEM discriminant $P(S|X)$ in the SR after the fit to data, for MEM discriminant values larger than $2.0\times10^{-4}$, after subtraction of all backgrounds. The fitted distribution for the simulation of the signal is shown together with the post-fit uncertainty in the backgrounds. The binning is the same as the optimised binning used in the fit, resulting in a non-linear horizontal scale.

Pre-fit and post-fit event yields in the SR, for MEM discriminant values larger than $2.0\times10^{-4}$. The central value of the event yield for each process is calculated by summing the values of the discriminant bin contents, using the nominal expected yield for the pre-fit value, and the best-fit estimate for the post-fit value. The error includes statistical and systematic uncertainties summed in quadrature. All sources of systematic uncertainties are included, taking into account correlations and anti-correlations in the post-fit case. The uncertainties in the $t\bar{t}$ and $W$+jets normalisation factors, as well as in the s-channel signal cross-section, are not defined pre-fit and therefore only included in the post-fit uncertainties.

Observed impact of the different sources of uncertainty on the measured s-channel signal cross-section, grouped by categories. The impact of each category is obtained by repeating the fit after having fixed the set of nuisance parameters corresponding to that category, subtracting the square of the resulting uncertainty from the square of the uncertainty found in the full fit, and calculating the square root. The 'Systematic uncertainties' category combines all sources of systematic uncertainties. The statistical uncertainty is obtained by repeating the fit after having fixed all nuisance parameters, including the $t\bar{t}$ and $W$+jets normalisation factors. 'Total' gives the total uncertainty on the measurement.

Observed impact of the different sources of $t\bar{t}$ modelling uncertainty on the measured s-channel signal cross-section. The impact of each category is obtained by repeating the fit after having fixed the set of nuisance parameters corresponding to that category, subtracting the square of the resulting uncertainty from the square of the uncertainty found in the full fit, and calculating the square root. 'PS & had.' refers to the parton shower and hadronisation model, and 'ME/PS matching' to the matching of the ME to the parton shower.

Observed impact of the different sources of s-channel modelling uncertainty on the measured s-channel signal cross-section. The impact of each category is obtained by repeating the fit after having fixed the set of nuisance parameters corresponding to that category, subtracting the square of the resulting uncertainty from the square of the uncertainty found in the full fit, and calculating the square root. 'PS & had.' refers to the parton shower and hadronisation model, as described in Section 7 in the paper.

Observed impact of the different sources of t-channel modelling uncertainty on the measured s-channel signal cross-section. The impact of each category is obtained by repeating the fit after having fixed the set of nuisance parameters corresponding to that category, subtracting the square of the resulting uncertainty from the square of the uncertainty found in the full fit, and calculating the square root. 'PS & had.' refers to the parton shower and hadronisation model, as described in Section 7 in the paper.

Observed impact of the different sources of $tW$ modelling uncertainty on the measured s-channel signal cross-section, grouped by categories. The impact of each category is obtained by repeating the fit after having fixed the set of nuisance parameters corresponding to that category, subtracting the square of the resulting uncertainty from the square of the uncertainty found in the full fit, and calculating the square root. 'PS & had.' refers to the parton shower and hadronisation model, and '$t\bar{t}$ overlap' to the algorithm removing the overlap between $tW$ and $t\bar{t}$ production at NLO, as described in Section 7 in the paper.

Observed impact of the different sources of PDF uncertainties on the measured s-channel signal cross-section, grouped by categories. The impact of each category is obtained by repeating the fit after having fixed the set of nuisance parameters corresponding to that category, subtracting the square of the resulting uncertainty from the square of the uncertainty found in the full fit, and calculating the square root.

Comparison between data and prediction after the fit to data in the signal region for the leading-jet $p_{T}$. The last bin includes the overflow. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the leading-jet $\eta$. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the subleading-jet $p_{T}$. The last bin includes the overflow. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the subleading-jet $\eta$. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the lepton $p_{T}$. The last bin includes the overflow. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the lepton $\eta$. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the ${E}_{T}^{miss}$. The last bin includes the overflow. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Comparison between data and prediction after the fit to data in the signal region for the $m_{T}^{W}$. The last bin includes the overflow. The uncertainty band includes all uncertainties and their correlations. The lower panel of the figure shows the ratio of the data to the prediction.

Nuisance parameters ranked according to their post-fit impacts on the best-fit value of the ratio $\mu$ of the measured cross-section to the predicted cross-section. In the figure, only the 20 nuisance parameters with the largest post-fit impacts are shown. The empty (solid) blue rectangles illustrate the pre-fit (post-fit) impact on $\mu$, corresponding to the upper axis. The pre-fit (post-fit) impact of each nuisance parameter, $\Delta\mu$, is calculated as the difference in the fitted value of $\mu$ between the nominal fit and the fit when fixing the corresponding nuisance parameter to $\hat{\theta}\pm\Delta\theta$ ($\hat{\theta}\pm\Delta\hat{\theta}$), where $\hat{\theta}$ is the best-fit value of the nuisance parameter and $\Delta\theta$ ($\Delta\hat{\theta}$) is its pre-fit (post-fit) uncertainty. Several systematic uncertainties are split into different nuisance parameters, which are indicated by NP. JES (JER) indicates jet energy scale (resolution), and $\gamma$ indicates a nuisance parameter associated to the MC statistics in one of the 18 bins numbered from 0 to 17. 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 wrt. its nominal value, $\theta_0$, and in units of its pre-fit uncertainty, except the free-floating normalisation factors of the $t\bar{t}$ and $W$+jets backgrounds, and the parameters associated to the MC statistics in each bin, for which the post-fit values and uncertainties are shown.

Correlation matrix of the nuisance parameters and of the ratio $\mu$ of the measured cross-section to the predicted cross-section. The correlations are given after the fit to data. In the figure, only the parameters which have a correlation of at least 0.2 with any other parameter are shown.

Distribution of the MEM discriminant $P(S|X)$ in the SR for MEM discriminant values larger than $2.0\times10^{-4}$, for the collision data used for the measurement, and for 1000 pseudo-data replicas, generated using a bootstrapping technique, in order to assess the statistical correlations between this measurement and others, for the purpose of combinations. The replicas are obtained by reweighting each observed data event by a random integer generated according to Poisson statistics, using the <a href="https://zenodo.org/record/5361038">BootstrapGenerator</a> software package , which implements a technique described in <a href="https://cds.cern.ch/record/2759945/">ATL-PHYS-PUB-2021-011</a>. The ATLAS event number and run number of each event are used as seed to uniquely but reproducibly initialise the random number generator for each event. Each pseudo-data replica is assigned an index, ranging from 0 to 999, corresponding to the random number index used consistently for each observed data event.

Measured values of the signal cross-section and of the $t\bar{t}$ and $W$+jets normalisation factors, obtained by statistical-only fits to the collision data used for the measurement, and to 1000 pseudo-data replicas, generated using a bootstrapping technique, in order to assess the statistical correlations between this measurement and others, for the purpose of combinations. The central values and their statistical uncertainties are obtained by repeating the fit after having fixed all nuisance parameters, except the $t\bar{t}$ and $W$+jets normalisation factors, which are let free-floating (unlike for the statistical uncertainty on the cross-section quoted in the paper). The replicas are obtained by reweighting each observed data event by a random integer generated according to Poisson statistics, using the <a href="https://zenodo.org/record/5361038">BootstrapGenerator</a> software package , which implements a technique described in <a href="https://cds.cern.ch/record/2759945/">ATL-PHYS-PUB-2021-011</a>. The ATLAS event number and run number of each event are used as seed to uniquely but reproducibly initialise the random number generator for each event. Each pseudo-data replica is assigned an index, ranging from 0 to 999, corresponding to the random number index used consistently for each observed data event.

Distribution of an input to the BDT classifier in the $3\ell$ category: The linear discrimminant $p_{T}^{miss,LD}$. The discriminant is defined by the relation $p_{T}^{miss,LD} = 0.6 p_{T}^{miss} + 0.4 H_{T}^{miss}$, where $H_{T}^{miss}$ corresponds to the magnitude of the vector $p_{T}$ sum of all electrons, muons, taus and small radius jets passing the selection critertia, and $p_{T}^{miss}$ represents the missing transverse momentum vector computed as the negative vector $p_{T}$ sum of all the particles reconstructed by the PF algorithm in the event.

Distributions of the transverse mass $m_{T}$ of the lepton that is not originating from the Z boson decay and the missing transverse momentum in the $3\ell$/WZ control region. The distributions expected for the WZ and ZZ as well as for other background processes are shown for the values of nuisance parameters obtained from the ML fit used in the signal extraction.

Distributions of the four lepton invariant mass $m_{4\ell}$ in the $4\ell$/ZZ control region. The distributions expected for the ZZ as well as for other background processes are shown for the values of nuisance parameters obtained from the ML fit used in the signal extraction.

Distributions of the transverse mass $m_{T}$ of the leading lepton and the missing transverse momentum in the $2\ell$(ss) control region. The distributions expected for the misidentified $\ell$ background as well as for other background processes is shown for the values of nuisance parameters obtained from the background-only ML fit, in which the HH signal is constrained to be zero.

Distributions of the reconstructed HH mass $m_{HH}$ in the $2\ell+2 au_{h}$ control region. The distributions expected for the misidentified $\ell/ au_{h}$ background as well as for other background processes is shown for the values of nuisance parameters obtained from the background-only ML fit, in which the HH signal is constrained to be zero.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $2\ell$(ss) category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate signal quantiles.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $3\ell$ category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate signal quantiles.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $4\ell$ category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate signal quantiles.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $3\ell+1 au_{h}$ category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate background quantiles.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $2\ell+2 au_{h}$ category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate background quantiles.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $1\ell+3 au_{h}$ category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate background quantiles.

Distribution in the output of the BDT trained for nonresonant HH production and evaluated for the benchmark scenario JHEP04 BM7 for the $4 au_{h}$ category. The SM HH signal is shown for a cross section amounting to $30$ times the value predicted in the SM. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate background quantiles.

Observed and expected $95\%$ CL upper limits on the SM HH production cross section, obtained for both individual search categories and from a simultaneous fit of all seven categories combined.

Observed and expected $95\%$ CL upper limits on the HH production cross section as a function of the Higgs boson self-coupling strength modifier $\kappa_\lambda$ for the combination of all seven categories. All Higgs boson couplings other than $\lambda$ are assumed to have the values predicted in the SM.

Observed and expected $95\%$ CL upper limits on the HH production cross section as a function of the Higgs boson self-coupling strength modifier $\kappa_\lambda$ for the seven different categories as well as their combination. All Higgs boson couplings other than $\lambda$ are assumed to have the values predicted in the SM.

Observed and expected $95\%$ CL upper limits on the HH production cross section for the twelve benchmark scenarios from doi:10.1007/JHEP04(2016)126, the additional benchmark scenario 8a from doi:10.1007/JHEP09(2018)057, the seven benchmark scenarios from doi:10.1007/JHEP03(2020)091, and for the SM.

Observed and expected $95\%$ CL upper limits on the HH production cross section for the twelve benchmark scenarios from doi:10.1007/JHEP04(2016)126, the additional benchmark scenario 8a from doi:10.1007/JHEP09(2018)057, the seven benchmark scenarios from doi:10.1007/JHEP03(2020)091, and for the SM. The limits are shown for all seven search categories as well as for their combination.

Observed and expected $95\%$ CL upper limits on the HH production cross section as a function of the effective coupling $c_{2}$ for the combination of all seven categories. All Higgs boson couplings other than $c_{2}$ are assumed to have the values predicted in the SM.

Observed and expected $95\%$ CL upper limits on the HH production cross section as a function of the effective coupling $c_{2}$ and the top Yukawa coupling modifier $\kappa_{t}$ for the combination of all seven categories. All Higgs boson couplings other than $c_{2}$ and $\kappa_{t}$ are assumed to have the values predicted in the SM. The position of the SM in the ($c_{2}-\kappa_{t}$) plane, as well as the best fit value of $(c_{2},\kappa_{t})=(1.05, 1.74)$ together with contours for the theory cross section are shown as well.

Observed and expected $95\%$ CL upper limits on the HH production cross section as a function of the Higgs boson self-coupling modifier $\kappa_{\lambda}$ and the top Yukawa coupling modifier $\kappa_{t}$ for the combination of all seven categories. All Higgs boson couplings other than $\kappa_{\lambda}$ and $\kappa_{t}$ are assumed to have the values predicted in the SM. The position of the SM in the ($\kappa_{\lambda}-\kappa_{t}$) plane, as well as the best fit value of $(\kappa_{\lambda},\kappa_{t})=(-3.54, -1.70)$ together with contours for the theory cross section are shown as well.

Observed and expected $95\%$ CL upper limits on the HH production cross section as a function of the effective coupling $c_{2}$ and the Higgs boson self-coupling modifier $\kappa_{\lambda}$ for the combination of all seven categories. All Higgs boson couplings other than $c_{2}$ and $\kappa_{\lambda}$ are assumed to have the values predicted in the SM. The position of the SM in the ($c_{2}-\kappa_{\lambda}$) plane, as well as the best fit value of $(c_{2},\kappa_{\lambda})=(-0.66, 5.33)$ together with contours for the theory cross section are shown as well.

Observed and expected $95\%$ CL upper limits on the production of new particles X of spin $0$ and mass $m_{X}$ in the range $250$-$1000$ GeV, which decay to Higgs boson pairs.

Observed and expected $95\%$ CL upper limits on the production of new particles X of spin $0$ and mass $m_{X}$ in the range $250$-$1000$ GeV, which decay to Higgs boson pairs. The Limit is shown for the seven different search categories as well as their combination.

Observed and expected $95\%$ CL upper limits on the production of new particles X of spin $2$ and mass $m_{X}$ in the range $250$-$1000$ GeV, which decay to Higgs boson pairs.

Observed and expected $95\%$ CL upper limits on the production of new particles X of spin $2$ and mass $m_{X}$ in the range $250$-$1000$ GeV, which decay to Higgs boson pairs. The Limit is shown for the seven different search categories as well as their combination.

Distribution in the output of the BDT trained for resonances of spin 2 and mass $750\,$GeV production for the $2\ell$(ss) category. The resonant HH signal is shown for a cross section amounting to $1\,$pb. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate signal quantiles.

Distribution in the output of the BDT trained for resonances of spin 2 and mass $750\,$GeV production for the $3\ell$ category. The resonant HH signal is shown for a cross section amounting to $1\,$pb. The distributions expected for the background processes are shown for the values of nuisance parameters obtained from the ML fit of the signal+background hypothesis to the data. The binning is chosen to approximate signal quantiles.