CERN Accelerating science

A schematic view of the typical Run~2 data flow during 2018 showing the data acquisition strategy with scouting and parking data streams, along with the standard data stream. A value of ${\Linst = \sci{1.2}{34}\invcms}$ over a typical 2018 fill, corresponding to an average pileup of 38, is considered.

Comparison of the typical HLT rates of the standard, parking, and scouting data streams from Run~1 to Run~3. The \Linst averaged over one typical fill of a given data-taking year is shown in pink.

The efficiency of the Run~2 Calo scouting (left) and standard (right) jet triggers as a function of the reconstructed mass of the dijet system. Figures taken from Ref.~\cite{EXO-16-056}.

The efficiency of the Run~2 Calo scouting (left) and standard (right) jet triggers as a function of the reconstructed mass of the dijet system. Figures taken from Ref.~\cite{EXO-16-056}.

The efficiency of the Run~2 PF scouting jet triggers as a function of \HT (left) and as a function of the leading large-radius jet \pt and trimmed jet mass (right).

The efficiency of the Run~2 PF scouting jet triggers as a function of \HT (left) and as a function of the leading large-radius jet \pt and trimmed jet mass (right).

The observed percent difference between the \pt of Calo jets at the HLT and the \pt of PF jets reconstructed offline (points), fitted to a smooth function (curve), \vs the Calo jet \pt. Both Calo and PF jets are calibrated with corrections derived from simulation. Figure taken from Ref.~\cite{EXO-16-056}.

The distribution of \mjjj for the resolved three-jet search (left), and average jet mass {($\tilde{m} = (m_1 + m_2)/ 2$)} for the merged three-parton search (right), adapted from Ref.~\cite{EXO-21-004}. Both analyses use PF jets. The peak around 170\GeV in both distributions corresponds to the all-hadronic decay of the top quark. The data (points) are compared to the background-only prediction (blue) and the full background fit including simulations of the top quark resonance (red).

The distribution of \mjjj for the resolved three-jet search (left), and average jet mass {($\tilde{m} = (m_1 + m_2)/ 2$)} for the merged three-parton search (right), adapted from Ref.~\cite{EXO-21-004}. Both analyses use PF jets. The peak around 170\GeV in both distributions corresponds to the all-hadronic decay of the top quark. The data (points) are compared to the background-only prediction (blue) and the full background fit including simulations of the top quark resonance (red).

Left: output of the QGD for quark (orange) and gluon (blue) jets. The corresponding receiver operating characteristic (ROC) curve is also shown. Right: observation of fully hadronic top quark decays in the invariant mass of three jets with a QCD multijet background, for an inclusive selection (no QGD), and for a selection including the QGD score. Figure adapted from Ref.~\cite{EXO-21-004}.

Left: output of the QGD for quark (orange) and gluon (blue) jets. The corresponding receiver operating characteristic (ROC) curve is also shown. Right: observation of fully hadronic top quark decays in the invariant mass of three jets with a QCD multijet background, for an inclusive selection (no QGD), and for a selection including the QGD score. Figure adapted from Ref.~\cite{EXO-21-004}.

Dimuon invariant mass distribution of events selected with the standard muon triggers (blue, dashed) and scouting muon triggers (pink, solid) in the mass range 11--240\GeV, normalized to ${\Lint = 96.6\fbinv}$, corresponding to the scouting data collected in 2017 and 2018. The selection applied to obtain each distribution is described in Ref.~\cite{EXO-19-018}.

Dimuon invariant mass spectrum and event rate of each L1 seed (legend) obtained with the scouting stream reconstructed at the HLT, using data collected in 2018 corresponding to ${\Lint = 60\fbinv}$. Well-known dimuon resonances from various meson decays or from \PZ boson decays are indicated above each peak.

Dimuon invariant mass distributions in bins of transverse displacement from the PV (\lxy). Figure adapted from Ref.~\cite{EXO-20-014}.

Efficiency of the dimuon scouting trigger and logical ``OR'' of all L1 triggers measured with 2017 and 2018 data. The efficiency is shown as a function of the angular separation between the two muons \drMM and the dimuon mass \mMM (\cmsLeft), and as a function of \drMM and the subleading muon \pt (\cmsRight). The selection in the \cmsRight plot also requires ${0.45 < \mMM < 0.65\GeV}$ to focus on the \PGh meson resonance region. The statistical uncertainty in the measured values is generally less than 3\% per bin on the left plot and less than 15\% per bin on the right plot.

Efficiency of the dimuon scouting trigger and logical ``OR'' of all L1 triggers measured with 2017 and 2018 data. The efficiency is shown as a function of the angular separation between the two muons \drMM and the dimuon mass \mMM (\cmsLeft), and as a function of \drMM and the subleading muon \pt (\cmsRight). The selection in the \cmsRight plot also requires ${0.45 < \mMM < 0.65\GeV}$ to focus on the \PGh meson resonance region. The statistical uncertainty in the measured values is generally less than 3\% per bin on the left plot and less than 15\% per bin on the right plot.

Background rejection \vs signal efficiency of the new MVA-based muon ID strategies evaluated on scouting data: \PgUa-trained MVA (black line), \PJGy-trained MVA (red line). A comparison with the performance of the previous cut-based selection, which was optimized for signals with masses higher than 11\GeV, is also shown for the \PgUa (blue triangle) and \PJGy (brown square) signals. Figure adapted from Ref.~\cite{EXO-21-005}.

Relative width of dimuon resonances as a function of mass, measured in 2017 and 2018 scouting data. The fits are performed separately for the 2017 and 2018 data sets. The values shown are the average width of each fit, weighted by the \Lint value corresponding to the data accumulated in each year. From left to right, the \PGh, \PGf, \PJGy, \Pgy, \PgUa, \PgUb, and \PgUc resonances are shown. The inserts display fits of the \PJGy and \PGU peaks obtained with scouting data (black markers) separately for the signal (green) and background (red) components, and for their sum (blue).

Left: dijet mass spectra (points) compared to a fitted parametrization of the background (solid curve) for the inclusive search performed in Ref.~\cite{EXO-16-056}. Right: dijet mass spectrum (points) compared to a fitted parametrization of the background (solid curve) for the three-jet analysis performed in Ref.~\cite{EXO-19-004}. The lower panel shows the difference between the data and the fitted parametrization, divided by the statistical uncertainty of the data. Examples of predicted signals from narrow gluon-gluon, quark-gluon, and quark-quark resonances are shown with cross sections equal to the observed upper limits at 95\%~\CL. Figures taken from Refs.~\cite{EXO-16-056} (left) and \cite{EXO-19-004} (right).

Left: dijet mass spectra (points) compared to a fitted parametrization of the background (solid curve) for the inclusive search performed in Ref.~\cite{EXO-16-056}. Right: dijet mass spectrum (points) compared to a fitted parametrization of the background (solid curve) for the three-jet analysis performed in Ref.~\cite{EXO-19-004}. The lower panel shows the difference between the data and the fitted parametrization, divided by the statistical uncertainty of the data. Examples of predicted signals from narrow gluon-gluon, quark-gluon, and quark-quark resonances are shown with cross sections equal to the observed upper limits at 95\%~\CL. Figures taken from Refs.~\cite{EXO-16-056} (left) and \cite{EXO-19-004} (right).

Observed limits on the universal coupling $g^\prime_\mathrm{q}$ between a leptophobic \PZpr boson and quarks~\cite{EXO-16-032} from various CMS dijet analyses. Regions above the lines are excluded at 95\%~\CL. The grey dashed lines show the $g^\prime_\mathrm{q}$ values at fixed values of ${\Gamma_{\PZpr}/m_{\PZpr}}$. Limits from scouting-based analyses are indicated with bold-dashed lines.

Comparison of limits from searches for RPV gluinos decaying to three partons. Regions above the lines are excluded at 95\%~\CL. The two CMS analyses that use data scouting are also indicated with bold-dashed lines.

Observed (points) and expected (dashes) limits on the product of production cross section, branching fraction, and acceptance for pair-produced merged two-quark resonances. The variations at the one and two standard deviation levels in the expected limits are displayed with shaded bands. A comparison with the theoretical predictions for top squark production (red) is also shown. Figure taken from Ref.~\cite{EXO-21-004}.

Expected and observed upper limits at 90\%~\CL on the square of the kinetic mixing coefficient ($\epsilon^2$) as a function of dark photon mass. Results obtained with scouting triggers are displayed to the left of the vertical purple line, while those obtained with standard triggers are shown to the right. Limits at 90\%~\CL obtained from the search performed by the LHCb Collaboration~\cite{EXO-19-018_LHCb_PhysRevLett.124.041801} are shown in red, and constraints at 95\%~\CL from the measurements of electroweak observables are shown in light blue~\cite{EXO-19-018_EW_Curtin:2014cca}. Figure taken from Ref.~\cite{EXO-19-018}.

The \mMM distribution obtained with the scouting data collected during 2017 and 2018 with two sets of selections: the \PJGy-trained (red) and the \PgUa-trained (blue) MVA-based muon identification algorithms. Figure taken from Ref.~\cite{EXO-21-005}.

Upper limits at 90\%~\CL on the square of the kinetic mixing coefficient ($\epsilon^2$) in the minimal dark photon model obtained as a recast of model-independent limits on the production rates of dimuon resonances for the inclusive category. The CMS limits (pink) are compared with the existing limits at 90\%~CL provided by LHCb~\cite{EXO-21-005_LHCb:2020ysn} (blue) and BaBar~\cite{BaBar:2012wey} (gray). In the CMS analysis, the background-model fit of the mass distribution becomes unreliable when the tails of \PJGy and \Pgy resonances enter the fit mass window, so the mass range 2.6–4.2 \GeV is excluded from the search. Figure taken from Ref.~\cite{EXO-21-005}.

Left: diagram illustrating an SM-like Higgs boson (\PH) decay to four leptons ($\ell$) via two intermediate dark photons (\PZD). Right: diagram illustrating the production of a scalar resonance \PGf in a \PQb hadron decay, through mixing with an SM-like Higgs boson. Figure taken from Ref.~\cite{EXO-20-014}.

Left: diagram illustrating an SM-like Higgs boson (\PH) decay to four leptons ($\ell$) via two intermediate dark photons (\PZD). Right: diagram illustrating the production of a scalar resonance \PGf in a \PQb hadron decay, through mixing with an SM-like Higgs boson. Figure taken from Ref.~\cite{EXO-20-014}.

Observed limits at 95\%~\CL on (upper) the branching fraction $\mathcal{B}(\PH \to \PZD\PZD)$ and (lower) the branching fraction product ${\mathcal{B}(\mathrm{h}_{\PQb}\to\PGf \mathrm{X})\cdot\mathcal{B}(\PGf\to\PGm\PGm)}$ as contours in the parameter space containing the signal mass ($m_{\PZD}$ or $m_{\PGf}$, respectively) and the signal lifetime $c\tau_0$. The vertical gray bands indicate mass ranges containing known SM resonances, which are masked for this search. The limits are obtained using the combination of all dimuon and four-muon event categories. Figures taken from Ref.~\cite{EXO-20-014}.

Observed limits at 95\%~\CL on (upper) the branching fraction $\mathcal{B}(\PH \to \PZD\PZD)$ and (lower) the branching fraction product ${\mathcal{B}(\mathrm{h}_{\PQb}\to\PGf \mathrm{X})\cdot\mathcal{B}(\PGf\to\PGm\PGm)}$ as contours in the parameter space containing the signal mass ($m_{\PZD}$ or $m_{\PGf}$, respectively) and the signal lifetime $c\tau_0$. The vertical gray bands indicate mass ranges containing known SM resonances, which are masked for this search. The limits are obtained using the combination of all dimuon and four-muon event categories. Figures taken from Ref.~\cite{EXO-20-014}.

The four-muon invariant mass (\mMMMM) distribution in the range 0.46--0.90\GeV, obtained with pp collision data collected during 2017--2018. The observed distribution (points) is compared to the background-only prediction (green dashed) and to the full background fit including simulations of the signal (solid blue). The peak observed in the mass window 0.53--0.57\GeV corresponds to the \PGh meson. The pull distribution in the lower panel is shown relative to the background component of the fit model and defined as ${(\mathrm{Data} - \mathrm{Fit})/ \mathrm{Uncertainty}}$, where the uncertainty is statistical only. Figure taken from Ref.~\cite{BPH-22-003}.

The product of acceptance ($A$) and efficiency as a function of the generated \PGh meson \pt for the two-muon (red circles and blue squares) and four-muon (orange up triangles and green down triangles) \PGh meson decays. The product is evaluated using simulated samples. Figure taken from Ref.~\cite{BPH-22-003}.

Relative rate of each L1 algorithm category, shown as the fraction of the total rate based on the 2022 (orange) and 2023 (blue) configurations. The proportional rate of each category with respect to the total is shown on the right. The values are computed using reference runs with average pileup of 60.

Trigger efficiency as a function of AK4 jet \pt (left), AK8 jet \pt (center), and \HT (right). The efficiency is computed from collision data recorded in 2022 by the scouting (black points) and standard (red points) streams.

Trigger efficiency as a function of AK4 jet \pt (left), AK8 jet \pt (center), and \HT (right). The efficiency is computed from collision data recorded in 2022 by the scouting (black points) and standard (red points) streams.

Trigger efficiency as a function of AK4 jet \pt (left), AK8 jet \pt (center), and \HT (right). The efficiency is computed from collision data recorded in 2022 by the scouting (black points) and standard (red points) streams.

The JES as a function of mean scouting jet \pt derived from simulated (left) and recorded (right) events. The red and black points correspond to events where the two leading jets have ${\abs{\eta} <1.3}$ and ${1.3 < \abs{\eta} < 2.5}$, respectively.

The JES as a function of mean scouting jet \pt derived from simulated (left) and recorded (right) events. The red and black points correspond to events where the two leading jets have ${\abs{\eta} <1.3}$ and ${1.3 < \abs{\eta} < 2.5}$, respectively.

The JER as a function of average \pt. The JER is computed from simulated (upper) and recorded (lower) events, by requiring the two leading jets to have ${\abs{\eta} < 1.3}$ (left) and ${1.3 < \abs{\eta} < 2.5}$ (right). The red and black data points denote 2022 collision data reconstructed by the scouting and offline algorithms, respectively.

The JER as a function of average \pt. The JER is computed from simulated (upper) and recorded (lower) events, by requiring the two leading jets to have ${\abs{\eta} < 1.3}$ (left) and ${1.3 < \abs{\eta} < 2.5}$ (right). The red and black data points denote 2022 collision data reconstructed by the scouting and offline algorithms, respectively.

The JER as a function of average \pt. The JER is computed from simulated (upper) and recorded (lower) events, by requiring the two leading jets to have ${\abs{\eta} < 1.3}$ (left) and ${1.3 < \abs{\eta} < 2.5}$ (right). The red and black data points denote 2022 collision data reconstructed by the scouting and offline algorithms, respectively.

The JER as a function of average \pt. The JER is computed from simulated (upper) and recorded (lower) events, by requiring the two leading jets to have ${\abs{\eta} < 1.3}$ (left) and ${1.3 < \abs{\eta} < 2.5}$ (right). The red and black data points denote 2022 collision data reconstructed by the scouting and offline algorithms, respectively.

The ratio of the JER derived from scouting events to the JER derived from offline events as a function of average \pt. The ratio is computed from simulated (left) and recorded (right) events, by requiring the two leading jets to have ${\abs{\eta} < 1.3}$ (red points) and ${1.3 < \abs{\eta} < 2.5}$ (black points).

The ratio of the JER derived from scouting events to the JER derived from offline events as a function of average \pt. The ratio is computed from simulated (left) and recorded (right) events, by requiring the two leading jets to have ${\abs{\eta} < 1.3}$ (red points) and ${1.3 < \abs{\eta} < 2.5}$ (black points).

Trigger efficiency for ggF boosted $\PH \to \PQb \PAQb$ events as a function of the highest particle-level Higgs boson \pt (left) and highest offline-reconstructed AK8 jet \pt (right), as determined from simulation. The black and red points correspond to the scouting and the standard trigger selection, respectively.

Trigger efficiency for ggF boosted $\PH \to \PQb \PAQb$ events as a function of the highest particle-level Higgs boson \pt (left) and highest offline-reconstructed AK8 jet \pt (right), as determined from simulation. The black and red points correspond to the scouting and the standard trigger selection, respectively.

Number of ggF boosted $\PH \to \PQb \PAQb$ events as a function of the highest particle-level Higgs boson \pt (left) and highest offline-reconstructed AK8 jet \pt (right). The large-radius jet with highest \pt in each event is required to have a maximum angular distance ${\deltar < 0.8}$ from the two final-state \PQb quarks (blue). The events are then required to pass either the standard (red) or scouting (black) trigger selection. The number of events is computed from simulation with projected ${\Lint= 100\fbinv}$.

Number of ggF boosted $\PH \to \PQb \PAQb$ events as a function of the highest particle-level Higgs boson \pt (left) and highest offline-reconstructed AK8 jet \pt (right). The large-radius jet with highest \pt in each event is required to have a maximum angular distance ${\deltar < 0.8}$ from the two final-state \PQb quarks (blue). The events are then required to pass either the standard (red) or scouting (black) trigger selection. The number of events is computed from simulation with projected ${\Lint= 100\fbinv}$.

Invariant mass distribution of opposite-sign muon pairs obtained with the scouting triggers, collected during 2022 with all Run~3 dimuon algorithms (blue curve), and with each individual algorithm (remaining colors).

Comparison between the \lxy distribution for Run~2 (orange) and Run~3 (blue) events in data that contain dimuon pairs with a common displaced vertex and a minimal selection on the vertex quality. The dashed vertical lines, placed at radii of 29, 68, 109, and 160\mm, correspond to the positions of the pixel layers where photons undergo conversion processes in the material, causing the observed peaks in the \lxy distribution.

Resolution on the transverse momentum of scouting muons compared to offline muons using data collected in 2022. Differences in muon momentum scale between the scouting and offline reconstruction algorithms are studied in bins of 1\GeV (10\GeV) for muon \pt smaller (larger) than 60\GeV. Values for the barrel (blue circles) and endcap (orange triangles) sections are shown separately.

Comparison of the dimuon spectra obtained with scouting (pink filled histogram) and offline (blue solid line) muons during the 2022 data-taking period. The ratio between the two distributions with a wider binning is also shown in the bottom panel as a gray band.

Trigger efficiencies for the scouting trigger paths seeded by L1 algorithms targeting either single-electron events (left) or single-photon events (right), as a function of the respective object \pt reconstructed offline. To be considered for scouting, the leading electron or photon must have ${\pt>30\GeV}$. Results are only shown for electrons or photons detected in the barrel region.

Trigger efficiencies for the scouting trigger paths seeded by L1 algorithms targeting either single-electron events (left) or single-photon events (right), as a function of the respective object \pt reconstructed offline. To be considered for scouting, the leading electron or photon must have ${\pt>30\GeV}$. Results are only shown for electrons or photons detected in the barrel region.

Dielectron mass distribution observed with Run~3 scouting data collected during the 2023 data-taking period. The \PJGy and two of the \PGU meson peaks are visible.

The L1 trigger rate and the amount of pileup as a function of time, shown for representative LHC fills during 2017 (\cmsLeft) and 2018 (\cmsRight). Occasional lower rates are observed due to transient effects, such as the throttling of the trigger system in response to subdetector dead time~\cite{CMS:2016ngn}. Changes in the trigger configuration are indicated by vertical green dashed lines.

The L1 trigger rate and the amount of pileup as a function of time, shown for representative LHC fills during 2017 (\cmsLeft) and 2018 (\cmsRight). Occasional lower rates are observed due to transient effects, such as the throttling of the trigger system in response to subdetector dead time~\cite{CMS:2016ngn}. Changes in the trigger configuration are indicated by vertical green dashed lines.

Left: an example scenario in which the \bparking data throughput per L1 trigger setting is adjusted to maintain an average of approximately 2\gbs throughout an LHC fill. The dotted red and dashed blue lines trace the \bparking data throughput and maximum allowed HLT rate, respectively, determined for each trigger configuration. Changes in the trigger configuration are indicated by vertical green dashed lines. The trigger logic is adjusted to operate close to the permitted HLT rate. Right: rate of \bbbar events in acceptance versus HLT rate for a parameter scan over \pt and \ipsig thresholds imposed in the HLT logic for the \Linst and L1 requirements indicated in the legend; each point (blue circle) represents a unique pair of thresholds and the red star indicates the optimal pairing of ${\pt > 12\GeV}$ and ${\ipsig > 6}$ at a peak ${\Linst = \sci{1.7}{34}\invcms}$ and an HLT rate close to the maximum allowed value of ${\approx}1.5\unit{kHz}$.

Left: an example scenario in which the \bparking data throughput per L1 trigger setting is adjusted to maintain an average of approximately 2\gbs throughout an LHC fill. The dotted red and dashed blue lines trace the \bparking data throughput and maximum allowed HLT rate, respectively, determined for each trigger configuration. Changes in the trigger configuration are indicated by vertical green dashed lines. The trigger logic is adjusted to operate close to the permitted HLT rate. Right: rate of \bbbar events in acceptance versus HLT rate for a parameter scan over \pt and \ipsig thresholds imposed in the HLT logic for the \Linst and L1 requirements indicated in the legend; each point (blue circle) represents a unique pair of thresholds and the red star indicates the optimal pairing of ${\pt > 12\GeV}$ and ${\ipsig > 6}$ at a peak ${\Linst = \sci{1.7}{34}\invcms}$ and an HLT rate close to the maximum allowed value of ${\approx}1.5\unit{kHz}$.

HLT trigger rates and the number of pileup events shown as a function of time during a representative LHC fill in 2018. The rates for the promptly reconstructed core physics (black solid markers) and \bparking (blue open markers) data streams are shown separately. Occasional lower rates are observed due to transient effects, such as the throttling of the trigger system in response to subdetector dead time~\cite{CMS:2016ngn}. Changes in the trigger configuration are indicated by vertical green dashed lines.

Mass difference between reconstructed \PDstp and \PDz candidates from the production mode $\PBz \to \PDstp \PGmm \PAGn$ and the subsequent decay chain $\PDstp \to \PDz\PGpp_{\text{soft}} \to (\PKm\PGpp)\PGpp_{\text{soft}}$. Events containing kaon and muon candidates with same-sign (opposite-sign) charges are indicated by solid (open) markers.

The \cmsLeft panel shows the \ptgen spectra of the leading and subleading electrons (dashed and solid green histograms) from \BKee decays, and the efficiency to identify genuine electrons as a function of \ptgen for PF (solid red squares) and low-\pt electron candidates (solid blue circles). Efficiencies for the low-\pt electron candidates that satisfy an ID score threshold, tuned to give the same misidentification probability as for PF electron candidates, are also shown (open markers). The \cmsRight panel shows the performance of the PF (solid red square) and low-\pt (solid blue circle) reconstruction algorithms and their corresponding ID algorithms (curves). The efficiencies and misidentification probabilities are determined relative to charged-particle tracks obtained from simulation, for both \BKee decays and background processes, and satisfying ${\pt > 2\GeV}$ and ${\abs{\eta} < 2.5}$.

The \cmsLeft panel shows the \ptgen spectra of the leading and subleading electrons (dashed and solid green histograms) from \BKee decays, and the efficiency to identify genuine electrons as a function of \ptgen for PF (solid red squares) and low-\pt electron candidates (solid blue circles). Efficiencies for the low-\pt electron candidates that satisfy an ID score threshold, tuned to give the same misidentification probability as for PF electron candidates, are also shown (open markers). The \cmsRight panel shows the performance of the PF (solid red square) and low-\pt (solid blue circle) reconstruction algorithms and their corresponding ID algorithms (curves). The efficiencies and misidentification probabilities are determined relative to charged-particle tracks obtained from simulation, for both \BKee decays and background processes, and satisfying ${\pt > 2\GeV}$ and ${\abs{\eta} < 2.5}$.

The pileup distribution obtained from the \bparking data set. Contributions from each trigger combination are shown, with the histogram areas normalized to the number of events recorded by each trigger.

The invariant mass distribution for pairs of oppositely charged muons originating from a common vertex, obtained from a subset of the \bparking data.

Results of an unbinned likelihood fit to the invariant mass distributions for the \BKJpmm (\cmsLeft) and the \BKmm (\cmsRight) channels. Various functions are used to parametrize the contributions from the signal and various background processes. Taken from Ref.~\cite{CMS:2023klk}.

Results of an unbinned likelihood fit to the invariant mass distributions for the \BKJpmm (\cmsLeft) and the \BKmm (\cmsRight) channels. Various functions are used to parametrize the contributions from the signal and various background processes. Taken from Ref.~\cite{CMS:2023klk}.

Results of an unbinned likelihood fit to the invariant mass distributions for the \BKJpee (\cmsLeft) and the \BKee (\cmsRight) channels. The upper (lower) panels are for candidates using the PF-PF (PF-LP) category. Various functions are used to parametrize the contributions from the signal and various background processes. Taken from Ref.~\cite{CMS:2023klk}.

Results of an unbinned likelihood fit to the invariant mass distributions for the \BKJpee (\cmsLeft) and the \BKee (\cmsRight) channels. The upper (lower) panels are for candidates using the PF-PF (PF-LP) category. Various functions are used to parametrize the contributions from the signal and various background processes. Taken from Ref.~\cite{CMS:2023klk}.

Results of an unbinned likelihood fit to the invariant mass distributions for the \BKJpee (\cmsLeft) and the \BKee (\cmsRight) channels. The upper (lower) panels are for candidates using the PF-PF (PF-LP) category. Various functions are used to parametrize the contributions from the signal and various background processes. Taken from Ref.~\cite{CMS:2023klk}.

Results of an unbinned likelihood fit to the invariant mass distributions for the \BKJpee (\cmsLeft) and the \BKee (\cmsRight) channels. The upper (lower) panels are for candidates using the PF-PF (PF-LP) category. Various functions are used to parametrize the contributions from the signal and various background processes. Taken from Ref.~\cite{CMS:2023klk}.

Comparison of the measured differential \BKmm branching fraction with the theoretical predictions obtained using \textsc{hepfit}, \textsc{superiso}, \textsc{flavio}, and \textsc{eos} packages. Reliable predictions are not available between the \PJGy and \Pgy resonance regions. The \textsc{hepfit} predictions are available only for ${\qsq < 8\GeV^2}$. Taken from Ref.~\cite{CMS:2023klk}.

Feynman diagram of a semileptonic decay of a \PB meson into the primary lepton ($\ell_{P}$), a hadronic system (X), and an SM neutrino, which contains a small admixture of a heavy neutrino (\PN). The \PN decays weakly into a charged lepton $\ell^\pm$ and a charged pion \PGpmp, forming a vertex displaced from the \pp interaction point. Taken from Ref.~\cite{CMS-PAS-EXO-22-019}.

Expected and observed 95\%~\CL limits on \VV as a function of \mN in the Majorana scenario, for the coupling hypotheses (\rehnl, \ruhnl, \rthnl) = (0, 1, 0) on the left and (\rehnl, \ruhnl, \rthnl) = (1/3, 1/3, 1/3) on the right. The mass range with no limits shown corresponds to the \PDz meson veto employed by the search. Taken from Ref.~\cite{CMS-PAS-EXO-22-019}.

Expected and observed 95\%~\CL limits on \VV as a function of \mN in the Majorana scenario, for the coupling hypotheses (\rehnl, \ruhnl, \rthnl) = (0, 1, 0) on the left and (\rehnl, \ruhnl, \rthnl) = (1/3, 1/3, 1/3) on the right. The mass range with no limits shown corresponds to the \PDz meson veto employed by the search. Taken from Ref.~\cite{CMS-PAS-EXO-22-019}.

Observed limits on $c\tau_{\PN}$ as a function of the coupling ratios (\rehnl, \ruhnl, \rthnl) for fixed \PN masses of 1\GeV (upper left), 1.5\GeV (upper right), and 2\GeV (lower center), in the Majorana scenario. A red cross indicates that no exclusion limit was set for that point. The tick orientation indicates the direction of reading. Taken from Ref.~\cite{CMS-PAS-EXO-22-019}.

Observed limits on $c\tau_{\PN}$ as a function of the coupling ratios (\rehnl, \ruhnl, \rthnl) for fixed \PN masses of 1\GeV (upper left), 1.5\GeV (upper right), and 2\GeV (lower center), in the Majorana scenario. A red cross indicates that no exclusion limit was set for that point. The tick orientation indicates the direction of reading. Taken from Ref.~\cite{CMS-PAS-EXO-22-019}.

Observed limits on $c\tau_{\PN}$ as a function of the coupling ratios (\rehnl, \ruhnl, \rthnl) for fixed \PN masses of 1\GeV (upper left), 1.5\GeV (upper right), and 2\GeV (lower center), in the Majorana scenario. A red cross indicates that no exclusion limit was set for that point. The tick orientation indicates the direction of reading. Taken from Ref.~\cite{CMS-PAS-EXO-22-019}.

Dimuon mass spectra obtained from data recorded in 2022 during Run~3, corresponding to ${\Lint = 3.2\fbinv}$. In the range ${2m_{\PGm}^{\text{PDG}} < \mMM < 8.5\GeV}$, the light blue distribution represents the subset of dimuon events triggered by the inclusive low-mass trigger algorithm, while the dark blue distribution shows the subset of dimuon events triggered by the displaced low-mass trigger path. In the range ${8.5 < \mMM < 11.5\GeV}$, dimuon events are instead triggered by the HLT paths targeting the \PGUPnS resonances, which are shown by the pink distribution.

Left: invariant mass distribution for candidate \bdtokshortjpsi decays. Right: proper decay length (ct) distribution obtained from candidate \BKJpmm decays. Both types of candidates are reconstructed from events recorded using the dimuon triggers.

Left: invariant mass distribution for candidate \bdtokshortjpsi decays. Right: proper decay length (ct) distribution obtained from candidate \BKJpmm decays. Both types of candidates are reconstructed from events recorded using the dimuon triggers.

Dimuon invariant mass distributions in the \PGh (\cmsLeft) and \PJGy (\cmsRight) mass regions, as obtained from data recorded by the inclusive low-mass dimuon trigger algorithm.

Dimuon invariant mass distributions in the \PGh (\cmsLeft) and \PJGy (\cmsRight) mass regions, as obtained from data recorded by the inclusive low-mass dimuon trigger algorithm.

Total L1 (\cmsLeft) and HLT (\cmsRight) trigger rates, and the number of pileup interactions, shown as a function of time for a representative LHC fill during 2022. The rates for the promptly reconstructed core physics (black solid markers) and \bparking (blue open markers) data streams are shown separately in the \cmsRight panel. Occasional lower rates are observed due to transient effects, such as the throttling of the trigger system in response to subdetector dead time~\cite{CMS:2016ngn}. Changes in the trigger configuration are indicated by vertical green dashed lines.

Total L1 (\cmsLeft) and HLT (\cmsRight) trigger rates, and the number of pileup interactions, shown as a function of time for a representative LHC fill during 2022. The rates for the promptly reconstructed core physics (black solid markers) and \bparking (blue open markers) data streams are shown separately in the \cmsRight panel. Occasional lower rates are observed due to transient effects, such as the throttling of the trigger system in response to subdetector dead time~\cite{CMS:2016ngn}. Changes in the trigger configuration are indicated by vertical green dashed lines.

Pileup distribution measured in the dielectron data set. Contributions from each trigger combination are shown, with the histogram areas normalized to the number of events recorded by each trigger.

The invariant mass distribution for pairs of oppositely charged electrons originating from a common vertex, reconstructed from the dielectron data set.

The invariant mass distribution for candidate \BKJpee decays, reconstructed from the dielectron data set. The histogram is normalized to unit area.

The L1 (blue) and L1+HLT (red) efficiencies as a function of \mjj for the VBF inclusive (left) and VBF+\ptmiss (right) parking triggers. In the right figure, \ptmiss(no-\PGm) refers to the event \ptmiss corrected for muons.

The L1 (blue) and L1+HLT (red) efficiencies as a function of \mjj for the VBF inclusive (left) and VBF+\ptmiss (right) parking triggers. In the right figure, \ptmiss(no-\PGm) refers to the event \ptmiss corrected for muons.

Distributions of \mjj for VBF $\PH \to \text{invisible}$ events passing the triggers used in the Run~2 analysis (blue), compared to events passing either one of the Run~2 triggers, the VBF+\ptmiss parking trigger, or the VBF inclusive parking trigger implemented in Run 3 (black). The Run~2 trigger selection includes the VBF+\ptmiss trigger algorithm introduced in Run~2, plus the standard trigger requiring $\ptmiss > 120\GeV$ and $\mht> 120\GeV$. In all cases, loose offline selections are applied to match the trigger-level requirements.

Trigger efficiency for selecting signal $\PH\PH$ events, plotted as a function of the reconstructed invariant mass of the two Higgs bosons, as measured in simulated ${\PH\PH \to 4\PQb}$ (left) and ${\PH\PH \to 2\PQb2\PGt}$ (right) samples corresponding to nominal Run~3 conditions.

Trigger efficiency for selecting signal $\PH\PH$ events, plotted as a function of the reconstructed invariant mass of the two Higgs bosons, as measured in simulated ${\PH\PH \to 4\PQb}$ (left) and ${\PH\PH \to 2\PQb2\PGt}$ (right) samples corresponding to nominal Run~3 conditions.