Search for anomalous production of prompt like-sign lepton pairs at √ s = 7 TeV with the ATLAS detector

: An inclusive search for anomalous production of two prompt, isolated leptons with the same electric charge is presented. The search is performed in a data sample corresponding to 4.7 fb − 1 of integrated luminosity collected in 2011 at √ s = 7 TeV with the ATLAS detector at the LHC. Pairs of leptons ( e ± e ± , e ± µ ± , and µ ± µ ± ) with large transverse momentum are selected, and the dilepton invariant mass distribution is examined for any deviation from the Standard Model expectation. No excess is found, and upper limits on the production cross section of like-sign lepton pairs from physics processes beyond the Standard Model are placed as a function of the dilepton invariant mass within a ﬁducial region close to the experimental selection criteria. The 95% conﬁdence level upper limits on the cross section of anomalous e ± e ± , e ± µ ± , or µ ± µ ± production range between 1.7 fb and 64 fb depending on the dilepton mass and ﬂavour combination.


Introduction
Events containing two prompt leptons with large transverse momentum (p T ) and equal electric charge are rarely produced in the Standard Model (SM) but occur with an enhanced rate in many models of new physics. For instance, left-right symmetric models [1][2][3][4], Higgs triplet models [5][6][7], the little Higgs model [8], fourth-family quarks [9], supersymmetry [10], universal extra dimensions [11], and the neutrino mass model of refs. [12][13][14] may produce final states with two like-sign leptons. In the analysis described here, pairs of isolated, high-p T leptons are selected, and the invariant mass of the dilepton system (e ± e ± , e ± µ ± , µ ± µ ± ) is examined for the inclusive final state and separately for positively-and negativelycharged pairs.
The ATLAS Collaboration has previously reported inclusive searches for new physics in the like-sign dilepton final state in a data sample corresponding to an integrated luminosity of 34 pb −1 [15] and in like-sign muon pairs with 1.6 fb −1 [16]. No significant deviation from SM expectations was observed, and fiducial production cross-section limits as well as limits on several specific models of physics beyond SM were derived. The CDF Collaboration has performed similar inclusive searches [17,18] without observing any evidence for new physics. Furthermore, the ATLAS and CMS Collaborations have performed several searches for likesign leptons produced in association with jets or missing transverse momentum where no evidence of non-SM physics was observed [19][20][21][22][23][24][25][26].
This article is organised as follows. A brief description of the ATLAS detector is given in section 2. The data and simulation samples, the event selection, and the background determination are explained in sections 3, 4, and 5, respectively. The systematic uncertainties on the background estimate (including theoretical uncertainties on the production cross sections) and signal acceptances are summarised in section 6. In section 7 the number of observed lepton pairs in data is compared to the background estimate, and in section 8 these results are used to derive upper limits on the fiducial cross section for like-sign dilepton production in a kinematic region closely related to the experimental event selection.

The ATLAS detector
The ATLAS detector [27] consists of an inner tracking system, calorimeters, and a muon spectrometer. The inner detector (ID), directly surrounding the interaction point, is composed of a silicon pixel detector, a silicon microstrip detector, and a transition radiation tracker, all immersed in a 2 T axial magnetic field. It covers the pseudorapidity 1 range |η| < 2.5 and is enclosed by a calorimeter system consisting of electromagnetic and hadronic sections. The electromagnetic part is a lead/liquid-argon sampling calorimeter, divided into a barrel (|η| < 1.475) and two end-cap sections (1.375 < |η| < 3.2). The barrel (|η| < 0.8) and extended barrel (0.8 < |η| < 1.7) hadronic calorimeter sections consist of iron and scintillator tiles, while the end-cap (1.5 < |η| < 3.2) and forward (3.1 < |η| < 4.9) calorimeters are composed of copper or tungsten, and liquid-argon.
The calorimeter system is surrounded by a large muon spectrometer (MS) built with air-core toroids. This spectrometer is equipped with precision tracking chambers (composed of monitored drift tubes and cathode strip chambers) to provide precise position measurements in the bending plane in the range |η| < 2.7. In addition, resistive plate chambers and thin gap chambers with a fast response time are used primarily to trigger muons in the rapidity ranges |η| ≤ 1.05 and 1.05 < |η| < 2.4, respectively. The resistive plate chambers and thin gap chambers also provide position measurements in the nonbending plane, which are used for the pattern recognition and the track reconstruction. The ATLAS trigger system has a hardware-based Level-1 trigger followed by a softwarebased high-level trigger [28]. The Level-1 muon trigger searches for hit coincidences between different muon trigger detector layers inside geometrical windows that define the muon transverse momentum and provide a rough estimate of its position. It selects high-p T muons in the pseudorapidity range |η| < 2.4. The Level-1 electron trigger selects local energy clusters of cells in the electromagnetic section of the calorimeter. For low-energy electron clusters, low activity in the hadronic calorimeter nearby in the η-φ plane is required. The high-level trigger selection is based on similar reconstruction algorithms as those used offline.

Data sample and Monte Carlo simulation
This analysis is carried out using a data sample corresponding to an integrated luminosity of 4.7 ± 0.2 fb −1 of pp collisions, recorded in 2011 at a centre-of-mass energy of 7 TeV. In this dataset, the average number of interactions per beam crossing ranges from about six in the first half of the year to about fifteen at the end of 2011.
The data were selected using single-muon and single-electron triggers with p T thresholds of 10 GeV and 16 GeV at Level-1, respectively. In the high-level trigger, a muon with p T > 18 GeV is required, while for electrons, the p T threshold is 20 GeV in the early 2011 data and 22 GeV in the later part of the year. To ensure no efficiency loss for electrons with very high-p T , a trigger with a p T threshold of 45 GeV is also used which has no requirement on the hadronic calorimeter energy deposits near the electron in the η-φ plane at Level-1.
Monte Carlo (MC) simulation is used to estimate some of the background contributions and to determine the selection efficiency and acceptance for possible new physics signals. The dominant SM processes that contribute to prompt like-sign dilepton production are W Z and ZZ, with smaller contributions from like-sign W pair production (W ± W ± ) and production of a W or Z boson in association with a top quark pair (ttW , and ttZ). These are all estimated using MC simulation. For processes with a Z boson, the contribution from γ * → ℓ + ℓ − due to internal or external bremsstrahlung of final-state quarks or leptons is also simulated for m(ℓ + ℓ − ) > 0.1 GeV. Sherpa [29] is used to generate W Z and ZZ events, while W ± W ± , ttW , and ttZ production is modelled using MadGraph [30] for the matrix element and Pythia [31] for the parton shower and fragmentation. The W ± W ± sample includes the W ± W ∓ W ± process.
The normalisation of the W Z and ZZ MC samples is based on cross sections determined at next-to-leading order (NLO) with MCFM [32]. The cross sections times branching ratios for W ± Z → ℓ ± νℓ ± ℓ ∓ and ZZ → ℓ ± ℓ ∓ ℓ ± ℓ ∓ after requiring two charged leptons (electrons, muons, or taus) with the same electric charge and with p T > 20 GeV and |η| < 2.5, are 372 fb and 91 fb, respectively.
For ttW and ttZ production, the higher-order corrections are calculated in ref. [33] and refs. [34][35][36], respectively. The full higher-order corrections for W ± W ± production have not been calculated. However, for parts of the process, the NLO QCD corrections have been shown to be small [37,38]. Based on this, no correction is applied to the LO cross section.
Opposite-sign dilepton events due to Z/γ * , tt, and W ± W ∓ production constitute a background if the charge of one of the leptons is misidentified. The Z/γ * process is generated with Pythia, and the cross section is calculated at next-to-next-to-leading order (NNLO) using PHOZPR [39]. The ratio of this cross section to the leading-order cross section is used to determine a mass dependent QCD K-factor which is applied to the result are chosen such that they are on the plateau of the trigger efficiency. In addition to the lepton selection, events must have a primary vertex reconstructed with at least three tracks with p T > 0.4 GeV. If more than one interaction vertex is reconstructed in an event, the one with the highest p 2 T , summed over the tracks associated with the vertex, is chosen as the primary vertex. For Z bosons decaying to electrons or muons, the efficiency of the primary vertex requirements is close to 100%.
Electrons are identified as compact showers in the electromagnetic sections of the calorimeters that are matched to a reconstructed track in the ID using the tight criteria described in ref. [60]. These criteria include requirements on the transverse shower shapes, the geometrical match between the track and shower, the number and type of hits in the ID, and they reject electrons associated with reconstructed photon conversion vertices. Electrons must have |η| < 2.47, excluding the transition region between the barrel and endcap calorimeters (1.37 < |η| < 1.52). The electron tracks are refitted with the Gaussian Sum Filter algorithm to account for radiative energy losses (bremsstrahlung) in the detector material [61].
Muon candidates are formed from tracks reconstructed in the ID combined with tracks reconstructed in the MS [62]. To reduce the charge mismeasurement rate the independent charge measurements from these two detectors are required to agree. Muons must have |η| < 2.5.
The ID tracks of electron and muon candidates are required to be consistent with originating from the primary interaction vertex. The requirements are |d 0 | < 1.0 mm, |z 0 sin θ| < 1.0 mm, and |d 0 /σ(d 0 )| < 3, where d 0 (z 0 ) is the transverse (longitudinal) distance between the primary vertex and the point of closest approach of the track and σ(d 0 ) is the estimated uncertainty on the d 0 measurement, typically 15 µm. To ensure precise impact-parameter measurements, a hit in the innermost pixel detector layer is required if the track traverses an active detector component. This requirement also reduces background due to electrons from photon conversions.
Both electrons and muons are required to be isolated from other activity in the event. The isolation energy, E cone∆R iso T , is the sum of transverse energies 2 in calorimeter cells (including electromagnetic and hadronic sections) in a cone ∆R = (∆η) 2 + (∆φ) 2 < ∆R iso around the lepton direction. This quantity is corrected for the energy of the electron object as well as energy deposits from pileup interactions. The track isolation, p cone∆R iso T , is analogously defined as the scalar sum of the transverse momentum of tracks with p T > 1 GeV in a cone of ∆R < ∆R iso around the lepton direction, excluding the lepton track. For muons, In addition to the isolation criteria, selected leptons are required to be well separated from jets to suppress leptons from hadronic decays. For this purpose, jet candidates are reconstructed from topological clusters in the calorimeter [63] using the anti-k t algorithm [64] with a radius parameter of 0.4. Their energies are corrected for calorimeter non-compensation, energy loss in upstream material, and other instrumental effects. Furthermore, quality criteria are applied to remove reconstructed jets not arising from hardscattering interactions [65]. Jets are required to have p T > 25 GeV, |η| < 2.8, and 75% of the momentum of the tracks within the jet must originate from the primary vertex. Any jet within a distance ∆R = 0.2 of a candidate electron is removed from the list of jets to avoid counting the electron as a jet. Electrons and muons are then required to be separated by ∆R > 0.4 from any jet with p T > 25 GeV + 0.05 × p T (ℓ).
Lepton pairs are selected if they contain two leptons with the same electric charge passing the above selection requirements with invariant mass m(ℓ ± ℓ ± ) > 15 GeV. Furthermore, for e ± e ± pairs, the mass range 70-110 GeV is vetoed due to large backgrounds from opposite-sign electron pairs produced by Z decays, where the charge of one lepton is misidentified. Any combination of two leptons is considered, allowing more than one lepton pair per event to be included.

Background determination
Backgrounds to this search arise from three principal sources: SM production of prompt, like-sign lepton pairs (prompt background); production of opposite-sign lepton pairs where the charge of one lepton is misidentified (charge-flip background) or where a photon produced in association with the opposite-sign leptons converts into an e + e − pair; and hadrons or leptons from hadronic decays (non-prompt background).

Prompt lepton backgrounds
SM processes resulting in two prompt leptons of the same electric charge are W Z, ZZ, W ± W ± , ttW , and ttZ production. The background due to these processes is determined from MC simulation using the samples and cross sections described in section 3. Other SM processes are not expected to contribute significantly to the background and are neglected.

Charge misidentification and photon conversions
Charge misidentification can occur for high-momentum tracks when the tracking detector is unable to determine the curvature of the track. This misidentification can occur for both electrons and muons. In the momentum range of a few hundred GeV relevant for this analysis, the background arising from this source is determined to be negligible based on MC simulation. The rate of misidentification is studied in data with muons from Z → µµ decays where the muons are selected based only on the MS (ID) measurements to study the ID (MS) mismeasurement. The combined probability to mismeasure the charge in both the ID and the MS is found to be consistent with zero and an upper limit ranging from 5 × 10 −10 at low p T to 7 × 10 −2 for p T ∼ 400 GeV is placed. This upper limit is used to determine the systematic uncertainty for muons.
Another source of charge misidentification occurs for electrons when a high-momentum photon is radiated and converts into an e + e − pair. Electron candidates are rejected if they are associated with conversion vertices [60]. However, for asymmetric conversions, sometimes only one of the tracks is reconstructed and its charge may be different from the charge of the original electron that radiated the photon. The probability for electron charge misidentification to occur, either by this mechanism or by track mismeasurement, is measured as the fraction of like-sign e ± e ± pairs with 80 GeV< m(ee) < 100 GeV as a function of the electron η following the procedure explained in ref. [66]. The simulation is found to overestimate this probability by about 15%. An η-dependent scaling factor is applied to the simulated rate to account for this overestimate, and the simulation is then used to predict the backgrounds from Z/γ * , tt, and W ± W ∓ production. In the central (forward) region the probability increases from about 0.01% (0.6%) at p T ≈ 50 GeV to 0.3% (4%) at p T ≈ 300 GeV. An uncertainty on this misidentification probability of ±(10-20)% (depending on η) is derived by comparing different methods to determine this factor as described in ref. [66] and by considering the statistical uncertainty on each method.
Production of W γ → ℓνγ and Zγ → ℓ + ℓ − γ can lead to like-sign lepton pairs if the photon converts. This background is determined using the MC samples described in section 3. Since this background is closely related to the charge misidentification for electrons described above, the MC-based estimate is also scaled down by 15% and the same systematic uncertainty of ±(10-20)% is applied.

Non-prompt lepton backgrounds
The non-prompt background includes lepton pairs where one or both of the leptons result from hadronic decay or misidentification. Processes contributing to this background include W +jet, Z+jet, multi-jet (including bb and cc), and tt production.
The non-prompt background is determined directly from data. For both electrons and muons, the determination relies on measuring a factor f that is the ratio of the number of selected leptons satisfying the analysis selection criteria (N S ) to the number of leptons failing these selection criteria but passing a less stringent set of requirements, referred to as anti-selected leptons (N A ): The factor f is determined as a function of p T and η in a data sample dominated by nonprompt leptons. Any contamination by prompt leptons in the numerator and denominator is subtracted using MC simulation. The details of the anti-selected definitions and the regions used for measuring f depend on the lepton flavour and are described in more detail below.
The primary sources of non-prompt muons are semi-leptonic decays of b-and c-hadrons. For the anti-selection, the isolation criterion described in section 4 is inverted, but a looser requirement of p cone0.4 T /p T (µ) < 1.0 is placed. The measurement of f is done in a sample dominated by non-prompt muons with |d 0 |/σ(d 0 ) > 5 and |d 0 | < 10 mm. These muons are selected from dimuon events with p T (µ) > 10 GeV, m(µµ) > 15 GeV, and where Z-boson candidates have been vetoed. Simulated muons from b-and c-hadron decays are used to determine the expected difference between f for muons with |d 0 |/σ(d 0 ) > 5 (used in the data-driven determination of f ) and muons with |d 0 |/σ(d 0 ) < 3 (used in the event selection). In simulation, muons with |d 0 |/σ(d 0 ) < 3 are more isolated, so a correction factor of 1.34 ± 0.34 is applied to f . The resulting value of f is about 0.10 at p T = 20 GeV and increases to about 0.25 at p T = 100 GeV. One source of uncertainty in this procedure is that the fraction of non-prompt muons from pion and kaon decays may be larger for muons passing the analysis-level impact parameter criteria than for those used in the measurement of f . The uncertainty due to the contribution of light-flavour decays is assessed by exploiting the difference in the track momentum measurement in the ID and the MS, which is expected to be large for kaons or pions decaying in the ID or calorimeters. The difference between f for light-flavour decays and heavy-flavour decays is combined with the fraction of non-prompt muons from light-flavour decays to give an uncertainty of ±15% on f . The total systematic uncertainty on f is the quadratic sum of the statistical error, the uncertainty on the prompt background subtraction, the correction for the isolation-dependence on the impact parameter significance criteria, and the uncertainty in the contribution of pion and kaon decays. It is about ±37% at low p T and ±100% for p T > 100 GeV.
For electrons, the main non-prompt backgrounds arise from heavy-flavour decays, charged pions that shower early in the calorimeter, and neutral pions decaying to two photons where one of the photons converts to an e + e − pair. The anti-selected electrons must have either |d 0 |/σ(d 0 ) > 3 or fail the medium electron identification criteria while passing the loose electron criteria, as defined in ref. [60]. The inversion of the impact parameter criteria enhances the heavy-flavour background, while the inversion of the medium identification criteria enhances the pion backgrounds. A sample dominated by non-prompt electrons is selected by triggering on a high-p T electron candidate and requiring that the event contains a jet with p T > 20 GeV. Contamination of this sample by prompt electrons is reduced by requiring that there be only one electron candidate and requiring the transverse mass 3 formed by the electron and the missing transverse momentum (E miss T ) to be below 40 GeV. The value of f is found to be about 0.18 at E T = 20 GeV and decreases to about 0.10 at E T = 100 GeV. A systematic uncertainty is determined accounting for the statistical uncertainty, the prompt background subtraction, the p T requirement of the jet in the event, and any trigger bias. The uncertainty is about ±10% at low E T and increases to ±100% for E T > 300 GeV. Alternate anti-selection definitions are used to assess the overlap between the non-prompt and charge misidentification backgrounds, as well as the contributions of light-and heavy-flavour jets to the non-prompt background. For the former, a correction is made to the non-prompt background, amounting to ±(11-25)%, and the full correction is taken as a systematic uncertainty. For the latter, an uncertainty of ±(4-84)% is found, depending on the invariant mass threshold considered.
A background prediction is derived from f using dilepton pairs where one or both leptons are anti-selected but pass all other event selection criteria. The non-prompt back- ground prediction is then given by The background estimates are verified in several control regions that are designed to probe specific sources of background. The understanding of prompt leptons is tested in a sample requiring two leptons of opposite charge. This sample is dominated by the Z/γ * process in all three final states, and the data agree with the background prediction to better than 5%. The kinematics of the electron charge misidentification background are verified in pairs of like-sign electrons with 80 GeV < m(e ± e ± ) < 100 GeV. The non-prompt background is tested by inverting the isolation criteria but requiring a looser isolation cut value. The heavy-flavour background is tested by inverting the |d 0 |/σ(d 0 ) criterion. Furthermore, for electrons, the light-flavour background is tested by inverting some of the shower shape criteria. The data agree with the background prediction in the various control regions within the systematic uncertainties.

Systematic uncertainties
Several systematic effects can change the signal acceptance and the background estimate. Uncertainties on the event selection efficiencies and on the luminosity affect the predicted yield of signal events as well as those backgrounds that are estimated from MC simulation. Uncertainties on the trigger efficiencies, lepton identification efficiencies, and lepton momentum scales are determined using W , Z, and J/ψ decays [67][68][69]. The uncertainties on the lepton identification efficiencies result in an uncertainty on the number of like-sign pairs of ±(3-4)%. Effects resulting from uncertainties in the muon momentum scale and resolution are negligible. For electrons the energy scale is known to ±1%, and its impact on the signal acceptance is below 0.1%, while the background estimate is affected by up to ±3.5%. The uncertainties due to the trigger efficiencies are < 1% in all channels. The uncertainty on the integrated luminosity is ±3.9% [70,71].
The uncertainties on the production cross sections of the SM processes affect the predicted yield of the prompt lepton background. For W Z and ZZ production, an uncertainty of ±10% is estimated due to higher-order QCD corrections by varying the renormalisation and factorisation scales by a factor of two. The uncertainty due to the parton distribution functions (PDFs) is evaluated using the eigenvectors provided by the CTEQ10 set [57] of PDFs, following the prescription given in ref. [72]. The difference between the central crosssection values obtained using this PDF set and that obtained with the MRST2008NLO PDFs [55] is also added in quadrature. This procedure gives a conservative estimate of the PDF uncertainty on the cross section. The resulting uncertainty on the cross section is ±7%. For ttW , ttZ and W ± W ± production a normalisation uncertainty of ±50% is assigned [34,37].
The production of Z/γ * , tt, and W ± W ∓ also constitutes a background in the ee and eµ channels due to charge misidentification. The PDF uncertainties for Z/γ * and tt are computed using the MSTW2008 NNLO PDF sets [73] and the renormalisation and factorisation scales are varied by factors of two to derive the uncertainties due to higherorder QCD corrections. For Z/γ * , an additional systematic uncertainty is attributed to electroweak corrections [40]. The total uncertainties on the Z/γ * and tt cross sections are ±7% and +10/-11%, respectively. The W ± W ∓ cross-section uncertainty is determined using the same scale and PDF variations as for W Z and ZZ, yielding a total uncertainty of ±12%.
The systematic uncertainty on the background from non-prompt leptons is discussed in section 5.3. It includes a component due to the systematic uncertainties associated with the measurement of the ratio f and a component due to the statistical uncertainty on the number of anti-selected leptons. For m(ℓ ± ℓ ± ) > 15 GeV it is ±33% for the µµ, ±31% for the eµ, and ±28% for the ee final states, and it increases to nearly ±100% at higher masses for all final states.
The probability of assigning the wrong charge to the lepton and its uncertainty is discussed in section 5.2. For the µµ final state, charge misidentification results in an uncertainty of +4.9 −0.0 events for m(ℓ ± ℓ ± ) > 15 GeV and +1.7 −0.0 events for m(ℓ ± ℓ ± ) > 400 GeV. For electrons, the uncertainty on the charge-flip background due to uncertainty in the charge mismeasurement probability ranges from ±15% to ±23% for the same mass thresholds. The uncertainty on the W γ and Zγ backgrounds from this source is ±(15-18)% depending on m(ℓ ± ℓ ± ).
Statistical uncertainties due to the limited size of the background MC samples are also considered. Systematic uncertainties on different physics processes due to a given source are assumed to be 100% correlated.

Comparison of data to the background expectation
The predicted numbers of background pairs are compared to the observed numbers of likesign lepton pairs in table 1. For the e ± e ± (e ± µ ± ) final state, about 30% (50%) of the background is from prompt leptons in all mass bins. The rest arises from leptons from hadronic decays, charge flips, or photon conversions. For the µ ± µ ± final state, the prompt background constitutes about 83% of the total background for m(µ ± µ ± ) > 15 GeV, rising to nearly 100% of the total background for m(µ ± µ ± ) > 400 GeV. The overall uncertainty on the background is about ±15% at low mass and ±30% at high mass. Table 2 shows the data compared to the background expectation separately for ℓ + ℓ + and ℓ − ℓ − events for the ee, µµ, and eµ final states. The background is higher for the ℓ + ℓ + final state due to the larger cross section in pp collisions for W + than W − bosons produced in association with γ, Z, or hadrons. Table 1. Expected and observed numbers of pairs of isolated like-sign leptons for various cuts on the dilepton invariant mass, m(ℓ ± ℓ ± ). The uncertainties shown are the quadratic sum of the statistical and systematic uncertainties. The prompt background contribution includes the W Z, ZZ, W ± W ± , ttW , and ttZ processes. When zero events are predicted, the uncertainty corresponds to the 68% confidence level upper limit on the prediction.

Sample
Number of electron pairs with m(e ± e ± ) > 15 GeV > 100 GeV > 200 GeV > 300 GeV > 400 GeV   The level of agreement between the data and the background expectation is evaluated using 1−CL b [74], defined as the one-sided probability of the background-only hypothesis to fluctuate up to at least the number of observed events. Statistical and systematic uncertainties and their correlations are fully considered for this calculation. The largest upward deviation, observed for m(µ − µ − ) > 100 GeV, occurs about 8% of the time in background-only pseudo-experiments for this mass bin. Figure 1 shows the dilepton invariant mass spectra for the e ± e ± , µ ± µ ± , and e ± µ ± final states, and figure 2 shows the p T of the leading leptons. Good agreement with the Table 2. Expected and observed numbers of positively-and negatively-charged lepton pairs for different lower limits on the dilepton invariant mass, m(ℓ ± ℓ ± ). The uncertainties shown are the quadratic sum of the statistical and systematic uncertainties.

Sample
Number of lepton pairs with m(ℓ ± ℓ ± ) > 15 GeV > 100 GeV > 200 GeV > 300 GeV > 400 GeV e + e + pairs Sum of backgrounds 208 ± 28 112 ± 14 28.6 ± 4.0 8.   Figure 1. Invariant mass distributions for (a) e ± e ± , (b) µ ± µ ± , and (c) e ± µ ± pairs passing the full event selection. The data are shown as closed circles. The stacked histograms represent the backgrounds composed of pairs of prompt leptons from SM processes, pairs with at least one nonprompt lepton, and for the electron channels, backgrounds arising from charge misidentification and photon conversions. Pairs in the ee channel with invariant masses between 70 GeV and 110 GeV are excluded because of the large background from charge misidentification in Z → e ± e ∓ decays. The last bin is an overflow bin. uncertainties are incorporated into the likelihoods as nuisance parameters with Gaussian probability density functions. For the inclusive ℓ ± ℓ ± final states, the upper limit ranges from 168 pairs for m(eµ) > 15 GeV to 4.8 pairs for m(µµ) > 400 GeV.
The limit on the number of lepton pairs can be translated to an upper limit on the cross section measured in a given region of phase space (referred to here as the fiducial region), σ fid 95 , via where ε fid is the efficiency for detecting events within the fiducial region and Ldt is the integrated luminosity of 4.7 fb −1 .
The fiducial region definition is based on MC generator information such that it is independent of the ATLAS detector environment. The value of ε fid generally depends on the new physics process, e.g. the number of leptons in the final state passing the kinematic  Figure 2. Leading lepton p T distributions for (a) e ± e ± , (b) µ ± µ ± , and (c) e ± µ ± pairs passing the full event selection. The data are shown as closed circles. The stacked histograms represent the backgrounds composed of pairs of prompt leptons from SM processes, pairs with at least one non-prompt lepton, and for the electron channels, backgrounds arising from charge misidentification and photon conversions. The last bin is an overflow bin. selection criteria or the number of jets that may affect the lepton isolation. In order to minimise this dependence, the definition of the fiducial region is closely related to the analysis selection. In the MC generator, electrons or muons are selected as stable particles that originate from a W or Z boson, from a τ lepton, or from an exotic new particle (e.g. a H ±± or a right-handed W ). The generated electrons and muons are required to satisfy the p T , η, and isolation requirements listed in table 3, which mirror the selection requirements described in section 4. The generator-level track isolation, p cone∆R iso T , is defined as the scalar sum of all stable charged particles with p T > 1 GeV in a cone of size ∆R iso around the electron or muon, excluding the lepton itself. The requirement on isolation energy (E cone∆R iso T ) placed in the electron selection is not emulated in the fiducial region definition because the measurement of isolation energy in the calorimeter has a poor resolution. Pairs of like-sign leptons must have m(ℓ ± ℓ ± ) > 15 GeV and for e ± e ± , the mass range 70-110 GeV is additionally excluded.
The fiducial efficiency, ε fid , is defined as the fraction of lepton pairs passing this se- lection that also satisfy the experimental selection criteria described in section 4. It is determined for a variety of new physics models: H ±± bosons with mass between 50 GeV and 1000 GeV [48], fourth generation quarks decaying to W t with masses of 300-500 GeV, like-sign top-quark production via a contact interaction [53], and right-handed W bosons decaying to ℓ ± ℓ ± via a right-handed neutrino with m(W R ) = 800-2500 GeV [49,50]. These models are chosen as they cover a broad range of jet multiplicities and lepton p T spectra.
For m(ℓ ± ℓ ± ) > 15 GeV in the e ± e ± channel, the fiducial efficiencies range from 43% for models with low-p T leptons to 65% for models with high-p T leptons. The primary reason for this dependence is that the electron identification efficiency varies by about 15% over the relevant p T range [60]. The model dependence introduced by not emulating the calorimeter isolation in the definition of the fiducial region is < 1%. For the e ± µ ± channel, ε fid ranges from 55% to 70%, and for the µ ± µ ± final state it varies between 59% and 72%. For the higher dilepton mass thresholds the efficiencies are slightly larger than for the lower mass thresholds. The efficiencies are also derived for ℓ + ℓ + and ℓ − ℓ − pairs separately and found to be independent of the charge. For the same new physics models, the fraction of events satisfying the experimental selection originating from outside the fiducial region ranges from < 1% to about 9%, depending on the final state and the model considered.
To derive the upper limit on the fiducial cross section, the lowest efficiency values are taken for all mass thresholds, i.e. 43% for the e ± e ± , 55% for the e ± µ ± , and 59% for the µ ± µ ± analysis. The 95% C.L. upper limits on the cross section, σ fid 95 , are calculated using equation 8.1 and given in table 4. The limits range from 64 fb to 1.7 fb for the inclusive analysis depending on the mass cut and the final state, and are generally within 1σ of the expected limits. Upper limits on the ℓ + ℓ + and ℓ − ℓ − cross sections are also derived, using the same efficiencies as for the inclusive limits.

Conclusions
A search for anomalous production of like-sign lepton pairs has been presented using 4.7 fb −1 of pp collision data at √ s = 7 TeV recorded by the ATLAS experiment at the LHC. The data are found to agree with the background expectation in e ± e ± , e ± µ ± , and µ ± µ ± final states both in overall rate and in the kinematic distributions. The data are used to constrain new physics contributions to like-sign lepton pairs within a fiducial region of Table 4. Upper limits at 95% C.L. on the fiducial cross section for ℓ ± ℓ ± pairs from non-SM physics. The expected limits and their 1σ uncertainties are given, as well as the observed limits in data, for the ee, eµ, and µµ final state inclusively and separated by charge.