Search for new physics in events with same-sign dileptons and jets in pp collisions at sqrt(s) = 8 TeV

A search for new physics is performed based on events with jets and a pair of isolated, same-sign leptons. The results are obtained using a sample of proton-proton collision data collected by the CMS experiment at a centre-of-mass energy of 8 TeV at the LHC, corresponding to an integrated luminosity of 19.5 inverse femtobarns. In order to be sensitive to a wide variety of possible signals beyond the standard model, multiple search regions defined by the missing transverse energy, the hadronic energy, the number of jets and b-quark jets, and the transverse momenta of the leptons in the events are considered. No excess above the standard model background expectation is observed and constraints are set on a number of models for new physics, as well as on the same-sign top-quark pair and quadruple-top-quark production cross sections. Information on event selection efficiencies is also provided, so that the results can be used to confront an even broader class of new physics models.


Introduction
In the standard model (SM), proton-proton collision events having a final state with isolated leptons of the same sign are extremely rare. Searches for anomalous production of same-sign dileptons can therefore be very sensitive to new physics processes that produce this signature copiously. These include supersymmetry (SUSY) [1][2][3], universal extra dimensions [4], pair production of T 5/3 particles (fermionic partners of the top quark) [5], heavy Majorana neutrinos [6], and same-sign top-quark pair production [7,8]. In SUSY, for example, same-sign dileptons occur naturally with the production of gluino pairs, when each gluino decays to a top quark and a top anti-squark, with the anti-squark further decaying into a top anti-quark and a neutralino.
In this paper we describe searches for new physics with same-sign dileptons (ee, eµ, and µµ) and hadronic jets, with or without accompanying missing transverse energy (E miss T ). Our choice of signatures is driven by the following considerations. New physics signals with large cross sections are likely to be produced by strong interactions, and we thus expect significant hadronic activity in conjunction with the two leptons. Astrophysical evidence for dark matter [9] suggests considering SUSY models with R-parity conservation, which provides an excellent dark matter candidate -a stable lightest supersymmetric particle (LSP) that escapes detection. Therefore, a search for this signature involves sizable E miss T due to undetected LSPs. Nevertheless, we also consider signatures without significant E miss T in order to be sensitive to SUSY models with R-parity violation (RPV) [10] which imply an unstable LSP. Beyond these general guiding principles, the choice of signatures is made independently of any particular physics model and, as a result, these signatures can be applied also to probe non-supersymmetric extensions of the SM.
The results reported in this document expand upon a previous search [11] and are based on the proton-proton collision dataset at √ s = 8 TeV collected with the Compact Muon Solenoid (CMS) detector at the Large Hadron Collider (LHC) during 2012, corresponding to an integrated luminosity of 19.5 fb −1 . We consider several final states, characterized by the scalar sum (H T ) of the transverse momenta (p T ) of jets, E miss T , the number of jets, and the number of jets identified as originating from b quarks (b-tagged jets). Additionally, in order to provide coverage for a wide range of generic signatures, we perform the analysis with two different requirements on the lepton p T : the high-p T analysis, where the leptons are selected with a p T requirement of at least 20 GeV, and the low-p T analysis, where the p T threshold is lowered to 10 GeV. While the low-p T leptons extend the sensitivity to scenarios with a compressed spectrum of SUSY particle masses, the high-p T analysis targets models where the leptons are produced via on-shell W or Z bosons, and is less subject to backgrounds with leptons originating from jets. The use of a lower threshold on lepton p T for the low-p T analysis is compensated by a tighter H T requirement. In this respect, the two searches are complementary, even if partially overlapping.
In contrast to the previous analysis [11], the signal regions within each of the low-and high-p T analyses are defined to be exclusive. Furthermore, we increase the number of search regions in order to improve the sensitivity to a wider class of beyond-standard-model (BSM) processes. The selection criteria for the analysis objects and the methods used to estimate the SM backgrounds are largely unchanged from those of our previous same-sign dilepton studies [11][12][13][14].
Tables of observed yields and estimated SM backgrounds are provided for both the high-p T and low-p T analyses in each exclusive signal region. Having found no evidence for a BSM contribution to the event counts, limits are set on a variety of SUSY-inspired models by performing a counting experiment in each exclusive search region. Additionally, results for the jets when selecting events. This number is, however, used in the categorization of events into various signal regions. Kinematic selections for jets, leptons, and b-tagged jets are summarized in Table 1. Events with a third lepton are rejected if the lepton forms an opposite-sign same-flavour pair with one of the first two leptons for which the invariant mass of the pair (m ) satisfies m < 12 GeV (p T > 5 GeV) or 76 < m < 106 GeV (p T > 10 GeV). These requirements are designed to minimize backgrounds from processes with a low-mass bound state or γ * → + − in the final state, as well as multiboson (WZ, ZZ, and triboson) production.
Monte Carlo (MC) simulations, which include pileup effects, are used to estimate some of the SM backgrounds (see Section 5), as well as to calculate the efficiency for various new physics scenarios. All SM background samples are generated with the MADGRAPH 5 [18] program and simulated using a GEANT4-based model [19] of the CMS detector. Signal samples are produced with MADGRAPH 5 using the CTEQ6L1 [20] parton distribution functions (PDF); up to two additional partons are present in the matrix element calculations. Version 6.424 of PYTHIA [21] is used to simulate parton showering and hadronization, as well as the decay of SUSY particles. A signal sample for an RPV model is produced with PYTHIA 6.424. For signal samples, the detector simulation is performed using the CMS fast simulation package [22]. Detailed cross checks are performed to ensure that the results obtained with fast simulation are in agreement with the ones obtained with GEANT-based detector simulation. Simulated events are processed with the same chain of reconstruction programs that is used for data.

Search strategy
The search is based on comparing the number of observed events with the expectation from SM processes in several signal regions (SR) that have different requirements on four discriminating variables: E miss T , H T , the number of jets, and the number of b-tagged jets. We define two sets of signal regions: baseline and final SRs. The former set imposes looser selection requirements, thereby forming a sample of events where the contributions of signal events are expected to be negligible, that is used to validate methods that are employed to predict the background in the final SRs; the latter set is based on tighter selection requirements, making it sensitive to many BSM processes. The interpretation of the results, discussed in Section 8, is primarily based on the final SRs.
Search regions defined in bins of the number of jets and b-tagged jets provide broad coverage of strongly produced SUSY particles, including signatures with low hadronic activity as well as signatures involving third-generation squarks. Additionally, as SUSY models with a small mass splitting between the parent sparticle and the LSP may result in low E miss T , we also define search regions with a looser requirement on E miss T . The high-p T search is ideal for BSM models with an on-shell W boson produced in a new-physics particle decay, but events with an off-

Search strategy
shell W boson can produce low-p T leptons, which is why leptons with transverse momenta as low as 10 GeV are included in this study. Table 2: Definition of the baseline signal regions for the three different requirements on the number of b-tagged jets (N b-jets ). N jets refers to the number of jets in the event. The same naming scheme is used for both the low-and high-p T analyses, which differ only in a looser requirement on H T (in parentheses) for the high-p T analysis.
We define the three baseline signal regions (BSR0, BSR1, and BSR2) for both the low-and highp T analyses, as described in Table 2. The event selection criteria are tightened and the granularity of the regions is increased to define the 24 final SRs described in Table 3 for the high-p T analysis. For the low-p T signal regions, the categories are equivalent to those of the high-p T analysis, but the selection differs in the requirement on H T and lepton p T . The threshold on H T is increased from 200 to 250 GeV in order to ensure 100% efficiency for the triggers used by the low-p T event selection. All 24 signal regions are mutually exclusive and may therefore be statistically combined within either high-p T or low-p T analysis. Table 3: Definition of the signal regions for the high-p T analysis. The low-p T analysis employs a tighter requirement H T > 250 GeV and uses the same numbering scheme, in which the first digit in the name represents the requirement on the number of b-tagged jets for that search region, e.g. SR01, SR11, and SR21 correspond to SRs with N b-jets 0, 1, and ≥2, respectively.
Additional (overlapping) signal regions, listed in Table 4, are defined with no or loose E miss T requirements in order to provide better sensitivity to scenarios such as RPV SUSY models and same-sign top-quark pair production. These search regions are formed using events that satisfy high-p T lepton selection and contain at least two jets. Because in RPV SUSY scenarios the LSP decays, mainly into detectable leptons and quarks, such events are not expected to have large E miss T , but they usually have substantial H T . Thus, in search regions designed for such models, the E miss T requirement is removed completely, while a relatively high H T > 500 GeV requirement is applied to reduce the level of SM background. These search regions are labelled as RPV0 and RPV2 for N b-jets ≥ 0 and ≥2, respectively. Same-sign top quark pair events in which the W bosons decay leptonically generally contain moderate E miss T , due to the accompanying neutrinos. Using events with E miss T > 30 GeV, we form four signal regions, denoted SStop1, SStop2, SStop1++, and SStop2++, where "++" refers to the selection of only positively charged dilepton pairs. Note that in most new physics scenarios, pp → tt is suppressed with respect to pp → tt because the PDF of the proton is dominated by quarks, rather than anti-quarks. For such scenarios, the SStop1++ and SStop2++ signal regions are expected to provide higher sensitivity. Table 4: Signal regions that are used in the search for same-sign top-quark pair production and RPV SUSY processes.

Backgrounds
There are three main sources of SM background in this analysis, which are described below. More details on the methods used to estimate these backgrounds can be found in Refs. [12,14].
• "Non-Prompt leptons", i.e. leptons from heavy-flavour decays, misidentified hadrons, muons from light-meson decays in flight, or electrons from unidentified photon conversions. The background caused by these non-prompt leptons, which is dominated by tt and W + jets processes, is estimated from a sample of events with at least one lepton that passes a loose selection but fails the full set of tight identification and isolation requirements described in Section 3. The background rate is obtained by scaling the number of events in this sample by a "tight-to-loose" ratio, i.e. the probability that a loosely identified non-prompt lepton also passes the full set of requirements. Various definitions of the loose lepton selection criteria are studied in detail, and combination of relaxed isolation and lepton-identification requirements is used. These probabilities are measured as a function of lepton p T and η, as well as event kinematics, in control samples of QCD multijet events that are enriched in non-prompt leptons.
• Rare SM processes that yield same-sign leptons, mostly from ttW, ttZ, and diboson production. We also include the contribution from the SM Higgs boson produced in association with a vector boson or a pair of top quarks in this category of background. All these backgrounds are estimated from MC simulation. The event yields are corrected for several effects, summarized in Section 6, to account for the differences between object selection efficiencies in data and simulation.
• Charge misidentification, i.e. events with opposite-sign isolated leptons where the charge of one of the leptons is misidentified because of severe bremsstrahlung in the tracker material. This background, which is relevant only for electrons and is negligible for muons, is estimated by selecting opposite-sign ee or eµ events passing the full kinematic selection and then weighting them by the p T -and η-dependent probability of electron charge misassignment. This probability, which varies between 10 −4 and 10 −5 , is obtained from simulation and is then validated with a control data sample of Z → ee events.
Backgrounds stemming from non-prompt leptons constitute the major contribution to the total background in most search regions. The rare SM processes dominate in the search regions with large numbers of b-tagged jets or high E miss T requirements. The contribution from charge misidentification is generally much smaller and stays below the few-percent level in all search regions.
The primary origin of the systematic uncertainty for the non-prompt lepton background estimate is differences between the QCD multijet sample, where the "tight-to-loose" ratio is determined, and the signal regions, where the method is applied, both for the event kinematics and for the relative rates of the various sources of non-prompt leptons. A systematic uncertainty also arises because tt and W + jets events, the two dominant components of the non-prompt background, differ themselves in the event kinematics and relative importance of the various sources, making it difficult to define a "tight-to-loose" ratio that is equally appropriate for both components. Based on the variation between true and predicted background yields when the background estimation method is applied to simulation, the systematic uncertainty of the estimate is assessed at 50%. This systematic part is the dominant uncertainty in the non-prompt lepton background estimate in most signal regions. The statistical uncertainty in the method is driven by the number of events in the sideband regions, defined with relaxed lepton requirements, that are used to estimate the non-prompt lepton background. As the kinematic selections are tightened, the statistical uncertainty becomes more important, becoming comparable in size to the systematic uncertainty in the search regions with the tightest selections.
For the rare SM processes, the next-to-leading-order (NLO) production cross sections are used to normalize the MC predictions. The cross section values used for the most relevant processes, ttW and ttZ, are 232 fb [23] and 208 fb [24,25], respectively. Because these and other rare processes are simulated using leading order (LO) generators, the systematic uncertainty for the rare SM background accounts both for the theoretical uncertainty in the cross sections and for the non-uniformity of the ratio between the LO and NLO cross sections as a function of jet multiplicity, H T , and E miss T [23]. The systematic uncertainties for each SM process that contributes to this background are assigned to be 50% and are considered to be 100% correlated across all signal regions.
The uncertainty associated with the charge-misidentification background estimate, which is estimated to be 30%, accounts for differences between data and simulation, and the limited momentum range of electrons probed in the control sample.
The total background in each search region is obtained by summing the yields from each of these background sources, and the total uncertainty is calculated by considering the individual uncertainties to be uncorrelated.

Efficiencies and associated uncertainties
The trigger efficiency is measured with data, using triggers that are orthogonal to those described in Section 3. The measured efficiencies are summarized in Table 5. Correction factors to take the trigger inefficiencies into account are applied to all acceptances calculated from MC simulation, for both signal and background samples. We assign a 6% uncertainty to these efficiencies, based on the statistical uncertainty of the measurement and deviations from the quoted numbers in Table 5 as a function of |η| and p T . Table 5: Summary of the trigger selection efficiencies for low-and high-p T analyses in each channel. The thresholds on |η| and p T correspond to the lower p T lepton of the dilepton pair.

Channel
Low The offline lepton selection efficiencies in data and simulation are measured using Z-boson events to derive simulation-to-data correction factors. The correction factors applied to simulation are 90 (96)% for p T < 20 GeV and 94 (98)% for p T > 20 GeV for electrons (muons). The uncertainty of the total efficiency is 5% (3%) for electrons (muons) with p T > 15 GeV, increasing to 10% (5%) for lower transverse momentum. An additional systematic uncertainty is assigned to account for potential mismodelling of the lepton isolation efficiency due to varying hadronic activity in signal events. This uncertainty is 3% for all leptons except muons with p T < 30 GeV, for which it is 5%.
Another source of systematic uncertainty is associated with the jet energy scale correction. This systematic uncertainty varies between 5% and 2% in the p T range 40-100 GeV for jets with |η| < 2.4 [26]. It is evaluated on a single-jet basis, and its effect is propagated to H T , E miss T , the number of jets, and the number of b-tagged jets. The importance of these effects depends on the signal region and the model of new physics. In general, models with high hadronic activity and large E miss T are less affected by the uncertainty in the jet energy scale. In addition, there is a contribution to the total uncertainty arising from limited knowledge of the resolution of the jet energy, but this effect is generally of less importance than the contribution from the jet energy scale.
The b-tagging efficiency for b-quark jets with |η| < 2.4, measured in data using samples enriched in tt and muon-jet events, has a p T -averaged value of 0.72. The false positive b-tagging probability for charm-quark jets is approximately 20%, while for jets originating from lightflavour quarks or gluons it is of the order of 1%. Correction factors, dependent on jet flavour and kinematics, are applied to simulated jets to account for the differences in the tagging efficiency in simulation with respect to data. The total uncertainty of the b-tagging efficiency is determined by simultaneously varying the efficiencies to tag a bottom, charm, or light quark up and down by their uncertainties [16]. The importance of this effect depends on the signal region and the model of new physics. In general, models with more than two b quarks in the final state are less affected by this uncertainty.
Additional uncertainties due to possible mismodelling of the pileup conditions or initial-state radiation (ISR) [27] are evaluated and found to be 5% and 3-15%, respectively. The uncertainty of the signal acceptance due to the PDF choice is found to be less than a few percent. Finally, there is a 2.6% uncertainty in the yield of events because of the uncertainty in the luminosity normalization [28].
A summary of the systematic uncertainties associated with the acceptance and signal efficiency for this analysis is provided in Table 6. While the uncertainties associated with the integrated luminosity, modelling of lepton selection, trigger efficiency, and pileup are taken to be constant across the parameter space of the new physics models considered in this paper, uncertainties arising from the remaining observables are estimated for each model separately on an event-by-event basis by varying those observables within their uncertainties. The total uncertainty in the computed acceptance is in the 13-25% range. The figures in Table 6 are representative values for these uncertainties and do not characterize the results for extreme kinematic regions, such as those near the diagonal of the parameter space of the SUSY simplified models discussed in Section 8, where the particle mass spectra are compressed.

Results
The distributions of E miss T versus H T for events in the three baseline signal regions are shown in Fig. 1. The results are shown separately for the low-and high-p T samples. The corresponding results for the four selection variables H T , E miss T , N jets , and N b-jets are shown in Fig. 2. For these latter results, the SM background prediction is also shown. There are no significant discrepancies observed between the observations and background predictions for any region.   [29] for each signal region in the low-and high-p T analyses are studied, and are found to be consistent with a uniform distribution between 0 and 1.

Limits on models of new physics and on rare SM processes
Given the lack of a significant excess over the expected SM background, the results of the search are used to derive limits on the parameters of various models of new physics and to derive limits on the cross sections of rare SM processes. The 95% confidence level (CL) upper limits on the signal yields are calculated using the LHC-type CL s method [30][31][32]. Lognormal nuisance parameters are used for the signal (Table 6) and background estimate (Tables 7 and 8) uncertainties. For each model considered, limits are obtained by performing a statistical combination of the most sensitive signal regions. Figure 4: Diagrams for the six SUSY models considered (A1, A2, B1, B2, C1, and RPV).

Model A1
The signal regions used to set limits on the new physics models explored in this paper are given in Table 9.
The number of events that are expected to satisfy the selection for a given signal model is obtained from MC simulation. The uncertainties for the event yields are computed as described in Section 6. For a given signal region, the different sources of uncertainties in the signal acceptance are considered to be uncorrelated, with correlations across signal regions taken into account. The uncertainties in the total background across the signal regions are considered to be fully correlated.
First, we present limits on the parameter spaces of various R-parity-conserving simplified SUSY models [33]. The exclusion contours are obtained with the gluino or bottom-squark pair production cross sections at the NLO+NLL (i.e. next-to-leading-logarithm) accuracy that are calculated in the limit where other sparticles are heavy enough to be decoupled [34][35][36][37][38][39]. The production of SUSY particles and the decay chains under consideration are shown schematically in Fig. 4.
Scenarios A1 and A2 represent models of gluino pair production resulting in the tttt χ 0 1 χ 0 1 final state, where χ 0 1 is the lightest neutralino [33,[40][41][42][43]. In model A1, the gluino undergoes a threebody decay g → tt χ 0 1 mediated by an off-shell top squark. In model A2, the gluino decays to a top quark and a top anti-squark, with the on-shell anti-squark further decaying into a top anti-quark and a neutralino. Both of these models produce four on-shell W bosons and four b quarks. Therefore, search regions SR21-SR28, which require at least two b-tagged jets and high-p T leptons, are used to derive the limits on the parameters of these models; the region with the best sensitivity is SR28. The 95% CL upper limits on the cross section times branching fraction, as well as the exclusion contours, are shown in Fig. 5. For model A1, the results are presented as a function of gluino mass and χ 0 1 mass, and for model A2 as a function of gluino mass and top squark mass with the χ 0 1 mass set to 50 GeV. In model A2, the limits do not depend on the top squark or χ 0 1 masses provided that there is sufficient phase space to produce on-shell top quarks with a moderate boost in the decay of both the gluino and the top squark. This range extends to approximately 600 GeV for the χ 0 1 mass. Model B1 is a model of bottom-squark pair production, followed by one of the most likely decay modes of the bottom squark, b 1 → t χ − 1 with χ − 1 → W − χ 0 1 , where b 1 and χ − 1 represent the lightest bottom squark and lightest chargino, respectively. We consider three cases in this decay mode. We either set the χ 0 1 mass to 50 GeV and present the limits in the (m χ ± 1 , m b 1 ) plane, or consider the (m χ 0 1 , m b 1 ) plane with the mass of the chargino set according to m χ 0 /m χ ± 1 = 0.5 or m χ 0 /m χ ± 1 = 0.8. The values 0.5 and 0.8 are representative choices that determine whether the top quark and W boson are on-shell or off-shell, which has a direct impact on the sensitivity of the analysis in this model. The limits for this model, obtained using search regions SR11 to SR28, are presented in Fig. 6. For m χ 0 /m χ ± 1 = 0.8, the low-p T lepton selection is used, while high-p T leptons are used for the other two scenarios. SR28 is again the most sensitive signal region, followed by the regions requiring one b-tagged jet: SR18, SR15, and SR13.
Model B2 consists of gluino pair production followed by g → b 1 b. The gluino decay modes in models A1 and A2 are expected to be dominant if the top squark is the lightest squark. Conversely, if the bottom squark is the lightest, the decay mode in model B2 would be the most probable. The limits on this model, calculated using search regions SR21-SR28 and the high-p T lepton selection, are presented in Fig. 6 as a function of m( b 1 ) and m( g) for two fixed masses of m χ ± 1 , 150 and 300 GeV. The region with the largest sensitivity to this model is SR28.
Model C1 is based on the production of a gluino pair where each gluino decays to light quarks and m g , giving rise to either high-or low-p T leptons. We examine x values of 0.5 and 0.8. The former value ensures that the W boson is on-shell in the sparticle mass range considered, while the latter yields mostly off-shell W bosons. In this model, no enrichment of heavy-flavour jets is expected. Therefore, the search regions SR01-SR08, with both the low-and high-p T lepton selection, are used for cross section upper limit calculation. The limits are presented in Fig. 7. In this model, gluino masses up to 900 GeV are probed. Most of the sensitivity to this model is obtained from signal region SR08.
These results extend the sensitivity obtained in the previous analysis [11] on gluino and sbottom masses. For the gluino-initiated models (A1, A2, B2, and C1), we probe gluinos with masses up to about 1050 GeV, with relatively small dependence on the details of the models. This is because the limits are driven by the common gluino pair production cross section. In the case of the direct bottom-squark pair production, model B1, our search shows sensitivity for bottom-squark masses up to about 500 GeV.
These models are also probed by other CMS new physics searches in different decay modes. Other searches are usually interpreted in the context of model A1 but not A2, B1, or B2. For model A1, the limits given here are complementary to the limits from the searches presented in Refs. [44][45][46][47]. In particular, they are less stringent at low m( χ 0 1 ) but more stringent at high   observed 95% CL Limits expected 95% CL Limits Figure 8: 95% CL upper limit on the gluino production cross section for an RPV simplified model, pp → g g, g → tbs(tbs). m( χ 0 1 ). A similar conclusion applies to model A2, since the final state is the same. For bottomsquark pair production, limits on m( b 1 ) of about 600 GeV have been presented [46], but assuming the decay mode b 1 → b χ 0 1 instead of the model B1 mode b 1 → t χ − 1 considered here. Comparable limits for model A1, as well as for similar models with top and bottom quarks from gluino decays, have been reported by the ATLAS Collaboration [48-51].
A single RPV scenario is considered in this analysis, one in which gluino pair production is followed by the decay of each gluino to three quarks, as is favoured in the SUSY model with minimal flavour violation [52]: g → tbs(tbs) (model RPV). Such decays lead to same-sign Wboson pairs in the final state in 50% of the cases. Compared with the decays g → tsd(tsd), which also yield same-sign W-boson pairs, the mode considered profits from having two extra b quarks in the final state, resulting in a higher signal selection efficiency. The model is governed by one parameter (m g ), which dictates the production cross section and the final state kinematics. The dedicated search region RPV2 with the high-p T lepton selection is used to place an upper limit on the production cross section. The result is shown in Fig. 8. In this scenario, the gluino mass is probed up to approximately 900 GeV.
The results for the signal regions SStop1, SStop1++, SStop2, and SStop2++ are used to set limits on the cross section for same-sign top-quark pair production, σ(pp → tt, tt) from SStop1 and SStop2, and σ(pp → tt) from SStop1++ and SStop2++. Here σ(pp → tt, tt) is shorthand for the sum σ(pp → tt) + σ(pp → tt). These limits are calculated using an acceptance obtained from simulated pp → tt events and an opposite-sign selection. This acceptance, including branching fractions, is 0.43% (0.26%) for the SStop1 (SStop2) search region. The relative uncertainty in this acceptance is 14%. The observed upper limits are σ(pp → tt, tt) < 720 fb and σ(pp → tt) < 370 fb at 95% CL. The median expected limits are 470 +180 −110 fb and 310 +110 −80 fb, respectively. Similarly, the results from signal regions SR21-SR28 with the high-p T lepton selection are used to set limits on the SM cross section for quadruple top-quark production. The observed upper limit is σ(pp → tttt) < 49 fb at 95% CL, compared to a median expected limit of 36 +16 −9 fb. The SM cross section as computed with the MC@NLO program [53] is σ SM = 0.914 ± 0.005 fb. The most sensitive signal regions, SR24 and SR28, have a signal acceptance of 0.52% and 0.49%, respectively, with relative uncertainties of 13% and 17%.

Information for additional model testing
We have described a signature-based search that finds no evidence for physics beyond the SM. In Section 8, the results are used to place bounds on the parameters of a number of models of new physics. Here, additional information is presented that can be used to confront other models of new physics in an approximate way through MC generator-level studies. The expected numbers of events can then be compared with an upper limit on the number of signal events that can be obtained using inputs from Tables 7 and 8 and a signal acceptance uncertainty estimated from the generator-level studies.
The E miss T and H T turn-on curves, shown in Fig. 9 as a function of the respective generator-level quantities, are parametrized as 0.5 · ∞ · erf (x − x 1/2 )/σ + 1 , with erf(z) the error function, and ∞ , x 1/2 , and σ the parameters of the fit. The generator H T is calculated using generator jets, obtained by clustering all stable particles from the hard collision, after showering and hadronization, except for neutrinos and other non-interacting particles. The parameters of the fitted functions are summarized in Tables 10 and 11 for E miss T and H T , respectively. Analogously to the offline selection, only generator jets that are separated from generator electrons and muons by ∆R ≡ ∆φ 2 + ∆η 2 > 0.4 are considered in the derivation and application of the efficiency   An additional turn-on curve, introduced since the publication of Ref. [11], has been added to parametrize the efficiency to reconstruct a jet with p T > 40 GeV. The curve, shown in Fig. 10 (left) as a function of the generator jet p T , is described by the same functional form as the H T turn-on. The parameters of the fit are ( ∞ , x 1/2 , σ) = (1.0, 29.8 GeV, 18.8 GeV). Figure 10 also shows the b-tagging efficiency, obtained from simulation, for b quarks with |η| < 2.4. The efficiency is fit with a third-order (first-order) polynomial for p T < 120 GeV (p T > 120 GeV). The parameters of the fit are given in Table 12.
The turn-on curves for the lepton selection are shown in Fig. 11. The lepton efficiency (ε)including the effects of reconstruction, identification, and isolation as well as relevant datato-simulation scale factors-is parametrized as ε(p T ) = ∞ · erf (p T − 10)/σ + 10 · 1 − erf (p T − 10)/σ . The results of the fit are summarized in Table 13.   The prescription to apply the efficiency model is similar to that described in Ref. [14], with some modifications needed to accommodate the use of exclusive signal regions. The efficiencies for the H T and E miss T selections in regions with upper and lower bounds are obtained by taking the difference between the relevant curves in Fig. 9. The jet reconstruction and b-tagging efficiencies are provided as per-jet quantities. Thus, one scale factor per jet should be obtained from the relevant curves. Additional combinatorial factors should be included, as dictated by the requirements of the signal region selection. The application of the lepton efficiency remains unchanged, with one factor per lepton obtained from the appropriate fit of Fig. 11. All the quoted efficiencies are multiplicative. The resulting signal yield, obtained by summing the contribution derived from the efficiency model over all events, is then compared to the calculated upper limit as described at the beginning of this section. The efficiency model presented was applied to a variety of the signal models and search regions considered in this analysis. Results from the efficiency model were found to agree with those obtained using the detector simulation and reconstruction to within approximately 30%. It should be emphasized that the efficiency model is approximate and is not universally applicable. Lepton isolation efficiency, for example, depends on the hadronic activity in the event and in some extreme cases on the event topology. For instance, in models giving rise to top quarks with a significant boost, the lepton isolation efficiency in Fig. 11 overestimates the true value.

Summary
We have presented the results of a search for physics beyond the standard model with samesign dilepton events using the CMS detector at the LHC. The study is based on a sample of pp collisions at √ s = 8 TeV corresponding to an integrated luminosity of 19.5 fb −1 . The data are analyzed in exclusive signal regions formed by placing different requirements on the discriminating variables H T , E miss T , number of jets, and number of b-tagged jets. The latter can assume values of 0, 1, and 2 or more, which allow us to probe signatures both with and without thirdgeneration squarks. No significant deviation from standard model expectation is observed.
Using sparticle production cross sections calculated in the decoupling limit, and assuming that gluinos decay exclusively into top or bottom squarks and that the top and bottom squarks decay as t 1 → t χ 0 1 and b 1 → t χ − 1 ( χ − 1 → W − χ 0 1 ), lower limits on gluino and sbottom masses are calculated. Gluinos with masses up to approximately 1050 GeV and bottom squarks with masses up to about 500 GeV are probed. In models where gluinos do not decay to thirdgeneration squarks, sensitivity for gluino masses up to approximately 900 GeV is obtained. A similar reach in the gluino masses is demonstrated in the scope of an R-parity violating model.
The results are used to set upper limits on the same-sign top-quark pair production cross section σ(pp → tt, tt) < 720 fb and σ(pp → tt) < 370 fb at 95% CL. An upper limit at 95% CL of σ(pp → tttt) < 49 fb is obtained for the cross section of quadruple top-quark production.