Measurement of the inclusive $\mathrm{t}\overline{\mathrm{t}}$ cross section in pp collisions at $\sqrt{s} =$ 5.02 TeV using final states with at least one charged lepton

The top quark pair production cross section ($\sigma_{\mathrm{t}\overline{\mathrm{t}}}$) is measured for the first time in pp collisions at a center-of-mass energy of 5.02 TeV. The data were collected by the CMS experiment at the LHC and correspond to an integrated luminosity of 27.4 pb$^{-1}$. The measurement is performed by analyzing events with at least one charged lepton. The measured cross section is $ \sigma_{\mathrm{t}\overline{\mathrm{t}}} = 69.5 \pm 6.1$ (stat) $\pm 5.6$ (syst) $\pm 1.6$ (lumi) pb, with a total relative uncertainty of 12%. The result is in agreement with the expectation from the standard model. The impact of the presented measurement on the determination of the gluon distribution function is investigated.


Introduction
The top quark, the heaviest elementary particle in the standard model (SM), has been the subject of numerous detailed studies using hadron-hadron collisions. The pair production (tt) cross section (σ tt ) as a function of center-of-mass energy can be of interest for the extraction of the top quark pole mass [1] and has been used to constrain the gluon distribution function [2] at large fractions x of the proton longitudinal momentum carried by the gluon, where the gluon distribution is poorly known. Precise measurements of σ tt in proton-proton (pp) collisions have been published at √ s values of 7 and 8 [3-6] and 13 TeV [7-10] by the ATLAS and CMS Collaborations at the LHC.
In November 2015, the LHC delivered pp collisions at √ s = 5.02 TeV. The fraction of tt events initiated by gluon-gluon collisions grows monotonically with √ s. It is around 73% at 5.02 TeV, as calculated with POWHEG (v2) [11][12][13] at next-to-leading order (NLO) using the NNPDF3.0 NLO [14] parton distribution functions (PDFs), and increases to around 86% at 13 TeV, making this new data set partially complementary to the higher-energy samples. Measurements of tt production at various √ s probe different values of x and thus can provide complementary information on the gluon distribution. In addition, future measurements of σ tt in nuclear collisions at the same nucleon-nucleon center-of-mass energy [15,16] would profit from the availability of a reference measurement in pp collisions at √ s = 5.02 TeV, without the need to extrapolate from measurements at different √ s. This has already been demonstrated with the first observation of the tt process using proton-nucleus collisions at a higher nucleon-nucleon center-of-mass energy [17].
In the SM, top quarks in pp collisions are mostly produced as tt pairs. Each top quark decays predominantly to a W boson and a bottom (b) quark. The tt events are categorized according to the decay of the two W bosons. In tt events where one W boson decays leptonically and the other hadronically ( +jets channel), the final state presents a typical signature of one isolated lepton, missing transverse momentum, two jets from the W boson hadronic decay, and two jets coming from the hadronization of the b quarks ("b jets"). On the other hand, in tt events where both W bosons decay leptonically (dilepton channel), the final state contains two leptons of opposite electric charge, missing transverse momentum, and at least two b jets. The +jets channel has a large branching ratio with a moderate amount of background, while the dilepton channel is characterized by a high purity. This analysis represents the first measurement of σ tt in pp collisions at √ s = 5.02 TeV using tt candidate events with +jets, where leptons are either electrons ( = e) or muons ( = µ), and dilepton (e ± µ ∓ or µ ± µ ∓ ) final states. In the former case, σ tt is extracted by a fit to the distribution of a kinematic variable for different categories of lepton flavor and jet multiplicity, while in the latter an event counting approach is used. The two results are then combined in the final measurement, which is used as input to a quantum chromodynamics (QCD) analysis at next-to-next-to-leading order (NNLO) to investigate the impact on the determination of the gluon distribution in the less-explored kinematic range of x 0.1. This paper is structured as follows. Section 2 describes the CMS detector. Section 3 gives a summary of the data and simulated samples used. After the discussion of the object reconstruction in Section 4, and of the trigger and event selection in Section 5, Section 6 describes the determination of the background sources. The systematic uncertainties are discussed in Section 7. The extraction of σ tt is presented in Section 8 and the impact of the presented measurement on the determination of the proton PDFs is discussed in Section 9. A summary of all the results is given in Section 10.
The MADGRAPH5 aMC@NLO (v5 2.2.2) generator [25] is used to simulate W boson production with additional jets (W+jets), and high-mass (>50 GeV) Drell-Yan quark-antiquark annihilation into lepton-antilepton pairs through Z boson or virtual-photon exchange (referred to as "Z/γ * "). The simulation includes up to two extra partons at matrix element level, and the FXFX merging procedure [26] is used to interface with PYTHIA. Low-mass Z/γ * events  are simulated with PYTHIA. The normalization of the W+jets and Z/γ * processes is either derived from data (in the dilepton channel) or estimated based on the NNLO cross sections (in the +jets channel) from the FEWZ program (v3.1.b2) [27]. Single top quark plus W boson events (tW) are simulated using POWHEG (v1) [28,29] interfaced with PYTHIA, and are normalized to the approximate NNLO cross sections [30]. The contributions from WW and WZ production (referred to as "WV") are simulated with PYTHIA, and are normalized to the NLO cross sections calculated with the MCFM (v8.0) program [31]. All generated events undergo a full GEANT4 [32] simulation of the detector response.
The expected signal yields are normalized to the value of the SM prediction for the tt production cross section: σ NNLO = 68.9 +1.9 −2.3 (scale) ± 2.3 (PDF) +1.4 −1.0 (α s ) pb, as calculated with the TOP++ program [33] at NNLO in perturbative QCD, including softgluon resummation at next-to-next-to-leading-logarithmic order [34], using the NNPDF 3.0 NNLO PDF set, with α s (M Z ) = 0.118 and m top = 172.5 GeV. The systematic uncertainties in the theoretical tt cross section are associated with the choice of the renormalization (µ R ) and factorization (µ F ) scales-nominally set at µ R = µ F = √ m 2 top + p 2 T,top with p T,top the top quark transverse momentum-as well as with the PDF set and the α s value. The uncertainty of 0.1% in the LHC beam energy [35] translates into an additional uncertainty of 0.22 pb in the expected cross section, with negligible impact on the acceptance of any of the channels included in this analysis.

Object reconstruction
The particle-flow (PF) algorithm [36] is used to reconstruct and identify individual particles using an optimized combination of information from the various elements of the CMS detector.
The electron momentum is calculated by combining the energy measurement in the ECAL with the momentum measurement in the tracker, taking into account the bremsstrahlung photons spatially compatible with originating from the electron track. The momentum resolution for electrons with transverse momentum p T ≈ 45 GeV from Z → ee decays ranges from 1.7% for nonshowering electrons in the barrel region to 4.5% for showering electrons in the endcaps [37]. Muon candidates are reconstructed from a combination of the information collected by the muon spectrometer and the silicon tracker. This results in a relative p T resolution of 1.3-2.0% in the barrel and better than 6% in the endcaps, for muons with 20 < p T < 100 GeV and within the range |η| < 2.4 [38,39]. The photon energy is directly obtained from the ECAL measurement, corrected for zero-suppression effects. The charged hadron energies are determined from a combination of their momenta measured in the tracker and the matching ECAL and HCAL energy deposits, corrected for zero-suppression effects and for the response function of the calorimeters to hadronic showers. Finally, the neutral hadron energies are obtained from the corresponding corrected ECAL and HCAL energies.
The missing transverse momentum vector is defined as the negative vector sum of the momenta of all reconstructed PF candidates in an event, projected onto the plane perpendicular to the direction of the proton beams. Its magnitude is referred to as p miss T and the corrections to jet momenta are propagated to the p miss T calculation [40].
The reconstructed vertex with the largest value of summed physics-object p 2 T is taken to be the primary pp interaction vertex. The physics objects are the jets, clustered using the jet finding algorithm [41,42] with the tracks assigned to the vertex as inputs, and the associated p miss T . The isolation of electron and muon candidates from nearby jet activity is then evaluated as follows. For electron and muon candidates, a cone of ∆R = 0.3 and 0.4, respectively, is constructed around the direction of the lepton track at the primary event vertex, where ∆R is defined as √ (∆η) 2 + (∆φ) 2 , and ∆η and ∆φ are the differences in pseudorapidity and azimuthal angle between the directions of the lepton and another particle. A relative isolation discriminant, I rel , is calculated by the ratio between the scalar p T sum of all particle candidates inside the cone consistent with originating from the primary vertex and the p T of the lepton candidate. In this sum, we exclude the p T of the lepton candidate. The neutral particle contribution to I rel is corrected for energy deposits from pileup interactions using different techniques for electrons and muons. For muons, half of the total p T of the charged hadron PF candidates not originating from the primary vertex is subtracted. The factor of one half accounts for the different fraction of charged and neutral particles in the cone. For electrons, the FASTJET technique [43] is used, in which the median of the energy-density distribution of neutral particles (within the area of any jet in the event) multiplied by the geometric area of the isolation cone-scaled by a factor that accounts for the residual η-dependence of the average energy deposition due to pileup-is subtracted.
The efficiency of the lepton selection is measured using a "tag-and-probe" method in sameflavor dilepton events enriched in Z boson candidates, following the method of Ref. [44]. The sample of Z → µ + µ − events used for muon efficiency extraction is selected by the same trigger requirement used by the main analysis (Section 5). The Z → e + e − sample for electron efficiency extraction makes use of events that satisfy a diphoton trigger with symmetric transverse energy, E T = ∑ i E i sin θ i , thresholds of E T = 15 GeV covering the full tracker acceptance, where E i is the energy seen by the calorimeters for the ith particle, θ i is the polar angle of particle i, and the sum is over all particles emitted into a fixed solid angle in the event. Pairs of photon candidates above the E T threshold are accepted only if their invariant mass is above 50 GeV. The trigger selection requires a loose identification using cluster shower shapes and a selection based on the ratio of the hadronic to the electromagnetic energy of the photon candidates. Based on a comparison of the lepton selection efficiency in data and simulation, the event yield in simulation is corrected using data-to-simulation scale factors.
Jets are reconstructed from the PF candidates using the anti-k T clustering algorithm [41] with a distance parameter of 0.4. Jets closer than ∆R = 0.3 to the nearest muon or electron are discarded. Jet energy corrections extracted from full detector simulation are also applied as a function of jet p T and η [45] to data and simulation. A residual correction to the data is applied to account for the discrepancy between data and simulation in the jet response.

Event selection
The event sample is selected by a loose online trigger and further filtered offline to remove noncollision events, such as beam-gas interactions or cosmic rays. Collision events containing one high-p T electron (muon) candidate are selected online by requiring values of E T (p T ) greater than 40 (15) GeV and of |η| less than 3.1 (2.5). The measured trigger efficiency for each decay channel, relative to the final selection, is higher than 90%.
In the +jets analysis, electron candidates are selected if they have p T > 40 GeV and |η| < 2.1. Further identification and isolation criteria are applied to the electron candidates. Electrons reconstructed in the ECAL barrel (endcap) are required to have I rel < 4 (5)%. Electron candidates in the 1.44 < |η| < 1.57 region, i.e., in the transition region between the barrel and endcap sections of the ECAL, are excluded because the reconstruction of an electron object in this region is less efficient. Muons are required to have p T > 25 GeV and |η| < 2.1. Additional identification criteria are applied and I rel is required to be < 15%. Events are rejected if they contain extra electrons or muons identified using a looser set of identification criteria and have p T > 10 or 15 GeV, respectively.
The distinct signature of two b jets, expected in tt decays, is rare in background events, and thus is exploited in the +jets analysis. Backgrounds from W+jets, QCD multijet, and Z/γ * events are controlled by counting the number of b jets in the selected events. In addition, two lightflavor jets are expected to be produced in the decay of one of the W bosons for signal events. The correlation in phase space of these light jets carries a distinctive hallmark with respect to the main backgrounds. To that end, jets are selected if they have p T > 30 GeV and |η| < 2.4. The flavor of the jets is identified using a combined secondary vertex algorithm [46] with an operating point that yields a b jet identification efficiency of about 70%, and misidentification (mistag) probabilities of about 1% and 15% for light-flavor (u, d, s, and gluons) and c jets, respectively. The event selection requires at least two non-b-tagged jets to be identified as candidates from the W boson hadronic decay. Additional jets passing the b quark identification criteria are counted and used to classify the selected events in none (0 b), exactly one (1 b), or at least two (≥2 b) tagged jet categories. The efficiency of the b jet identification algorithm is measured in situ, simultaneously with the signal cross section.
Dilepton events are required to contain at least one muon candidate at trigger level. No requirement on the presence of electron candidates is made at trigger level owing to the relatively high-E T threshold (40 GeV) of the trigger. Electrons are selected if they have p T > 20 GeV, |η| < 2.4, and I rel < 9 (or 12)% if in the barrel (or one of the endcaps). As in the +jets channel, electrons detected in the transition region between the barrel and endcap sections of the ECAL are excluded. Muons are required to have p T > 18 GeV, |η| < 2.1, and I rel < 15%. At least two jets satisfying the criteria p T > 25 GeV and |η| < 3 are required. Events are subsequently selected if they have a pair of leptons with opposite charge (e ± µ ∓ or µ ± µ ∓ ) passing the requirements listed above. In events with more than one pair of leptons passing the above selection, the two leptons of opposite charge that yield the highest scalar p T sum are selected.
Candidate events with dilepton invariant masses of M < 20 GeV are removed to suppress events from decays of heavy-flavor resonances and low-mass Z/γ * processes. Dilepton events with two muons in the final state are still dominated by the Z/γ * background. In order to suppress this contribution, events in the Z boson mass window of 76 < M < 106 GeV are vetoed in this channel. To further suppress the Z/γ * events, a requirement on p miss T of >35 GeV is imposed.
In both the +jets and dilepton analyses, events with τ leptons are considered as signal if they decay to electrons or muons that satisfy the selection requirements, and are included in the simulation.
6 Background estimation

The +jets final state
In the +jets analysis, the contributions of all background processes are estimated from simulation, with the exception of the QCD multijet background. Due to its large cross section, there is a nonnegligible contribution from the latter faking a tt event with +jets in the final state. Both the contribution from hard fragmentation of c and b quarks whose hadrons decay semileptonically, and the contribution from misidentified leptons, such as from either punch-through hadrons or collimated jets with a high electromagnetic fraction, can yield +jets-like topologies.
The estimation of the QCD multijet background is separately performed for the events with 0, 1, or ≥2 b jets using a control region where either the muon candidate fails a looser isolation requirement (I rel < 20%) or the electron candidate fails the identification criteria. The choice of the QCD multijet control region has been made in such a way as to minimize the contamination due to the signal and W+jets events, while retaining a large number of events in the sample for the estimation of this type of background. The initial normalization of the QCD multijet contribution in the signal region is derived from events with p miss T < 20 GeV ("reduced-signal" region). Events in both the reduced-signal and control regions fulfilling this requirement are counted. After subtracting the expected contributions from non-QCD processes, the ratio between the numbers of events observed in the reduced-signal region and in the control region, is used as a transfer factor to normalize the QCD multijet background estimate. In both the electron and muon channels, a 30% uncertainty is assigned to the estimate of the expected contribution from non-QCD processes, estimated after varying the QCD scales in the W+jets simulation. This uncertainty propagates as both a normalization and a shape uncertainty in the predicted distributions for the QCD multijet processes. The variations are applied independently in the reduced-signal and control regions in order to determine an uncertainty envelope. A more accurate normalization for this contribution is obtained by the fit performed to extract the final cross section, described in Section 8.1.

The dilepton final state
Final states with two genuine leptons can originate from background processes, primarily from Z/γ * → τ + τ − (where the τ leptonic decays can yield e ± µ ∓ or µ ± µ ∓ plus p miss T due to the neutrinos), tW, and WV events. Other background sources, such as W+jets events or tt production in the +jets final state, can contaminate the signal sample if a jet is misidentified as a lepton, or if an event contains a lepton from the decay of b or c hadrons. These are included in the "non-W/Z" category, since genuine leptons are defined as originating from decays of W or Z bosons. The yields from tW and WV events are estimated from simulation, while the contribution of the Z/γ * background is evaluated using control samples in data. The rate of non-W/Z backgrounds is extracted from control samples in data for the e ± µ ∓ channel and is estimated from simulation for the µ ± µ ∓ channel.
A scale factor for the Z/γ * background normalization is estimated, as in Ref.
[47], from the number of events within the Z boson mass window in data, which is extrapolated to the number of events outside the window. A scale factor of 0.91 ± 0.14 (stat) is obtained in the e ± µ ∓ channel, and 0.96 ± 0.78 (stat) in the µ ± µ ∓ channel. The estimation is performed using events with at least two jets, and the dependence on different jet multiplicities is discussed in Section 7.
The non-W/Z background in the e ± µ ∓ channel is estimated using an extrapolation from a control region of same-sign (SS) dilepton events to the signal region of opposite-sign (OS) dileptons. The SS control region is defined using the same criteria as for the nominal signal region, except requiring dilepton pairs of the same charge. The muon isolation requirement is relaxed in order to enhance the number of events. The SS dilepton events predominantly contain at least one misidentified lepton. Other SM processes produce genuine SS or charge-misidentified dilepton events with significantly smaller rates; these are estimated using simulation and subtracted from the observed number of events in data. The scaling from the SS control region in data to the signal region is performed using an extrapolation factor extracted from MC simulation, given by the ratio of the number of OS events with misidentified leptons to the number of SS events with misidentified leptons. The resulting estimate for the non-W/Z background is 1.0 ± 0.9 (stat) events, where the central value comes from the estimation using events with at least two jets. No particular dependence of this scale factor is observed for different jet multiplicities within the large statistical uncertainty.

Systematic uncertainties
The integrated luminosity has been estimated offline using a pixel cluster counting method [20]. The estimation takes into account normalization uncertainties and uncertainties related to the different conditions during typical physics periods relative to the specially tailored beamseparation scans, adding up to a total uncertainty of ±2.3%.
The uncertainties in the electron trigger efficiency (1.5%) and the identification and isolation efficiency (2.5%) are estimated by changing the values of the data-to-simulation scale factors within their uncertainties, as obtained from the "tag-and-probe" method. The uncertainty in the muon identification and isolation efficiency, including the trigger efficiency, is 3% and covers one standard deviation of the scale factor from unity. . Finally, a 30% uncertainty is conservatively assigned to the jet misidentification probability in the +jets analysis, as no dedicated measurement of this quantity has been performed for the considered data set.
Theoretical uncertainties in the simulation of tt production cause a systematic bias related to the missing higher-order diagrams in POWHEG, which is estimated through studies of the signal modeling by modifying the µ R , µ F scales within a factor of two with respect to their nominal value. In the +jets analysis, the impact of the µ R , µ F variations are examined independently, while in the dilepton analysis they are varied simultaneously. In both analyses, these variations are applied independently at the matrix element (ME) and parton shower (PS) levels. The uncertainty arising from the hadronization model mainly affects the JES and the fragmentation of jets. The hadronization uncertainty is determined by comparing samples of events generated with POWHEG, where the hadronization is either modeled with PYTHIA or HER-WIG++ (v2.7.1) [49]. This also accounts for differences in the PS model and the underlying event. The uncertainty from the choice of PDF is determined by reweighting the sample of simulated tt events according to the root-mean-square (RMS) variation of the NNPDF3.0 replica set. Two extra variations of α s are added in quadrature to determine the total PDF uncertainty.
In the +jets analysis, the uncertainty in the choice of the µ R , µ F scales in the W+jets simulation is taken into account by considering alternative shapes and yields after varying independently the µ R , µ F scales, following a similar procedure to that described above for the signal. Due to the finite event count in the W+jets simulated sample, an additional bin-by-bin uncertainty is assigned by generating an alternative shape to fit (see Section 8.1), where the bin prediction is varied by ±1 standard deviation, while keeping all the other bins at their nominal expectation. The uncertainty assigned to the QCD multijet background includes the statistical uncertainty in the data, and the uncertainty from the non-QCD multijet contributions subtracted from the control region, as described in Section 6.1, and an additional 30%-100% normalization uncertainty. The latter depends on the event category and stems from the measured difference with respect to an alternative estimate of the QCD normalization based on the transverse mass, m T , of the lepton and p miss T system. The magnitude of m T equals √ 2p T p miss T (1 − cos ∆φ), where p T is the lepton transverse momentum and ∆φ is the azimuthal angle between the lepton and the direction of p miss T . Finally, a 30% normalization uncertainty in the theoretical tW, Z/γ * , and WV background cross sections is assigned [5], given the previously unexplored √ s value and that the final states contain several jets.
In the dilepton channel, an uncertainty of 30% is assumed [5] for the cross sections of the tW and WV backgrounds to cover the theoretical uncertainties and the effect of finite simulated samples. The uncertainty in the Z/γ * estimation is calculated by combining in quadrature the statistical uncertainty and an additional 30% from the variation of the scale factor in the different levels of selection, resulting in uncertainties of about 30 and 80% in the e ± µ ∓ and µ ± µ ∓ channels, respectively. The systematic uncertainty in the non-W/Z background is estimated to be 90% in the e ± µ ∓ channel and is dominated by the statistical uncertainty in the method. Owing to the limited sample size in the data, the method cannot be applied in the µ ± µ ∓ channel. The estimation is therefore based on MC simulation, and an uncertainty of 100% is conservatively assigned.
8 Measurement of the tt cross section

The +jets final state
In the +jets analysis, the tt cross section is measured in a fiducial phase space by means of a fit. Two variables were independently considered for the fit, which are sensitive to the resonant behavior of the light jets produced from the W boson hadronic decay in a tt event. Given that these light jets, here denoted by j and j , are correlated during production, they are also expected to be closer in phase space when compared to pairs of other jets in the event. The angular distance ∆R can thus be used as a metric to rank all pairs of non-b-tagged jets in the event, maximizing the probability of selecting those from the W boson hadronic decay in cases where more than two non-b-tagged jets are found. From simulation we expect that the signal peaks at low ∆R, while the background is uniformly distributed up to ∆R ≈ 3. Above that value, fewer events are expected and background processes are predicted to dominate. The invariant mass M(j, j ) of jets j and j also has a distinctive peaking feature for the signal in contrast with a smooth background continuum. From simulation we expect that the minimum angular distance ∆R between all pairs of jets j and j , ∆R min (j, j ), is robust against signal modeling uncertainties such as the choice of the µ R , µ F scales and jet energy scale and resolution, while the M(j, j ) variable tends to be more affected by such uncertainties. Owing to its more robust systematic uncertainties and signal-to-background discrimination power, the ∆R min (j, j ) variable is used to extract the tt cross section.
In order to maximize the sensitivity of the analysis, the ∆R min (j, j ) distributions are categorized according to the number of jets-in addition to the ones assigned to the W boson hadronic decay-passing the b quark identification criteria. In total, 6 categories are used, corresponding to electron or muon events with 0, 1, or ≥2 b jets. The expected number of signal and background events in each category prior to the fit and the observed yields are given in Table 1. Good agreement is observed between data and expectations.
The M(j, j ) and ∆R min (j, j ) distributions are shown in Fig. 1. The distributions have been combined for the e+jets and µ+jets channels to maximize the statistical precision and are shown for events with different b-tagged jet multiplicities. Fair agreement is observed between data and the pre-fit expectations.
A profile likelihood ratio (PLR) method, similar to the one employed in Ref.
[10], is used to perform the fit. In addition, a scale factor for the b tagging efficiency (SF b ) is included as a parameter of interest in the fit. The PLR is written as:  where µ = σ/σ theo is the signal strength (ratio of the observed tt cross section to the expectation from theory) and Θ is a set of nuisance parameters that encode the effect on the expectations due to variations in the sources of the systematic uncertainties described in Section 7. The Figure 2 (left) shows the two-dimensional contours at the 68% confidence level (CL) obtained from the scan of −2 ln(λ), as functions of µ and SF b . The expected results, obtained using the Asimov data set [50], are compared to the observed results and found to be in agreement well within one standard deviation. The signal strength is obtained after profiling SF b and the result is µ = 1.00 +0.10 −0.09 (stat) +0.09 −0.08 (syst). As a cross-check, the signal strength is also extracted by fitting only the total number of events observed in each of the six categories. The observed value µ = 1.03 +0.10 −0.10 (stat) +0.21 −0.11 (syst) is in agreement with the analysis using the ∆R min (j, j ) distributions. Figure 2 (right) summarizes the results obtained for the signal strength fit in each channel separately from the analysis of the distributions and from event counting. In both cases, a large contribution to the uncertainty is systematic in nature, although the statistical component is still significant. In the +jets combination, the µ+jets channel is expected and observed to carry the largest weight.
In order to estimate the impact of the experimental systematic uncertainties in the measured signal strength, the fit is repeated after fixing one nuisance parameter at a time at its postfit uncertainty (±1 standard deviation) values. The impact on the signal strength fit is then evaluated from the difference induced in the final result from this procedure. By repeating the fits, the effect of some nuisance parameters being fixed may be reabsorbed by a variation of the ones being profiled, owing to correlations. As such, the individual experimental uncertainties obtained and summarized in Table 2 can only be interpreted as the observed post-fit values, Table 2: Estimated impact of each source of uncertainty in the value of µ extracted from the analysis of distributions, and in the cross-check from event counting. The "Other background" component includes the contributions from Z/γ * , tW, and WV events. The total uncertainty is obtained by adding in quadrature the statistical, experimental systematic, and theoretical uncertainties. The individual experimental uncertainties are obtained by repeating the fit after fixing one nuisance parameter at a time at its post-fit uncertainty (±1 standard deviation)  and not as an absolute, orthogonalized breakdown of uncertainties. With respect to the event counting, the analysis of the distributions is less prone to the uncertainties in the QCD multijet background, jet energy resolution, and signal modeling. In both cases, the signal modeling uncertainties and the b tagging efficiency are among the largest sources of uncertainty.
The combined acceptance in the e+jets and µ+jets channels is estimated using the NLO POWHEG simulation to be A = 0.301 ± 0.007, with the uncertainty being dominated by the variation of the µ R , µ F scales at ME and PS levels and the hadronization model used for the tt signal. The uncertainty due to the PDFs is included but verified to be less important. Taking into account the acceptance of the analysis and its uncertainty, the inclusive tt cross section is determined to be σ tt = 68.9 ± 6.5 (stat) ± 6.1 (syst) ± 1.6 (lumi) pb, in agreement with the SM prediction and attaining a 13% total relative uncertainty.

The dilepton final state
In the dilepton analysis, the tt cross section is extracted from an event counting measurement. Figure 3 shows the distributions of the jet multiplicity and the scalar p T sum of all jets (H T ), for events passing the dilepton criteria in the e ± µ ∓ channel. In addition, it displays the lepton-pair invariant mass and p T distributions, after requiring at least two jets in the event in the e ± µ ∓ channel. Figure 4 shows the p miss T and the lepton-pair invariant mass distributions in the µ ± µ ∓ channel for events passing the dilepton criteria, and the Z boson veto with the p miss T > 35 GeV requirement, in the second case. The predicted distributions take into account the efficiency corrections described in Section 5 and the background estimations discussed in Section 6.2. Good agreement is observed between the data and predictions for both signal and background.
The fiducial tt production cross section is measured by counting events in the visible phase space (defined by the same p T , |η|, and multiplicity requirements for leptons and jets as described in Section 5, but including the transition region for electrons) and is denoted by σ fid . It is extrapolated to the full phase space in order to determine the inclusive tt cross section using the expression  where N is the total number of dilepton events observed in data, N B the number of estimated background events, ε the selection efficiency, A the acceptance, and L the integrated luminosity. Table 3 gives the total number of events observed in data, together with the total number of signal and background events expected from simulation or estimated from data, after the full set of selection criteria. The total detector, trigger, and reconstruction efficiency is estimated from data to be ε = 0.55 ± 0.02 (0.57 ± 0.04) in the e ± µ ∓ (µ ± µ ∓ ) channel. Using the definitions above, the yields from Table 3, and the systematic uncertainties from Table 4, the measured fiducial cross section for tt production is σ fid = 41 ± 10 (stat) ± 2 (syst) ± 1 (lumi) pb in the e ± µ ∓ channel and σ fid = 22 ± 11 (stat) ± 4 (syst) ± 1 (lumi) pb in the µ ± µ ∓ channel.
The acceptance, as estimated from MC simulation, is found to be A = 0.53 ± 0.01 (0.37 ± 0.01) in the e ± µ ∓ (µ ± µ ∓ ) channel. The statistical uncertainty (from MC simulation) is included in Table 4: Summary of the individual contributions to the systematic uncertainty in the σ tt measurements for the dilepton channels. The relative uncertainties ∆σ tt /σ tt (in %), as well as absolute uncertainties in σ tt , ∆σ tt (in pb), are presented. The statistical and total uncertainties are also given, where the latter are the quadrature sum of the statistical and systematic uncertainties. the uncertainty in A. By extrapolating to the full phase space, the inclusive tt cross section is measured to be σ tt = 77 ± 19 (stat) ± 4 (syst) ± 2 (lumi) pb in the e ± µ ∓ channel and σ tt = 59 ± 29 (stat) ± 11 (syst) ± 1 (lumi) pb in the µ ± µ ∓ channel. Table 4 summarizes the relative and absolute statistical and systematic uncertainties from different sources contributing to σ tt . The separate total systematic uncertainty without the uncertainty in the integrated luminosity, the part attributed to the integrated luminosity, and the statistical contribution are added in quadrature to obtain the total uncertainty. The cross sections, measured with a relative uncertainty of 25 and 52%, are in agreement with the SM prediction (Eq. (1)) within the uncertainties in the measurements.

Combination
The three individual σ tt measurements are combined using the BLUE method [51,52] to determine an overall tt cross section. All systematic uncertainties are considered as fully correlated across all channels, with the following exceptions: the uncertainty associated with the finite event size of the simulated samples is taken as uncorrelated; the electron identification is not relevant for the µµ channel; and the b tagging and QCD multijet background uncertainties are only considered for the +jets channel. In the +jets channel, the WV and Z/γ * backgrounds are not considered separately but as part of the "Other backgrounds" component, which is dominated by tW events. The uncertainty associated with this category is therefore treated as fully correlated with the tW uncertainty in the dileptonic channels and uncorrelated with the WV and Z/γ * uncertainties.
The combined inclusive tt cross section is measured to be: σ tt = 69.5 ± 6.1 (stat) ± 5.6 (syst) ± 1.6 (lumi) pb = 69.5 ± 8.4 (total) pb, where the total uncertainty is the sum in quadrature of the individual uncertainties. The weights of the individual measurements, to be understood in the sense of Ref. [52], are 81.8% for +jets, 13.5% for e ± µ ∓ , and 4.7% for µ ± µ ∓ channels.
The combined result is found to be robust by performing an iterative variant of the BLUE method [53] and varying some assumptions on the correlations of different combinations of systematic uncertainties. Also, the post-fit correlations between the nuisance parameters in the +jets channel have been checked and found to have negligible impact.

QCD analysis
To illustrate the impact of the σ tt measurements at √ s = 5.02 TeV on the knowledge of the proton PDFs, the results are used in a QCD analysis at NNLO, together with the combined measurements of neutral-and charged-current cross sections for deep inelastic electron-and positron-proton scattering (DIS) at HERA [57], and the CMS measurement [58] of the muon charge asymmetry in W boson production at √ s = 8 TeV. The latter data set is used in order to improve the constraint on the light-quark distributions.
Version 2.0.0 of XFITTER [59,60], the open-source QCD-analysis framework for PDF determination, is employed, with the partons evolved using the Dokshitzer-Gribov-Lipatov-Altarelli-Parisi equations [61][62][63][64][65][66] at NNLO, as implemented in the QCDNUM 17-01/13 program [67]. The treatment and the choices for the central values and variations of the c and b quark masses, the strong coupling, and the strange-quark content fraction of the proton follow that of earlier CMS analyses, e.g., Ref. [58]. The µ R , µ F scales are set to the four-momentum transfer in the case of the DIS data, the W boson mass for the muon charge asymmetry results, and the top quark mass in the case of σ tt .
The systematic uncertainties in all three measurements of σ tt and their correlations are treated the same way as in the combination described in Section 8.3. The theoretical predictions for σ tt are obtained at NNLO using the HATHOR calculation [68], assuming m top = 172.5 GeV. The bin-to-bin correlations of the experimental uncertainties in the muon charge asymmetry and DIS measurements are taken into account. The theoretical predictions for the muon charge asymmetry are obtained as described in Ref. [58]. The procedure for the determination of the PDFs follows the approach used in the QCD analysis of Ref. [58] and results in a 14-parameter fit. The parametrized PDFs are the gluon distribution, xg, the valence quark distributions, xu v , xd v , and the u-type and d-type antiquark distributions, xU, xD. The relations xU = xu and xD = xd + xs are assumed at the initial scale of the QCD evolution Q 2 0 = 1.9 GeV 2 . At this scale, the parametrizations are of the form: The normalization parameters A u v , A d v , and A g are determined by the QCD sum rules, the B parameters are responsible for the small-x behavior of the PDFs, and the C parameters describe the shape of the distribution as x → 1. Additional constraints B U = B D and A U = A D (1 − f s ) are imposed, with f s being the strangeness fraction, s/(d + s), which is set to 0.31 ± 0.08 as Table 5: Partial χ 2 per number of data points, n dp , and the global χ 2 per degrees of freedom, n dof , as obtained in the QCD analysis of DIS data, the CMS muon charge asymmetry measurements, and the σ tt results at √ s = 5.02 TeV from this analysis. For the HERA measurements, the energy of the proton beam (E p ) is listed for each data set, with the electron/positron energy of 27.5 GeV. The correlated part of the global χ 2 value is also given.

Data sets
Partial χ 2 /n dp Global χ 2 /n dof 1387/1145 in Ref. [69], consistent with the value obtained using the CMS measurements of W+c production [70]. Using the measured values for σ tt allows the addition of a new free parameter, D u v , in Eq. (5), as compared to the analysis in Ref. [58].
The predicted and measured cross sections for all the data sets, together with their corresponding uncertainties, are used to build a global χ 2 , minimized to determine the PDF parameters [59,60]. The results of the fit are given in Table 5. The quality of the overall fit can be judged based on the global χ 2 divided by the number of degrees of freedom, n dof . For each data set included in the fit, the partial χ 2 divided by the number of the measurements (data points), n dp , is also provided. The correlated part of χ 2 , also given in Table 5, quantifies the influence of the correlated systematic uncertainties in the fit. The global and partial χ 2 values indicate a general agreement among all the data sets. The somewhat high χ 2 /n dp values for the combined DIS data are very similar to those observed in Ref. [57], where they are investigated in detail.
The experimental uncertainties in the measurements are propagated to the extracted QCD fit parameters using the MC method [71,72]. In this method, 400 replicas of pseudo-data are generated, with measured values for σ tt allowed to vary within the statistical and systematic uncertainties. For each of them, the PDF fit is performed and the uncertainty is estimated as the RMS around the central value. In Fig. 6, the ratio and the relative uncertainties in the gluon distributions, as obtained in the QCD analyses with and without the measured values for σ tt , are shown. A moderate reduction of the uncertainty in the gluon distribution at x 0.1 is observed, once the measured values for σ tt are included in the fit. The uncertainties in the valence quark distributions remain unaffected. All changes in the central values of the PDFs are well within the fit uncertainties.
Possible effects from varying the model input parameters and the initial PDF parametrization are investigated in the same way as in the similar analysis of Ref. [58]. The two cases when the measured values for σ tt are included or excluded from the fit are considered, resulting in the same associated model and parametrization uncertainties.
In conclusion, the σ tt measurements at √ s = 5.02 TeV provide improved uncertainties in the gluon PDF at high x, though the impact is small, owing to the large experimental uncertainties.  Figure 6: The relative uncertainties in the gluon distribution function of the proton as a function of x at µ 2 F = 10 5 GeV 2 from a QCD analysis using the HERA DIS and CMS muon charge asymmetry measurements (hatched area), and also including the CMS σ tt results at √ s = 5.02 TeV (solid area). The relative uncertainties are found after the two gluon distributions have been normalized to unity. The solid line shows the ratio of the gluon distribution function found from the fit with the CMS σ tt measurements included to that found without.

Summary
The first measurement of the top quark pair (tt) production cross section in pp collisions at √ s = 5.02 TeV is presented for events with one or two leptons and at least two jets, using a data sample collected by the CMS experiment, corresponding to an integrated luminosity of 27.4 ± 0.6 pb −1 . The final measurement is obtained from the combination of the measurements in the individual channels. The result is σ tt = 69.5 ± 6.1 (stat) ± 5.6 (syst) ± 1.6 (lumi) pb, with a total relative uncertainty of 12%, which is consistent with the standard model prediction. The impact of the measured tt cross section in the determination of the parton distribution functions of the proton is studied in a quantum chromodynamics analysis at next-to-next-toleading order. A moderate decrease of the uncertainty in the gluon distribution is observed at high values of x, the fractional momentum of the proton carried by the gluon.
Grid for delivering so effectively the computing infrastructure essential to our analyses. Finally, we acknowledge the enduring support for the construction and operation of the LHC and the CMS detector provided by the following funding agencies: BMWFW and FWF (Aus [2] CMS Collaboration, "Measurement of double-differential cross sections for top quark pair production in pp collisions at √ s = 8 TeV and impact on parton distribution functions", Eur. Phys. J. C 77 (2017) 459, doi:10.1140/epjc/s10052-017-4984-5, arXiv:1703.01630. [4] ATLAS Collaboration, "Measurement of the tt production cross-section using eµ events with b-tagged jets in pp collisions at √ s = 7 and 8 TeV with the ATLAS detector", Eur.