Transverse dilepton-$p_{T}^{miss}$ mass pre-fit distribution for selected events in SR2 of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Dilepton mass pre-fit distribution for selected events in SR3 of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Dilepton $p_T$ pre-fit distribution for selected events in SR3 of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Dijet $p_T$ pre-fit distribution for selected events in the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Lepton-dijet $p_T$ pre-fit distribution for selected events in the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Maximum $p_T$ of the visible particles divided by the missing $p_T$, for selected events in the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between jets, for selected events in the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the lepton and missing $p_T$, for selected events in the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the lepton and the dijet system, for selected events in the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta R$ distance between the leptons, for selected events in the non-prompt validation region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the di-lepton system and the missing $p_T$, for selected events in the non-prompt validation region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the lepton and the di-jet system, for selected events in the non-prompt validation region of the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the lepton-dijet system and the missing $p_T$, for selected events in the non-prompt validation region of the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Leading lepton $p_T$ pre-fit distribution for selected events in the top control region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Trailing lepton $p_T$ pre-fit distribution for selected events in the top control region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Dilepton mass pre-fit distribution for selected events in the Drell-Yan control region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Missing $p_T$ pre-fit distribution for selected events in the Drell-Yan control region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Missing $p_T$ pre-fit distribution for selected events in the $WW$ control region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Dilepton $p_T$ pre-fit distribution for selected events in the $WW$ control region of the di-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Minimum $p_T$ of the visible particles divided by the missing $p_T$, for selected events in the top control region of the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between jets, for selected events in the top control region of the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the lepton and missing $p_T$, for selected events in the $W$+jets control region of the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Pre-fit distribution of the $\Delta \phi$ angle between the lepton and the dijet system, for selected events in the $W$+jets control region of the semi-leptonic channel for the full dataset. The error bars on the data points represent the statistical uncertainty of the data, and the error bars on the predicted yields represent the combined systematic and statistical uncertainty in each bin. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$. The last bin includes the overflow.

Unrolled $m_{ll}-m_{T}^{l min, p_{T}^{miss}}$ post-fit distributions in the di-leptonic channel in SR1 for the full data set. The histogram bins are spaced uniformly. Each group of five bins (from left to right) corresponds to $m_{T}^{l min, p_{T}^{miss}}$ distribution in a $m_{ll}$ region, placed in ascending order. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$.

Unrolled $m_{ll}-m_{T}^{l min, p_{T}^{miss}}$ post-fit distributions in the di-leptonic channel in SR2 for the full data set. The histogram bins are spaced uniformly. Each group of five bins (from left to right) corresponds to $m_{T}^{l min, p_{T}^{miss}}$ distribution in a $m_{ll}$ region, placed in ascending order. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$.

Unrolled $m_{ll}-m_{T}^{l min, p_{T}^{miss}}$ post-fit distributions in the di-leptonic channel in SR3 for the full data set. The histogram bins are spaced uniformly. Each group of five bins (from left to right) corresponds to $m_{T}^{l min, p_{T}^{miss}}$ distribution in a $m_{ll}$ region, placed in ascending order. The signal prediction represents the mass point: $m_s = 160 GeV, m_{\chi} = 100 GeV, m_{Z'} = 500 GeV$.

Post-fit BDT distributions in the semi-leptonic channel for the full data set in the top control region. The signal prediction represents the mass point: $m_{s} = 160$ GeV, $m_{\chi} = 100$ GeV, $m_{Z'} = 500$ GeV

Post-fit BDT distributions in the semi-leptonic channel for the full data set in the $W$+jets control region. The signal prediction represents the mass point: $m_{s} = 160$ GeV, $m_{\chi} = 100$ GeV, $m_{Z'} = 500$ GeV

Post-fit BDT distributions in the semi-leptonic channel in the 2016 signal region. The signal prediction represents the mass point: $m_{s} = 160$ GeV, $m_{\chi} = 100$ GeV, $m_{Z'} = 500$ GeV

Post-fit BDT distributions in the semi-leptonic channel in the 2017+2018 signal region. The signal prediction represents the mass point: $m_{s} = 160$ GeV, $m_{\chi} = 100$ GeV, $m_{Z'} = 500$ GeV

Contour of the 95% CL observed upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

Contour of the 95% CL expected + 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

Contour of the 95% CL expected + 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

Contour of the 95% CL expected upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

Contour of the 95% CL expected - 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

Contour of the 95% CL expected - 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

Contour of the 95% CL observed upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

Contour of the 95% CL expected + 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

Contour of the 95% CL expected + 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

Contour of the 95% CL expected upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

Contour of the 95% CL expected - 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

Contour of the 95% CL expected - 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

Contour of the 95% CL observed upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

Contour of the 95% CL expected + 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

Contour of the 95% CL expected + 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

Contour of the 95% CL expected upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

Contour of the 95% CL expected - 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

Contour of the 95% CL expected - 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

Contour of the 95% CL observed upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

Contour of the 95% CL expected + 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

Contour of the 95% CL expected + 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

Contour of the 95% CL expected upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

Contour of the 95% CL expected - 1 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

Contour of the 95% CL expected - 2 std. upper limits representing $\sigma / \sigma_{theory} = 1$ for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

The $\sigma / \sigma_{theory}$ 95% CL upper limit values for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 100 GeV$.

The $\sigma / \sigma_{theory}$ 95% CL upper limit values for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 150 GeV$.

The $\sigma / \sigma_{theory}$ 95% CL upper limit values for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 200 GeV$.

The $\sigma / \sigma_{theory}$ 95% CL upper limit values for the dark Higgs model in the ($m_{s}$, $m_{Z'}$) plane for $m_{\chi} = 300 GeV$.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the CRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is negligible and therefore is not shown. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the SL selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

Background-only post-fit $p_{T}^{miss}$ distributions for the SRs of the AH selection. The total theory signal (t/t+DM and tt+DM summed together) is presented by the red solid lines for a scalar mediator mass of 100 GeV. The last bin contains overflow events.

The expected and observed 95% CL limits on the DM production cross sections, relative to the theory predictions, shown for the scalar models.

The expected and observed 95% CL limits on the DM production cross sections, relative to the theory predictions, shown for the pseudoscalar models.

Cross sections of the t+DM and tt+DM processes, for different mediator and dark matter masses. The scalar hypothesis is considered. The t+DM processes are split by production mode (t-, s-, and tW-channels). The sum of the three is also provided.

Cross sections of the t+DM and tt+DM processes, for different mediator and dark matter masses. The pseudoscalar hypothesis is considered. The t+DM processes are split by production mode (t-, s-, and tW-channels). The sum of the three is also provided.

Expected yields for different processes after several selection stages. The preselection requires two leptons and at least one photon with $\mathrm{p_\mathrm{T}}$ larger than 25, 20, and 25 GeV, respectively; in addition the dilepton $\mathrm{p_\mathrm{T}}$ must be larger than 60 GeV, and the $\mathrm{p_\mathrm{T}}^{\mathrm{miss}}$ larger than 70 GeV. The signal prediction corresponds to $0.1 \sigma_{\mathrm{\mathrm{ZH}}}$ at $m_{H}$ = 125 GeV.

Expected and observed upper limits at 95\% confidence level on the product of $\sigma_{\mathrm{\mathrm{ZH}}}$ and $\mathcal{B}$($\mathrm{H}$ -> $\mathrm{invisible}+\gamma$) as a function of $m_{\mathrm{\mathrm{H}}}$.

Expected and observed upper limits at 95\% confidence level on the product of $\sigma_{\mathrm{\mathrm{ZH}}}$ and $\mathcal{B}$($\mathrm{H}$ -> $\mathrm{invisible}+\gamma$) as a function of $m_{\mathrm{\mathrm{H}}}$.

$\mathrm{M_{\mathrm{T}}}$ distribution for the $\mathrm{e}\mu$ control region.

$\mathrm{M_{\mathrm{T}}}$ distribution for the $\mathrm{e}\mu$ control region.

$\mathrm{M_{\mathrm{T}}}$ distribution for the WZ control region.

$\mathrm{M_{\mathrm{T}}}$ distribution for the WZ control region.

$\mathrm{M_{\mathrm{T}}}$ distribution for the ZZ control region.

$\mathrm{M_{\mathrm{T}}}$ distribution for the ZZ control region.

$\mathrm{M_{\mathrm{T}}}$ distribution for the signal region, events with $|\eta_{\gamma}|<1$.

$\mathrm{M_{\mathrm{T}}}$ distribution for the signal region, events with $|\eta_{\gamma}|<1$.

$\mathrm{M_{\mathrm{T}}}$ distribution for the signal region, events with $|\eta_{\gamma}|>1$.

$\mathrm{M_{\mathrm{T}}}$ distribution for the signal region, events with $|\eta_{\gamma}|>1$.

Muon reconstruction, identification, and isolation efficiency as a function of $|\eta|$ and $\mathrm{p_\mathrm{T}}$.

Muon reconstruction, identification, and isolation efficiency as a function of $|\eta|$ and $\mathrm{p_\mathrm{T}}$.

Electron reconstruction, identification, and isolation efficiency as a function of $|\eta|$ and $\mathrm{p_\mathrm{T}}$.

Electron reconstruction, identification, and isolation efficiency as a function of $|\eta|$ and $\mathrm{p_\mathrm{T}}$.

Photon efficiency as a function of $|\eta|$ and $\mathrm{p_\mathrm{T}}$.

Photon efficiency as a function of $|\eta|$ and $\mathrm{p_\mathrm{T}}$.

Covariance matrix for all bins used in the analysis. There are 45 bins in total, 15 for every data-taking year. For every year, the first bin corresponds to events in the $\mathrm{e}\mu$ control region, the following five bins correspond to events with $|\eta^\gamma|< 1$ in the signal region, the next five bins correspond to events with $|\eta^\gamma|> 1$ in the signal region, the next two bins correspond to events in the WZ control region, and finally the last two bins correspond to events in the ZZ control region.

Covariance matrix for all bins used in the analysis. There are 45 bins in total, 15 for every data-taking year. For every year, the first bin corresponds to events in the $\mathrm{e}\mu$ control region, the following five bins correspond to events with $|\eta^\gamma|< 1$ in the signal region, the next five bins correspond to events with $|\eta^\gamma|> 1$ in the signal region, the next two bins correspond to events in the WZ control region, and finally the last two bins correspond to events in the ZZ control region.

The $\varepsilon_{\text{reco}}$ values as a function of $E_{g}$, for $\tilde{g}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the two-body gluino decay, when $m_{\tilde{g}}$ is 1200 GeV.

The $\varepsilon_{\text{reco}}$ values as a function of $E_{g}$, for $\tilde{g}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the two-body gluino decay, when $m_{\tilde{g}}$ is 1800 GeV.

The $\varepsilon_{\text{reco}}$ values as a function of $E_{t}$, for $\tilde{t}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the two-body top squark decays, when $m_{\tilde{t}}$ is 400 GeV.

The $\varepsilon_{\text{reco}}$ values as a function of $E_{t}$, for $\tilde{t}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the two-body top squark decays, when $m_{\tilde{t}}$ is 600 GeV.

The $\varepsilon_{\text{reco}}$ values as a function of $E_{t}$, for $\tilde{t}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the two-body top squark decays, when $m_{\tilde{t}}$ is 1000 GeV.

The $\varepsilon_{\text{reco}}$ values as a function of $m_{\tilde{g}} - m_{\tilde{\chi}^{0}}$, for $\tilde{g}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the three-body gluino decay, when $m_{\tilde{g}}$ is 800 and 1000 GeV.

The $\varepsilon_{\text{reco}}$ values as a function of $m_{\tilde{g}} - m_{\tilde{\chi}^{0}}$, for $\tilde{g}$ R-hadrons that stop in the EB or HB, in the MC simulation, for the calorimeter search. The $\varepsilon_{\text{reco}}$ values are plotted for the three-body gluino decay, when $m_{\tilde{g}}$ is 1800 GeV.

The background extrapolation for the muon search. The integral of the fit function to $\Delta t_{DT}$ with the sum of two Gaussian distributions and a Crystal Ball function, for $\Delta t_{DT}$>−20 ns, is plotted as a function of the lower $\Delta t_{RPC}$ selection, for 2015 and 2016 data. The uncertainties correspond to when the fit function parameters are varied by one standard deviation.

The 95% CL upper limits on $\mathcal{B}\sigma$ for gluino pair production, using the cloud model of R-hadron interactions, as a function of lifetime, for combined 2015 and 2016 data for the calorimeter search. We show gluinos that undergo a two-body decay. The discontinuous structure observed between $10^{-7}$ and $10^{-5}$ s is due to the increase of the number of observed events in the search window as the lifetime increases.

The 95% CL upper limits on $\mathcal{B}\sigma$ for top squark pair production, using the cloud model of R-hadron interactions, as a function of lifetime, for combined 2015 and 2016 data for the calorimeter search. We show top squarks that undergo a two-body decay. The discontinuous structure observed between $10^{-7}$ and $10^{-5}$ s is due to the increase of the number of observed events in the search window as the lifetime increases.

The 95% CL upper limits on $\mathcal{B}\sigma$ for gluino pair production, using the cloud model of R-hadron interactions, as a function of lifetime, for combined 2015 and 2016 data for the calorimeter search. We show gluinos that undergo a three-body decay. The discontinuous structure observed between $10^{-7}$ and $10^{-5}$ s is due to the increase of the number of observed events in the search window as the lifetime increases.

The 95% CL upper limits on the gluino mass, using the cloud model of R-hadron interactions, as a function of lifetime, for combined 2015 and 2016 data for the calorimeter search. We show gluinos that undergo a two-body decay. The discontinuous structure observed between $10^{-7}$ and $10^{-5}$ s is due to the increase of the number of observed events in the search window as the lifetime increases.

The 95% CL upper limits on the top squark mass, using the cloud model of R-hadron interactions, as a function of lifetime, for combined 2015 and 2016 data for the calorimeter search. We show top squarks that undergo a two-body decay. The discontinuous structure observed between $10^{-7}$ and $10^{-5}$ s is due to the increase of the number of observed events in the search window as the lifetime increases.

The 95% CL upper limits on the gluino mass, using the cloud model of R-hadron interactions, as a function of lifetime, for combined 2015 and 2016 data for the calorimeter search. We show gluinos that undergo a three-body decay. The discontinuous structure observed between $10^{-7}$ and $10^{-5}$ s is due to the increase of the number of observed events in the search window as the lifetime increases.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded observed region. We show gluinos that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected region. We show gluinos that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected +1 s.d. region. We show gluinos that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected -1 s.d. region. We show gluinos that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~top squark mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded observed region. We show top squarks that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~top squark mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected region. We show top squarks that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~top squark mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected +1 s.d. region. We show top squarks that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~top squark mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected -1 s.d. region. We show top squarks that undergo a two-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded observed region. We show gluinos that undergo a three-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected region. We show gluinos that undergo a three-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected +1 s.d. region. We show gluinos that undergo a three-body decay.

The 95% CL upper limits in the neutralino mass vs.~gluino mass plane, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the calorimeter search. The mostly triangular region shows the excluded expected -1 s.d. region. We show gluinos that undergo a three-body decay.

The 95% CL upper limits on $\mathcal{B}\sigma$ for 1000 GeV gluino pair production as a function of lifetime, for combined 2015 and 2016 data for the muon search.

The 95% CL upper limits on $\mathcal{B}\sigma$ for 400 GeV MCHAMP pair production as a function of lifetime, for combined 2015 and 2016 data for the muon search.

95% CL upper limits on $\mathcal{B}\sigma$ for gluino pair production as a function of mass, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the muon search. The theory curves assume $\mathcal{B}=100\%$.

95% CL upper limits on $\mathcal{B}\sigma$ for MCHAMP pair production as a function of mass, for lifetimes between 10 $\mu$s and 1000 s, for combined 2015 and 2016 data for the muon search. The theory curves assume $\mathcal{B}=100\%$.

The $p_T ^{miss}$ distributions for muon channel comparing the data and background model with systematic uncertainty.

The $m_T$ distributions for electron channel comparing the data and background model with systematic uncertainty, after fitting the background-only model to the data.

The $m_T$ distributions for muon channel comparing the data and background model with systematic uncertainty, after fitting the background-only model to the data.

Expected and observed limits on the product of cross section and branching fraction of a new spin-2 heavy resonance X$\rightarrow$ZZ, assuming zero width, based on the combined analysis of the electron and muon channels.

Expected and observed limits on the product of cross section and branching fraction of a new spin-2 heavy resonance X$\rightarrow$ZZ, assuming zero width, based on the combined analysis of the electron channels.

Expected and observed limits on the product of cross section and branching fraction of a new spin-2 heavy resonance X$\rightarrow$ZZ, assuming zero width, based on the combined analysis of the muon channels.

Expected and observed limits on the product of cross section and branching fraction of a new spin-2 heavy resonance X$\rightarrow$ZZ based on a combined analysis of the electron and muon channels. The more generic version of the bulk graviton model takes 0/10%/20%/30% mass width of the resonance into consideration, assuming gluon-gluon fusion production.

Expected and observed limits on the product of cross section and branching fraction of a new spin-2 heavy resonance X$\rightarrow$ZZ based on a combined analysis of the electron and muon channels. The more generic version of the bulk graviton model takes 0/10%/20%/30% mass width of the resonance into consideration, assuming qq annihilation production.

Acceptance for a vector resonance decaying into a dijet as a function of the resonance mass. The acceptance is calculated using signal simulations for the analysis selection, namely three wide jets with $p_{\mathrm{T}}$ > 72 GeV and |$\eta$| < 2.5, and |$\eta_{1}$ − $\eta_{2}$|<1.1. The errors are dominated by the uncertainties in the parton shower modeling used in signal simulations.

Comparison of the dijet mass distribution of the nominal selection (black), control selection without correction (red), and control selection with correction (blue) for data (solid histograms) and a simulation of a signal from a resonance with a mass of 400 GeV (dashed histograms). The correction, applied as a function of the product of the two largest jet transverse momenta in the event, is obtained as the ratio of the numbers of events passing the signal selection to those passing the control selection. For ease of readability, the dijet mass distributions of the signal have been scaled by an arbitrary factor. The bottom plot shows the relative difference between the dijet mass distributions of the data from the control and nominal selections. The yield of the uncorrected control selection for data is 95.7% of the nominal selection. The overlap between the two data selections is about 35% of the nominal selection. The yield of the uncorrected control selection for signal simulation of a 400-GeV resonance is about 50% of the nominal selection.

Fraction of data and simulated signal events passing different stages of the analysis selection. The selection corresponding to each line includes the selection of the upper lines. Data are required to pass the trigger selection. The uncertainties quoted are statistical only.

Observed events and post-fit background yields for all 31 SRs

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the degenerate stau scenario for m(LSP)=1 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the degenerate stau scenario for m(LSP)=10 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the degenerate stau scenario for m(LSP)=20 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the degenerate stau scenario for m(LSP)=50 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the left-handed stau scenario for m(LSP)=1 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the left-handed stau scenario for m(LSP)=10 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the left-handed stau scenario for m(LSP)=20 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the left-handed stau scenario for m(LSP)=50 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the right-handed stau scenario for m(LSP)=1 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the right-handed stau scenario for m(LSP)=10 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the right-handed stau scenario for m(LSP)=20 GeV

Expected and observed 95% CL cross section upper limits as a function of the stau mass in the right-handed stau scenario for m(LSP)=50 GeV

Observed 95% CL cross section upper limits for degenerate stau pair production in the stau-LSP mass plane

Expected exclusion limits assuming NLO+NLL predictions for signal cross sections for degenerate stau pair production in the stau-LSP mass plane

Observed exclusion limits assuming NLO+NLL predictions for signal cross sections for degenerate stau pair production in the stau-LSP mass plane

Observed 95% CL cross section upper limits for purely left-handed stau pair production in the stau-LSP mass plane

Expected exclusion limits assuming NLO+NLL predictions for signal cross sections for purely left-handed stau pair production in the stau-LSP mass plane

Observed exclusion limits assuming NLO+NLL predictions for signal cross sections for purely left-handed stau pair production in the stau-LSP mass plane

Expected and observed 95% CL cross section upper limits as a function of the stau mass for long-lived staus in the maximally-mixed stau scenario for m(LSP)=1 GeV and ctau_0(stau)=0.01mm

Expected and observed 95% CL cross section upper limits as a function of the stau mass for long-lived staus in the maximally-mixed stau scenario for m(LSP)=1 GeV and ctau_0(stau)=0.05mm

Expected and observed 95% CL cross section upper limits as a function of the stau mass for long-lived staus in the maximally-mixed stau scenario for m(LSP)=1 GeV and ctau_0(stau)=0.1mm

Expected and observed 95% CL cross section upper limits as a function of the stau mass for long-lived staus in the maximally-mixed stau scenario for m(LSP)=1 GeV and ctau_0(stau)=0.5mm

Expected and observed 95% CL cross section upper limits as a function of the stau mass for long-lived staus in the maximally-mixed stau scenario for m(LSP)=1 GeV and ctau_0(stau)=1mm

Expected and observed 95% CL cross section upper limits as a function of the stau mass for long-lived staus in the maximally-mixed stau scenario for m(LSP)=1 GeV and ctau_0(stau)=2.5mm

Observed upper limits at 90% CL on the square of the kinetic mixing coefficient $\epsilon$ in the minimal model of a dark photon

Observed upper limits at 90% CL on the mixing angle $\theta_H$ for the type IV 2HDM+S scenario with fixed $tan\beta=0.5$

The combined efficiency of the dimuon scouting trigger and the MVA muon selection, averaged between 2017 and 2018, weighted by the integrated luminosity of each year. The trigger and selection efficiencies are also included in the table.

Theory cross-section times branching fraction to muons times acceptance for the dark photon and 2HDM+S models.

The expected and observed constraints on chargino lifetime and mass. The ratio of the vacuum expectation values of the two Higgs doublets, $\tan \beta$, is fixed to 5 with $\mu > 0$, where $\mu$ is the higgsino mass parameter. The direct chargino production cross section includes both $\widetilde{\chi}^{0}_{1}\widetilde{\chi}^\pm_{1}$ and $\widetilde{\chi}^\pm_{1}\widetilde{\chi}^\mp_{1}$ production in roughly a 2:1 ratio for all chargino masses considered, and the branching fraction of $\widetilde{\chi}^\pm_{1} \rightarrow \widetilde{\chi}^{0}_{1}\mathrm{\pi^{\pm}}$ is set to 100%. Chargino masses that are less than those on the curve are excluded at 95% CL.

The observed 95% CL upper limits on the product of the cross section for direct production of charginos and their branching fraction to $\widetilde{\chi}^{0}_{1}\mathrm{\pi^{\pm}}$ as a function of chargino mass and lifetime. The ratio of the vacuum expectation values of the two Higgs doublets, $\tan \beta$, is fixed to 5 with $\mu > 0$, where $\mu$ is the higgsino mass parameter. The direct chargino production cross section includes both $\widetilde{\chi}^{0}_{1}\widetilde{\chi}^\pm_{1}$ and $\widetilde{\chi}^\pm_{1}\widetilde{\chi}^\mp_{1}$ production in roughly a 2:1 ratio for all chargino masses considered, and the branching fraction of $\widetilde{\chi}^\pm_{1} \rightarrow \widetilde{\chi}^{0}_{1}\mathrm{\pi^{\pm}}$ is set to 100%.

Signal acceptance for each of the generated chargino masses and lifetimes for $\mathrm{p}\mathrm{p} \rightarrow \widetilde{\chi}^\pm_{1} \widetilde{\chi}^\mp_{1}$ events. The denominator for the acceptance is the total number of $\widetilde{\chi}^\pm_{1} \widetilde{\chi}^\mp_{1}$ events generated, with no generator-level filtering, while the numerator is the number of those events that are selected by the disappearing track signal selection, after being processed with the same reconstruction software used on the data. The acceptance corresponds to the 2015 data-taking period. The uncertainties are only the statistical uncertainties resulting from the sizes of the generated samples.

Signal acceptance for each of the generated chargino masses and lifetimes for $\mathrm{p}\mathrm{p} \rightarrow \widetilde{\chi}^\pm_{1} \widetilde{\chi}^\mp_{1}$ events. The denominator for the acceptance is the total number of $\widetilde{\chi}^\pm_{1} \widetilde{\chi}^\mp_{1}$ events generated, with no generator-level filtering, while the numerator is the number of those events that are selected by the disappearing track signal selection, after being processed with the same reconstruction software used on the data. The acceptance corresponds to the 2016A data-taking period. The uncertainties are only the statistical uncertainties resulting from the sizes of the generated samples.

Signal acceptance for each of the generated chargino masses and lifetimes for $\mathrm{p}\mathrm{p} \rightarrow \widetilde{\chi}^\pm_{1} \widetilde{\chi}^\mp_{1}$ events. The denominator for the acceptance is the total number of $\widetilde{\chi}^\pm_{1} \widetilde{\chi}^\mp_{1}$ events generated, with no generator-level filtering, while the numerator is the number of those events that are selected by the disappearing track signal selection, after being processed with the same reconstruction software used on the data. The acceptance corresponds to the 2016B data-taking period. The uncertainties are only the statistical uncertainties resulting from the sizes of the generated samples.

Signal acceptance for each of the generated chargino masses and lifetimes for $\mathrm{p}\mathrm{p} \rightarrow \widetilde{\chi}^\pm_{1} \widetilde{\chi}^0_{1}$ events. The denominator for the acceptance is the total number of $\widetilde{\chi}^\pm_{1} \widetilde{\chi}^0_{1}$ events generated, with no generator-level filtering, while the numerator is the number of those events that are selected by the disappearing track signal selection, after being processed with the same reconstruction software used on the data. The acceptance corresponds to the 2015 data-taking period. The uncertainties are only the statistical uncertainties resulting from the sizes of the generated samples.

Signal acceptance for each of the generated chargino masses and lifetimes for $\mathrm{p}\mathrm{p} \rightarrow \widetilde{\chi}^\pm_{1} \widetilde{\chi}^0_{1}$ events. The denominator for the acceptance is the total number of $\widetilde{\chi}^\pm_{1} \widetilde{\chi}^0_{1}$ events generated, with no generator-level filtering, while the numerator is the number of those events that are selected by the disappearing track signal selection, after being processed with the same reconstruction software used on the data. The acceptance corresponds to the 2016A data-taking period. The uncertainties are only the statistical uncertainties resulting from the sizes of the generated samples.

Signal acceptance for each of the generated chargino masses and lifetimes for $\mathrm{p}\mathrm{p} \rightarrow \widetilde{\chi}^\pm_{1} \widetilde{\chi}^0_{1}$ events. The denominator for the acceptance is the total number of $\widetilde{\chi}^\pm_{1} \widetilde{\chi}^0_{1}$ events generated, with no generator-level filtering, while the numerator is the number of those events that are selected by the disappearing track signal selection, after being processed with the same reconstruction software used on the data. The acceptance corresponds to the 2016B data-taking period. The uncertainties are only the statistical uncertainties resulting from the sizes of the generated samples.

Predicted signal yields for the 2015 data-taking period, corresponding to $2.7\,\text{fb}^{-1}$, after the application of each of the disappearing track selections for three chargino lifetime hypotheses ($\tau = 0.3$, $3.3$, and $33\,\text{ns}{}$) with a chargino mass of $700\,\text{GeV}{}$. The selections listed are cumulative, i.e., only the events and objects passing a given selection are considered in subsequent selections. The uncertainties shown include only the statistical uncertainty resulting from the limited sizes of the generated samples.

Predicted signal yields for the 2016A data-taking period, corresponding to $8.3\,\text{fb}^{-1}$, after the application of each of the disappearing track selections for three chargino lifetime hypotheses ($\tau = 0.3$, $3.3$, and $33\,\text{ns}{}$) with a chargino mass of $700\,\text{GeV}{}$. The selections listed are cumulative, i.e., only the events and objects passing a given selection are considered in subsequent selections. The uncertainties shown include only the statistical uncertainty resulting from the limited sizes of the generated samples.

Predicted signal yields for the 2016B data-taking period, corresponding to $27.4\,\text{fb}^{-1}$, after the application of each of the disappearing track selections for three chargino lifetime hypotheses ($\tau = 0.3$, $3.3$, and $33\,\text{ns}{}$) with a chargino mass of $700\,\text{GeV}{}$. The selections listed are cumulative, i.e., only the events and objects passing a given selection are considered in subsequent selections. The uncertainties shown include only the statistical uncertainty resulting from the limited sizes of the generated samples.