Comparison of the $m_{miss}$ shapes for the simulated signal events within the fiducial region and those outside it, after including the effect of PU protons as describe in the text, for a generated $m_{X}$ mass of 1000 GeV. The distributions are shown for single(+z)-single(−z) proton reconstruction categories.

Product of the acceptance times the combined reconstruction and identification efficiency, as a function of $m_{X}$ , for events generated inside the fiducial volume defined in Table 1. The curves shown panel display the different final states. The definition of the fiducial region and signal model used to estimate the acceptance is provided in the text.

Product of the acceptance times the combined reconstruction and identification efficiency, as a function of $m_{X}$ , for different proton reconstruction categories in the $Z \rightarrow \mu\mu $ analysis, and for events generated inside the fiducial volume defined in Table 1. The curves shown panel display the different final states. The definition of the fiducial region and signal model used to estimate the acceptance is provided in the text.

Distributions of the reconstructed proton $\xi$ in the negative arm from the proton mixing procedure with simulated MC events, are compared to data. The ee final states are shown.The ee events are displayed without the Z boson $p_T$ requirement.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed proton $\xi$ in the positive arm from the proton mixing procedure with simulated MC events, are compared to data. The ee final states are shown.The ee events are displayed without the Z boson $p_T$ requirement.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed di-proton rapidity from the proton mixing procedure with simulated MC events, are compared to data. The ee final states are shown.The ee events are displayed without the Z boson $p_T$ requirement.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed proton $\xi$ in the negative arm from the proton mixing procedure with simulated MC events, are compared to data. The $\mu\mu$ final states are shown.The $\mu\mu$ events are displayed without the Z boson $p_T$ requirement.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed proton $\xi$ in the positive arm from the proton mixing procedure with simulated MC events, are compared to data. The $\mu\mu$ final states are shown.The $\mu\mu$ events are displayed without the Z boson $p_T$ requirement.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed di-proton rapidity from the proton mixing procedure with simulated MC events, are compared to data. The $\mu\mu$ final states are shown.The $\mu\mu$ events are displayed without the Z boson $p_T$ requirement.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed proton $\xi$ in the negative arm from the proton mixing procedure with simulated MC events, are compared to data. The $\gamma$ final states are shown.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed proton $\xi$ in the positive arm from the proton mixing procedure with simulated MC events, are compared to data. The $\gamma$ final states are shown.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Distributions of the reconstructed di-proton rapidity from the proton mixing procedure with simulated MC events, are compared to data. The $\gamma$ final states are shown.For illustration, the simulated signal distributions are superimposed for various choices of $m_{X}$ , normalised to a generated fiducial cross section of 100 pb.

Validation of the background modelling method, using the $e\mu$ control sample. Selected $e\mu$ events are mixed with protons from $Z\rightarrow \mu \mu$ events with $p_{T}(Z)<10$ GeV to simulate the combinatorial background shape, while the data points are unaltered $e\mu$ events. The proton $\xi$ distribution for the positive CT-PPS arm is shown.

Validation of the background modelling method, using the $e\mu$ control sample. Selected $e\mu$ events are mixed with protons from $Z\rightarrow \mu \mu$ events with $p_{T}(Z)<10$ GeV to simulate the combinatorial background shape, while the data points are unaltered $e\mu$ events. The proton $\xi$ distribution for the negative CT-PPS arm is shown.

Validation of the background modelling method, using the $e\mu$ control sample. Selected $e\mu$ events are mixed with protons from $Z\rightarrow \mu \mu$ events with $p_{T}(Z)<10$ GeV to simulate the combinatorial background shape, while the data points are unaltered $e\mu$ events. The di-proton invariant mass is shown.

Validation of the background modelling method, using the $e\mu$ control sample. Selected $e\mu$ events are mixed with protons from $Z\rightarrow \mu \mu$ events with $p_{T}(Z)<10$ GeV to simulate the combinatorial background shape, while the data points are unaltered $e\mu$ events. The $m_{miss}$ is shown.

$m_{miss}$ distributions in the $pp\rightarrow ppZeeX$, with the protons reconstructed with the multi(+z)-multi(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZeeX$, with the protons reconstructed with the multi(+z)-single(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZeeX$, with the protons reconstructed with the single(+z)-multi(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZeeX$, with the protons reconstructed with the single(+z)-single(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZ\mu\mu X$, with the protons reconstructed with the multi(+z)-multi(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZ\mu\mu X$, with the protons reconstructed with the multi(+z)-single(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZ\mu\mu X$, with the protons reconstructed with the single(+z)-multi(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow ppZ\mu\mu X$, with the protons reconstructed with the single(+z)-single(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 1 pb.

$m_{miss}$ distributions in the $pp\rightarrow pp\gamma X$, with the protons reconstructed with the multi(+z)-multi(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 10 pb.

$m_{miss}$ distributions in the $pp\rightarrow pp\gamma X$, with the protons reconstructed with the multi(+z)-single(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 10 pb.

$m_{miss}$ distributions in the $pp\rightarrow pp\gamma X$, with the protons reconstructed with the single(+z)-multi(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 10 pb.

$m_{miss}$ distributions in the $pp\rightarrow pp\gamma X$, with the protons reconstructed with the single(+z)-single(-z)method. The background distributions are shown after the fit. The expectations for a signal with $m_{X} = 1000$ GeV are superimposed, where the fiducial production cross section is normalised to 10 pb.

Upper limits on the $pp\rightarrow ppZ/\gamma X$ cross section at 95% CL, as a function of $m_X$. The 68 and 95% central intervals of the expected limits are represented by the green and yellow bands, respectively, while the observed limit is superimposed as a curve. The plot corresponds to the $Z\rightarrow ee$ analysis.

Upper limits on the $pp\rightarrow ppZ/\gamma X$ cross section at 95% CL, as a function of $m_X$. The 68 and 95% central intervals of the expected limits are represented by the green and yellow bands, respectively, while the observed limit is superimposed as a curve. The plot corresponds to the $Z\rightarrow \mu\mu$ analysis.

Upper limits on the $pp\rightarrow ppZ/\gamma X$ cross section at 95% CL, as a function of $m_X$. The 68 and 95% central intervals of the expected limits are represented by the green and yellow bands, respectively, while the observed limit is superimposed as a curve. The plot corresponds to the $Z$ analysis.

Upper limits on the $pp\rightarrow ppZ/\gamma X$ cross section at 95% CL, as a function of $m_X$. The 68 and 95% central intervals of the expected limits are represented by the green and yellow bands, respectively, while the observed limit is superimposed as a curve. The plot corresponds to the $\gamma$ analysis.

Expected and observed upper limits on the AQGC operators $a^Z_C/\Lambda^2$, with no unitarization. The $y$ axis shows the limit on the ratio of the observed cross section to the cross section predicted for each anomalous coupling value ($\sigma_\mathrm{AQGC}$).

Limits on LEP-like dimension-6 anomalous quartic gauge coupling parameters, with and without unitarization via a clipping procedure.

Conversion of limits on $a^W_0$ to dimension-8 $f_{M,i}$ operators, using the assumption of vanishing $WWZ\gamma$ couplings to eliminate some parameters. When quoting limits on one of the operators, the other is fixed to zero. The results for $|f_{M,0}/\Lambda^{4}|$ and $|f_{M,4}/\Lambda^{4}|$ are shown with and without clipping of the signal model at 1.4 TeV, when the other parameter is fixed to the SM value of zero.

Conversion of limits on $a^W_0$ and $a^W_C$ to dimension-8 $f_{M,i}$ operators, using the assumption that all $f_{M,i}$ except one are equal to zero. The results are shown with and without clipping of the signal model at 1.4 TeV.

{Expected and observed limits in the two-dimensional plane of $a^W_0/\Lambda^2$ vs. $a^W_C/\Lambda^2$. The limits are described by analytical ellipses of equation $(x-x0)^2/a^2 + (y-y0)^2/b^2 = 1$ and rotated counter-clockwise by $\theta$ degrees, where $x$ and $y$ in the equation correspond to the $a_0^W$ and $a_C^W$ couplings, respectively.

{Expected and observed limits in the two-dimensional plane of $a^W_0/\Lambda^2$ vs. $a^W_C/\Lambda^2$ with unitarization imposed by clipping the signal model at 1.4 TeV. The limits are described by analytical ellipses of equation $(x-x0)^2/a^2 + (y-y0)^2/b^2 = 1$ and rotated counter-clockwise by $\theta$ degrees, where $x$ and $y$ in the equation correspond to the $a_0^W$ and $a_C^W$ couplings, respectively.

{Expected and observed limits in the two-dimensional plane of $a^Z_0/\Lambda^2$ vs. $a^Z_C/\Lambda^2$. The limits are described by analytical ellipses of equation $(x-x0)^2/a^2 + (y-y0)^2/b^2 = 1$ and rotated counter-clockwise by $\theta$ degrees, where $x$ and $y$ in the equation correspond to the $a_0^Z$ and $a_C^Z$ couplings, respectively.

Upper limits on the production cross section times branching fraction of the b* RH hypothesis at a 95% CL. Dashed colored lines show the expected limits from the l+jets and all-hadronic channels, where the latter start at resonance masses of 1.4 TeV. The observed and expected limits from the combination are shown as solid and dashed black lines, respectively. The green and yellow bands show the 68 and 95% confidence intervals on the combined expected limits.

Upper limits on the production cross section times branching fraction of the b* VL hypothesis at a 95% CL. Dashed colored lines show the expected limits from the l+jets and all-hadronic channels, where the latter start at resonance masses of 1.4 TeV. The observed and expected limits from the combination are shown as solid and dashed black lines, respectively. The green and yellow bands show the 68 and 95% confidence intervals on the combined expected limits.

Upper limits on the production cross section times branching fraction of the B+b hypothesis at a 95% CL. Dashed colored lines show the expected limits from the l+jets and all-hadronic channels, where the latter start at resonance masses of 1.4 TeV. The observed and expected limits from the combination are shown as solid and dashed black lines, respectively. The green and yellow bands show the 68 and 95% confidence intervals on the combined expected limits.

Upper limits on the production cross section times branching fraction of the B+t hypothesis at a 95% CL. Dashed colored lines show the expected limits from the l+jets and all-hadronic channels, where the latter start at resonance masses of 1.4 TeV. The observed and expected limits from the combination are shown as solid and dashed black lines, respectively. The green and yellow bands show the 68 and 95% confidence intervals on the combined expected limits.

Dijet signal efficiencies as a function of the signal mass and lifetime for events satisfying all event and vertex requirements, with corrections based on systematic differences in the vertex reconstruction efficiency between data and simulation.

The distribution of $d_{\mathrm{BV}}$ for $\geq$5-track one-vertex events in data and three simulated multijet signal samples each with a mass of 1600 GeV. The production cross section for each signal model is assumed to be the lower limit excluded by CMS-EXO-17-018, corresponding to values of 0.8, 0.25, and 0.15 fb for the samples with $c\tau =$ 0.3, 1.0, and 10 mm, respectively. The last bin includes the overflow events. This bin includes one event in data with a vertex with large $d_{\mathrm{BV}}$ that appears to arise from tracks originating from separate pp interaction vertices, consistent with background.

Distribution of the $x$-$y$ distances between vertices, $d_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution $d_{\mathrm{VV}}^{\kern 0.15em\mathrm{C}}$ (blue continuous line) is constructed from one-vertex events in data, and is normalized to the number of two-vertex events in data with two 3-track vertices. The two vertical red dashed lines separate the regions used in the fit.

Distribution of the $x$-$y$ distances between vertices, $d_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution $d_{\mathrm{VV}}^{\kern 0.15em\mathrm{C}}$ (blue continuous line) is constructed from one-vertex events in data, and is normalized to the number of two-vertex events in data which have exactly one 4-track vertex and one 3-track vertex. The two vertical red dashed lines separate the regions used in the fit.

Distribution of the $x$-$y$ distances between vertices, $d_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution $d_{\mathrm{VV}}^{\kern 0.15em\mathrm{C}}$ (blue continuous line) is constructed from one-vertex events in data, and is normalized to the number of two-vertex events in data with two 4-track vertices. The two vertical red dashed lines separate the regions used in the fit.

Distribution of the $x$-$y$ distances between vertices, $d_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution $d_{\mathrm{VV}}^{\kern 0.15em\mathrm{C}}$ (blue continuous line) is constructed from one-vertex events in data, and is normalized using $\geq$5-track one-vertex event information. The two vertical red dashed lines separate the regions used in the fit.

Predicted yields for the background-only normalized template, predicted yields for three simulated multijet signals each with a mass of 1600 GeV, and the observed yield in each $d_{\mathrm{VV}}$ bin. The production cross section for each signal model is assumed to be the lower limit excluded by CMS-EXO-17-018, corresponding to values of 0.8, 0.25, and 0.15 fb for samples with $c\tau =$ 0.3, 1.0, and 10 mm, respectively. The uncertainties in the signal yields and the systematic uncertainties in the background prediction reflect the systematic uncertainties given in the text.

Observed 95% CL upper limits on the product of cross section and branching fraction squared for the multijet signals, as a function of mass and $c\tau$. The overlaid mass-lifetime exclusion curves assume pair-production cross sections for the neutralino (red) and gluino (purple) with 100% branching fraction to each model's respective decay mode specified. The solid black (dashed colored) lines represent the observed (median expected) limits at 95% CL. The thin black lines represent the variation of the observed limit within theoretical uncertainties of the signal cross section. The thin dashed colored lines represent the region containing 68% of the expected limit distribution under the background-only hypothesis. The observed limits from the CMS displaced jets search (CMS-EXO-19-021) are also shown in teal for comparison.

Observed 95% CL upper limits on the product of cross section and branching fraction squared for the multijet signals, as a function of mass and $c\tau$. The overlaid mass-lifetime exclusion curves assume pair-production cross sections for the neutralino (red) and gluino (purple) with 100% branching fraction to each model's respective decay mode specified. The solid black (dashed colored) lines represent the observed (median expected) limits at 95% CL. The thin black lines represent the variation of the observed limit within theoretical uncertainties of the signal cross section. The thin dashed colored lines represent the region containing 68% of the expected limit distribution under the background-only hypothesis. The observed limits from the CMS displaced jets search (CMS-EXO-19-021) are also shown in teal for comparison.

Observed 95% CL upper limits on the product of cross section and branching fraction squared for the multijet signals, as a function of mass and $c\tau$. The overlaid mass-lifetime exclusion curves assume pair-production cross sections for the neutralino (red) and gluino (purple) with 100% branching fraction to each model's respective decay mode specified. The solid black (dashed colored) lines represent the observed (median expected) limits at 95% CL. The thin black lines represent the variation of the observed limit within theoretical uncertainties of the signal cross section. The thin dashed colored lines represent the region containing 68% of the expected limit distribution under the background-only hypothesis. The observed limits from the CMS displaced jets search (CMS-EXO-19-021) are also shown in teal for comparison.

Observed 95% CL upper limits on the product of cross section and branching fraction squared for the dijet signals, as a function of mass and $c\tau$. The overlaid mass-lifetime exclusion curves assume pair-production cross sections for the top squark with 100% branching fraction to each model's respective decay mode specified. The solid black (dashed colored) lines represent the observed (median expected) limits at 95% CL. The thin black lines represent the variation of the observed limit within theoretical uncertainties of the signal cross section. The thin dashed colored lines represent the region containing 68% of the expected limit distribution under the background-only hypothesis. The observed limits from the CMS displaced jets search (CMS-EXO-19-021) are also shown in teal for comparison.

Observed 95% CL upper limits on the product of cross section and branching fraction squared for the dijet signals, as a function of mass and $c\tau$. The overlaid mass-lifetime exclusion curves assume pair-production cross sections for the top squark with 100% branching fraction to each model's respective decay mode specified. The solid black (dashed colored) lines represent the observed (median expected) limits at 95% CL. The thin black lines represent the variation of the observed limit within theoretical uncertainties of the signal cross section. The thin dashed colored lines represent the region containing 68% of the expected limit distribution under the background-only hypothesis. The observed limits from the CMS displaced jets search (CMS-EXO-19-021) are also shown in teal for comparison.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of mass for multijet signals, for a fixed $c\tau$ of 300um in the full Run-2 data set. The neutralino and gluino pair production cross sections are shown for the multijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of mass for dijet signals, for a fixed $c\tau$ of 300um in the full Run-2 data set. The top squark pair-production cross section is shown for the dijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of mass for multijet signals, for a fixed $c\tau$ of 1 mm in the full Run-2 data set. The neutralino and gluino pair production cross sections are shown for the multijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of mass for dijet signals, for a fixed $c\tau$ of 1 mm in the full Run-2 data set. The top squark pair-production cross section is shown for the dijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of mass for multijet signals, for a fixed $c\tau$ of 10 mm in the full Run-2 data set. The neutralino and gluino pair production cross sections are shown for the multijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of mass for dijet signals, for a fixed $c\tau$ of 10 mm in the full Run-2 data set. The top squark pair-production cross section is shown for the dijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of $c\tau$ for multijet signals, for a fixed mass of 800 GeV in the full Run-2 data set. The neutralino and gluino pair production cross sections are shown for the multijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of $c\tau$ for dijet signals, for a fixed mass of 800 GeV in the full Run-2 data set. The top squark pair-production cross section is shown for the dijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of $c\tau$ for multijet signals, for a fixed mass of 1600 GeV in the full Run-2 data set. The neutralino and gluino pair production cross sections are shown for the multijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of $c\tau$ for dijet signals, for a fixed mass of 1600 GeV in the full Run-2 data set. The top squark pair-production cross section is shown for the dijet signals.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of $c\tau$ for multijet signals, for a fixed mass of 2400 GeV in the full Run-2 data set. The neutralino and gluino pair production cross sections are shown for the multijet signals. For $m$ = 2400 GeV, the expected neutralino cross section is $\approx 8\times 10^{-5}$ fb and is not shown.

Observed and expected 95% CL upper limits on the product of cross section and branching fraction squared, as a function of $c\tau$ for dijet signals, for a fixed mass of 2400 GeV in the full Run-2 data set. The top squark pair-production cross section is shown for the dijet signals.

Data-to-simulation efficiency correction factors, shown for multijet and dijet signal topologies in several ranges of $c\tau$. Note that these correction factors account for the two long-lived particles in the simulated events, and are therefore the total correction factors used to scale event yields rather than the correction factors one would apply to individual vertices.

Distribution of the azimuthal angle between vertices, $\Delta\phi_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution (blue continuous line) is constructed from 3-track one-vertex events in data, and is normalized to the number of 3-track two-vertex events in data.

Distribution of the azimuthal angle between vertices, $\Delta\phi_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution (blue continuous line) is constructed from 4-track and 3-track one-vertex events in data, and is normalized to the number of two-vertex events in data which have exactly one 4-track vertex and one 3-track vertex.

Distribution of the azimuthal angle between vertices, $\Delta\phi_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution (blue continuous line) is constructed from 4-track one-vertex events in data, and is normalized to the number of 4-track two-vertex events in data.

Distribution of the azimuthal angle between vertices, $\Delta\phi_{\mathrm{VV}}$, for 2017 and 2018 data. The background distribution (blue continuous line) is constructed from $\geq$5-track one-vertex events in data, and is normalized using one-vertex event information. No $\geq$5-track two-vertex data events pass the selection.